Datasets:

bzantium

/

LongPerplexity

like 0

this is the second in the series of four papers initiated by @xcite and devoted to geometry behind directed networks on surfaces , with a particular emphasis on their poisson properties . in @xcite , we concentrated on postnikov s construction @xcite that uses weighted directed planar graphs to parametrize cells in grassmannians . we found that the space of edge weights of networks in a disk can be endowed with a natural family of poisson brackets ( that we called _ universal _ ) that `` respects '' the operation of concatenation of diagrams . we have shown that , under postnikov s parametrization , these poisson brackets induce a two - parameter family of poisson brackets on the grassmannian . every poisson bracket in this family is compatible ( in the sense of @xcite ) with the cluster algebra on the grassmannian described in @xcite and , on the other hand , endows the grassmannian with a structure of a poisson homogeneous space with respect to the natural action of the general linear group equipped with an r - matrix poisson - lie structure . as was announced in @xcite , the current paper builds a parallel theory for directed weighted networks in an annulus ( or , equivalently , on a cylinder ) . first , we have to modify the definition of the boundary measurement map , whose image now consists of rational valued matrix functions of an auxiliary parameter @xmath0 associated with the notion of a _ cut _ ( see section 1 ) . we then show that the analogue of postnikov s construction leads to a map into the space of loops in the grassmannian . universal poisson brackets for networks in an annulus are defined in exactly the same way as in the case of a disk . we show that they induce a two - parameter family of poisson brackets on rational - valued boundary measurement matrices . in particular , when sources and sinks belong to distinct circles bounding the annulus , one of the generators of these family coincides with the sklyanin r - matrix bracket associated with the trigonometric solution of the classical yang - baxter equation in @xmath1 . moreover , we prove that the two - parameter family of poisson brackets can be further pushed - forward to the space of loops in the grassmannian . in proving the latter , we departed from the approach of @xcite where the similar result was obtained via a more or less straightforward calculation . such an approach would have been too cumbersome in our current setting . instead , we found a way to utilize so - called _ face weights _ and their behavior under _ path - reversal maps_. the forthcoming third paper in this series @xcite focuses on particular graphs in an annulus that can be used to introduce a cluster algebra structure on the coordinate ring of the space of normalized rational functions in one variable . this space is birationally equivalent , via the moser map @xcite , to any minimal irreducible coadjoint orbit of the group of upper triangular matrices associated with a coxeter element of the permutation group . in this case , the poisson bracket compatible with the cluster algebra structure coincides with the quadratic poisson bracket studied in @xcite in the context of toda flows on minimal orbits . we show that cluster transformations serve as bcklund - darboux transformations between different minimal toda flows . the fourth paper @xcite solves , in the case of graphs in an annulus with one source and one sink , the inverse problem of restoring the weights from the image of the generalized postnikov map . in the case of arbitrary planar graphs in a disk , this problem was completely solved by postnikov @xcite who proved that for a fixed minimal graph , the space of weights modulo gauge action is birational to its image . to the contrary , already for simplest graphs in an annulus , the corresponding map can only be shown to be finite . the original application of directed weighted planar networks was in the study of total positivity , both in @xmath2 @xcite and in grassmannians @xcite . we do not address this issue for networks in an annulus . it has been studied in a recent preprint @xcite . the paper is organized as follows . in section [ netsbm2 ] , we introduce a notion of a perfect network in an annulus and associate with every such network a _ matrix of boundary measurements_. each boundary measurement is shown to be a rational function in edge weights and in an auxiliary parameter @xmath0 , see corollary [ sfree3 ] . besides , we define the space of face and trail weights , a generalization of the space of face weights studied in @xcite for the case of networks in a disk , and provide its cohomological interpretation , see section [ cohomol ] . in section [ psrat ] , we characterize all _ universal _ poisson brackets on the space of edge weights of a given network that respect the natural operation of concatenation of networks , see proposition [ 6param ] . furthermore , we establish that the family of universal brackets induces a linear two - parameter family of poisson brackets on the space of face and trail weights , see theorem [ psviay ] , and hence on boundary measurement matrices , see theorem [ psr ] . this family depends on a mutual location of sources and sinks , but not on the network itself . we provide an explicit description of this family in propositions [ psre ] and [ psre2 ] . an important tool in the proof of theorem [ psr ] is the realization theorem , see theorem [ realiz ] , that claims that any rational matrix function can be realized as the boundary measurement matrix of a network with a given set of sources and sinks . finally , if the sources and the sinks are separated , that is , all sources belong to one of the bounding circles of the annulus , and all sinks to the other bounding circle , one of the generators of the 2-parametric family can be identified with the r - matrix sklyanin bracket corresponding to the trigonometric r - matrix , see theorem [ sklyaninbr ] . in section [ psongrloop ] , the boundary measurement map defined by a network with @xmath3 sources , @xmath4 sinks and @xmath5 boundary vertices on the outer boundary circle is extended to the grassmannian boundary measurement map into the space @xmath6 of grassmannian loops . the poisson family on boundary measurement matrices allows us to equip @xmath6 with a two - parameter family of poisson brackets @xmath7 in such a way that for any choice of a universal poisson bracket on edge weights there is a unique member of @xmath7 that makes the grassmannian boundary measurement map poisson , see theorem [ pslgr ] . this latter family depends only on the number of sources and sinks and on the distribution of the boundary vertices between the bounding circles of the annulus . the main tool in the proof of theorem [ pslgr ] is the path reversal operation on networks and its properties , see theorem [ pathrev ] . let @xmath8 be a directed planar graph drawn inside an annulus with the vertex set @xmath9 and the edge set @xmath10 . exactly @xmath11 of its vertices are located on the boundary circles of the annulus and are called _ boundary vertices _ ; @xmath12 of them lie on the outer circle , and @xmath13 on the inner circle . the graph is considered up to an isotopy relative to the boundary ( with fixed boundary vertices ) . each boundary vertex is marked as a source or a sink . a _ source _ is a vertex with exactly one outcoming edge and no incoming edges . _ sinks _ are defined in the same way , with the direction of the single edge reversed . all the internal vertices of @xmath14 have degree @xmath15 and are of two types : either they have exactly one incoming edge , or exactly one outcoming edge . the vertices of the first type are called ( and shown in figures ) _ white _ , those of the second type , _ black_. a _ cut _ @xmath16 is an oriented non - selfintersecting curve starting at a _ base point _ on the inner circle and ending at a base point on the outer circle considered up to an isotopy relative to the boundary ( with fixed endpoints ) . we assume that the base points of the cut are distinct from the boundary vertices of @xmath14 . for an arbitrary oriented curve @xmath17 with endpoints not lying on the cut @xmath16 we denote by @xmath18 the algebraic intersection number of @xmath17 and @xmath16 . recall that each transversal intersection point of @xmath17 and @xmath16 contributes to this number @xmath19 if the oriented tangents to @xmath17 and @xmath16 at this point form a positively oriented basis , and @xmath20 otherwise . non - transversal intersection points are treated in a similar way . let @xmath21 be independent variables . perfect planar network in an annulus _ @xmath22 is obtained from a graph @xmath14 equipped with a cut @xmath16 as above by assigning a weight @xmath23 to each edge @xmath24 . below we occasionally write `` network '' instead of `` perfect planar network in an annulus '' . each network defines a rational map @xmath25 ; the _ space of edge weights _ @xmath26 is defined as the intersection of the image of @xmath27 with @xmath28 . in other words , a point in @xmath26 is a graph @xmath14 as above with edges weighted by nonzero real numbers obtained by specializing the variables @xmath21 in the expressions for @xmath29 . a _ path _ @xmath30 in @xmath31 is an alternating sequence @xmath32 of vertices and edges such that @xmath33 for any @xmath34 $ ] . sometimes we omit the names of the vertices and write @xmath35 . a path is called a _ cycle _ if @xmath36 and a _ simple cycle _ if additionally @xmath37 for any other pair @xmath38 . to define the weight of a path we need the following construction . consider a closed oriented polygonal plane curve @xmath39 . let @xmath40 and @xmath41 be two consequent oriented segments of @xmath39 , and let @xmath42 be their common vertex . we assume for simplicity that for any such pair @xmath43 , the cone spanned by @xmath40 and @xmath41 is not a line ; in other words , if @xmath40 and @xmath41 are collinear , then they have the same direction . observe that since @xmath39 is not necessary simple , there might be other edges of @xmath39 incident to @xmath42 . let @xmath44 be an arbitrary oriented line . define @xmath45 in the following way : @xmath46 if the directing vector of @xmath44 belongs to the interior of the cone spanned by @xmath40 and @xmath41 and @xmath47 otherwise . define @xmath48 as the sum of @xmath49 over all pairs of consequent segments in @xmath39 . it follows immediately from theorem 1 in @xcite that @xmath48 does not depend on @xmath44 , provided @xmath44 is not collinear to any of the segments in @xmath39 . the common value of @xmath48 for different choices of @xmath44 is denoted by @xmath50 and called the _ concordance number _ of @xmath39 . in fact , @xmath50 equals @xmath51 the rotation number of @xmath39 ; the definition of the latter is similar , but more complicated . in what follows we assume without loss of generality that @xmath31 is drawn in such a way that all its edges and the cut are smooth curves . moreover , any simple path in @xmath31 is a piecewise - smooth curve with no cusps , at any boundary vertex of @xmath31 the edge and the circle intersect transversally , and the same holds for the cut at both of its base points . given a path @xmath30 between a source @xmath52 and a sink @xmath53 , we define a closed piecewise - smooth curve @xmath54 in the following way : if both @xmath52 and @xmath53 belong to the same circle , @xmath54 is obtained by adding to @xmath30 the path between @xmath53 and @xmath52 that goes counterclockwise along the boundary of the corresponding circle . otherwise , if @xmath52 and @xmath53 belong to distinct circles , @xmath54 is obtained by adding to @xmath30 the path that starts at @xmath53 , goes counterclockwise along the corresponding circle to the base point of the cut , follows the cut to the other base point and then goes counterclockwise along the other circle up to @xmath52 . clearly , the obtained curve @xmath54 does not have cusps , so its concordance number @xmath55 can be defined in a straightforward manner via polygonal approximation . finally the weight of @xmath30 is defined as @xmath56 where @xmath0 is an auxiliary independent variable . occasionally , it will be convenient to assume that the internal vertices of @xmath14 do not lie on the cut and to rewrite the above formula as @xmath57 where @xmath58 are _ modified edge weights_. observe that the weight of a path is a relative isotopy invariant , while modified edge weights are not . the weight of an arbitrary cycle in @xmath31 is defined in the same way via the concordance number of the cycle . if edges @xmath59 and @xmath60 in @xmath30 coincide and @xmath61 , the path @xmath30 can be decomposed into the path @xmath62 and the cycle @xmath63 . clearly , @xmath64 , and hence @xmath65 an example of a perfect planar network in an annulus is shown in fig . [ fig : pergrann ] on the left . it has two sources , @xmath66 on the outer circle and @xmath53 on the inner circle , and one sink @xmath52 on the inner circle . each edge @xmath59 is labeled by its weight . the cut is shown by the dashed line . the same network is shown in fig . [ fig : pergrann ] on the right ; it differs from the original picture by an isotopic deformation of the cut . consider the path @xmath67 from @xmath66 to @xmath52 . its algebraic intersection number with the cut equals @xmath68 ( in the left picture , two intersection points contribute @xmath19 each , and two other intersection points contribute @xmath20 each ; in the right picture there are no intersection points ) . the concordance number of the corresponding closed curve @xmath69 equals @xmath19 . therefore , ( [ weightviapath ] ) gives @xmath70 . on the other hand , the modified weights for the left picture are given by @xmath71 , @xmath72 , @xmath73 , @xmath74 , and hence , computation via ( [ weightviaedge ] ) gives @xmath75 . finally , the modified weights for the relevant edges in the right picture coincide with the original weights , and we again get the same result . consider the path @xmath76 from @xmath53 to @xmath52 . its algebraic intersection number with the cut equals @xmath20 , and the concordance number of the corresponding closed curve @xmath77 equals @xmath19 . therefore , @xmath78 . the same result can be obtained by using modified weights . finally , consider the path @xmath79 from @xmath66 to @xmath52 . clearly , @xmath80 . the path @xmath81 can be decomposed into the path @xmath82 as above and a cycle @xmath83 with the weights @xmath70 and @xmath84 , hence relation ( [ offcycle ] ) yields the same expression for @xmath85 as before . let us see how moving a base point of the cut affects the weights of paths . let @xmath22 and @xmath86 be two networks with the same graph and the same weights , and assume that the cuts @xmath16 and @xmath87 are not isotopic . more exactly , let us start moving a base point of the cut in the counterclockwise direction . assume that @xmath66 is the first boundary vertex in the counterclockwise direction from the base point of @xmath16 that is being moved . clearly , nothing is changed while the base point and @xmath66 do not interchange . let @xmath87 be the cut obtained after the interchange , and assume that no other interchanges occured . then the relation between the weight @xmath88 of a path @xmath30 in @xmath31 and its weight @xmath89 in @xmath90 is given by the following proposition . [ movecut ] for @xmath31 and @xmath90 as above , @xmath91 where @xmath92 equals @xmath68 if the endpoints of @xmath30 lie on the same circle and @xmath19 otherwise , and @xmath93 the proof is straightforward . for example , consider the networks @xmath31 and @xmath90 shown in fig . [ fig : pergrann2 ] . the base point of the cut lying on the inner circle interchanges positions with the sink @xmath52 . the path @xmath82 from the previous example goes from the outer circle to the inner circle , so @xmath94 ; besides , @xmath52 is the sink of @xmath82 , so @xmath95 . therefore , by proposition [ movecut ] , @xmath96 , which coincides with the value obtained via ( [ weightviaedge ] ) . the path @xmath97 from the previous example starts and ends at the inner circle , so @xmath98 . besides , @xmath99 , and hence , by proposition [ movecut ] , @xmath100 ; once again , this coincides with the value obtained via ( [ weightviaedge ] ) . given a perfect planar network in an annulus as above , we label its boundary vertices @xmath101 in the following way . the boundary vertices lying on the outer circle are labeled @xmath102 in the counterclockwise order starting from the first vertex that follows after the base point of the cut . the boundary vertices lying on the inner circle are labeled @xmath103 in the clockwise order starting from the first vertex that follows after the base point of the cut . for example , for the network @xmath31 in fig . [ fig : pergrann2 ] , the boundary vertices are labeled as @xmath104 , @xmath105 , @xmath106 , while for the network @xmath90 , one has @xmath104 , @xmath107 , @xmath108 . the number of sources lying on the outer circle is denoted by @xmath109 , and the corresponding set of indices , by @xmath110 $ ] ; the set of the remaining @xmath111 indices is denoted by @xmath112 . similarly , the number of sources lying on the inner circle is denoted by @xmath113 , and the corresponding set of indices , by @xmath114 $ ] ; the set of the remaining @xmath115 indices is denoted by @xmath116 . finally , we denote @xmath117 and @xmath118 ; the cardinalities of @xmath119 and @xmath120 are denoted @xmath121 and @xmath122 . given a source @xmath123 , @xmath124 , and a sink @xmath125 , @xmath126 , we define the _ boundary measurement _ @xmath127 as the sum of the weights of all paths starting at @xmath123 and ending at @xmath125 . assume first that the weights of the paths are calculated via ( [ weightviaedge ] ) . the boundary measurement thus defined is a formal infinite series in variables @xmath128 , @xmath24 . the following proposition holds true in @xmath129 $ ] . [ sfree2 ] let @xmath31 be a perfect planar network in an annulus , then each boundary measurement in @xmath31 is a rational function in the modified weights @xmath128 , @xmath24 . the proof by induction on the number of internal vertices literally follows the proof of the similar statement in @xcite . the only changes are that modified weights are used instead of original weights and that the counterclockwise cyclic order @xmath130 is replaced by the cyclic order @xmath131 induced by the labeling . taking into account the definition of the modified weights , we immediately get the following corollary . [ sfree3 ] let @xmath31 be a perfect planar network in an annulus , then each boundary measurement in @xmath31 is a rational function in the parameter @xmath0 and the weights @xmath29 , @xmath24 . for example , the boundary measurement @xmath132 in the network @xmath90 shown on fig . [ fig : pergrann2 ] equals @xmath133 boundary measurements can be organized into a @xmath134 _ boundary measurement matrix _ @xmath135 exactly as in the case of planar networks in the disk : let @xmath136 and @xmath137 , then @xmath138 , @xmath139 $ ] , @xmath140 $ ] , where @xmath141 . let @xmath142 stand for the space of real rational @xmath134 matrix functions in one variable . then each network @xmath31 defines a map @xmath143 given by @xmath135 and called the _ boundary measurement map _ corresponding to @xmath31 . the boundary measurement matrix has a block structure @xmath144 where @xmath145 is @xmath146 and @xmath147 is @xmath148 . moving a base point of the cut changes the weights of paths as described in proposition [ movecut ] , which affects @xmath135 in the following way . define @xmath149 then interchanging the base point of the cut with @xmath150 implies transformation @xmath151 if @xmath150 is a source and @xmath152 if @xmath150 is a sink , while interchanging the base point of the cut with @xmath153 yields @xmath154 if @xmath153 is a source and @xmath155 if @xmath150 is a sink . note that @xmath156 and @xmath157 are @xmath158 in the first case , @xmath159 in the second case , @xmath160 in the third case and @xmath161 in the fourth case . let @xmath22 be a perfect network . consider the @xmath162-module @xmath163 generated by the edges of @xmath14 . clearly , points of @xmath26 can be identified with the elements of @xmath164 via @xmath165 , where @xmath166 is the abelian multiplicative group @xmath167 . further , consider the @xmath162-module @xmath168 generated by the vertices of @xmath14 and its @xmath162-submodule @xmath169 generated by the internal vertices . an arbitrary element @xmath170 acts on @xmath26 as follows : if @xmath171 then @xmath172 therefore , the weight of a path between the boundary vertices @xmath123 and @xmath125 is multiplied by @xmath173 . it follows that the _ gauge group _ @xmath174 , which preserves the weights of all paths between boundary vertices , consists of all @xmath170 such that @xmath175 for any boundary vertex @xmath66 , and can be identified with @xmath176 . thus , the boundary measurement map @xmath177 factors through the quotient space @xmath178 as follows : @xmath179 , where @xmath180 is the projection and @xmath181 is a map @xmath182 . the space @xmath183 is called the _ space of face and trail weights _ for the following reasons . first , by considering the cochain complex @xmath184 with the coboundary operator @xmath185 defined by @xmath186 for @xmath171 , we can identify @xmath183 with the first relative cohomology @xmath187 of the complex , where @xmath188 is the set of all boundary vertices of @xmath14 . second , consider a slightly more general situation , when the annulus is replaced by an arbitrary riemann surface @xmath189 with the boundary @xmath190 , and @xmath14 is embedded into @xmath189 in such a way that all vertices of degree 1 belong to @xmath191 ( boundary vertices ) . then the exact sequence of relative cohomology with coefficients in @xmath166 gives @xmath192 evidently , @xmath193 . next , @xmath194 , since each connected component of @xmath14 is connected to at least one connected component of @xmath190 , and hence @xmath195 the space @xmath196 can be described as follows . graph @xmath14 divides @xmath189 into a finite number of connected components called _ faces_. the boundary of each face consists of edges of @xmath14 and , possibly , of several arcs of @xmath190 . a face is called _ bounded _ if its boundary contains only edges of @xmath14 and _ unbounded _ otherwise . given a face @xmath197 , we define its _ face weight _ @xmath198 as the function on @xmath26 that assigns to the edge weights @xmath29 , @xmath24 , the value @xmath199 where @xmath200 if the direction of @xmath201 is compatible with the counterclockwise orientation of the boundary @xmath202 and @xmath203 otherwise . it follows immediately from the definition that face weights are invariant under the gauge group action , and hence are functions on @xmath196 , and , moreover , form a basis in the space of such functions . consider now the space @xmath204 . if @xmath189 is a disk , then @xmath205 , and hence @xmath183 and @xmath196 coincide . this case was studied in @xcite , and the space @xmath183 was called there the _ space of face weights_. if @xmath189 is an annulus , there are two possible cases . indeed , @xmath206 . the dimension of @xmath207 is either @xmath208 or @xmath19 , depending on the existence of a trail connecting the components of @xmath190 . here trail _ is a sequence @xmath209 of vertices such that either @xmath210 or @xmath211 is an edge in @xmath14 for all @xmath212 $ ] an the endpoints @xmath213 and @xmath214 are boundary vertices of @xmath14 . given a trail @xmath215 , the _ trail weight _ @xmath216 is defined as @xmath217 where @xmath218 clearly , the trail weights are invariant under the action of the gauge group . if @xmath14 does not contain a trail connecting the inner and the outer circles , then @xmath219 , and hence @xmath205 . otherwise , @xmath220 , and hence @xmath221 . the functions on @xmath204 are generated by the weight of any connecting trail . the construction of a poisson structure on the space @xmath26 for perfect planar networks in an annulus is a straightforward extension of the corresponding construction for the case of the disk studied in @xcite . let g be a directed planar graph in an annulus as described in section [ pandw2 ] . a pair @xmath222 is called a _ flag _ if @xmath42 is an endpoint of @xmath201 . to each internal vertex @xmath42 of @xmath14 we assign a 3-dimensional space @xmath223 with coordinates @xmath224 . we equip each @xmath223 with a poisson bracket @xmath225 . it is convenient to assume that the flags involving @xmath42 are labeled by the coordinates , as shown on figure [ fig : trefoils ] . , height=56 ] besides , to each boundary vertex @xmath125 of @xmath14 we assign a 1-dimensional space @xmath226 with the coordinate @xmath227 ( in accordance with the above convention , this coordinate labels the unique flag involving @xmath125 ) . define @xmath228 to be the direct sum of all the above spaces ; thus , the dimension of @xmath228 equals twice the number of edges in @xmath14 . note that @xmath228 is equipped with a poisson bracket @xmath229 , which is defined as the direct sum of the brackets @xmath225 ; that is , @xmath230 whenever @xmath231 and @xmath232 are not defined on the same @xmath223 . we say that the bracket @xmath229 is _ universal _ if each of @xmath225 depends only on the color of the vertex @xmath42 . define the weights @xmath29 by @xmath233 provided the flag @xmath222 is labeled by @xmath234 and the flag @xmath235 is labeled by @xmath236 . in other words , the weight of an edge is defined as the product of the weights of the two flags involving this edge . therefore , in this case the space of edge weights @xmath237 coincides with the entire @xmath28 , and the weights define a _ weight map _ @xmath238 . we require the pushforward of @xmath229 to @xmath28 by the weight map to be a well defined poisson bracket ; this can be regarded as an analog of the poisson lie property for groups . [ 6param ] universal poisson brackets @xmath229 such that the weight map @xmath27 is poisson form a @xmath239-parametric family defined by relations @xmath240 , i\ne j,\ ] ] at each white vertex @xmath42 and @xmath241 , i\ne j,\ ] ] at each black vertex @xmath42 . indeed , let @xmath42 be a white vertex , and let @xmath242 and @xmath243 be the two outcoming edges . by definition , there exist @xmath244 $ ] , @xmath38 , such that @xmath245 , @xmath246 . therefore , @xmath247 where @xmath248 stands for the pushforward of @xmath229 . recall that the poisson bracket in @xmath223 depends only on @xmath249 , @xmath250 and @xmath251 . hence the only possibility for the right hand side of the above relation to be a function of @xmath29 and @xmath252 occurs when @xmath253 , as required . black vertices are treated in the same way . the 6-parametric family of universal poisson brackets described in proposition [ 6param ] induces a 6-parametric family of poisson brackets @xmath254 on @xmath26 . our next goal is to study the pushforward of this family to @xmath183 by the map @xmath232 . [ psviay ] the 6-parametric family @xmath254 induces a 2-parametric family of poisson brackets @xmath255 on @xmath183 with parameters @xmath256 and @xmath257 given by @xmath258 in what follows it will be convenient to assume that boundary vertices are colored in _ gray_. let @xmath171 be a directed edge . we say that the flag @xmath235 is _ positive _ , and the flag @xmath222 is _ negative_. the color of a flag is defined as the color of the vertex participating in the flag . consider first a bracket of two face weights . let @xmath197 and @xmath259 be two faces of @xmath31 . we say that a flag @xmath222 is _ common _ to @xmath197 and @xmath259 if both @xmath42 and @xmath201 belong to @xmath260 . clearly , the bracket @xmath261 can be calculated as the sum of the contributions of all flags common to @xmath197 and @xmath259 . assume that @xmath222 is a positive white flag common to @xmath197 and @xmath259 , see fig . [ fig : comflag ] . then @xmath262 and @xmath263 , where @xmath264 are the weights of flags involving @xmath42 and @xmath265 . therefore , by ( [ 6parw ] ) , the contribution of @xmath222 equals @xmath266 , which by ( [ cond ] ) equals @xmath267 . assume now that @xmath222 is a negative white flag common to @xmath197 and @xmath259 , see fig . [ fig : comflag ] . in this case @xmath268 and @xmath263 , so the contribution of @xmath222 equals @xmath269 . in a similar way one proves that the contribution of a positive black flag common to @xmath197 and @xmath259 equals @xmath270 , and the contribution of a negative black flag common to @xmath197 and @xmath259 equals @xmath271 . finally , the contributions of positive and negative gray flags are clearly equal to zero . therefore , the brackets @xmath261 form a 2-parametric family with parameters @xmath256 and @xmath257 defined by ( [ cond ] ) . the case of a bracket @xmath272 is treated in a similar way . fix an arbitrary pair of partitions @xmath273 $ ] , @xmath274 , @xmath275 $ ] , @xmath276 , and denote @xmath277 , @xmath278 . let @xmath279 stand for the set of all perfect planar networks in an annulus with the sources @xmath123 , @xmath280 and sinks @xmath125 , @xmath281 , on the outer circle , sources @xmath123 , @xmath282 and sinks @xmath125 , @xmath283 , on the inner circle , and edge weights @xmath29 defined by ( [ connect ] ) . we assume that the space of edge weights @xmath284 is equipped with the poisson bracket @xmath248 obtained as the pushforward of the 6-parametric family @xmath285 described in proposition [ 6param ] . [ psr ] there exists a @xmath208-parametric family of poisson brackets @xmath286 on @xmath142 with the following property : for any choice of parameters @xmath287 , @xmath288 in ( [ 6parw ] ) , ( [ 6parb ] ) this family contains a unique poisson bracket on @xmath142 such that for any network @xmath289 the map @xmath290 is poisson . first of all , we use the factorization @xmath291 to decrease the number of parameters . by theorem [ psviay ] , it suffices to consider the 2-parametric family @xmath292 and @xmath293 with @xmath256 and @xmath257 defined by ( [ cond ] ) , instead of the 6-parametric family ( [ 6parw ] ) , ( [ 6parb ] ) the rest of the proof consists of two major steps . first , we compute the induced poisson bracket on the image of the boundary measurement map . more exactly , we show that the bracket @xmath254 of any pair of pullbacks of coordinate functions on the image can be expressed in terms of pullbacks of other coordinate functions , and that for fixed @xmath294 these expressions do not depend on the network @xmath289 . second , we prove that any rational matrix function belongs to the image of the boundary measurement map ( for a sufficiently large @xmath289 ) , and therefore @xmath254 induces @xmath286 on @xmath142 . this approach allows us to circumvent technical difficulties one encounters when attempting to check the jacobi identity in the image in a straightforward way . to compute the induced poisson bracket on the image of the boundary measurement map , we consider coordinate functions @xmath295 that assign to any @xmath296 its value at point @xmath215 . given a pair of two matrix entries , it suffices to calculate the bracket between arbitrary pair of functions @xmath297 and @xmath298 defined on two copies of @xmath299 representing these entries . since the pullback of @xmath297 is the corresponding component of @xmath300 , we have to deal with expressions of the form @xmath301 . to avoid overcomplicated formulas we consider separately two particular representatives of the family ( [ 2parw ] ) , ( [ 2parb ] ) : 1 ) @xmath302 , and 2 ) @xmath303 . any member of the family can be represented as a linear combination of the above two . denote by @xmath304 the member of the 2-parametric family ( [ 2parw ] ) , ( [ 2parb ] ) corresponding to the case @xmath302 . besides , define @xmath305 ; clearly , @xmath306 is closely related to @xmath307 defined and studied in @xcite . the bracket induced by @xmath304 on the image of the boundary measurement map is completely described by the following statement . [ psre ] ( i ) let @xmath308 $ ] and @xmath309 , then @xmath310 where @xmath311 \(ii ) let @xmath312 $ ] and @xmath313 , then @xmath314 where @xmath315 \(iii ) let @xmath316 , then @xmath317 \(iv ) let @xmath318 , then @xmath319 let us first make sure that relations ( [ psre1])([psre4 ] ) indeed allow to compute @xmath320 for any @xmath321 $ ] and @xmath322 $ ] . this is done by employing the following three techniques : moving a base point of the cut ; reversing the direction of the cut ; reversing the orientation of boundary circles . the first of the above techniques has been described in detail in section [ pandw2 ] . for example , let @xmath323 . this case is not covered explicitly by relations ( [ psre1])([psre4 ] ) . consider the network @xmath90 obtained from @xmath31 by moving the base point of the cut on the outer circle counterclockwise and interchanging it with @xmath324 . in this new network one has @xmath325 , so the conditions of proposition [ psre](i ) are satisfied and ( [ psre1 ] ) yields @xmath326 by lemma [ movecut ] , @xmath327 finally , @xmath328 for any pair of edge weights , so we get @xmath329 reversing the direction of the cut transforms the initial network @xmath31 to a new network @xmath90 ; the graph @xmath14 remains the same , while the labeling of its boundary vertices is changed . namely , the @xmath330 boundary vertices lying on the inner circle are labeled @xmath331 in the clockwise order starting from the first vertex that follows after the base point of the cut . the boundary vertices lying on the outer circle are labeled @xmath332 in the counterclockwise order starting from the first vertex that follows after the base point of the cut . the transformation @xmath333 is better visualized if the network is drawn on a cylinder , instead of an annulus . the boundary circles of a cylinder are identical , and reversing the direction of the cut simply interchanges them . clearly , the boundary measurements in @xmath31 and @xmath90 are related by @xmath334 for any @xmath335 , @xmath336 . besides , @xmath328 for any pair of edge weights . therefore , an expression for @xmath337 via @xmath338 , @xmath339 , @xmath340 , @xmath341 is transformed to the expression for @xmath342 via @xmath343 , @xmath344 , @xmath345 , @xmath346 by the substitution @xmath347 and @xmath348 in coefficients . for example , let @xmath349 . this case is not covered explicitly by relations ( [ psre1])([psre4 ] ) . consider the network @xmath90 obtained from @xmath31 by reversing the direction of the cut . in this new network one has @xmath350 , so the conditions of proposition [ psre](ii ) are satisfied and ( [ psre2 ] ) yields @xmath351 applying the above described rule one gets @xmath352 finally , reversing the orientation of boundary circles also retains the graph @xmath14 and changes the labeling of its boundary vertices . namely , the @xmath353 boundary vertices of @xmath90 lying on the outer circle are labeled @xmath102 in the clockwise order starting from the first vertex that follows after the base point of the cut . the boundary vertices lying on the inner circle are labeled @xmath103 in the counterclockwise order starting from the first vertex that follows after the base point of the cut . the transformation @xmath333 may be visualized as a mirror reflection . clearly , the boundary measurements in @xmath31 and @xmath90 are related by @xmath334 for any @xmath335 , @xmath336 . besides , @xmath354 for any pair of edge weights . therefore , the transformation of the expressions for the brackets differs from the one for the case of cut reversal by factor @xmath20 . for example , let @xmath323 . this case is not covered explicitly by relations ( [ psre1])([psre4 ] ) . consider the network @xmath90 obtained from @xmath31 by reversing the orientation of boundary circles . in this new network one has @xmath355 , so the conditions of proposition [ psre](iv ) are satisfied and ( [ psre4 ] ) yields @xmath356 applying the above described rule one gets @xmath357 elementary , though tedious , consideration of all possible cases reveals that indeed any quadruple @xmath358 can be reduced by the above three transformations to one of the quadruples mentioned in the statement of proposition [ psre ] . it is worth to note that cases ( i ) and ( ii ) are not independent . first , they both apply if @xmath359 ; the expressions prescribed by ( [ psre1 ] ) and ( [ psre2 ] ) are distinct , but yield the same result : @xmath360 besides , the expression for @xmath361 in case ( ii ) can be obtained from the expressions for @xmath362 in case ( i ) by reversing the direction of the cut . however , we think that the above presentation , though redundant , better emphasizes the underlying symmetries of the obtained expressions . the proof of relations ( [ psre1])([psre4 ] ) is similar to the proof of theorem 3.3 in @xcite and is based on the induction on the number of internal vertices in @xmath31 . the key ingredient of the proof is the following straightforward analog of lemma 3.5 from @xcite . consider an arbitrary boundary vertex @xmath123 ( without loss of generality we may assume that @xmath123 lies on the outer circle of the annulus ) and suppose that the neighbor of @xmath123 is a black vertex @xmath363 . denote by @xmath364 the unique vertex in @xmath14 such that @xmath365 , and by @xmath366 the neighbor of @xmath363 distinct from @xmath364 and @xmath123 . create a new network @xmath367 by deleting @xmath123 and the edge @xmath368 from @xmath14 , splitting @xmath363 into one new source @xmath369 and one new sink @xmath370 placed on the outer circle ( so that either @xmath371 or @xmath372 ) and replacing the edges @xmath373 and @xmath374 by new edges @xmath375 and @xmath376 , see figure [ fig : blackforkan ] . we may assume without loss of generality that the cut @xmath16 in @xmath31 does not intersect the edge @xmath377 , and hence @xmath16 remains a valid cut in @xmath367 . ( upper part ) and @xmath378 ( lower part),height=302 ] [ recalcan ] boundary measurements in the networks @xmath31 and @xmath367 are related by @xmath379 in the second formula above , sign @xmath380 corresponds to the cases @xmath381 and sign @xmath382 corresponds to the cases @xmath383 where @xmath130 is the cyclic order @xmath384 . we leave the details of the proof to the interested reader . denote by @xmath385 the member of the 2-parametric family ( [ 2parw ] ) , ( [ 2parb ] ) corresponding to the case @xmath303 . besides , define @xmath386 ; clearly , @xmath387 is closely related to @xmath388 defined and studied in @xcite . the bracket induced by @xmath385 on the image of the boundary measurement map is completely described by the following statement . [ psre2 ] ( i ) let @xmath309 , then @xmath389 where @xmath390 and @xmath391 , @xmath392 are the derivatives of @xmath393 and @xmath394 . \(ii ) let @xmath395 and either @xmath396 or @xmath397 , then @xmath398 \(iii ) let @xmath399 and either @xmath400 or @xmath401 , then @xmath402 the proof is similar to the proof of proposition [ psre ] . we leave the details to the interested reader . it is worth to mention that the bracket induced on @xmath134 matrices via perfect planar networks in a disk , which was studied in @xcite , can be considered as a particular case of ( [ psre1])([psre4 ] ) ( for @xmath302 ) or ( [ psre21])([psre23 ] ) ( for @xmath303 ) . to see this it suffices to consider only networks without edges that intersect the cut @xmath16 , and to cut the annulus along @xmath16 in order to get a disk . to conclude the proof of theorem [ psr ] we need the following statement . we say that @xmath403 is _ represented by _ a network @xmath31 if @xmath404 belongs to the image of @xmath135 . [ realiz ] for any @xmath403 there exists a network @xmath405 such that @xmath404 is represented by @xmath31 . we preface the proof by the following simple observation concerning perfect planar networks in a disk . [ transpos ] let @xmath406 , @xmath407 , @xmath408 . there exists a network @xmath409 such that the @xmath410 identity matrix is represented by @xmath411 . the proof is furnished by the network depicted on fig . [ fig : transpos ] . the corresponding boundary measurement matrix is given by @xmath412 which yields the identity matrix for @xmath413 and @xmath414 for @xmath415 . , height=170 ] effectively , lemma [ transpos ] says that the planarity restriction can be omitted in the proof of theorem [ realiz ] . indeed , if @xmath403 is represented by a nonplanar perfect network in an annulus , one can turn it to a planar perfect network in annulus by replacing each intersection by a copy of @xmath411 . in what follows we make use of the concatenation of planar networks in an annulus . similarly to the case of networks in a disk , the most important particular case of concatenation arises when the sources and the sinks are separated , that is , all sources lie on the outer circle , and all sinks lie on the inner circle . we can concatenate two networks of this kind , one with @xmath3 sources and @xmath416 sinks and another with @xmath416 sources and @xmath44 sinks , by gluing the sinks of the former to the sources of the latter . more exactly , we glue together the inner circle of the former network and the outer circle of the latter in such a way that the corresponding base points of the cuts are identified , and the @xmath417th sink of the former network is identified with the @xmath418th source of the latter . the erasure of the common boundary and the identification of edges are performed exactly as in the case of a disk . let us start with representing any rational function @xmath419 by a network with the only source on the outer circle and the only sink on the inner circle . [ 1 - 1realiz ] any rational function @xmath419 can be represented by a network @xmath420 . first , if networks @xmath421 represent functions @xmath422 and @xmath423 respectively , their concatenation @xmath424 represents @xmath425 . second , define the _ direct sum _ @xmath426,\varnothing,\varnothing,[3,4]}$ ] as shown in the left part of fig . [ fig : dirsum ] . the shadowed annuli contain networks @xmath427 and @xmath428 . the intersections of the dashed parts of additional edges with the edges of @xmath427 and @xmath428 are resolved with the help of @xmath411 ( not shown ) . note that this direct sum operation is not commutative . clearly , @xmath429 represents the @xmath410 matrix @xmath430 . the direct sum of networks is used to represent the sum @xmath431 as shown in the right part of fig . [ fig : dirsum ] . third , if @xmath420 represents @xmath404 , the network shown in fig . [ fig : netdenom ] represents @xmath432 , and , with a simple adjustment of weights , can also be used to represent @xmath433 . taking the direct sum with the trivial network representing @xmath19 , we get a representation for @xmath434 . , height=188 ] finally , functions @xmath435 for any integer @xmath3 can be represented by networks in @xmath436 . the cases @xmath437 and @xmath438 are shown in fig . [ fig : rep1 ] . other values of @xmath3 are obtained in the same way . for @xmath437 ( left ) and @xmath438 ( right),height=151 ] we now have all the ingredients for the proof of the lemma . any rational function @xmath404 can be represented as @xmath439 , where @xmath440 is an integer and @xmath441 is a polynomial satisfying @xmath442 . therefore , it suffices to represent each of the summands , and to use the direct sum construction . each summand , in its turn , is represented by the concatenation of a network representing @xmath443 with a network representing @xmath444 . the latter network is obtained as explained above from a network representing @xmath445 via the direct sum construction . to get an analog of lemma [ 1 - 1realiz ] for networks with the only source and the only sink on the outer circle , one has to use once again lemma [ transpos ] . [ 2 - 0realiz ] any rational function @xmath419 can be represented by a network @xmath446 . such a representation is obtained from the one constructed in the proof of lemma [ 1 - 1realiz ] by replacing the edge incident to the sink with a new edge sharing the same tail . the arising intersections , if any , are resolved with the help of @xmath411 . for example , representation of @xmath447 obtained this way is shown in fig . [ fig : rep6 ] on the right . it makes use of the network @xmath411 described in lemma [ transpos ] ; the latter is shown in thin lines inside a dashed circle . note that the network on the left , which represents @xmath447 in @xmath436 , is not the one built in the proof of lemma [ 1 - 1realiz ] . by a network in @xmath436 ( left ) and @xmath448 ( right),height=188 ] representation of rational functions by networks in @xmath449 and @xmath450 is obtained in a similar way . in the latter case one has to replace also the edge incident to the source with a new one sharing the head . the next step is to prove theorem [ realiz ] in the case when all sources lie on the outer circle and all sinks lie on the inner circle . [ seprealiz ] for any rational matrix @xmath403 there exists a network @xmath451,\varnothing,\varnothing,[k+1,k+m]}$ ] such that @xmath404 is represented by @xmath31 . first of all , we represent @xmath404 as @xmath452 , where @xmath453 is the @xmath454 constant matrix given by @xmath455 @xmath456 is the @xmath457 diagonal matrix @xmath458 and @xmath459 is the @xmath460 constant matrix @xmath461 with @xmath462 . similarly to the case of networks in a disk , the concatenation of networks representing matrices @xmath422 and @xmath423 produces a network representing @xmath463 . therefore , in order to get @xmath404 as above , we have to represent matrices @xmath464 , @xmath465 and @xmath466 . the first representation is achieved trivially as the disjoint union of networks representing the @xmath467 matrix @xmath468 ; in fact , since @xmath464 is constant , it can be represented by a network in a disk . the second representation is obtained as the direct sum of networks representing each of @xmath469 . finally , the third representation can be also achieved by a network in a disk , via a repeated use of the network @xmath411 . observe that in order to represent a @xmath134 matrix we have to use intermediate matrices of a larger size . to complete the proof of theorem [ realiz ] we rely on lemma [ seprealiz ] and use the same idea of replacing edges incident to boundary vertices as in the proof of lemma [ 2 - 0realiz ] . so , theorem [ realiz ] has been proved , and hence the proof of theorem [ psr ] is completed . as we already mentioned in the proof , the 2-parametric family of poisson brackets on @xmath142 induced by @xmath470 , where @xmath471 and @xmath472 are described in propositions [ psre ] and [ psre2 ] , respectively , is denoted @xmath286 . as an application of results obtained in section [ psonrat ] , consider the set @xmath474,\varnothing,\varnothing,[k+1,2k]}$ ] of perfect networks with @xmath3 sources on the outer circle and @xmath3 sinks on the inner circle . clearly , in this case the boundary measurement map takes @xmath237 to @xmath473 . just as we did in the proof of theorem [ realiz ] , we can replace @xmath135 with @xmath475 and observe that the concatenation @xmath31 of networks @xmath476,\varnothing,\varnothing , [ k+1,2k]}$ ] leads to @xmath477 . we would like to take a closer look at the bracket @xmath478 in this case . first , recall ( see , e.g. @xcite ) that the space @xmath473 can be equipped with an r - matrix ( sklyanin ) poisson bracket @xmath479 , \label{ansklyanin}\ ] ] where the left - hand side should be understood as @xmath480 and the r - matrix @xmath481 is an operator acting in @xmath482 that depends on parameters @xmath483 and solves the classical yang - baxter equation . of interest to us is a bracket ( [ ansklyanin ] ) that corresponds to the so - called _ trigonometric r - matrix _ @xcite @xmath484 bracket ( [ ansklyanin ] ) , ( [ trig - matrix ] ) can be re - written in terms of matrix entries of @xmath485 as follows ( we only list non - zero brackets ) : for @xmath486 and @xmath487 , @xmath488 @xmath489 @xmath490 @xmath491 it is now straightforward to check that for @xmath492,\varnothing,\varnothing , [ k+1,2k]}$ ] , the poisson algebra satisfied by the entries of @xmath493 coincides with that of the sklyanin bracket ( [ ansklyanin ] ) , ( [ trig - matrix ] ) . more exactly , relations ( [ brack_ij_ab ] ) and ( [ brack_ij_ib ] ) are equivalent to ( [ psre1 ] ) with @xmath494 calculated according to the third case , while ( [ brack_ib_aj ] ) and ( [ brack_ij_aj ] ) are equivalent to ( [ psre2 ] ) with @xmath495 calculated according to the first case . finally , the brackets that vanish identically , correspond exactly to the situations listed in the second case in ( [ psre3 ] ) and in the third case in ( [ psre4 ] ) . to summarize , we obtained the following statement . [ sklyaninbr ] for any @xmath492,\varnothing,\varnothing , [ k+1,2k]}$ ] and any choice of parameters @xmath496 in ( [ 6parw ] ) , ( [ 6parb ] ) such that @xmath497 and @xmath498 in ( [ cond ] ) , the map @xmath499 is poisson with respect to the sklyanin bracket ( [ ansklyanin ] ) associated with the r - matrix ( [ trig - matrix ] ) . equations ( [ brack_ij_ab])([brack_ij_aj ] ) can be also used to define a poisson bracket in the `` rectangular '' case of @xmath500 . in this case , a concise description ( [ ansklyanin ] ) of the bracket should be modified as follows : @xmath501 where @xmath502 denotes the r - matrix ( [ trig - matrix ] ) acting in @xmath503 . let @xmath289 be a perfect planar network in an annulus . similarly to the case of a disk , we are going to provide a grassmannian interpretation of the boundary measurement map defined by @xmath31 . to this end , we extend the boundary measurement matrix @xmath135 to a @xmath504 matrix @xmath505 as follows : \(i ) the @xmath506 submatrix of @xmath505 formed by @xmath3 columns indexed by @xmath117 is the identity matrix @xmath507 ; \(ii ) for @xmath508 $ ] and @xmath509 , the @xmath510-entry of @xmath505 is @xmath511 , where @xmath512 is the number of elements in @xmath119 lying strictly between @xmath513 and @xmath514 in the linear ordering ; note that the sign is selected in such a way that the minor @xmath515}^{i(i_p\to j)}$ ] coincides with @xmath516 , where @xmath517 . we will view @xmath505 as a matrix representative of an element @xmath518 in the space @xmath6 of rational functions @xmath519 . the latter space is called the _ space of grassmannian loops _ , and the corresponding rational map @xmath520 is called the _ grassmannian boundary measurement map_. given a network @xmath31 and a simple path @xmath30 from a source @xmath123 to a sink @xmath125 in @xmath31 , we define the _ reversal _ of @xmath30 as follows : for every edge @xmath521 , change its direction and replace its weight @xmath29 by @xmath522 ; equivalently , the modified weight @xmath128 is replaced by @xmath523 . clearly , after the reversal of @xmath30 all vertices preserve their color . denote by @xmath524 the network obtained from @xmath31 by the reversal of @xmath30 , and by @xmath525 the corresponding _ path reversal map _ @xmath526 . besides , put @xmath527 if both endpoints of @xmath30 belong to the same boundary circle , and @xmath528 otherwise . define two maps from @xmath6 to itself : @xmath529 is the identity map , while @xmath530 takes any @xmath531 to @xmath532 . our next goal is to prove that the path reversal map for paths not intersecting the cut commutes with the grassmannian boundary measurement map up to @xmath533 . [ pathrev ] let @xmath30 be a simple path from a source @xmath123 to a sink @xmath125 in @xmath31 such that @xmath127 does not vanish identically and @xmath30 does not intersect the cut . then @xmath534 let @xmath119 be the index set of the sources in @xmath31 . the statement of the theorem is equivalent to the equality @xmath535 for any subset @xmath536 $ ] of size @xmath3 . here and in what follows the superscript @xmath30 means that the corresponding value is related to the network @xmath524 . the signs of the elements @xmath537 are chosen in such a way that @xmath538 , so we have to prove @xmath539 the proof of ( [ pathreveq ] ) relies on the induction on the number of inner vertices in @xmath31 . let us start with the case when @xmath31 does not have inner vertices . in this case it suffices to prove with @xmath540 for all edges @xmath541 . assume first that the intersection index of each edge with the cut @xmath16 equals @xmath542 or @xmath68 ; consequently , after a suitable isotopy , each edge either intersects @xmath16 exactly once , or does not intersect it at all . let @xmath543 be the edge to be reversed , and hence @xmath544 . if both @xmath123 and @xmath125 belong to the same boundary circle , then exactly the following two cases are prohibited : @xmath545 consequently , reversing @xmath546 does not change @xmath547 , which corresponds to the map @xmath529 . if @xmath123 and @xmath125 belong to distinct boundary circles , then the above two cases are prohibited whenever @xmath201 does not intersect the cut . if @xmath201 intersects the cut then the above two cases are the only possibilities . consequently , reversing @xmath546 does not change @xmath547 for the edges not intersecting the cut and reverses it for the edges intersecting the cut , which corresponds to the map @xmath530 . it remains to lift the restriction on the intersection index of edges with @xmath16 . if there exists an edge @xmath40 such that @xmath548 then the endpoints of @xmath40 belong to distinct boundary circles , and for any other edge with the endpoints on distinct boundary circles , the intersection index with @xmath16 does not vanish . consequently , only edges with the endpoints on the same boundary circle can be reversed , and the above reasoning applies , which leads to @xmath529 . let now @xmath31 have inner vertices , and assume that the first inner vertex @xmath42 on @xmath30 is white . denote by @xmath201 and @xmath40 the first two edges of @xmath30 , and by @xmath41 the third edge incident to @xmath42 . in what follows we assume without loss of generality that the cut in @xmath31 does not intersect @xmath201 . to find @xmath549 consider the network @xmath550 that is related to @xmath524 exactly in the same way as the network @xmath551 defined immediately before lemma [ recalcan ] is related to @xmath31 . similarly to the first relation in lemma [ recalcan ] , we find @xmath552 taking into account that @xmath553 , @xmath554 , @xmath555 , we finally get @xmath556 to find @xmath557 we proceed as follows . [ xkpwhite ] let the first inner vertex @xmath42 of @xmath30 be white , then @xmath558 the proof utilizes explicit formulas ( similar to those provided by lemma [ recalcan ] ) that relate boundary measurements in the networks @xmath524 and @xmath550 . what is important , the sign @xmath559 in the second formula in lemma [ recalcan ] and the sign @xmath560 defined at the beginning of this section interplay in such a way that any submatrix of @xmath561 is the sum of the corresponding submatrix of @xmath562 and a submatrix of the rank @xmath19 matrix that is equal to the tensor product of the @xmath417th column of @xmath561 and the @xmath563th row of @xmath562 . to find @xmath564 , create a new network @xmath565 by deleting @xmath123 and the edge @xmath201 from @xmath14 , splitting @xmath42 into @xmath208 sources @xmath566 ( so that either @xmath567 or @xmath568 ) and replacing the edges @xmath569 and @xmath570 by @xmath571 and @xmath572 , respectively . [ xkwhite ] let the first inner vertex @xmath42 of @xmath30 be white , then @xmath573 the proof is a straightforward computation . by ( [ xipwhite ] ) and lemmas [ xkpwhite ] and [ xkwhite ] , relation ( [ pathreveq ] ) boils down to @xmath574 for @xmath575 and @xmath576 for @xmath577 . to prove these two equalities , we identify @xmath578 with @xmath579 and @xmath580 with @xmath581 . under this identification we have @xmath582 , where @xmath583 is the path from @xmath578 to @xmath125 in @xmath584 induced by @xmath30 . observe that @xmath584 has less inner vertices than @xmath31 , and that the index set of the sources in @xmath584 is @xmath585 . therefore , by the induction hypothesis , @xmath586 for any @xmath587 of size @xmath588 . besides , @xmath589 and @xmath590 . therefore , using ( [ indhyp ] ) for @xmath591 , @xmath592 and @xmath593 we get both equalities above . assume now that the first inner vertex @xmath42 on @xmath30 is black . denote by @xmath201 and @xmath40 the first two edges of @xmath30 , and by @xmath41 the third edge incident to @xmath42 . to find @xmath549 , consider the network @xmath594 similar to the one defined immediately before lemma [ xkwhite ] , the difference being that the two new boundary vertices @xmath595 and @xmath596 are sinks rather than sources . clearly , @xmath597 to find @xmath557 we proceed as follows . [ xkpblack ] let the first inner vertex @xmath42 of @xmath30 be black , then @xmath598 the proof is a straightforward computation . to find @xmath564 we consider the network @xmath551 defined immediately before lemma [ recalcan ] . [ xkblack ] let the first inner vertex @xmath42 of @xmath30 be black , then @xmath599 the proof is similar to the proof of lemma [ xkpwhite ] . by ( [ xipblack ] ) and lemmas [ xkpblack ] and [ xkblack ] , relation ( [ pathreveq ] ) boils down to @xmath600 for @xmath575 and @xmath601 for @xmath577 . to prove these two equalities , we identify @xmath602 with @xmath581 and @xmath603 with @xmath579 . under this identification we have @xmath604 , where @xmath605 is the path from @xmath581 to @xmath125 in @xmath551 induced by @xmath30 . observe that @xmath551 has less inner vertices than @xmath31 , and that the index set of the sources in @xmath551 is @xmath606 . therefore , by the induction hypothesis , @xmath607 for any @xmath608 of size @xmath3 . taking into account that @xmath609 , @xmath610 and using ( [ indhypb ] ) for @xmath611 , @xmath612 , @xmath613 and @xmath614 we get both equalities above . theorem [ pathrev ] is proved in @xcite for networks in a disk . observe that in this case @xmath615 vanishes identically , and hence @xmath518 and @xmath525 always commute . consider a subspace @xmath616 consisting of all @xmath617 such that the plcker coordinate @xmath618 does not vanish identically ; clearly , @xmath619 . therefore , we can identify @xmath620 with the space @xmath142 equipped with the 2-parametric family of poisson brackets @xmath286 . the following result says that for any fixed @xmath621 , the families of brackets @xmath286 on different subspaces @xmath622 can be glued together to form the unique 2-parametric family of poisson brackets on @xmath6 that makes all maps @xmath518 poisson . [ pslgr ] ( i ) for any fixed @xmath353 , @xmath623 , and any choice of parameters @xmath256 and @xmath257 there exists a unique poisson bracket @xmath624 on @xmath6 such that for any network @xmath31 with @xmath353 boundary vertices on the outer circle , @xmath625 boundary vertices on the inner circle , @xmath3 sources , @xmath4 sinks and weights defined by ( [ connect ] ) , the map @xmath626 is poisson provided the parameters @xmath287 and @xmath288 defining the bracket @xmath248 on @xmath28 satisfy relations ( [ cond ] ) . \(ii ) for any @xmath627 $ ] , @xmath628 , and any @xmath353 , @xmath623 , the restriction of @xmath624 to the subspace @xmath622 coincides with the bracket @xmath629 with @xmath630 $ ] , @xmath631\setminus i_1 $ ] , @xmath632 , @xmath633\setminus i_2 $ ] . this result is an analog of theorem 4.3 proved in @xcite , and one may attempt to prove it in a similar way . the main challenge in implementing such an approach is to check that the poisson structures defined for two distinct subspaces @xmath620 and @xmath634 coincide on the intersection @xmath635 . for the case of networks in an annulus , the direct check becomes too cumbersome . we suggest to bypass this difficulty in the following way . assume first that @xmath636 and take @xmath637 , @xmath638 . denote by @xmath639 the set of networks in @xmath279 satisfying the following two conditions : @xmath127 does not vanish identically and there exists a path from @xmath123 to @xmath125 that does not intersect the cut . the set @xmath640 is defined similarly , with the roles of @xmath417 and @xmath641 interchanged . clearly , the path reversal introduced in section [ pathrev ] establishes a bijection between @xmath639 and @xmath640 . moreover , a suitable modification of theorem [ realiz ] remains true for networks in @xmath639 : these networks represent all rational matrix function such that corresponding component of the matrix does not vanish identically . to see that we use the following construction . let @xmath42 be the neighbor of @xmath123 in @xmath31 and @xmath363 be the neighbor of @xmath125 in @xmath31 . add two new white vertices @xmath642 and @xmath643 and two new black vertices @xmath644 and @xmath645 . replace edge @xmath646 by the edges @xmath647 and @xmath648 so that the weight of the obtained path is equal to the weight of the replaced edge . in a similar way , replace @xmath649 by @xmath650 and @xmath651 . besides , add edges @xmath652 and @xmath653 of weight 1 and two parallel edges @xmath654 , one of weight 1 , and the other of weight @xmath20 . finally , resolve all the arising intersections with the help of the network @xmath411 . since the set of functions representable via networks in @xmath639 is dense in the space of all rational matrix functions , the 2-parametric family @xmath629 is defined uniquely already by the fact that @xmath135 is poisson for any @xmath655 . recall that the boundary measurement map @xmath135 factors through @xmath183 ; clearly , the same holds for the grassmannian boundary measurement map @xmath518 . besides , the path reversal map @xmath525 commutes with the projection @xmath656 and commutes with @xmath518 up to @xmath533 . finally , poisson brackets satisfying relations - and - commute with @xmath533 . therefore , poisson structures @xmath629 and @xmath657 coincide on @xmath635 . if @xmath658 , we consider a sequence @xmath659 such that @xmath660 for all @xmath661 and apply to each pair @xmath662 the same reasoning as above . we wish to express gratitude to a. postnikov who explained to us the details of his construction and to v. fock , s. fomin and n. reshetikhin for stimulating discussions . m. g. was supported in part by nsf grants dms # 0400484 and dms # 0801204 . m. s. was supported in part by nsf grants dms # 0401178 and phy # 0555346 . a. v. was supported in part by isf grant # 1032/08 . j. moser , _ finitely many mass points on the line under the influence of the exponential potential - an integrable system_. dynamical systems , theory and applications , 467497 , lecture notes in physics * 38 * , springer , berlin , 1975 .

since the foundation of quantum physics there has been huge interest in the nontrivial transition from the quantum to the classical regime . an important observation in this context is the ehrenfest theorem stating that the dynamics of quantum mechanical expectation values is determined by the classical equations of motion @xcite . going beyond expectation values , an _ ehrenfest time _ scale has been identified as the time below which the quantum time evolution is well approximated by the corresponding classical dynamics @xcite . the ehrenfest time is the time it takes for the chaotic classical dynamics to stretch an initially narrow wave packet to some relevant classical length scale such as the system size @xmath5 . since the stretching is exponential in classically chaotic systems , one has @xmath6 with the lyapunov exponent @xmath7 of the classical dynamics , the initial spread @xmath8 of the wave packet and the fermi momentum @xmath9 . in recent years , there has been much interest in determining the influence of the ehrenfest time on stationary transport quantities such as the conductance @xcite , its variance @xcite and its behavior under decoherence @xcite , shot noise @xcite and higher moments of the current @xcite , and on time dependent quantities such as the spectral form factor @xcite , the survival probability @xcite and the fidelity @xcite . most of these papers used the trajectory - based semiclassical approach to transport @xcite which currently is the method of choice for investigating ehrenfest - time dependences of quantum observables @xcite . the leading order quantum correction to the conductance was found in refs . @xcite to decay exponentially with the ehrenfest time . qualitatively speaking this can be understood by noting first that this contribution originates from loop diagrams [ see fig . [ fig : orbt](b ) ] , and as such depends on the return probability , and second that , neglecting system - dependent nongeneric processes , this return probability contribution essentially vanishes for times shorter than the ehrenfest time . such an intuitive interpretation of the ehrenfest - time dependence does not always work : though inherently of nonclassical nature , the leading order contribution to the conductance variance of systems ideally coupled to external leads turns out to be independent of the ehrenfest time @xcite . ( gray boxes ) . a classical trajectory traversing the system is shown by a solid ( blue ) line . the second dashed ( red ) line on the left indicates a trajectory backreflected at the barrier . this occurs with probability @xmath10 . ] in this paper , we determine semiclassically the ehrenfest - time dependence of the variance @xmath11 of the conductance for a chaotic system coupled to external leads via nonideal contacts modeled by tunnel barriers of transparency @xmath12 . the situation is depicted in fig . the presence of tunnel barriers has the dramatic effect that @xmath11 increases or decreases with @xmath0 , depending on the value of @xmath3 . for @xmath4 , we even observe an enhancement of the variance above the universal value in the presence of time reversal symmetry for equal lead widths of @xmath13 @xcite upon increasing @xmath0 . this is very surprising , given the quantal nature of the conductance fluctuations . in the range @xmath14 we find a reduction of @xmath11 as @xmath0 increases that is strongest around @xmath15 , and recover the @xmath0-independent behavior of @xmath16 at @xmath17 observed in refs . @xcite . the precise dependence on @xmath3 is depicted in fig . [ condfluc ] . . the function @xmath18 , defined in eqs . ( [ li22.9 ] ) and ( [ li23 ] ) , measures the deviation from the universal rmt variance . in this graph the parameter @xmath19 is plotted as a function of the tunnel barrier transparency @xmath3 for time - reversal symmetric systems . the solid red line gives the analytical prediction , eq . ( [ li23 ] ) , and the blue dots are results of numerical simulations . error bars indicate the standard deviation over the ensemble of calculated data . , scaledwidth=47.0% ] to obtain these results we first , in sec . [ semiapprox ] , introduce the semiclassical approximation to the conductance variance . then , in sec . [ part1 ] , we analytically calculate the @xmath3-dependence of the diagrammatic contributions to the variance , to leading order in the total number of open channels and linear in the ehrenfest time . we list all relevant diagrams and calculate their contributions . given their number , we also identify the most relevant ones and specify the range in @xmath3 where they are particularly important . in sec . [ part2 ] we present numerical results that confirm our analytical results and finally conclude in sec . [ concs ] . within the scattering approach to quantum transport @xcite , the energy averaged conductance @xmath20 ( in units of @xmath21 ) can be expressed in terms of the transmission @xmath22 as @xmath23 with @xmath24 denoting an average over an energy window that is classically small but quantum mechanically large . this leads to the following expression for the variance @xmath25 ^ 2\right\rangle-\left\langle{\rm tr}\left(\textbf{tt}^\dagger\right)\right\rangle^{2}.\ ] ] the scattering matrix elements are related by the fisher - lee relation @xcite to the projection of the green function onto the transverse directions in the leads . performing the projection to leading order in @xmath26 and approximating the green function semiclassically , one obtains @xmath27 with the heisenberg time @xmath28 , the time conjugate to the mean level spacing . here the sum is over the scattering trajectories @xmath29 which connect channel @xmath30 in the entrance ( or say left ) lead and channel @xmath31 in the exit ( or right ) lead in fig . the summands contain rapidly oscillating phases depending on the classical actions @xmath32 of the considered classical trajectories , and classical stability prefactors @xmath33 whose precise form is given for example in @xcite . inserting eq . ( [ in6 ] ) into ( [ in4.9 ] ) we obtain the semiclassical expression for the conductance , @xmath34 using ( [ in6 ] ) in ( [ in5 ] ) yields the semiclassical expression its variance , @xmath35 with the channel sums in eqs . ( [ in6.9 ] ) and ( [ in7 ] ) running over all open lead channels ( @xmath36 in the left and @xmath37 in the right lead ) . if we consider contributions in the first term in eq . ( [ in7 ] ) where @xmath29 and @xmath38 form a correlated pair ( with self - encounters ) and @xmath39 and @xmath40 form a separate correlated pair , we simply recreate the second term . we can thus remove the second term in the above equation by removing such pairs from the semiclassical treatment of the first term . in terms of trajectories we then obtain @xmath41 where the trajectories @xmath42 go from channel @xmath30 in the entrance lead to channel @xmath31 in the exit lead . likewise trajectories @xmath43 go from channel @xmath44 to channel @xmath45 . because we have removed terms from correlated trajectories where @xmath46 and @xmath47 [ this restriction is denoted by the double bracket in eq . ( [ condvartraj ] ) ] we are left with quadruplets where all four trajectories interact through encounters . before performing the energy average the approximations for @xmath20 and @xmath16 in eqs . ( [ in6.9 ] ) and ( [ condvartraj ] ) are rapidly fluctuating as a function of energy for @xmath48 . thus only contributions from very similar trajectories survive the average . the classical contribution to eq . ( [ in6.9 ] ) results from equal trajectories @xmath49 , the so called diagonal approximation @xcite , for an illustration see fig . [ fig : orbt](a ) . ( b).,scaledwidth=45.0% ] here eq . ( [ in6.9 ] ) yields @xmath50}=\frac{1}{\th}\sum_{a , b } \sum_{\gamma(a\to b)}\left|a_\gamma\right|^2.\ ] ] from here on , superscripts refer to the corresponding figure . to evaluate the remaining @xmath29-summation the sum rule for open systems @xcite that transforms the sum over orbits into an integral over their durations is applied @xmath51 here @xmath1 is the dwell time of the cavity , i.e. the typical time a classical particle remains inside the chaotic system . this is given by @xmath52 with @xmath53 . this finally yields @xmath54}\approx\frac{n_{\rm l}n_{\rm r}}{n_{\rm l}+n_{\rm r}}.\ ] ] quantum corrections to this result are obtained from pairs of slightly different trajectories . the considered trajectories are almost identical differing only in how they are connected within self encounters , for the pair considered in @xcite , see fig . [ fig : orbt](b ) . here the orbits possess close self encounters with one orbit crossing and the other anticrossing there leading to a different direction of traversal of the closed loop . considering this pair in eq . ( [ in6.9 ] ) leads to the leading order quantum correction to the conductance . to determine its contribution the action difference between the partner trajectories and the number of crossings needs to be determined . the calculation is done here within the phase - space approach , in the context of the conductance it was first performed in ref . we will follow the latter approach throughout this paper . there a poincar surface of section is considered inside the encounter region and the difference along the stable and unstable directions of the piercing points of the two stretches , @xmath55 and @xmath56 , respectively , is used to characterize an encounter . in terms of these coordinates the action difference for the orbit pair in fig . [ fig : orbt ] is given by @xmath57 @xcite . the weight @xmath58 that additionally depends on the duration of the orbit @xmath59 measuring the number of encounters is obtained from the ergodicity of the flow as @xcite @xmath60 where @xmath61 is the duration of the encounter . in general , the action difference @xmath62 and the weight @xmath63 depend on the trajectory configuration considered . in total we obtain for the quantum correction @xmath64 resulting from the diagram in fig . [ fig : orbt](b ) @xmath65}&\approx&\int_{-c}^c\!\!{\rm d}s{\rm d}u \!\!\int_{2t_{\rm enc}}^\infty\!\!{\rm d}t \,w({s},{u}){\rm e}^{\frac{i\delta s}{\hbar } } { \rm e}^{-\frac{\left(t - t_{\rm enc}\right)}{\tau_d}}\nonumber\\&\approx&-\frac{n_{\rm l}n_{\rm r}}{\left(n_{\rm l}+n_{\rm r}\right)^2}{\rm e}^{-\te/\td}\end{aligned}\ ] ] with @xmath66 . in the first line additionally the survival probability correction during the encounter @xcite is taken into account . during the encounter the stretches are so close that the orbit escapes either during the first stretch or does not escape at all leading to the enhanced survival probability in eq . ( [ quantumco ] ) . the @xmath67-integrals in eq . ( [ quantumco ] ) are performed as described in @xcite , by substituting @xmath68 and @xmath69 . the @xmath70-integral then essentially cancels the @xmath71 in the denominator and the @xmath72-integral yields ( after a partial integration ) the contribution @xmath73 . to treat orbits differing at several places , ref . @xcite introduces the splitting of the orbit into encounters and links . during the encounters the orbits are close to themselves but the orbit and its partner are differently connected . due to the exponential separation of neighboring trajectories in the chaotic case these last essentially an ehrenfest time , as will become clear from the calculation below . the stretches are connected by the links , where ` links ' denote the long parts of the trajectory where the trajectory and its partner are essentially identical ( up to time reversal symmetry ) . with a suitable change of variables in the calculation , in the rmt limit @xmath74 , one can treat different encounters as distinct and separate the semiclassical contribution into a product of contributions over each of the links and encounters . the total contribution can therefore be obtained by diagrammatic rules @xcite . away from this limit , and for the ehrenfest - time dependence , the encounters may start to overlap and for the conductance variance the trajectories can be seen to meet and surround periodic orbits trapped inside the system , i.e. they have encounters with periodic orbits , and must be treated as part of a continuous family @xcite . when we include tunnel barriers ( as in fig . [ sys ] ) , three main changes occur that were originally described in @xcite : 1 . the particles enter and leave the cavity with the probability @xmath3 , which leads to a factor @xmath75 for each trajectory pair . 2 . while without tunnel barriers every particle that hits the lead leaves the system , now only the ratio @xmath3 of the particles hitting the lead leaves the cavity . for the links , the effective dwell time is therefore @xmath76 and the dwell time in the exponential in ( [ in8 ] ) should be replaced by this effective dwell time . however , if trajectory stretches are correlated , as they are during encounters , then the whole configuration is lost if just one stretch of the encounter leaves the system . this happens with a probability @xmath77 for @xmath78 correlated stretches . the dwell time in such a situation is therefore replaced by @xmath79 $ ] . additional encounter diagrams become possible ; for an example which contributes to the energy averaged conductance @xmath20 , see fig . [ fig : orbt](c ) . in this case one encounter stretch can be moved into the lead forcing the other to be backreflected at the opening . note that configurations where both stretches are backreflected at the opening are already taken into account by the modified dwell time explained above . although the effective dwell times are altered by the tunnel barriers , the action difference and weight functions are unaffected so that in the rmt limit ( @xmath74 ) contributions can still be obtained by diagrammatic rules : the contribution of each link is now given by @xmath80 . the stretches of an encounter of @xmath78 orbits yield @xmath81 . for the ehrenfest - time dependence , however , these changes render the calculation of the contributions to the conductance variance much more difficult compared to @xmath82 : due to the discontinuous form of the effective dwell time , the contributions from diagrams with a different number of surroundings of trapped periodic orbits need to be split and treated separately . also the possibility for encounter stretches to be backreflected at the tunnel barriers increases the number of diagrams considerably . here we calculate the leading order in @xmath83 contributions to the variance of the conductance for non - zero ehrenfest time in the presence of tunnel barriers . we show all the relevant diagrams and calculate their contributions . the results given here are valid in the unitary case ; results for the orthogonal case can be obtained by multiplying the total by a factor of @xmath84 . we start with the contributions important in the rmt limit which allow us to recover the rmt result . the corresponding rmt calculation was performed in @xcite by brouwer and beenakker . first we consider two 2-encounters ( encounters involving two trajectory stretches ) in a row , see fig . [ ana1](a ) . this diagram also occurred in the @xmath82 treatment of @xcite , but we will explain how for @xmath85 other diagrams with backreflected stretches can be derived from this one . when the encounters are inside the cavity and not touching the tunnel barriers , as depicted in fig . [ ana1](a ) , we start with eq . ( [ condvartraj ] ) and use the sum rule ( [ in8 ] ) , the action difference @xmath86 ( where the subscripts refer to the different encounters ) and the weight function @xcite @xmath87 containing the durations of the two trajectories indicated by solid lines in fig . [ ana1](a ) , @xmath88 and @xmath89 , respectively and the durations of the two encounters of the trajectories @xcite @xmath90 with a classical constant @xmath44 again of the order one . from eqs . ( [ in7 ] ) and ( [ in8 ] ) we write @xmath91}&= & \frac{\gamma^4n_{\rm l}^2n_{\rm r}^2}{\th^2}\prod_{i=1}^6 \int_0^\infty { \rm d}t_i\,{\rm e}^{-\gamma t_i/\td } \int_{-c}^c{\rm d}{\boldsymbol s}{\rm d}{\boldsymbol u}\frac{1}{\omega^2}\nonumber\\ & & \times\frac{{\rm e}^{\frac{{\rm i}}{\hbar}{\boldsymbol s}{\boldsymbol u } } } { t_{{\rm enc},1}t_{{\rm enc},2}}{\rm e}^{-p_2\left(t_{{\rm enc},1}+t_{{\rm enc},2}\right)/\td } , \end{aligned}\ ] ] with the superscript again referring to the corresponding figure : fig . [ ana1 ] . as explained in sec . [ semiapprox ] , the trajectory quadruplet leads to the overall factor @xmath92 while the sum over possible channels provides @xmath93 . the six links have an effective dwell time of @xmath76 , while each 2-encounter experiences the dwell time @xmath94 as explained at the end of the last section . the @xmath95-integrals are performed , as described after eq . ( [ quantumco ] ) , by substituting @xmath96 and @xmath97 . each @xmath98-integral yields ( after a partial integration ) the contribution @xmath99 as already obtained for the conductance for @xmath82 . finally the @xmath100-integrals yield @xmath101 } = \frac{n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4 } \frac{p_2 ^ 2}{\gamma^2}{\rm e}^{-2p_2\te/\td},\ ] ] which generalizes the result for @xmath82 from @xcite . the representation of eq . ( [ li0.2 ] ) , and the integrals to arrive at ( [ li0.3 ] ) , also nicely illustrate how the diagrammatic rules introduced above arise in this context . we now turn to the new diagrams that arise due to the tunnel barriers where some of the links are shrunk until an encounter touches a barrier . first we consider the case in which just one link connecting the encounter to the opening is removed and the corresponding encounter stretch now starts in the opening ; as in the example in fig . [ fig : orbt](b ) . as the stretches during an encounter lie very close to each other , the other encounter stretch has to be backreflected at the opening ( so only one link is lost ) . this contribution can therefore only exist for @xmath85 . the changes in the analytical calculation mainly affect the weight function , see also @xcite . if we shrink a link on trajectory 1 then , compared to eq . ( [ li0 ] ) , the orbit of duration @xmath88 now only involves two links so the factor @xmath102 is replaced by @xmath103 . moreover , for the encounter which touches the barrier , @xmath104 is replaced by an integration variable @xmath105 that is integrated from zero to the @xmath104 defined in ( [ li0.1 ] ) . this variable measures the length of the encounter that remains inside the system , i.e. which has not yet been moved into the lead . performing again the steps described after ( [ li0.1 ] ) yields an expression similar to ( [ li0.2 ] ) but with @xmath106 instead of @xmath107 link factors and @xmath104 in the exponential in the second line replaced by an integration variable @xmath105 that is integrated from zero to @xmath104 . because half of the encounter is backreflected at the tunnel barrier , additionally this contribution is multiplied by @xmath108 . the contribution from the lower limit of the @xmath105-integral is zero @xcite , leading to @xmath109 } = -\frac{4n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4 } \frac{(1-\gamma)p_2}{\gamma}{\rm e}^{-2p_2\te/\td},\ ] ] with the ` @xmath110 ' denoting that one link was removed . the prefactor 4 is due to the fact that there are four such links we can remove . analogously , the contribution where two links that connect two _ different _ encounters to the opening are removed is @xmath111}= \frac{4n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4 } ( 1-\gamma)^2{\rm e}^{-2p_2\te/\td},\ ] ] where there are again four possibilities of picking two links to remove . additionally we can , and this is a possibility which also exists for @xmath82 , remove both links connecting the _ same _ encounter to the opening . this means that the trajectories tunnel straight into the encounter . as the encountering orbits are so close together this means that @xmath112 or @xmath113 so that there is only one channel summation in the lead where the encounter touches . also we have two links fewer and one integral over the part of the encounter that remains inside the system , yielding @xmath114}= -\frac{\left(n_{\rm l}^2n_{\rm r}+n_{\rm l}n_{\rm r}^2\right)}{\left(n_{\rm l}+n_{\rm r}\right)^3 } p_2{\rm e}^{-2p_2\te/\td},\ ] ] where the additional ` ( s ) ' in the superscript indicates that two links were removed at the same encounter . we can further remove one link from the other encounter to obtain @xmath115}= \frac{2\left(n_{\rm l}^2n_{\rm r}+n_{\rm l}n_{\rm r}^2\right)}{\left(n_{\rm l}+n_{\rm r}\right)^3 } \gamma(1-\gamma){\rm e}^{-2p_2\te/\td}.\ ] ] finally , when all four links connecting the encounter to the leads are removed , we have @xmath116}= \frac{n_{\rm l}n_{\rm r}}{\left(n_{\rm l}+n_{\rm r}\right)^2}\gamma^2 { \rm e}^{-2p_2\te/\td}.\ ] ] with time reversal symmetry , however , we can also have fig . [ ana1](b ) where , because channels @xmath30 and @xmath44 are in the left lead and channels @xmath31 and @xmath45 in the right lead , we can not shrink more than two links simultaneously . similarly , we can not remove two links from the same encounter . since @xmath117 , the contributions ( [ li0.6][li0.7 ] ) actually cancel so that the diagram in fig . [ ana1](b ) and the ones obtained by cutting links provide the same contribution as the diagram in fig . [ ana1](a ) . time reversal symmetry then still gives a factor 2 in this case , while for all the following cases , diagrams related by time reversal symmetry provide the same contributions directly . next we treat the diagrams with two 2-encounters which lie along a trapped periodic orbit ; depicted in fig . although the links connect the encounter stretches in a different way compared to fig . [ ana1 ] these diagrams again contain two 2-encounters and the same number of links . the possibilities for shrinking links are identical as for fig . [ ana1](b ) and so each diagram provides the same contributions as above and the same total contribution as the configuration in fig . [ ana1](a ) . they are multiplied , however , by a factor @xmath84 since the orbits here have two possibilities to traverse the enclosed periodic orbits ( schematically we can also reflect the diagrams horizontally ) . the last relevant diagram type to obtain the rmt result is shown in fig . this diagram contains one 3-encounter with a periodic orbit . its contribution is calculated in an analogous way to eq . ( [ li0.2 ] ) . here we have one link fewer , one encounter instead of two and involving three rather than two orbit stretches . therefore we have @xmath118}= -\frac{2n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4 } \frac{p_3}{\gamma}{\rm e}^{-p_3\te/\td}.\ ] ] alternatively this result can be obtained from the contribution of a 3-encounter , @xmath119 , in the appendix of @xcite by including the correct dwell times as well as the contributions from the links . also in this case encounter stretches can be shrunk and removed : first removing one link connecting the encounter to the opening yields @xmath120}= \frac{8n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4 } ( 1-\gamma)^2{\rm e}^{-p_3\te/\td}\ ] ] and second removing both links : @xmath121}= \frac{2\left(n_{\rm l}n_{\rm r}^2+n_{\rm l}^2n_{\rm r}\right)}{\left(n_{\rm l}+n_{\rm r}\right)^3 } \gamma(1-\gamma){\rm e}^{-p_3\te/\td}.\ ] ] having calculated all the contributions in the rmt limit , we can obtain the rmt result by setting @xmath122 . when summing the the results in eqs . ( [ li0.3][li0.91 ] ) , we obtain @xmath123 , \end{aligned}\ ] ] which agrees with the rmt prediction in @xcite . in appendix [ diffprobs ] , we use these diagrams to obtain the rmt result when each channel has a different tunneling probability . having gone through all the diagrams that contribute at zero ehrenfest time , we now turn to those diagrams whose contribution vanishes at zero ehrenfest time . for these contributions , the periodic orbits in figs . [ ana2 ] and [ ana3 ] become important and we now view those diagrams as trajectories which have an encounter with the periodic orbit , rather than with each other . for example , in fig . [ ana3 ] we could start with the solid trajectory which passes from channel @xmath44 to @xmath45 and the dashed trajectory from @xmath30 to @xmath31 and build the rest of the diagram from those starting points and the periodic orbit . both of those trajectories encounter the periodic orbit once . in the semiclassical treatment of fig . [ ana3 ] above , it was implicitly assumed that these encounters occur at the same point along the periodic orbit . the resulting 3-encounter can therefore be considered as an ` aligned ' 3-encounter , but for the further ehrenfest time dependence we also need to consider the situation where the two encounters with the periodic orbit occur at different points along the periodic orbit but still overlap . in this case we have a ` non - aligned ' 3-encounter , while when the encounters no longer overlap we return to the two separate 2-encounters of fig . [ ana2 ] . we now derive the ehrenfest - time dependence of a non - aligned 3-encounter , whose base trajectories are depicted in fig . [ ana4](a ) . in order to obtain the complete trajectory quadruplet that contributes in ( [ condvartraj ] ) we first need to include an additional traversal of the periodic orbit with one of the base trajectories to obtain the original trajectory pair [ which has positive action in ( [ condvartraj ] ) ] . the partner pair ( with negative action ) is then created by including the extra traversal of the periodic orbit with the other base trajectory . in this way we recover a diagram like fig . [ ana3](a ) from fig . [ ana4](a ) and a small action difference ( the action of the periodic orbit itself cancels ) . after writing the contribution of a non - aligned 3-encounter in an analogous way as for the aligned 3-encounter , as explained in eqs . ( [ li0][li0.2 ] ) , as product of link and encounter contributions , it can be evaluated by making use of the results for @xmath124 in the appendix of @xcite . of course , with the tunnel barriers , the dwell times must be modified compared to @xcite : the link dwell time is modified by the factor @xmath125 , while the parts of the encounter where only one trajectory is correlated to the periodic orbit these contributions are called fringes in @xcite have the factor @xmath126 since we have two stretches close to each other . likewise , when both trajectories are correlated with the periodic orbit we have three stretches in total and the corresponding factor @xmath127 . with these corrections , the contribution is @xmath128}&=&\frac{4n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4}\frac{p_2 ^ 2 } { \gamma\left(2p_2-p_3\right)}\nonumber\\ & & \times\left({\rm e}^{-p_3\te/\td}-{\rm e}^{-2p_2\te/\td}\right ) . \end{aligned}\ ] ] in order to proceed to the additional diagrams that arise from touching the tunnel barriers , we first reconsider this contribution in detail along the lines of @xcite . as for eqs . ( [ li0][li0.2 ] ) , we start with the weight function for two base trajectories encountering a periodic orbit , as explained in @xcite @xmath129 here @xmath130 and @xmath131 are the durations of the links of the base trajectories that connect the periodic orbit encounters to the lead , while @xmath132 is the period of the periodic orbit . the integral over the period corresponds to summing over all periodic orbits which can be encountered . the encounter times are as in ( [ li0.1 ] ) , but using the stable and unstable distances between the encounter stretches and the periodic orbit itself . finally , @xmath105 measures the time difference between the midpoints of the two encounter stretches in fig . [ ana4 ] , and the @xmath105-integral covers the different arrangements of the stretches relative to each other . the limits of the two last integrals in ( [ li1.1 ] ) are not given as they depend on the specific configuration considered below . the first two integrals in ( [ li1.1 ] ) can again be transformed , together with the integrals from the sum rule ( [ in8 ] ) , into a product of link and encounter contributions . the general expression for the contributions to the variance from diagrams containing an enclosed periodic orbit that does not touch the lead becomes @xmath133 like the limits of the @xmath132- and @xmath105-integrals , the function @xmath134 determining the survival probability along the periodic orbit ( and the encountering trajectory stretches ) will be specified for each contribution separately below . for example , for the non - aligned 3-encounter , we can arrange the different alignments in terms of the durations of the two encounters . we let @xmath135 denote the longer encounter and @xmath136 the shorter . in fig . [ ana4](a ) , we do not yet allow the encounter stretches to surround the periodic orbit so we impose the restriction @xmath137 . then we can first consider the case that the shorter encounter lies inside the longer . we will also refer to this case later as a ` generalized ' 3-encounter . the time difference between the midpoints of the encounters therefore satisfies @xmath138 , while the survival probability function in ( [ li1.2 ] ) is given by @xmath139 the different terms simply correspond to the parts of the periodic orbit which are followed by one , two and three trajectory stretches respectively . as this survival probability is independent of @xmath105 , the @xmath105-integral simply yields @xmath140 . performing the remaining integrals , we obtain @xmath141 of course , the range of @xmath105 includes 0 , the case in which the encounters are perfectly aligned . ( [ li1.755 ] ) therefore includes the contribution of the 3-encounter in fig . [ ana3](a ) , and for just the non - aligned 3-encounter we subtract ( [ li0.8 ] ) to give @xmath142 now we consider the case in which the shorter encounter is no longer fully inside the longer , but where the two encounters still overlap . this can be further separated according to the total length of the encounters , @xmath143 , compared to the periodic orbit . when @xmath144 , we have the range @xmath145 .\ ] ] and the survival probability @xmath146 where the terms are now expressed as corrections due to additional correlated stretches . integrating the exponential depending on this survival probability in ( [ li1.2 ] ) with respect to @xmath105 yields @xmath147/\td}\right . \\ & & \left.-{\rm e}^{-\left[p_3t_{{\rm enc , min}}+p_2\left(t_{{\rm enc , max}}-t_{{\rm enc , min}}\right)+\gamma\left(\tau_p - t_{{\rm enc , max}}\right)\right]/\tau_d}\right ) .\nonumber\end{aligned}\ ] ] performing the remaining integrals in ( [ li1.2 ] ) leads to @xmath148 \nonumber\end{aligned}\ ] ] if the encounters are longer than the periodic orbit , @xmath149 , but we still do not allow them to overlap at both ends , we have the restriction @xmath150 , \ ] ] while the survival probability remains as in ( [ li1.3 ] ) . the @xmath105-integral yields @xmath151/\td } \right . \\ & & \left.-{\rm e}^{-\left[p_3t_{{\rm enc , min}}+p_2 \left(t_{{\rm enc , max}}-t_{{\rm enc , min}}\right ) + \gamma\left(\tau_p - t_{{\rm enc , max}}\right)\right]/\td}\right ) , \nonumber \end{aligned}\ ] ] which finally leads to @xmath152\nonumber\\ & & -\frac{1}{\gamma\left(p_3 - 2p_2+\gamma\right ) } \left[\left(p_3-p_2\right)p_2{\rm e}^{-p_3\te/\td}\right.\nonumber\\ & & \left.\left .- p_2\left(p_3-p_2+\gamma\right){\rm e}^{-\left(p_3+\gamma\right ) \te/\td}\right ] \right\rbrace .\end{aligned}\ ] ] as can be easily checked , the sum of the contributions of the non - aligned 3-encounter in eqs . ( [ li1.756],[li1.6],[li1.7 ] ) equals the contribution calculated directly in ( [ li1 ] ) . the reason why we have discussed this more complicated route , is that we can use it to easily calculate the contributions when we start to shrink links and allow the encounter to touch the tunnel barriers . for example , if we remove one link , then by performing analogous steps as explained before ( [ li0.4 ] ) , we obtain @xmath153}&= & \frac{8n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}^2+n_{\rm r}^2\right)^4 } \frac{p_2\left(1-\gamma\right ) } { \left(2p_2-p_3\right)}\left[1+\left(1-\gamma\right)\right ] \nonumber \\ & & \times\left({\rm e}^{-2p_2\te/\td}-{\rm e}^{-p_3\te/\td}\right ) , \end{aligned}\ ] ] where in the square brackets , the @xmath154 results from configurations where the enclosed periodic orbit touches the tunnel barrier where only one encounter stretch is correlated with the orbit , and the @xmath155 from configurations where the periodic orbit touches the tunnel barrier while both encounter stretches are correlated with it . likewise , if two links connecting _ different _ stretches to the lead are removed , we obtain the contribution @xmath156}&= & \frac{4n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}^2+n_{\rm r}^2\right)^4 } \frac{\gamma\left(1-\gamma \right)^2\left(2-\gamma\right)^2 } { \left(2p_2-p_3\right ) } \nonumber \\ & & \times\left({\rm e}^{-p_3\te/\td}-{\rm e}^{-2p_2\te/\td}\right ) . \end{aligned}\ ] ] next we turn to the configuration in fig . [ ana4](b ) with encounter stretches overlapping at both ends . we still have @xmath149 , but a different restriction on @xmath105 : @xmath157 .\ ] ] the survival probability is again independent of @xmath105 , and given by @xmath158 so that the @xmath105-integral yields @xmath159 . performing the remaining integrals we obtain @xmath160}&= & \frac{2n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4 } \left[\frac{p_2 ^ 2}{\left(2p_2-p_3\right)^2}\right . \nonumber\\ & & \times \left({\rm e}^{-2p_2\te/\td}-{\rm e}^{-p_3\te/\td}\right ) \nonumber\\ & & \left . + \frac{p_2\left(p_3-p_2\right)\te}{\left(2p_2-p_3\right)\td } { \rm e}^{-p_3\te/\td } \right].\end{aligned}\ ] ] again the encounter stretches can be brought to the lead by first removing one link @xmath161}&= & \frac{8n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4 } \left[\frac{p_2\gamma(1-\gamma)^2}{\left(2p_2-p_3\right)^2}\right.\nonumber\\ & & \times \left({\rm e}^{-p_3\te/\td}-{\rm e}^{-2p_2\te/\td}\right ) \nonumber \\ & & \left.-\frac{p_3\gamma(1-\gamma)^2\te}{2\left(2p_2-p_3\right)\td } { \rm e}^{-p_3\te/\td } \right ] , \end{aligned}\ ] ] and second by removing two links connecting _ different _ stretches to the opening @xmath162}&= & \frac{8n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4 } \left[\frac { \gamma^2(1-\gamma)^4}{\left(2p_2-p_3\right)^2}\right.\nonumber\\ & & \times\left({\rm e}^{-2p_2\te/\td}-{\rm e}^{-p_3\te/\td}\right ) \nonumber\\ & & \left.+\frac{\gamma^2(1-\gamma)^4\te}{\left(2p_2-p_3\right)\td } { \rm e}^{-p_3\te/\td}\right ] . \end{aligned}\ ] ] as the durations of the encounters are of the order of the ehrenfest time , for vanishing ehrenfest time we only considered the situation where the encounter stretches could be partially reflected from the tunnel barriers in _ one _ lead . however , for increasing ehrenfest time configurations where encounter stretches are partially reflected from the tunnel barriers in _ both _ leads , i.e. where they touch the opening at both ends , become important . if one of the 2-encounters in fig . [ ana1 ] were to be partially reflected at both ends then one of the links between the two encounters would need to tunnel through the barrier and exit the system so that the rest of the diagram could not be completed . in figs . [ ana2 ] and [ ana3 ] , however , as long as the trajectory stretches which follow the periodic orbit are reflected at the tunnel barriers and remain in the system , we can allow the other links to tunnel through the barrier and exit the system . for the base trajectories in fig . [ ana4](a ) this means we can allow both links of one of the base trajectories , on either side of the _ same _ encounter , to shrink into the start and end lead . we start with the configuration in fig . [ ana2 ] and for the calculation we note that the contributions of the different orbit parts in ( [ condvartraj ] ) are multiplicative @xcite . we can therefore reconnect the orbits in such a way that they split into parts whose contributions have previously been calculated . first there is the remaining base trajectory , with a 2-encounter with the enclosed periodic orbit and two links connecting it to the opening . this contributes the factor @xmath163 the rest of the diagram involves the periodic orbit itself and the encountering stretch that tunnels through to start and end in the leads . the contribution is calculated by summing over all enclosed periodic orbits , using the sum rule ( [ in8 ] ) . remember , that when allowing the encounter to move into the lead an additional time integral occurred measuring the part of the stretch that lies still inside the system . now there are two time integrations representing the amount of the encounter which is cut short in each of the two leads . the first time integral cancels @xmath104 , while the second essentially yields a factor @xmath164 $ ] . for the details of the calculation we refer to @xcite . this part of the diagram then contributes @xmath165 so that this configuration of fig . [ ana2 ] altogether yields @xmath166}&= & -\frac{4n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4}(1-\gamma)^2 { \rm e}^{-p_2\te/\td}\nonumber\\ & & \times\left(1-{\rm e}^{-p_2\te/\td}\right),\end{aligned}\ ] ] where the ` ( s ) ' in the superscript denotes that the two links were removed along the same base trajectory from the same encounter stretch . one factor 2 in the last equation derives again from the mirror symmetry of this configuration explained after eq . ( [ li0.7 ] ) , the other from the fact that each of the two stretches can touch the opening at both ends . furthermore , one of the links of the other base trajectory may also be shrunk so that the other encounter tunnels into the lead at one end . this contribution is @xmath167}&= & \frac{8n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4}\frac{\gamma(1-\gamma)^3}{p_2 } { \rm e}^{-p_2\te/\td}\nonumber\\ & & \times\left(1-{\rm e}^{-p_2\te/\td}\right).\end{aligned}\ ] ] removing all 4 links , so that both encounter stretches on the base trajectories tunnel into the leads at both ends , is also possible . the contribution is simply given by the square of ( [ li1.13 ] ) . we can repeat this treatment for the aligned 3-encounter of fig . [ ana3 ] . if one encounter stretch tunnels into the leads at both ends we obtain @xmath168}&= & \frac{8n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4}\frac{\gamma(1-\gamma)^4}{p_3 } \left(1-{\rm e}^{-p_3\te/\td}\right),\nonumber\\\end{aligned}\ ] ] where , because of the alignment and proximity of the two encounter stretches , the periodic orbit and the other encounter stretch must be backreflected at the tunnel barriers . allowing the ends of the second encounter stretch to progressively tunnel through into the leads as well , we have , if it tunnels into the lead at one end @xmath169}&= & \frac{4\left(n_{\rm l}n_{\rm r}^2+n_{\rm l}^2n_{\rm r}\right)}{\left(n_{\rm l}+n_{\rm r}\right)^3 } \frac{\gamma^2(1-\gamma)^3}{p_3}\nonumber\\ & & \times\left(1-{\rm e}^{-p_3\te/\td}\right)\end{aligned}\ ] ] and at both ends @xmath170}= \frac{2n_{\rm l}n_{\rm r}}{\left(n_{\rm l}+n_{\rm r}\right)^2}\frac{\gamma^3(1-\gamma)^2}{p_3 } \left(1-{\rm e}^{-p_3\te/\td}\right).\nonumber\\\end{aligned}\ ] ] such configurations can be also considered for the diagrams in fig.[ana4 ] and their contributions can be calculated analogously . to simplify the results , however , we will later perform an expansion of the contributions in powers of the ehrenfest time and only retain terms up to linear order . as the results for the diagrams of fig . [ ana4 ] are all of higher order in @xmath0 than the linear one , we will not focus on their explicit form here . a further effect , and one that actually causes the independence of the conductance variance of the ehrenfest time for @xmath82 , are correlations during fringes near periodic orbits . encounter fringes refer to regions where the two base trajectories which encounter the periodic orbit are correlated with each other , but are no longer correlated with the periodic orbit itself ; see fig . [ ana5 ] for a schematic depiction . as the two base trajectories leave the periodic orbit correlated with each other , we can consider the encounters with the periodic orbit to be aligned . the encounter of both trajectories with the periodic orbit has length @xmath171 while the fringes , during which the base trajectories are correlated have lengths @xmath172 and @xmath173 where the subscripts refer to the fact that the stable and unstable distances between the base trajectories must be small for them to remain correlated . here , we treat the case in which the encounter length is shorter than the periodic orbit . by this we generalize the calculation of @xcite to @xmath174 . there , it was shown that the contribution derived from ( [ li1.2 ] ) can be expressed as @xmath175}&= & \frac{2n_{\rm l}^2n_{\rm r}^2c^2\lambda}{\left(\pi\hbar\right)^2 \left(n_{\rm l}+n_{\rm r}\right)^4}\int_{1-b}^{b-1}{\rm d}s'{\rm d}u ' \int_0 ^ 1{\rm d}u\,u\nonumber\\ & & \times\int_{t_{\rm enc}}^\infty { \rm d}\tau_p { \rm e}^{-\left[\left(p_3-\gamma\right){t_{\rm enc}}+\gamma\tau_p\right]/\td } \nonumber\\ & & \times\left(\left|\frac{s'}{b-1}\right|^{\frac{p_2}{\lambda\td}}-1\right ) \left(\left|\frac{u'}{b-1}\right|^{\frac{p_2}{\lambda\td}}-1\right)\nonumber\\ & & \times\cos\left[\frac{c^2u\left(s'-u'\right)}{\hbar}\right]\end{aligned}\ ] ] where @xmath31 is again a classical constant of order unity . the stable and unstable coordinates are defined as in @xcite : as correlation effects away from the periodic orbits get only important when the encountering orbits approach and leave this at classically close points one first characterize the correlation of all three orbit pieces ( i.e. the periodic orbit and the two encountering orbits ) by the coordinate @xmath176 . furthermore the difference between the stable and the unstable coordinates of the encountering orbits at the beginning and the end of the encounter is denoted by @xmath177 and @xmath178 . these coordinates characterize the time the encountering orbits remain correlated before and after they are correlated with the periodic orbit , respectively . in ( [ li1.19 ] ) , we included the correct dwell times for the different parts of the trajectory : @xmath76 for the isolated links , @xmath94 for the fringes and @xmath179 for the encounter stretches correlated with the periodic orbit . following the steps in @xcite , this contribution evaluates to @xmath180}&= & \frac{2n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4 } \frac{p_2 ^ 2}{\gamma\left(2p_2-p_3\right)}\nonumber\\ & & \times\left({\rm e}^{-p_3\te/\td}-{\rm e}^{-2p_2\te/\td}\right).\end{aligned}\ ] ] without tunnel barriers , as shown in @xcite , further contributions can be obtained when the fringes start in the leads . similarly , with tunnel barriers , we can have the fringes tunnel into the leads . this leads to an additional time integral over the duration of the fringe that remains inside the system . when one fringe tunnels into the lead , two links are removed and the contribution becomes @xmath181}&= & -\frac{2\left(n_{\rm l}^2n_{\rm r}+n_{\rm l}n_{\rm r}^2\right)}{\left(n_{\rm l}+n_{\rm r}\right)^3 } \frac{\gamma p_2}{\left(2p_2-p_3\right)}\nonumber\\ & & \times\left({\rm e}^{-p_3\te/\td}-{\rm e}^{-2p_2\te/\td}\right).\end{aligned}\ ] ] the ` ( s ) ' in the superscript refers to the fact that the links were removed from the same side of the encounter . when both fringes , and all four links are removed , we have @xmath182}&= & \frac{2n_{\rm l}n_{\rm r}}{\left(n_{\rm l}+n_{\rm r}\right)^2 } \frac{\gamma^3}{\left(2p_2-p_3\right)}\nonumber\\ & & \times\left({\rm e}^{-p_3\te/\td}-{\rm e}^{-2p_2\te/\td}\right).\end{aligned}\ ] ] with the tunnel barriers though , new possibilities also arise . for example , when a fringe tunnels into the lead , one of the trajectories could be backscattered , as for the 2-encounters previously . similarly to ( [ li0.4 ] ) , we obtain here @xmath183}&= & -\frac{8n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4 } \frac{(1-\gamma)p_2}{\left(2p_2-p_3\right)}\nonumber\\ & & \times\left({\rm e}^{-p_3\te/\td}-{\rm e}^{-2p_2\te/\td}\right).\end{aligned}\ ] ] furthermore we can move additionally one fringe into the lead yielding @xmath184}&= & \frac{4\left(n_{\rm l}n_{\rm r}^2+n_{\rm l}^2n_{\rm r}\right)}{\left(n_{\rm l}+n_{\rm r}\right)^3 } \frac{(1-\gamma)\gamma^2}{\left(2p_2-p_3\right)}\nonumber\\ & & \times\left({\rm e}^{-p_3\te/\td}-{\rm e}^{-2p_2\te/\td}\right).\end{aligned}\ ] ] if both fringes are partially backscattered at the leads , we have the contribution @xmath185}&= & \frac{8n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4}\frac{\gamma(1-\gamma)^2 } { \left(2p_2-p_3\right)}\nonumber\\ & & \times\left({\rm e}^{-p_3\te/\td}-{\rm e}^{-2p_2\te/\td}\right).\end{aligned}\ ] ] next we can allow the periodic orbit itself , rather than the fringes , to touch the lead and be reflected from the tunnel barriers . the encounter stretches may then both tunnel into the lead , or one may be reflected . we recall that the case in which both are reflected is already included in the survival probability of the encounter . one fringe is therefore removed from the systems , while the fringe that remains could then be inside the system , tunnel into the lead or be partially backreflected at the opening . if the periodic orbit touches both leads during the encounter , we return to the situation in the previous subsection , so here we only consider the cases in which at least some of the second fringe remains . the removal of one fringe entirely implies the following changes in ( [ li1.19 ] ) : with only one fringe time left , we only have one of the factors in the third line of ( [ li1.19 ] ) . also the encounter time is replaced by an integration variable ( with a range from @xmath186 to @xmath187 ) which accounts for the amount of the encounter left after the periodic orbit is reflected from the tunnel barriers . performing , with these changes , again the steps in @xcite analogous to those following ( [ li1.19 ] ) above , we obtain the contributions for the various different cases . first , when the remaining fringe lies inside the system , we could have the encounter ( which touches the lead ) partially reflected @xmath188}&= & -\frac{4n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4}\frac{(1-\gamma)^2p_2}{p_3}\nonumber\\ & & \times\left(1-{\rm e}^{-p_3\te/\td}\right){\rm e}^{-p_2\te/\td},\end{aligned}\ ] ] where ` -1f-1l ' refers to the fact that one fringe and one link were removed . likewise , both encounter stretches could tunnel directly into the lead so that two links are removed on the same side of the encounter @xmath189}&= & -\frac{\left(n_{\rm l}n_{\rm r}^2+n_{\rm l}^2n_{\rm r}\right)}{\left(n_{\rm l}+n_{\rm r}\right)^3 } \frac{(1-\gamma)p_2\gamma}{p_3}\nonumber\\ & & \times\left(1-{\rm e}^{-p_3\te/\td}\right){\rm e}^{-p_2\te/\td}.\end{aligned}\ ] ] the remaining fringe could also be paritally reflected from the tunnel barriers . for the encounter stretches have the same two possibilities as above . this leads to @xmath190}&= & \frac{8n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4}\frac{(1-\gamma)^3\gamma}{p_3}\nonumber\\ & & \times\left(1-{\rm e}^{-p_3\te/\td}\right){\rm e}^{-p_2\te/\td } , \end{aligned}\ ] ] and @xmath191}&= & \frac{2\left(n_{\rm l}^2n_{\rm r}+n_{\rm l}n_{\rm r}^2\right)}{\left(n_{\rm l}+n_{\rm r}\right)^3 } \frac{(1-\gamma)^2\gamma^2}{p_3}\nonumber\\ & & \times\left(1-{\rm e}^{-p_3\te/\td}\right){\rm e}^{-p_2\te/\td}.\end{aligned}\ ] ] finally the remaining fringe can tunnel directly into the lead , giving the contributions @xmath192}&= & \frac{2\left(n_{\rm l}^2n_{\rm r}+n_{\rm l}n_{\rm r}^2\right)}{\left(n_{\rm l}+n_{\rm r}\right)^3 } \frac{(1-\gamma)^2\gamma^2}{p_3}\nonumber\\ & & \times\left(1-{\rm e}^{-p_3\te/\td}\right){\rm e}^{-p_2\te/\td } , \end{aligned}\ ] ] and @xmath193}&= & \frac{2n_{\rm l}n_{\rm r}}{\left(n_{\rm l}+n_{\rm r}\right)^2 } \frac{(1-\gamma)^2\gamma^3}{p_3}\nonumber\\ & & \times\left(1-{\rm e}^{-p_3\te/\td}\right){\rm e}^{-p_2\te/\td}.\end{aligned}\ ] ] having treated all encounter configurations with periodic orbits where the encounter stretches are shorter than the enclosed periodic orbit we now turn to the corresponding contributions where the encounter stretches are longer than the enclosed periodic orbit . in this context only diagrams with stretches that both possess the same number of traversals around the enclosed periodic orbit yield a contribution , for a justification see @xcite . in this context we consider that each encounter stretch has @xmath194 full windings around the enclosed periodic orbit . we take here the two encounter times @xmath195 , @xmath196 and the corresponding primitive times @xmath197 . these primitive times are again shorter than the enclosed periodic orbit and we can now also consider the different cases depicted in figs . [ ana2][ana5 ] , and how they can be arranged . we start with the case in which the two primitive encounter times , @xmath198 , do not overlap . when @xmath199 , the corresponding diagram is depicted in fig . [ ana2 ] . to explain the calculation we begin with eq . ( [ li1.2 ] ) for @xmath199 and perform the integrals over the links @xmath200 , and over @xmath105 where the latter integral leads to the factor @xmath201 : @xmath202}_{k=0}&= & \frac{n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^{4 } } \int_{-c}^c { \rm d}\boldsymbol{s}{\rm d}\boldsymbol{u } \frac{\th^2}{\omega^2}\frac{{\rm e}^{\frac{{\rm i}}{\hbar}\boldsymbol{s}\boldsymbol{u } } } { t_{{\rm enc},1 } t_{{\rm enc},2 } } \nonumber \\ & & \times\int_{t_{{\rm encs}}}^\infty { \rm d}\tau_p\left(\tau_p - t_{{\rm encs}}\right ) \nonumber \\ & & \times { \rm e}^{-\left[\left(p_2-\gamma\right)t_{{\rm encs } } + \gamma \tau_p\right]/\td } .\end{aligned}\ ] ] to turn to @xmath203 , @xmath104 is then replaced only inside the integrand of the @xmath132-integral by @xmath198 , these times are shorter than @xmath132 by definition . the @xmath104 before the @xmath132-integral compensate the overcounting of equivalent positions of the poincar surface of sections by the @xmath204-integrals and thus do not need to be altered . the possibilities for placing these stretches around the periodic orbit are the same as they were for @xmath104 for @xmath199 . this allows us to treat @xmath205 in essentially the same way we treated @xmath199 before . as @xmath206 , we reexpress the primitive encounter times in terms of @xmath104 and @xmath132 . the limits of the @xmath132-integration are also altered for @xmath203 : as the primitive encounters do not ( yet ) overlap , we have the condition @xmath207 so that the lower limit is now @xmath208 . since we remove all @xmath194 complete windings of the periodic orbit from both encounter times , the shorter encounter must be @xmath209 so that the upper limit is @xmath210 . finally @xmath211 is replaced by @xmath212 ( where in this context @xmath3 is defined as @xmath213 ) since during the primitive encounter stretches we have a total of @xmath214 correlated stretches while elsewhere we have @xmath215 . all these replacements finally yield @xmath216}_{{\rm mt}}&= & \frac{n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^{4 } } \int_{-c}^c { \rm d}\boldsymbol{s}{\rm d}\boldsymbol{u } \frac{\th^2}{\omega^2}\frac{{\rm e}^{\frac{{\rm i}}{\hbar}\boldsymbol{s}\boldsymbol{u } } } { t_{{\rm enc},1 } t_{{\rm enc},2 } } \\ & & \times\int_{\frac{t_{{\rm encs}}}{2k+1}}^{\frac{t_{{\rm enc , min}}}{k } } { \rm d}\tau_p\left[(2k+1)\tau_p - t_{{\rm encs}}\right]\nonumber\\ & & \times{\rm e}^{-\left[\left(p_{2k+2}-p_{2k+1}\right ) \left(t_{{\rm encs}}-2k\tau_p\right)+p_{2k+1}\tau_p\right]/\td } , \nonumber\end{aligned}\ ] ] where the additional index ` mt ' indicates that this contribution results from multiple traversals @xmath205 . performing now the remaining integrals in the same way as for example in eqs . ( [ li1.4][li1.7 ] ) we obtain this contribution . as the full expression is rather involved we just give here terms up to linear order in @xmath0 : @xmath217}_{\rm mt } & = & -\frac{2n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4}\sum_{k=1}^\infty \frac{\left(p_{2k+2}-p_{2k+1}\right)\te}{k^2(2k+1)\td } \nonumber \\ & & { } + { \mathcal o}\left(\te^2\right ) .\end{aligned}\ ] ] we also omitted here terms that are also nonzero for @xmath218 as they all cancel with the contributions below . also in this case one link can be removed leading to a configuration where one encounter stretch touches the opening . this implies that in ( [ li1.23.5 ] ) one factor @xmath219 is removed , the duration of the encounter stretch which touches the tunnel barrier is replaced by an integration variable integrated from @xmath186 to @xmath104 , and additionally a factor @xmath220 arises to account for the probability that all the @xmath215 orbital parts surrounding the periodic orbit are backreflected when they hit the lead . this contribution then yields @xmath221}_{\rm mt}&= & \frac{4n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4}\sum_{k=1}^\infty \frac{\left(p_{2k+2}-p_{2k+1}\right)\te}{k^2(2k+1)\td}\nonumber\\ & & { } + { \mathcal o}\left(\te^2\right ) . \end{aligned}\ ] ] shrinking two links connecting two _ different _ encounter stretches to the opening , we obtain ( following the same steps as just described also for the other @xmath104 ) no contribution which is linear in @xmath222 . shrinking two links connecting the _ same _ encounter we need again to introduce in ( [ li1.23.5 ] ) two additional time integrals as described before ( [ li1.13 ] ) . this also finally yields zero contribution linear in @xmath0 , because the terms resulting from the two limits of the @xmath132-integral cancel . the same also holds for contributions following . similar contributions are also obtained in the other cases . for a generalized 3-encounter introduced around eq . ( [ li1.75 ] ) we obtain the integrals given in ( [ ap2 ] ) which provide the contribution @xmath223}_{\rm mt}&= & \frac{2n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4}\sum_{k=1}^\infty \frac{\te}{k(k+1)\td } \\ & & \times \left(2p_{2k+2}-p_{2k+1}-p_{2k+3}\right ) + { \mathcal o}\left(\te^2\right ) , \nonumber\end{aligned}\ ] ] when the encounters lie inside the system and @xmath224}_{\rm mt}&= & -\frac{4n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4}\sum_{k=1}^\infty \frac{\te}{k(k+1)\td } \\ & & \times\left(2p_{2k+2}-p_{2k+1}-p_{2k+3}\right ) + { \mathcal o}\left(\te^2\right ) , \nonumber \end{aligned}\ ] ] when one link is removed . additionally we can shrink both links connecting the 3-encounter to the opening on one side . this amounts to removing two links , and introducing an additional integral over the part of the encounter that remains inside the system , in ( [ ap2 ] ) , yielding @xmath225}_{\rm mt}&= & \frac{2\left(n_{\rm l}n_{\rm r}^2+n_{\rm l}^2n_{\rm r}\right)}{\left(n_{\rm l}+n_{\rm r}\right)^3 } \sum_{k=1}^\infty\frac{\te}{k(k+1)\td } \\ & & \times\left(2p_{2k+2}-p_{2k+1}-p_{2k+3}\right ) + { \mathcal o}\left(\tau_e^2\right ) . \nonumber\end{aligned}\ ] ] again zero contribution is obtained when more links are removed . for a non - aligned 3-encounter , when the encounter lies inside the system we obtain from eqs . ( [ ap31],[ap32 ] ) @xmath226}_{\rm mt}&= & \frac{4n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4}\sum_{k=1}^\infty \left[\frac{\left(p_{2k+3}-p_{2k+2}\right)\te}{(k+1)(2k+1)\td}-\right.\nonumber\\ & & \left.-\frac{\left(p_{2k+2}-p_{2k+1}\right)\te}{k(2k+1)\td}\right ] + { \mathcal o}\left(\te^2\right ) \end{aligned}\ ] ] and when one link is removed : @xmath227}_{\rm mt}&= & -\frac{8n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4}\sum_{k=1}^\infty \left[\frac{\left(p_{2k+3}-p_{2k+2}\right)\te}{(k+1)(2k+1)\td}\right.\nonumber\\ & & + \frac{\left(2p_{2k+2}-p_{2k+1}-p_{2k+3}\right)\te}{2k(k+1)\td}\nonumber\\ & & \left.-\frac{\left(p_{2k+2}-p_{2k+1}\right)\te}{k(2k+1)\td}\right ] + { \mathcal o}\left(\te^2\right).\end{aligned}\ ] ] finally we treat the configuration in which the encounter stretches overlap at both ends . when @xmath199 this configuration is depicted in fig . [ ana4](b ) . with multiple traversals of the periodic orbit , the contribution given in eq . ( [ ap4 ] ) provides , when the stretches lie inside the system , @xmath228}_{\rm mt}&= & -\frac{2n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4}\sum_{k=1}^\infty \frac{\left(p_{2k+3}-p_{2k+2}\right)\te}{(k+1)^2(2k+1)\td}\nonumber\\ & & + { \mathcal o}\left(\te^2\right)\end{aligned}\ ] ] and when one link is removed @xmath229}_{\rm mt}&= & \frac{4n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4}\sum_{k=1}^\infty \frac{\left(p_{2k+3}-p_{2k+2}\right)\te}{(k+1)^2(2k+1)\td}\nonumber\\ & & + { \mathcal o}\left(\te^2\right).\end{aligned}\ ] ] no further contributions are obtained when taking into account fringes as in fig . [ ana5 ] . this can be checked from eq . ( [ li1.19 ] ) by adjusting the limits of the @xmath132-integral appropriately . then , the terms from the two limits of the @xmath132-integral cancel to linear order in @xmath0 . having calculated all the contributions we can now sum them to obtain the overall contributions to the conductance variance , to leading order in the number of open channels in the leads . first we can check that all the terms after sec . [ discrete ] are zero for @xmath122 , so that the rmt result ( [ li22.1 ] ) is preserved . the case that is especially interesting for the comparison with the numerics is the contribution to the conductance variance when @xmath230 , for arbitrary @xmath3 and of linear order in @xmath0 . we therefore consider for @xmath231 , in the presence of time reversal symmetry , @xmath232+{\alpha}\frac{\te}{\td}.\end{aligned}\ ] ] we obtain @xmath233 \end{aligned}\ ] ] by summing the contributions linear in @xmath222 from eqs . ( [ li0.3]-[li0.91 ] , [ li1 ] , [ li1.76]-[li1.8 ] , [ li1.10]-[li1.12 ] , [ li1.14]-[li1.16 ] , [ li1.20]-[li1.2297 ] , [ li1.24]-[li1.32 ] ) . here @xmath234 denotes the polylogarithmic function , which arises when performing the @xmath194-summations above . the function @xmath235 is shown as the solid ( red ) line in fig . [ condfluc ] . having derived the complete set of contributions , we now discuss the main terms that lead to the two extrema in fig . [ condfluc ] . first we note that the contributions from multiply - traversed periodic orbits are quite small . when @xmath82 they provide no contribution linear in @xmath0 , as can be checked from eqs . ( [ li1.24]-[li1.32 ] ) . for most of the contributions for @xmath85 the index @xmath194 of the @xmath236 s in the sum is much larger than one . in this case the differences of the @xmath236 s in eqs . ( [ li1.24]-[li1.32 ] ) tend to one and thus the contributions from these equations tend to zero . then the result is almost equal to the vanishing value for @xmath82 . in fig . [ condfluc ] , the maximum lies in a region of small @xmath3 . in this regime , most of the diagrams approximately cancel each other as can be seen by expanding their contributions in a taylor series for small @xmath3 : the ones from diagrams containing encounters inside the system [ e.g. the contribution ( [ li0.3 ] ) ] are canceled by the contributions obtained when one and two links are shrunk [ e.g. the contributions ( [ li0.4][li0.7 ] ) ] . however , this cancellation does not hold for the 3-encounter with the periodic orbit ( see fig . [ ana3 ] ) . in this case we have two possibilities : first that the encounter lies inside the system or touches the lead at one end [ see eqs . ( [ li0.8][li0.91 ] ) ] . in total this leads to an ehrenfest - time dependent factor of @xmath237 . additionally , we also need to take into account the contributions from encounters touching the leads on both sides leading to an increasing contribution proportional to @xmath238 with a larger prefactor than the first one [ see eqs . ( [ li1.18][li1.16 ] ) ] . together these cause the first peak in fig . [ condfluc ] . in the case of the dip the factor @xmath108 , i.e. the probability for the particle to be backreflected at the opening , is quite small . this implies that we get the main contributions from diagrams with encounters inside the system ( i.e. the ones also obtained for @xmath82 ) depicted in figs . [ ana1][ana5 ] . these together yield a negative contribution . we numerically confirm our main analytical prediction , eqs . ( [ li22.9 ] ) and ( [ li23 ] ) . the model we use is the open kicked rotator with time - dependent hamiltonian @xcite @xmath239 with @xmath240 the free flight time . depending on the kicking strength @xmath241 the dynamics changes from integrable for @xmath242 to fully chaotic for @xmath243 . in the latter regime , local exponential instability is characterized by the lyapunov exponent @xmath244 as a function of @xmath245 for various values of @xmath3 . the straight lines are linear fits constrained to go to @xmath246 at @xmath122 . their slope give the parameter @xmath19 defined in eqs . ( [ li22.9 ] ) and ( [ li23]).,scaledwidth=47.0% ] the quantities @xmath247 and @xmath248 are introduced to break the hamiltonian s two symmetries @xcite , and investigate different symmetry classes with ( @xmath247 or @xmath249 ) and without ( @xmath250 ) time reversal symmetry . the ehrenfest time in this system is determined , within a constant of little relevance , by @xmath251 with @xmath252 the size of the hilbert space , determined by the quantization of the hamiltonian via discretization of the coordinates as , e.g. @xmath253 , @xmath254 . here @xmath44 is a system - dependent constant of order one that is of classical origin , and as such does not depend on @xmath252 . a quantum representation of the hamiltonian ( [ num0 ] ) is then provided by the unitary @xmath255 floquet matrix @xmath256 , giving the time evolution for one iteration of the map defined by @xmath257 in a time interval @xmath258 . for our specific choice of the kicked rotator , the floquet operator has matrix elements @xmath259 } \\ & \times & \sum_m e^{2 \pi i m(l - l')/m } e^{-(imk/4\pi ) \cos(2\pi ( m+m_0)/m ) } \nonumber\end{aligned}\ ] ] with @xmath260 and @xmath261 . transport can finally be investigated once absorbing phase - space strips are introduced to model contacts to leads . this is achieved by means of projection operators @xmath134 . in our case of tunnel - coupled leads , the latter are @xmath262 diagonal matrices with entries @xmath263 , assuming that the system is coupled to all @xmath264 channels with the same transparency @xmath265 . the scattering matrix is finally defined as @xcite @xmath266^{-1 } u p^t \ , .\ ] ] the kicked rotator model is particularly performant to investigate the semiclassical limit with @xmath267 , as it allows for a rather large variation of the system size and hence the ehrenfest time . previous such investigations of the open kicked rotator are presented e.g. in refs . @xcite , and we refer the reader to these works for further details of the model . the variance of the conductance is calculated by varying the quasienergies @xmath268 at given lead position . averages of @xmath16 are further performed with different lead positions . we first show in fig . [ condflucraw ] the behavior of @xmath11 as a function of @xmath269 . we clearly see @xmath3-dependent behaviors , as @xmath16 increases for small @xmath3 and decreases for larger @xmath3 until it becomes independent of @xmath0 for @xmath82 . we extract the linear slope of the @xmath11 vs. @xmath269 curve via a linear fit , setting @xmath122 where the curves intersect the rmt universal value @xmath270 @xcite . further dividing this slope by @xmath3 gives us @xmath19 . there is still an uncertainty in @xmath19 due to an uncertainty in the precise value of the lyapunov exponent the latter has been found numerically to deviate in open systems from its exact value ( [ num1 ] ) , which has been attributed to finite - time effects and because of the arbitrary constant @xmath44 of order one in the definition of @xmath0 . we remove this uncertainty by forcing numerical and analytical data to agree for @xmath271 . once this is done , there is no free parameter left . we compare the so obtained numerical values for @xmath19 with the analytical prediction of eq . ( [ li23 ] ) in fig . [ condfluc ] . the excellent agreement between numerical data and the analytical curve fully confirms our theory . we also checked , but do not show , that the linear @xmath0-dependence of @xmath272 is halved when time - reversal symmetry is broken . in this article we determined the dependence of the conductance variance of a chaotic cavity with tunnel barriers on the tunneling probability @xmath3 . in particular we find an ehrenfest - time dependence for the general case @xmath273 of non - perfect coupling . we focused on the contribution which is linear in the ehrenfest time and at leading order in the total number of open channels @xmath83 . we predict a nonmonotonous sinusoidal behavior with one maximum at @xmath274 and a ( roughly 50% ) deeper minimum at @xmath275 . this analytical result was derived semiclassically by systematically considering a rather large number of possible configurations , but the general behavior derives from a much smaller set . we discussed the main contributions which lead to each of the two peaks to provide a better intuitive understanding of the overall structure . finally , we compared these analytical predictions with numerical simulations performed for the kicked rotor ( in the chaotic regime ) . there was good agreement between the two , with the analytical curve within the error bars of all the numerically determined datapoints . although , even at leading order , we treated a large number of possible semiclassical diagrams , they could be constructed , as we showed in this article , in a controlled way from the diagrams that exist when the tunnel barriers are absent . this calculation could therefore possibly be extended to higher order contributions in @xmath276 and @xmath0 , but a more natural next step would be to extend the ehrenfest - time dependence of the ( leading order in @xmath83 ) results for the full counting statistics of transport moments @xcite to include tunnel barriers in the leads and to account for the extra diagrammatic possibilities that then arise . on the numerical side , we emphasize that the agreement observed in this article is the first observed for the conductance properties of systems with tunnel barriers . neither the results for the conductance nor for shot noise obtained in @xcite could yet be confirmed by numerical simulations . further numerical investigation in this area would therefore be highly desirable . finally it would be interesting to check the observed phenomena experimentally . although ehrenfest - time effects in antidot superlattices were observed more than ten years ago @xcite , none of the other ehrenfest - time dependencies predicted for chaotic systems could be checked so far experimentally . this has been mainly due to the small range through which the system size can be varied in quantum dots . however , with the additional parameter of the tunneling probability and in view of the expected double peak structure , the prediction in this article could be amenable to experimental verification . this work was funded by the deutsche forschungsgemeinschaft ( research unit for 760 ) ( dw and kr ) and by the alexander von humboldt foundation ( jk ) . pj has been partially supported by the nsf under grant no dmr-0706319 and by the swiss center of excellence manep . it is a great pleasure to thank c. petitjean for discussions and for drawing our attention to this problem . furthermore we thank r. whitney for discussions at the early stage of this project . here we give the necessary integrals for obtaining the contributions from diagrams containing enclosed periodic orbits that are surrounded @xmath194-times by each encounter stretch . in these expressions the link and @xmath105-integrals in eq . ( [ li1.2 ] ) have already been performed , but not yet the @xmath132- and the @xmath277-integrals . these expressions are especially useful because they not only allow one to obtain the contributions from encounters inside the system but ( via integrations with respect to the corresponding encounter times ) also the contributions when encounters touch the openings . the corresponding expression obtained for two independent 2-encounters was already given in the main text in eq . ( [ li1.23.5 ] ) . in the case of a generalized 3-encounter we obtain @xmath278}_{\rm mt}&= & \frac{2n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4 } \int_{-c}^c{\rm d}\boldsymbol{s}{\rm d}\boldsymbol{u } \int_{\frac{t_{{\rm enc , max}}}{k+1}}^{\frac{t_{{\rm enc , min}}}{k}}{\rm d}\tau_p\frac{\th^2}{\omega^2}\nonumber\\ & & \times\frac{{\rm e}^{\frac{{\rm i}}{\hbar}\boldsymbol{s}\boldsymbol{u } } } { t_{{\rm enc},1}t_{{\rm enc},2}}\left(t_{\rm enc , max}-t_{\rm enc , min}\right)\nonumber\\ & & \times{\rm e}^{-\left(p_{2k+3}-p_{2k+2}\right)\left(t_{\rm enc , min}-k\tau_p\right)/\td}\\ & & \times{\rm e}^{-\left[\left(p_{2k+2}-p_{2k+1}\right)\left(t_{\rm enc , max}-k\tau_p\right)+p_{2k+1}\tau_p\right]/\td}\nonumber.\end{aligned}\ ] ] for a non - aligned 3-encounter , similar to ( [ li1.4],[li1.5 ] ) , it is given first for @xmath279 by @xmath280}_{\rm mt(1)}&= & \frac{2n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4}\frac{2\td}{\left(p_{2k+3}+p_{2k+1}-2p_{2k+2}\right)}\nonumber \\ & & \times \int_{-c}^c{\rm d}\boldsymbol{s}{\rm d}\boldsymbol{u } \frac{\th^2}{\omega^2 } \frac{{\rm e}^{\frac{{\rm i}}{\hbar}\boldsymbol{s}\boldsymbol{u } } } { t_{{\rm enc},1}t_{{\rm enc},2 } } \int_{\frac{t_{{\rm encs}}}{2k+1}}^{\frac{t_{{\rm enc , min}}}{k } } { \rm d}\tau_p\nonumber\\ & & \times{\rm e}^{-p_{2k+1}\tau_p/\tau_d}\nonumber\\&&\times\left[{\rm e}^{-\left(p_{2k+2}-p_{2k+1}\right ) \left(t_{{\rm encs}}-2k\tau_p\right)/\td}\right.\nonumber\\ & & { } -{\rm e}^{-\left(p_{2k+3}-p_{2k+2}\right ) \left(t_{\rm enc , min}-k\tau_p\right)/\td } \nonumber\\ & & \left.\times{\rm e}^{-\left(p_{2k+2}-p_{2k+1}\right ) \left(t_{\rm enc , max}-k\tau_p\right)/\td}\right ] , \end{aligned}\ ] ] and second for @xmath281 by @xmath282}_{\rm mt(2)}&= & \frac{2n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4}\frac{2\td}{\left(p_{2k+3}+p_{2k+1}-2p_{2k+2}\right)}\nonumber \\ & & \times \int_{-c}^c{\rm d}\boldsymbol{s}{\rm d}\boldsymbol{u } \frac{\th^2}{\omega^2 } \frac{{\rm e}^{\frac{{\rm i}}{\hbar}\boldsymbol{s}\boldsymbol{u } } } { t_{{\rm enc},1}t_{{\rm enc},2 } } \int_{\frac{t_{{\rm enc , max}}}{k+1}}^{\frac{t_{{\rm encs}}}{2k+1 } } { \rm d}\tau_p \nonumber\\ & & \times\left[{\rm e}^{-\left ( 2p_{2k+2}-p_{2k+3}\right)\tau_p/\td}\right.\nonumber\\ & & \times{\rm e}^{-\left(p_{2k+3}-p_{2k+2}\right ) \left(t_{{\rm encs}}-2k\tau_p\right)/\td } \nonumber\\ & & { } -{\rm e}^{-\left(p_{2k+3}-p_{2k+2}\right ) \left(t_{\rm enc , min}-k\tau_p\right)/\td } \nonumber \\ & & \times{\rm e}^{-\left(p_{2k+2}-p_{2k+1}\right ) \left(t_{\rm enc , max}-k\tau_p\right)/\td } \nonumber \\ & & \left.\times{\rm e}^{-p_{2k+1}\tau_p/\td}\right].\end{aligned}\ ] ] in the case of the two encounters overlapping at both ends the contribution is obtained to be @xmath283}_{\rm mt}&=&\frac{2n_{\rm l}^2n_{\rm r}^2}{\left(n_{\rm l}+n_{\rm r}\right)^4 } \int_{-c}^c{\rm d}\boldsymbol{s}{\rm d}\boldsymbol{u } \int_{\frac{t_{{\rm enc , max}}}{k+1}}^{\frac{t_{{\rm encs}}}{2k+1 } } { \rm d}\tau_p \frac{\th^2}{\omega^2 } \nonumber\\ & & \times\frac{{\rm e}^{\frac{{\rm i}}{\hbar}\boldsymbol{s}\boldsymbol{u } } } { t_{{\rm encs}}}\left[t_{{\rm encs}}-\left(2k+1\right)\tau_p\right]\nonumber\\ & & \times{\rm e}^{-\left(p_{2k+3}-p_{2k+2}\right ) \left(t_{{\rm encs}}-2k\tau_p\right)/\td}\nonumber\\ & & \times{\rm e}^{-\left(2p_{2k+2}-p_{2k+3}\right)\tau_p/\td}.\end{aligned}\ ] ] here we generalize eq . ( [ li22.1 ] ) to the case of different @xmath284 for the different lead modes . factors like @xmath285 are then replaced by a sum over the @xmath284 with respect to the @xmath83 open channels . an analogous replacement is made for the @xmath211 for @xmath286 . in order to keep the notation compact we introduce @xmath287 for example , the contribution from the diagram in fig . [ ana1](a ) , previously given in ( [ li0.3 ] ) , becomes @xmath288 } = \frac{g_{(1)}^{2}g_{(2)}^{2}g_{2}^{2}}{g_{1}^{6}}.\ ] ] where the ehrenfest - time dependence is the same as in ( [ li0.3 ] ) although we set @xmath122 here . the first two terms in the numerator result from the channel summations . the third term replaces the factor @xmath289 in ( [ li0.3 ] ) and the denominator replaces the previous @xmath290 . to keep the contributions from configurations where one link connecting the encounter to the opening is removed in a compact form , we define @xmath291 considering again the corresponding contribution from fig.[ana1](a ) , previously given in ( [ li0.4 ] ) , we have @xmath292 } = -\frac{2\left[h_{(1),2}g_{(2)}+g_{(1)}h_{(2),2}\right]g_{2 } g_{(1)}g_{(2)}}{g_{1}^{5}}.\ ] ] the first term derives from when the link connecting the encounter to lead 1 is shrunk while the second term corresponds to when the link connecting an encounter to lead 2 is removed . compared to ( [ ap6 ] ) , the factor @xmath293 takes into account that the particle is entering the system in a certain channel and returns to the same channel after traversing a link . removing two links from different encounters likewise leads to @xmath294 } = \frac{4g_{(1)}h_{(1),2}h_{(2),2}g_{(2)}}{g_{1}^{4}}.\ ] ] for the diagram in fig . [ ana1](a ) we can also remove two links from the same encounter , and for this we define @xmath295 so that we obtain @xmath296}\!&=&\ ! -\frac{\left[i_{(1),1}g_{(2)}^{2}+g_{(1)}^{2}i_{(2),1}\right]g_{2}}{g_{1}^{4 } } , \\ \label{ap8.3 } { \rm var}\,g^{[{\ref{ana1}(a)-3{\rm l}}]}\!&=&\ ! \frac{2\left[g_{(1)}h_{(1),2}i_{(2),1}+i_{(1),1}h_{(2),2}g_{(2)}\right]}{g_{1}^{3 } } , \nonumber\\ \\ \label{ap8.4 } { \rm var}\,g^{[{\ref{ana1}(a)-4{\rm l}}]}\!&=&\ ! \frac{i_{(1),1}i_{(2),1}}{g_{1}^{2 } } .\end{aligned}\ ] ] for the diagram in fig . [ ana1](b ) we can not shrink two links attached to the same encounter ( or more than two links ) and because of the way the encounters are arranged when we shrink two links we obtain @xmath297 } & = & \frac{2g_{(1)}h_{(1),2}h_{(2),2}g_{(2)}}{g_{1}^{4 } } \\ & & { } + \frac{h_{(1),2}^{2}g_{(2)}^{2}+g_{(1)}^{2}h_{(2),2}^{2}}{g_{1}^{4 } } . \nonumber\end{aligned}\ ] ] using the definitions above , this is equal to ( [ ap8.1 ] ) plus the results from ( [ ap8.2][ap8.4 ] ) so that time reversal symmetry still leads simply to a factor of 2 . in an analogous manner , we can show that each diagram in fig . [ ana2 ] gives twice the contribution of the diagram in fig . [ ana1](b ) while for the diagrams in fig . [ ana3 ] we obtain @xmath298 } & = & -\frac{2g_{(1)}^{2}g_{(2)}^{2}g_{3}}{g_{1}^{5}},\\ \label{ap8.7 } { \rm var}\,g^{[{\ref{ana3}-1{\rm l } } ] } & = & \frac{4\left[g_{(1)}h_{(1),3}g_{(2)}^{2}+g_{(1)}^{2}h_{(2),3}g_{(2)}\right ] } { g_{1}^{4}},\nonumber\\ \\ \label{ap8.8 } { \rm var}\,g^{[{\ref{ana3}-2{\rm l } } ] } & = & \frac{2\left[i_{(1),2}g_{(2)}^{2}+g_{(1)}^{2}i_{(2),2}\right]}{g_{1}^{3 } } .\end{aligned}\ ] ] summing all these contributions , we obtain the rmt result @xcite for the leading order in @xmath83 contribution to the conductance variance @xmath299 , \end{aligned}\ ] ] for systems without time reversal symmetry , and twice this result for those with . here the notation introduced in @xcite @xmath300 was used . here we calculate the shot noise power . it can be written as @xcite @xmath301\right\rangle - \left\langle { \rm tr}\left[\left(\mathbf{tt}^{\dagger}\right)^2\right]\right\rangle.\ ] ] the average conductance @xmath20 has previously been calculated @xcite , and the first few terms for systems with time reversal symmetry can be written as @xmath302 using the notation of appendix [ diffprobs ] . the second two terms derive from the diagrams in fig . [ fig : orbt ] so that without time reversal symmetry only the first term in ( [ gdiffprobeq ] ) remains . this is included as @xmath303 for systems without time reversal symmetry and @xmath304 for those with time reversal symmetry . the function @xmath305 is given semiclassically in terms of 4 trajectories @xmath306 where the main difference from the conductance variance ( [ condvartraj ] ) is that the trajectories connect different channels and the sum is unrestricted . we start with the diagram from fig . [ trajectoryquad ] , where since the partner trajectories cross over in the encounter they automatically travel from and to the correct channels and all four channels are unrestricted . this means that this structure is now lower order ( in inverse channel number ) than it was for the conductance variance ( it was there contained by taking into account the diagram in fig . [ ana1](a ) and removing two links connected to the same encounter ) . for the diagram in fig . [ trajectoryquad ] , we therefore have @xmath307}&=&-\frac{g_{(1)}^{2}{g_{(2)}}^{2}g_2}{g_{1}^4 } \nonumber \\ & & { } + \frac{2\left[g_{(1)}h_{(1),2}g_{(2)}^2+g_{(1)}^2h_{(2),2}g_{(2)}\right]}{g_{1}^3 } \nonumber \\ & & { } + \frac{g_{(1)}^{2}i_{(2),1}+i_{(1),1}g_{(2)}^{2}}{g_{1}^{2}}.\end{aligned}\ ] ] this contribution , along with its ehrenfest - time dependence , was previously calculated in @xcite . moving to the next order term , which has recently been calculated using rmt @xcite , we again simply need to look at the corresponding structures and consider the possible ways of shrinking links . without time reversal symmetry there are no possible structures at this order , so we therefore consider systems with time reversal symmetry . the structures which contribute are depicted in fig . 2 of @xcite . there is also , in fig . 2(d ) there , a quadruplet involving two independent pairs , one of which is simply a diagonal pair , while the other involves a single 2-encounter , as in fig . [ fig : orbt ] of this paper . because the start ( or end ) channels must coincide @xmath112 ( or @xmath113 ) there is a subtlety when we consider shrinking one link . if we shrink the link on the left ( or right ) side of the 2-encounter so that it moves into the lead where the diagonal pair emanate ( or terminate ) we actually have the same case as when the 3-encounter from fig . 2(c ) ( of @xcite ) moves into the lead . as such , we include this case there , leaving a contribution of @xmath308}&=&-\frac{2\left[g_{(1)}^{2}i_{(2),1}+i_{(1),1}g_{(2)}^{2}\right]g_{2 } } { g_{1}^{4 } } \\ & & { } + \frac{2\left[g_{(1)}h_{(1),2}i_{(2),1}+i_{(1),1}h_{(2),2}g_{(2)}\right ] } { g_{1}^{3}}. \nonumber\end{aligned}\ ] ] the diagram in fig . 2(a ) of @xcite is similar to the diagram in fig . [ ana1](b ) here and we obtain @xmath309}&=&\frac{4g_{(1)}^{2}g_{(2)}^{2}g_{2}^{2}}{g_{1}^{6 } } + \frac{8 g_{(1)}h_{(1),2}h_{(2),2}g_{(2)}}{g_{1}^{4 } } \nonumber \\ & & { } -\frac{8\left[g_{(1)}h_{(1),2}g_{(2)}^{2}+g_{(1)}^{2}h_{(2),2}g_{(2)}\right]g_{2 } } { g_{1}^{5 } } \nonumber \\ & & { } + \frac{2\left[g_{(1)}^{2}h_{(2),2}^{2}+h_{(1),2}^{2}g_{(2)}^{2}\right ] } { g_{1}^{4 } } . \end{aligned}\ ] ] for the diagram in fig . 2(b ) of @xcite we obtain twice the result of fig . [ ana2](a ) here [ or four times the result of fig . [ ana1](b ) ] for the conductance variance . for the diagram in fig . 2(c ) of @xcite we likewise obtain twice the result of fig . [ ana3](a ) here for the conductance variance . with all the next to leading order shot noise contributions , we can combine them and indeed find the same result as in @xcite . ehrenfest , z. physik * 45 * , 455 ( 1927 ) . b. v. chirikov , f. m. izrailev , and d. l. shepelyansky , sov . rev.sect . c * 2 * , 209 ( 1981 ) . i. l. aleiner , a. i. larkin , phys . b * 54 * , 14423 ( 1996 ) . o. yevtushenko , g. ltjering , d. weiss , and k. richter , phys . rev.lett . * 84 * , 542 ( 2000 ) . i. adagideli , phys . b * 68 * , 233308 ( 2003 ) . p. w. brouwer and s. rahav , phys . b * 74 * , 075322 ( 2006 ) . whitney and ph . jacquod , phys . lett . * 94 * , 116801 ( 2005 ) ; ph . jacquod and r. s. whitney , phys . b * 73 * , 195115 ( 2006 ) . d. waltner and j. kuipers , phys . e * 82 * , 066205 ( 2010 ) . jacquod and e. v. sukhorukov , phys . lett . * 92 * , 116801 ( 2004 ) . j. tworzydo , a. tajic , h. schomerus , p. w. brouwer , and c. w. j. beenakker , phys . lett . * 93 * , 186806 ( 2004 ) . a. altland , p. w. brouwer , and c. tian , phys . rev . lett . * 99 * , 036804 ( 2007 ) . c. petitjean , p. jacquod , and r. s. whitney , jetp lett . * 86 * , 736 ( 2007 ) ; r. s. whitney , ph . jacquod and c. petitjean , phys . b * 77 * , 045315 ( 2008 ) . o. agam , i. l. aleiner , and a. i. larkin , phys . lett . * 85 * , 3153 ( 2000 ) . r. s. whitney and ph . jacquod , phys . lett . * 96 * , 206804 ( 2006 ) . p. braun , s. heusler , s. mller , and f. haake , j. phys . a * 39 * , l159 ( 2006 ) . g. berkolaiko , j. m. harrison , and m. novaes , j. phys . a * 41 * , 365102 ( 2008 ) . d. waltner , j. kuipers , and k. richter , phys . b * 83 * , 195315 ( 2011 ) . i. l. aleiner and a. i. larkin , phys . e * 55 * , r1243 ( 1997 ) . c. tian and a. i. larkin , phys . b * 70 * , 035305 ( 2004 ) . p. w. brouwer , s. rahav , and c. tian , phys . e * 74 * , 066208 ( 2006 ) . d. waltner , m. gutirrez , a. goussev , and k. richter , phys . rev , 174101 ( 2008 ) . m. gutirrez , d. waltner , j. kuipers , and k. richter , phys . rev . e * 79 * , 046212 ( 2009 ) . b. gutkin , d. waltner , m. gutirrez , j. kuipers , and k. richter , phys.rev . e * 81 * , 036222 ( 2010 ) . h. u. baranger , r. a. jalabert , and a. d. stone , chaos * 3 * , 665 ( 1993 ) . k. richter and m. sieber , phys . lett . * 89 * , 206801 ( 2002 ) . k. richter , _ semiclassical theory of mesoscopic quantum systems _ , springer , berlin , heidelberg ( 2000 ) . p. w. brouwer and s. rahav , phys . b * 74 * , 085313 ( 2006 ) . p. w. brouwer and c. w. j. beenakker , j. math . phys . * 37 * , 4904 ( 1996 ) . m. bttiker , phys . lett . * 57 * , 1761 ( 1986 ) . y. imry , in _ directions in condensed matter physics _ , g. grinstein and g. mazenko ( eds . ) ( singapore : world scientific ) ( 1986 ) . d. s. fisher and p. a. lee , phys . b * 23 * , 6851 ( 1981 ) . s. heusler , s. mller , p. braun , and f. haake , phys . lett . * 96 * , 066804 ( 2006 ) ; s. mller , s. heusler , p. braun , and f. haake , new j. phys . * 9 * , 1 ( 2007 ) . s. mller , s. heusler , p. braun , f. haake , and a. altland , phys . rev.lett . , 014103 ( 2004 ) ; phys . e * 72 * , 046207 ( 2005 ) . r. s. whitney , phys . b * 75 * , 235404 ( 2007 ) . j. tworzydo , a. tajic , h. schomerus , and c. w. j. beenakker , phys.rev . b * 68 * , 115313 ( 2003 ) . h. schomerus and ph . jacquod , j. phys . a * 38 * , 10663 ( 2005 ) . f. m. izrailev , phys . rep . * 196 * , 299 ( 1990 ) . y. v. fyodorov and h .- j . sommers , jetp letters * 72 * , 422 ( 2000 ) ; r. o. vallejos and a. m.ozorio de almeida , ann . phys . * 278 * , 86 ( 1999 ) . m. bttiker , phys . * 65 * , 2901 ( 1990 ) . j. kuipers , j. phys . a * 42 * , 425101 ( 2009 ) . j. g. g. s. ramos , a. l. r. barbosa , and a. m. s. macdo , phys.rev . b * 78 * , 235305 ( 2008 ) .

the quantum ising chain is a useful theoretical laboratory in which fundamental issues concerning quantum many - body systems can be thoroughly investigated , exploiting the exact knowledge of several features of its phase diagram and quantum correlations . many results for its low - energy properties have been derived in the quantum ordered and disordered phases , and , in particular , at the quantum critical point separating the two phases , both in the thermodynamic limit and in the finite - size scaling ( fss ) limit with several types of boundary conditions , see , e.g. , refs . @xcite and references therein . in this paper we extend these analytic results to the case in which boundary fields are present . we present a detailed study of the quantum ising chain in a transverse magnetic field @xcite in the presence of magnetic fields coupled with the order - parameter spin operator , located at the boundaries of the chain . we assume the two magnetic fields to have the same strength and consider two cases : ( i ) the two fields are parallel ; ( ii ) the two fields are oppositely aligned . as expected , in the quantum paramagnetic phase the boundary fields do not change the large - size behavior of low - energies quantities . at the critical point , bulk behavior is independent of boundary conditions . however , the magnetic fields induce a surface phase transition with a corresponding scaling behavior . the quantum ferromagnetic phase is more interesting . if the boundary fields have opposite direction , one observes two different bulk phases . for small magnetic fields , the ground state is ferromagnetic as it occurs in the absence of boundary interactions . on the other hand , kink states @xcite are the relevant low - energy excitations for large boundary fields . the two different phases are separated by a continuous transition that is only characterized by the nature of the two coexisting phases and , indeed , the same transition occurs in ising rings in the presence of a localized link defect @xcite . because of the quantum - to - classical mapping , our results can also be applied to the two - dimensional ising model , and , more generally , to any model in the ising universality class , in a strip geometry . in the two - dimensional case the quantum transition between the ferromagnetic and the kink phase corresponds to the wetting transition , which , in the context of a strip geometry , is sometimes identified as an interface localization / delocalization transition @xcite . our results , therefore , provide new analytic expressions for two - dimensional ising systems at the wetting transition in the presence of boundary magnetic fields . the paper is organized as follows . in sec . [ model ] we introduce the one - dimensional quantum ising chain with boundary fields . in sec . [ fermmodel ] we compute the low - energy spectrum by exploiting the equivalent quadratic fermionic formulation of the hamiltonian @xcite . explicit calculations are reported in the following sections . [ paramag ] , [ critpoint ] and [ magnphase ] report results for the quantum paramagnetic phase , at the quantum critical point and for the ordered magnetized phase , respectively . [ fktra ] is devoted to the study of the magnet - to - kink transition driven by the boundary fields in the ordered phase . we obtain exact results for the gap , the magnetization profile , the two - point spin - spin correlation function , and the entanglement entropy . in sec . [ comp-2d ] these results are compared with the existing ones for two - dimensional classical models . finally , in sec . [ conclu ] we summarize the main results and draw our conclusions . a number of appendices report the derivations of some of the results . the hamiltonian of the quantum ising chain in a transverse magnetic field is given by @xmath1 where @xmath2 are the pauli matrices . in the following we assume ferromagnetic nearest - neighbor interactions with @xmath3 , and @xmath4 . the ising chain undergoes a continuous transition at @xmath5 @xcite , separating a quantum ferromagnetic phase ( @xmath6 ) from a quantum paramagnetic phase ( @xmath7 ) . in this paper we investigate the effects of boundary magnetic fields aligned along the @xmath8 axis . they give rise to an additional energy term @xmath9 that is added to hamiltonian ( [ isc ] ) . in the following we shall consider the model in the specific case of equal parallel boundary fields ( pbf ) @xmath10 and of equal oppositely - aligned boundary fields ( obf ) @xmath11 it is not restrictive to assume @xmath12 in both cases . we will often use a basis in which @xmath13 is diagonal . states will be labelled as @xmath14 , where @xmath15 is the eigenvalues of @xmath16 . signs will be fixed so that @xmath17 has the form @xmath18 in this basis . in the absence of boundary fields , the hamiltonian commutes with the generator @xmath19 of the @xmath20 transformations @xmath21 and with the operator @xmath22 of the reflection transformations defined by @xmath23 . these operators do not generally commute with @xmath24 . however , note that @xmath25=0 $ ] for obf . in this paper we analyze the low - energy spectrum of the model . in particular , we compute the energy differences between the lowest states and the ground state @xmath26 ( here @xmath27 are the energy eigenvalues ordered so that @xmath28 , and @xmath29 is the gap ) , the local magnetization and the two - point correlation function , @xmath30 for obf , because of the symmetry under @xmath31 , we have @xmath32 and , as a consequence , the average magnetization @xmath33 always vanishes . we also define the integrated correlation @xmath34 and the correlation length @xmath35 with respect to the center of the chain : @xmath36 to determine the spectrum of hamiltonian ( [ isc ] ) , we extend the model , considering two additional spins located in 0 and @xmath37 and the hamiltonian @xmath38 this is the ising hamiltonian with two different couplings on the boundary links and zero transverse field on the boundaries . let us note that the hamiltonian @xmath39 commutes with both @xmath40 and @xmath41 , which can therefore be simultaneously diagonalized . the hilbert space can be divided into four sectors , which we label as @xmath42 , @xmath43 , @xmath44 and @xmath45 , where @xmath46 are the eigenvalues of @xmath40 and @xmath41 . the restriction of @xmath39 to each sector gives rise to the hamiltonian @xmath47 , defined in eq . ( [ isc ] ) , with a boundary term of the form ( [ hb ] ) . hence , the spectrum of @xmath39 also provides the spectrum of @xmath48 . let us also note that @xmath39 is @xmath49 symmetric . indeed , if @xmath50 , then @xmath51 = 0 $ ] . since @xmath52 does not commute with @xmath40 and @xmath41 , the spectrum is necessarily degenerate . moreover , since @xmath52 anticommutes with these two boundary operators , @xmath52 maps sector @xmath42 to @xmath45 and @xmath44 to @xmath43 , so that the restriction of @xmath39 to @xmath53 allows us to compute the full spectrum of @xmath39 . to compute the spectrum of hamiltonian ( [ def - hextended ] ) , we follow ref . @xcite . we first perform a jordan - wigner transformation , defining fermionic operators @xmath54 and @xmath55 @xmath56 where @xmath57 . these relations can be inverted , obtaining @xmath58 thus , hamiltonian ( [ def - hextended ] ) becomes @xmath59 with @xmath60 for @xmath61 . in this formalism @xmath62 and @xmath63 . to go further , let us rewrite @xmath39 as @xmath64,\ ] ] with @xmath65 and @xmath66 symmetric and antisymmetric matrices , respectively . then , we perform a bogoliubov transformation . we introduce new canonical fermionic variables @xmath67 where @xmath68 and @xmath69 are fixed by the requirement that @xmath47 takes the form @xmath70 with @xmath71 following ref . @xcite , we define the vectors @xmath72 the variables @xmath73 satisfy canonical anticommutation relations if the vectors @xmath74 form an orthonormal basis , and so does the set @xmath75 . the vectors @xmath75 satisfy @xmath76 which turns the determination of the energies @xmath77 into an eigenvalue problem . if @xmath77 does not vanish , @xmath74 is given by @xmath78 if @xmath77 is zero , @xmath74 is the null eigenvector of @xmath79 . it is also possible to evaluate the constant @xmath80 in eq . ( [ h1-diag ] ) , which provides the energy of the ground state : @xmath81 the matrix @xmath82 is given by ( we write the @xmath83 matrix @xmath84 for @xmath85 , the generalization to any @xmath0 being obvious ) @xmath86 this matrix @xmath84 has a zero eigenvalue , @xmath87 , with eigenvector @xmath88 . correspondingly , we obtain @xmath89 and @xmath90 which depends only on the boundary fermion operators . note that it can be rewritten as a combination of @xmath40 , @xmath41 , and @xmath52 , hence it does not represent an additional symmetry of the hamiltonian . the presence of a zero eigenvalue was expected , as the spectrum is degenerate . if @xmath91 and @xmath92 do not vanish , since there is only one fermion operator with zero energy , the spectrum is doubly degenerate . let us now consider the nonzero eigenvalues . since the vectors @xmath75 are an orthonormal set , orthogonality with @xmath93 implies @xmath94 . it follows analogously @xmath95 . therefore , we have for @xmath96 @xmath97 where the dots represent a polynomial in @xmath98 and @xmath99 with @xmath100 . it is then easy to verify that @xmath101 = 0.\ ] ] it follows that , if a state @xmath102 satisfies @xmath103 , then we have @xmath104 . analogously , if @xmath105 , we have latexmath:[$\sigma_{l+1}^{(1)}\eta_k |\psi\rangle = s_1 \eta_k sector @xmath46 , then @xmath107 belongs to sector @xmath108 . to conclude , we should determine to which sectors the two degenerate ground states belong . if @xmath91 and @xmath92 are both positive , we expect the ground - state configurations to be ordered . therefore , if on one boundary the spin is directed in the positive @xmath8 direction , we expect the same to occur at the other end of the chain . therefore , we conclude that the ground states belong to sectors @xmath109 and @xmath45 . this identification is supported by exact diagonalization . we can now classify the states . we only consider the states with @xmath110 , to avoid the double degeneracy . then , if @xmath111 we have : * in the sector @xmath112 , @xmath113 , the lowest energy state is the ground state of the hamiltonian and all states are obtained as @xmath114 , @xmath115 , with @xmath116 even . the first excited state is @xmath117 and the energy gap is @xmath118 . * in the sector @xmath112 , @xmath119 , the lowest energy state is the first excited state @xmath120 of the hamiltonian @xmath39 and all states are obtained as @xmath121 , @xmath115 , with @xmath116 odd . in particular , the first excited state in the sector is @xmath122 , so that the energy gap is @xmath123 . the first case is relevant when considering @xmath124 in the boundary hamiltonian @xmath24 , which are identified with @xmath91 and @xmath92 , respectively . the second case is relevant when @xmath125 and @xmath126 have opposite signs : the corresponding gap is obtained by identifying @xmath127 and @xmath128 . the energy spectrum of hamiltonian @xmath39 can be computed exactly , as discussed in [ app.a ] . here we will focus on the case @xmath129 , which allows us to derive the spectrum for parallel or opposite equal magnetic fields at the boundary , cf . ( [ pbfdef ] ) and ( [ obfdef ] ) , respectively . neglecting the irrelevant zero eigenvalue , we obtain @xmath37 elementary excitations whose energy @xmath130 is given by @xmath131 where @xmath132 are the @xmath37 solutions of the equation @xmath133 \sin k l = \label{secular - equation}\\ & & \quad g ( 1 - 2 g \cos k + g^2 - \zeta^4 ) \sin [ k ( l+1 ) ] . \nonumber\end{aligned}\ ] ] we should consider real solutions in @xmath134 ( correspondingly @xmath135 ) , purely imaginary solutions @xmath136 with @xmath137 ( correspondingly @xmath138 ) , and solutions @xmath139 ( correspondingly @xmath140 ) . equation ( [ secular - equation ] ) has also a spurious solution for @xmath141 for all values of the parameters , which should be discarded . for @xmath142 , there is also a second solution with @xmath143 , which corresponds to a true excitation of the model . for @xmath144 we obtain the equation appropriate for open boundary conditions ( obc ) @xcite , @xmath145 and @xmath146 , which implies an additional zero mode for @xmath39 . this is not surprising as the spectrum has a fourfold degeneracy for @xmath147 : the four sectors are equivalent in the absence of boundary fields . the real solutions of eq . ( [ secular - equation ] ) are obtained by solving @xmath148\cos k + 2 g^2 \cos^2 k}. \label{realtankl}\ ] ] the imaginary solutions are obtained by setting @xmath149 . ( [ secular - equation ] ) then becomes @xmath150 ^ 2 \over ( g - e^h ) ( e^h g - 1 + \zeta^2)^2 } . \label{equazione - j1eqj2-small}\ ] ] the corresponding energies @xmath151 are always smaller than those corresponding to real momenta , hence the localized states are the most important ones in the determination of the low - energy spectrum . for @xmath152 ( paramagnetic phase ) the lowest energy state is a localized state , i.e. , a solution of eq . ( [ equazione - j1eqj2-small ] ) with @xmath137 . for large values of @xmath0 we must consider the poles in the right - hand side of the equation . they are @xmath153 for @xmath152 , the second equation does not have any solution . the relevant pole is the first one , so that the relevant solution can be written as @xmath154 . for large values of @xmath0 , substituting this expression in eq . ( [ equazione - j1eqj2-small ] ) , we obtain @xmath155 correspondingly , we find @xmath156 the energy @xmath157 vanishes with exponentially small corrections . the energy of the second excited state corresponds to the lowest value @xmath158 of @xmath159 satisfying eq . ( [ realtankl ] ) . we obtain @xmath160 and @xmath161 the system is therefore gapped , and the same gap is obtained for both pbf and obf modulo exponentially small corrections , since @xmath162 : @xmath163 note that the same result would have been obtained for obc , confirming that boundary conditions and/or boundary fields are irrelevant in the paramagnetic quantum phase . at the critical point @xmath165 , there are no localized solutions , hence the two lowest energies @xmath166 are obtained by considering eq . ( [ realtankl ] ) . the equation becomes @xmath167 the lowest - energy solutions correspond to momenta that scale as @xmath168 . therefore , we can expand the right - hand side in powers of @xmath159 , obtaining @xmath169 which shows that @xmath170 must diverge for @xmath171 . this fixes @xmath172 with @xmath173 including the corrections , we obtain for the two lowest momenta @xmath174 so that @xmath175 these results imply that the pbf and obf gaps are @xmath176 respectively . note that the amplitude of the @xmath177 term differs from that obtained in the obc case ( corresponding to @xmath178 ) @xcite , @xmath179 indicating that @xmath144 is a surface critical point . as we have discussed in ref . @xcite in the obc case , the corrections of order @xmath177 to the leading behavior can be interpreted as the effect of a nonlinear scaling field associated with @xmath0 : the leading irrelevant boundary operator gives corrections that scale as @xmath180 . the same holds in the presence of boundary fields . indeed , we can define a new length scale @xmath181 so that @xmath182 using the previously reported expressions we obtain @xmath183 note that the same rescaling applies both to pbf and obf , a result which is not obvious in the original formulation , but which has a natural explanation in the present setting in which we consider @xmath39 . indeed , the rescaling should cancel the leading correction in the whole low - energy spectrum of @xmath39 , hence in all low - energy excitation energies @xmath77 . therefore , it should apply to both pbf and obf . since @xmath144 is a surface critical point and the boundary term is a relevant perturbation , we can study the crossover behavior close to @xmath144 . the relevant scaling variable is @xmath184 . therefore , in the limit @xmath185 , @xmath171 at fixed @xmath186 we have @xmath187 where the function @xmath188 depends on the type , obf or pbf , of boundary fields . in the scaling limit defined above , momenta scale as @xmath189 . using eq . ( [ seceq - g1 ] ) , we obtain an implicit equation for @xmath190 : @xmath191 therefore , to compute the scaling function , one determines the two lowest values of @xmath190 satisfying eq . ( [ scaling - b - boundary ] ) and uses them to determine the gap as before . the scaling curves are reported in fig . [ fss - crossover - etab ] . let us now consider the scaling behavior in a neighborhood of the critical point @xmath165 . the corresponding scaling variable is @xmath192 the relevant equations are obtained from eqs . ( [ realtankl ] ) and ( [ equazione - j1eqj2-small ] ) by expanding the right - hand sides in the limit @xmath193 , @xmath171 , @xmath194 at fixed @xmath195 and @xmath196 . as expected , as long as @xmath197 , we obtain a result that is independent of @xmath198 : the fss behavior only depends on the nature of the boundary conditions , but not on the specific values of the boundary fields . a localized state exists only for @xmath199 . if we set @xmath200 , the parameter @xmath201 is a solution of the equation @xmath202 the corresponding excitation energy is @xmath203 note that @xmath204 for @xmath205 , so that @xmath206 in the limit . on the other hand , for @xmath207 we obtain @xmath208 so that @xmath209 , consistently with eq . ( [ cale1_para ] ) . the propagating modes can be written as @xmath210 , where @xmath211 are the positive solutions of the equation @xmath212 with @xmath213 ( the solution with @xmath214 only exists for @xmath215 ) . the corresponding excitation energies are then @xmath216 . using these results we obtain for the energy gap in the fss limit : @xmath217 the two signs refer to the case of parallel ( plus sign ) and opposite ( minus sign ) boundary fields . the curves are reported in fig . [ fss - delta1 ] together with the corresponding obc scaling function . in the paramagnetic phase all curve have a similar shape they coincide asymptotically while in the phase @xmath218 ( @xmath219 ) , the behavior is clearly different . for @xmath220 , we have @xmath221 , @xmath222 , and @xmath223 . it is interesting to discuss the scaling corrections , that can be easily shown to decay as @xmath177 , as it occurs for @xmath144 . in ref . @xcite we argued that such corrections are not associated with irrelevant boundary operators , which give rise to corrections of order @xmath180 , but that they should be interpreted as due to nonlinear scaling fields . the same holds for @xmath197 . first , one should consider the effective length @xmath181 defined in eq . ( [ leff - def ] ) . then , one should consider the nonlinear scaling field @xmath224 associated with @xmath225 and , finally , rescale the energies with the _ sound velocity _ @xmath226 . the nonlinear scaling fields do not depend on the boundary conditions , hence we can use the expressions reported in ref . @xcite : @xmath227 then , we define the new scaling variable @xmath228 and express @xmath211 and @xmath201 as functions of @xmath229 . a somewhat lengthy calculation shows that @xmath230 can be written as in eq . ( [ gap - gcritico ] ) , with @xmath229 replacing @xmath195 , with corrections that scale as @xmath231 . as expected , the leading corrections cancel out . [ magnphase ] let us now consider the case @xmath218 . in the infinite - volume limit , there exist two degenerate ground states @xmath232 and @xmath233 , which differ by the value of the magnetization @xcite @xmath234 where @xmath235 is a bulk magnetic field applied along the @xmath8 direction . for a chain of finite size @xmath0 , the degeneracy is lifted . in the obc case , the energy difference between the two lowest - energy states vanishes exponentially as @xmath0 increases @xcite : @xmath236 . \label{deltaobc}\end{aligned}\ ] ] moreover , @xmath237 by symmetry , and @xmath238 these are the standard features of a quantum ferromagnetic phase with a spontaneously broken @xmath49 symmetry . we now extend the discussion to the case in which boundary fields are present , considering equal pbf and obf . to identify localized states , we consider eq . ( [ equazione - j1eqj2-small ] ) . the relevant solutions are associated with the poles in the right - hand side of the equation , see eq . ( [ equazione - poli ] ) . it is immediate to verify that there are no poles for @xmath239 , while a pole exists in the opposite case . therefore , we distinguish three different cases , depending on whether @xmath198 is smaller , larger , or equal to @xmath240 for @xmath241 , the two lowest - energy excitations corresponding to @xmath157 and @xmath242 are localized . we have @xmath243 ( @xmath137 ) with @xmath244 where @xmath245 and @xmath201 is a correction term . let us first obtain a large-@xmath0 expression for @xmath201 that is uniform as @xmath246 . this requires particular care as the right - hand side of eq . ( [ equazione - j1eqj2-small ] ) has a different singular behavior for @xmath144 and @xmath247 . for @xmath144 , it has a simple pole as @xmath248 . on the other hand , for @xmath249 , it has a double pole for @xmath250 . to compute @xmath201 we set @xmath251 in all terms of eq . ( [ equazione - j1eqj2-small ] ) that are not relevant for the singular behavior as @xmath171 , i.e. , we write @xmath252 with @xmath253 equation ( [ eq - magn-1 ] ) is a second - order algebraic equation of @xmath254 , whose solutions are @xmath255 the energies @xmath256 are correspondingly @xmath257 and @xmath258 , @xmath259 . there are now two interesting cases . first , we take the limit @xmath260 at fixed @xmath261 . we can then simplify eq . ( [ epm - general ] ) , obtaining @xmath262 these results imply @xmath263 in the pbf case the gap is finite and proportional to @xmath198 for small fields . instead , @xmath264 vanishes exponentially , as in the obc case ( which corresponds to vanishing boundary fields ) . a second possibility consists in considering the fss limit around @xmath144 . as discussed in ref . @xcite , the ratio @xmath265 is expected to become a function of @xmath266 ( to compare with the formulae of ref . @xcite note that the energy associated with the perturbation is here proportional to @xmath198 , without additional factors of @xmath0 ) . in this limit @xmath267 is proportional to @xmath268 . using the expression for @xmath269 reported in eq . ( [ deltaobc ] ) , we obtain @xmath270 so that @xmath271 note that the scaling function agrees with that predicted in ref . we can also consider the ratio @xmath272 in the same limit , but in this case we obtain that the ratio is simply equal to 1 , i.e. , independent of @xmath268 . the reason can be easily understood . for obf the energy associated with the perturbation is zero the contributions of the two boundaries cancel hence @xmath273 identically . for @xmath274 there are no localized states . the secular equation ( [ realtankl ] ) can be rewritten as @xmath275 for @xmath171 , the solutions can be parametrized as @xmath276 it follows that @xmath277 for large values of @xmath0 we obtain finally @xmath278 note that the correction term @xmath279 diverges as @xmath280 and has a finite limit as @xmath281 . moreover , the explicit form of the hamiltonian is only relevant for the corrections . it follows @xmath282 we obtain finally @xmath283 for obf , the gap vanishes for @xmath171 . however , while the approach is exponential for @xmath241 , we have @xmath284 for large fields . for pbf , the system is gapped as for @xmath285 . however , in this case corrections are of order @xmath180 . the gap equation is particularly simple . we obtain a solution @xmath287 and @xmath288 hence @xmath289 with no @xmath0 dependence . as for the other states we obtain @xmath290 with @xmath291 . it follows @xmath292 , as for @xmath293 . however , corrections , of order @xmath180 , are a factor of five smaller compared to the large - field case . for obf we obtain the asymptotic behavior @xmath294 as in the large - field case , @xmath264 scales as @xmath180 . however , the prefactor differs by a factor of 3 from that obtained for @xmath295 , cf . ( [ deltalk ] ) . we wish now to compute the ground - state magnetization and the corresponding two - point spin - spin correlation function in the magnetized phase , i.e. , for @xmath296 . in the pbf case , the system is magnetized , hence we expect @xmath297 and @xmath298 to be both independent of @xmath299 and @xmath300 , except possibly close to the boundaries , and to be equal to @xmath301 and @xmath302 , respectively , where @xmath301 is given in eq . ( [ modef ] ) . thus , the boundary magnetic fields have the only role of lifting the degeneracy . the obf case is more interesting . in [ app.b ] we have fully characterized the ground state of the model for small values of @xmath303 . we find that for it is ferromagnetic for @xmath285 ( this is consistent with the results for the gap discussed above ) . for @xmath293 kink states @xcite are the relevant low - energy excitations , while for @xmath304 the ground state is a superposition of kink and ferromagnetic states . the exact knowledge of the ground state allows us to compute the ground - state magnetization and the correlation function perturbatively in @xmath303 . for obf we find that @xmath305 vanishes in the low - field case except at the boundaries , depends linearly on @xmath299 for @xmath286 , and varies in a simple fashion for @xmath274 . we expect the same space dependence for all finite values of @xmath306 in the large-@xmath0 universal limit , modulo a multiplicative renormalization . if we set @xmath307 and @xmath308 , so that the center of the chain corresponds to @xmath309 , we predict @xmath310 \displaystyle{m(x)\over m_0 } = \displaystyle{x\over \ell } & \qquad { \zeta = \zeta_c } , \\[2 mm ] \displaystyle{m(x)\over m_0 } = 0 & \qquad { \zeta < \zeta_c}. \end{array } \label{magnetization - scaling}\ ] ] these relation should hold in the limit @xmath311 , @xmath312 at fixed @xmath313 , except possibly at the boundary , i.e. , for @xmath314 . here @xmath301 is the bulk magnetization of the system in the infinite - volume limit defined in eq . ( [ modef ] ) . it is interesting to note that the linear behavior of the magnetization for @xmath315 is due to the fact that the ground state is translationally invariant : the space probability of the kinks is independent of @xmath8 . due to the presence of the boundary fields , the translation invariance of the kink space distribution is not trivial and , indeed , it does not hold for @xmath295 . to check these predictions and study the approach to the asymptotic behavior , we perform numerical simulations using the density matrix renormalization - group ( dmrg ) method @xcite . for convenience , we consider chains with odd @xmath0 , setting @xmath316 ( since we are considering ferromagnetic interactions , the low - energy properties do not depend on this feature , unlike the antiferromagnetic case , see , e.g. , ref . the results for the magnetization are shown in fig . [ magnetization - fasi ] . for @xmath285 , the magnetization vanishes in a large interval around @xmath317 , assuming positive and negative values only close to the boundaries . note that the region in which @xmath318 shrinks as @xmath0 increases , conferming the validity of eq . ( [ magnetization - scaling ] ) in the large-@xmath0 limit at fixed @xmath313 . for @xmath293 we show results for two values of @xmath303 . once we renormalize @xmath319 by using @xmath301 , i.e. , we consider the ratio @xmath320 , results fall one on top of the other , confirming universality . the results are in full agreement with the expressions ( [ magnetization - scaling ] ) . the same arguments apply to the correlation function @xmath298 . extending the small-@xmath303 results of [ app.b ] to the whole low-@xmath303 phase , we predict @xmath321 in the large-@xmath0 limit . therefore , the correlation function should be constant in the magnetized phase , @xmath322 , it should depend linearly on @xmath323 for @xmath304 , while it should have a sinusoidal dependence in the kink phase . again , we check these predictions by comparing them with dmrg data . they are reported in fig . [ gx - fasi ] . the dmrg data approach the theoretical predictions with increasing @xmath0 , confirming the theoretical predictions . as already discussed in ref . @xcite , in the case of obf and for @xmath218 , an interesting universal scaling behavior is observed for @xmath280 . such a transition is uniquely characterized by the nature of two phases occurring for @xmath285 and @xmath324 : on one side the ground state is ferromagnetic , while on the other side kink states are the relevant low - energy excitations . this characterization is confirmed by the analysis of the ising chain on a ring with a bond defect : at the boundary of the two phases one observes @xcite a behavior analogous to the one discussed here . here , we will report the computation of the energy gap the result was already presented in ref . @xcite and additional results for the local magnetization , correlation function , and entanglement entropy . moreover , we will discuss the density of the fermionic modes in the equivalent fermionic picture of the model @xcite , showing that the transition can be interpreted as a localization transition of the fermionic states at the boundaries . the crossover behavior around the transition is parametrized in terms of the scaling variable @xcite @xmath325 then , in the large-@xmath0 limit we have @xmath326 in general , the scaling functions @xmath327 and @xmath328 are expected to be universal modulo a @xmath303-dependent normalization ( which is independent of the observable considered ) of the argument @xmath329 . the @xmath303 dependent prefactor appearing in eq . ( [ etasdef ] ) is the required nonuniversal factor . it is determined in sec . [ sec7.1 ] , analyzing the scaling behavior of the gap @xmath264 . according to renormalization - group theory , this normalization factor is independent of the observable considered , hence also the scaling functions of other observables should be independent of @xmath303 apart from a multiplicative factor ( not present if renormalization - group invariant ratios are considered ) , once @xmath329 is used as scaling variable . the results we will present for the magnetization and the two - point correlation function confirm this prediction . beside the gap @xmath330 , we also consider the integrated correlation @xmath34 defined in eq . ( [ chixidef ] ) . it satisfies the scaling relation @xmath331 where @xmath332 is universal , i.e. , @xmath303 independent . these results allow us to define the rg dimensions of the different operators . in general , we expect @xmath333,\qquad \chi = a_\chi l^{d - y_\sigma } f_\chi [ l^{y_\zeta } ( \zeta-\zeta_c ) ] . \label{scalzetas}\ ] ] comparison with our results allows us to identify @xmath334 , @xmath335 , and @xmath336 . the latter result is consistent with the fact that a bulk magnetic field coupled to @xmath337 drives a first - order quantum transition , see , e.g. , ref . @xcite . by defining @xmath329 as in eq . ( [ etasdef ] ) we have implicitly assumed that we are fixing @xmath303 and varying @xmath198 close to the critical value @xmath338 . in some contexts , however , it is more natural to consider a different point of view . the boundary field strength @xmath198 is fixed ( it should satisfy @xmath339 ) and @xmath303 is varied in a neighborhood of the critical value @xmath340 . in this case we have @xmath341 where the @xmath198 dependent prefactor guarantees that scaling functions are the same for all values of @xmath198 . to determine the scaling behavior of the energy gap @xmath342 , we first simplify the secular equation ( [ secular - equation ] ) , by taking the limit @xmath343 at fixed @xmath329 . in this limit @xmath159 scales as @xmath177 , so that we set @xmath344 . for @xmath171 the secular equation becomes @xmath345 expressing @xmath346 as a function of @xmath347 , we find that the solutions of eq . ( [ sec - etac-1 ] ) satisfy one of the two equations @xmath348 if @xmath349 and @xmath350 are the positive solutions of eqs . ( [ eq - tanz2 - 1 ] ) and ( [ eq - tanz2 - 2 ] ) , respectively , it is easy to verify that @xmath351 for @xmath352 and @xmath353 . for @xmath354 we have instead @xmath355 . for @xmath356 there are also localized solutions with @xmath357 . the parameter @xmath358 satisfies one of the two equations @xmath359 eq . ( [ eq - tanw2 - 2 ] ) has a positive solution @xmath360 for @xmath356 , while eq . ( [ eq - tanw2 - 1 ] ) has a positive solution @xmath361 satisfying @xmath362 only for @xmath363 . these results allow us to determine the scaling function defined in eq . ( [ sca ] ) . we obtain @xmath364 for @xmath365 , @xmath366 for @xmath367 and @xmath368 for @xmath363 . we can easily determine the asymptotic behaviors . for small @xmath329 we have @xmath369 while for @xmath370 we have @xmath371 finally , for @xmath372 we have @xmath373 the theoretical curve is shown in fig . [ deetac ] . the dmrg data confirm the theoretical calculation : the numerical estimates of the ratio @xmath374 clearly approach the scaling curve @xmath375 with increasing @xmath0 . let us now consider the behavior of the correlations close to @xmath376 . the local magnetization @xmath297 and the correlation function @xmath298 are computed exactly in the limit @xmath377 in [ app.b4 ] . as we have already done in sec . [ sec6.2 ] , we extend these expressions to all values of @xmath303 satisfying @xmath6 , by replacing @xmath297 with @xmath378 and @xmath298 with @xmath379 . therefore , we predict for @xmath365 @xmath380,\ ] ] where @xmath381 is the smallest positive solution ( @xmath382 ) of eq . ( [ eq - tanz2 - 2 ] ) , @xmath383 , and @xmath308 , as before . for @xmath356 we have instead @xmath384,\ ] ] where @xmath358 is the solution ( @xmath385 ) of eq . ( [ eq - tanw2 - 2 ] ) . as for the correlation function @xmath298 it satisfies eq . ( [ gij - m ] ) . it is important to stress that the previous expressions , although derived for small values of @xmath303 , are expected to be _ exact _ for any @xmath218 . for instance , since @xmath320 is a dimensionless renormalization - group invariant ratio , it should scale as @xmath386,\ ] ] where @xmath387 is a universal , hence @xmath303 independent , function . the specific features of the model enter in this expression only through the normalization nonuniversal factor @xmath388 and the critical value @xmath338 . the factor @xmath388 does not depend on the observable , hence it can be determined by using any quantity . we use the gap @xmath264 , for which we are able to obtain exact results for any @xmath218 , defining @xmath389 . therefore , @xmath387 as a function of @xmath329 and @xmath390 is @xmath303 independent . therefore , our results , obtained in the limit @xmath391 , hold for any value of @xmath303 satisfying @xmath218 . the previous expressions for @xmath319 hold in the limit @xmath311 , @xmath171 at fixed @xmath313 and smoothly interpolate between the expressions valid in the different phases . note that we have @xmath392 for @xmath393 in all cases . this result does not however necessarily apply to the boundary points , i.e. , for @xmath394 , as the magnetization is not continuous at the boundary . for instance , in the magnetized phase , @xmath395 in all interior points , while @xmath396 and @xmath397 are nonuniversal functions of the field strength @xmath198 . a graph of @xmath320 for several values of @xmath329 as a function of @xmath313 is reported in fig . [ magnxcr ] . as expected , for @xmath398 the magnetization approaches zero exponentially in an interval that is centered in the middle of the chain and that widens as @xmath399 increases . more precisely , for @xmath372 we have @xmath400 and @xmath401 which shows that @xmath319 vanishes except in two tiny intervals close to the boundaries of width @xmath402 . to verify the prediction ( [ gij - m ] ) for @xmath298 we compute the scaling functions associated with @xmath34 and @xmath35 . they are reported in [ app.b4 ] and compared with dmrg data in figs . [ chietacr ] and [ xietacr ] . results with different values of @xmath303 fall on top of the theoretical curve when plotted versus @xmath329 . this confirms our predictions for @xmath298 and the magnetization , as well as the correctness of the nonuniversal prefactor appearing in the definition of @xmath329 . it is interesting to reinterpret our results in the equivalent fermionic picture of the model . in the ferromagnetic phase , i.e. , for @xmath403 , the lowest eigenstates are superpositions of majorana fermionic states localized at the boundaries @xcite . in finite systems , their overlap does not vanish , giving rise to the splitting @xmath404 . the coherence length @xmath405 diverges at the kink - to - magnet transitions as @xmath406 , a behavior analogous to that observed at the order - disorder transition @xmath407 where @xmath408 . within this picture , it is natural to consider @xmath409 which can interpreted as the particle density in the fermionic representation of the ising chain , see eq . ( [ sigma - c - relation ] ) . its large-@xmath0 limit is given by @xcite @xmath410 we consider the difference @xmath411 where @xmath383 , as before . such a quantity has been computed for small values of @xmath303 in [ app.b4 ] . the universality argument we have used for @xmath297 and @xmath298 can be applied to @xmath412 . we therefore predict @xmath413 for @xmath365 and @xmath414 for @xmath356 , where @xmath381 and @xmath358 are determined as in the case of the local magnetization . to verify this prediction we compute @xmath415 numerically for @xmath416 and @xmath417 . the numerical results are compared with theory in fig . [ fig : deltan0 ] . the agreement confirms the analytic prediction . the subtracted fermionic density @xmath418 is reported in fig . [ deltanxcr ] as a function of @xmath313 . for positive values of @xmath329 it has a maximum at the origin and shows that fermions are delocalized . for @xmath419 , @xmath420 is constant , while for negative values of @xmath329 , there is a significant enhancement at the boundary . for @xmath372 , as in the case of the magnetization , @xmath418 is different from 0 only close to the boundary , in an interval of width @xmath421 . it is interesting to consider the behavior of @xmath422 at the boundaries , i.e. , for @xmath423 , which is shown in fig . [ deltanb ] . in the kink phase , such a quantity scales as @xmath424 for @xmath370 . on the other hand it diverges as @xmath425 for @xmath372 , that is when the magnetized phase is approached . as such , it represents a physically relevant order parameter for the transition , which distinguishes the two different phases . finally , we discuss the behavior of the half - chain entanglement entropy @xcite . we divide the chain into two connected parts @xmath426 $ ] and @xmath427 $ ] and consider the von neumann entropy @xmath428 } \ln \rho_{[-\ell,0 ] } , \label{vnen}\ ] ] where @xmath429}$ ] is the reduced density matrix of the subsystem @xmath426 $ ] , i.e. , @xmath430}={\rm tr}_{[1,\ell ] } \;\rho , \ ] ] and @xmath431 is the density matrix of the ground state . the entanglement entropy @xmath432 approaches a constant in the large-@xmath0 limit , except for the critical case @xmath5 where it increases logarithmically @xcite . its large-@xmath0 limit for obc is given by @xcite @xmath433 , \nonumber\\ & & i(z ) = \int_0 ^ 1 { dx\over \sqrt{(1-x^2)(1-z^2 x^2 ) } } . \label{sobc}\end{aligned}\ ] ] in fig . [ entaen ] we plot the difference @xmath434 which shows the scaling behavior @xmath435 where @xmath436 is a universal function apart from a multiplicative nonuniversal constant . these results imply that the large-@xmath0 limit of the half - chain entanglement entropy does not depend on the boundary conditions , being the same for obc and obf . however , the @xmath437 corrections are affected by the boundary term ( [ hb ] ) , giving rise to a nontrivial universal behavior around @xmath376 . the results we have obtained allow us to derive exact scaling functions for the classical two - dimensional ising model in a strip @xmath438 in the presence of two equal magnetic fields on the boundaries . this model has been extensively studied , especially in the case of opposite magnetic fields . in this case , the wetting ( or interface localization - delocalization ) transition , which is the analog of the magnet - to - kink transition we are considering here , has been extensively studied @xcite . we can therefore check some of our results , comparing them with the existing ones in the literature . the results for the gap @xmath264 obtained in sec . [ sec7.1 ] can be used to derive the scaling behavior of the longitudinal correlation length @xmath439 , defined by @xmath440 . such a quantity was computed in the restrictive solid - on - solid model in ref . their result is completely consistent with ours . if our scaling quantities @xmath381 and @xmath358 defined in sec . [ sec7.1 ] are identified with the quantities @xmath441 , @xmath442 appearing in ref . @xcite , we obtain exactly the same expression for @xmath443 . several papers study the magnetization profile , which gives information on the behavior of the interface between the two coexisting phases . the behavior in the large - field phase was determined in ref . @xcite by using the restrictive solid - on - solid model . the result is in agreement with our prediction ( [ magnetization - scaling ] ) for @xmath293 . the linear behavior observed at @xmath304 was also derived at the wetting transition by using the solid - on - solid model @xcite . no exact results are instead available for the magnetization profiles in the whole crossover region , which , in the classical case , is usually parametrized in terms of @xmath444l$ ] , where @xmath235 is the boundary field and @xmath445 the corresponding wetting temperature ( in the quantum - to - classical mapping @xmath303 corresponds to @xmath22 , hence this scaling corresponds to varying @xmath303 at fixed @xmath198 ) . however , the numerical results for the magnetization reported in refs . @xcite ( and , in particular , fig . 2 of ref . @xcite ) appear to be in full agreement with our prediction shown in fig . [ magnxcr ] . we consider the one - dimensional quantum ising chain in the presence of a transverse magnetic field @xmath303 @xcite and of boundary magnetic fields aligned with the order - parameter spin operator . in particular , we consider fields with equal strength @xmath198 , which have either the same ( pbf ) or the opposite ( obf ) direction , cf . ( [ pbfdef ] ) and ( [ obfdef ] ) . we assume the chain to have length @xmath0 and derive the finite - size behavior of several quantities in the limit @xmath171 as a function of @xmath198 . we derive analytic predictions for the gap in all phases as a function of the boundary field strength . in the paramagnetic quantum phase , the leading behavior is independent of the boundary conditions . at the critical transition @xmath165 , finite - size scaling depends on the boundary fields . the point @xmath144 , corresponding to open boundary conditions , is a surface critical point . we derive analytic expressions for the scaling functions that parametrize the crossover behavior for @xmath193 and any @xmath198 ( bulk critical behavior ) , and for @xmath165 and @xmath246 ( surface critical behavior ) . in the quantum ferromagnetic phase @xmath218 , if the boundary fields are oppositely aligned , the finite - size behavior drastically changes as @xmath198 varies @xcite . for small @xmath198 , the system is in a ferromagnetic phase in which the gap decreases exponentially with the system size . on the other hand , for large fields , kink propagating modes of momenta @xmath177 are the relevant low - energy excitations , so that the gap decreases as @xmath446 . the two phases are separated by a critical transition at @xmath286 . close to this transition , low - energy properties show a universal scaling behavior in terms of the scaling variable @xmath447 . the transition is only characterized by the nature of the coexisting phases . indeed , the same transition occurs in ising rings in the presence of a localized bond defect @xcite . it is interesting to interpret our results in the context of two - dimensional classical ising models , using the quantum - to - classical mapping . as is well known @xcite , the quantum ising chain of length @xmath0 corresponds to a classical ising model on a strip of width @xmath0 , the transverse field playing the role of the temperature . therefore , our results provide exact predictions for the ising model in an @xmath448 geometry with boundary fields . our results for the gap at the critical point can be used to predict the fss behavior of the longitudinal exponential correlation function @xmath449 as a function of @xmath450 and of @xmath451 , where @xmath65 and @xmath66 are nonuniversal model - dependent constants . moreover , our extensive results at the magnet - to - kink transition for @xmath218 for obf provide predictions for the finite - size scaling functions at the classical wetting transition ( sometimes also called interface localization / delocalization transition @xcite ) in two dimensions . in particular , we have exact predictions for the fss behavior of the magnetization profile and of the correlation function of the layer magnetization in the whole crossover region close to the transition . in the context of the wetting transition , one may also consider a second important surface parameter , which is usually referred to as surface enhancement @xcite . while a boundary magnetic field is the most relevant magnetic perturbation that breaks the @xmath20 symmetry at the boundary , surface enhancement is the most relevant ( energy - like ) perturbation that is @xmath20 invariant . in the context of the quantum chain , such an interaction can be mimicked by adding a term ( @xmath452 , localized at the boundary , to the general ising - chain hamiltonian . it is not difficult to generalize the results presented here to this more general case ( results for the spectrum in the absence of boundary magnetic fields appear in refs . @xcite ) . at the critical point , this would allow us to study the behavior at the ordinary and extraordinary surface transitions ( in the absence of surface enhancement the point @xmath165 , @xmath144 corresponds to the so - called special transition @xcite ) . moreover , a complete understanding of the phase behavior for @xmath218 in the absence of a bulk magnetic field should be possible . in the quantum case , these results would also be relevant for ising quantum rings with localized site defects , a distinct case with respect to that considered in ref . in this appendix we collect results on the spectrum of some matrices that are relevant for the discussion . we begin by considering the @xmath453 matrix ( we write it for @xmath454 ) @xmath455 to determine its eigenvalues , we compute @xmath456 if we set @xmath457 , the determinant @xmath458 satisfies the recursion relation @xmath459 the solution is @xmath460 where @xmath461 and @xmath462 are arbitrary constant . if we require @xmath463 and @xmath464 , we obtain @xmath465 it follows @xmath466 the eigenvalues are solutions of the equation @xmath467 . we obtain @xmath468 eigenvalues given by @xmath469 with @xmath470 . let us now compute the eigenvectors . let @xmath471 be the eigenvector corresponding to eigenvalue @xmath472 . it satisfies the relations @xmath473 the solution of the recursion relation ( [ eigenvectors - relation-2 ] ) is @xmath474 where @xmath475 and @xmath476 are two arbitrary constants . requiring @xmath477 , we obtain @xmath478 , so that @xmath479 where we have rescaled @xmath475 for convenience . requiring the eigenvector to have unit norm , we obtain @xmath480 we also determine the eigenvalues of the matrix @xmath481 which differs from @xmath65 only for two matrix elements : @xmath482 and @xmath483 are equal to @xmath484 . setting again @xmath485 and @xmath457 , we obtain @xmath486 using the recursion relation ( [ recursion - an ] ) , we obtain finally @xmath487 where @xmath488 . to determine the spectrum of hamiltonian ( [ def - hextended ] ) , we compute the eigenvalues of the matrix @xmath84 defined in eq . ( [ matrixc - def ] ) . we write them as @xmath489 note that nothing forbids @xmath490 to be larger than @xmath491 or smaller that @xmath492 , hence @xmath493 is not necessarily bounded between @xmath494 and @xmath495 . hence , we must consider real values of @xmath159 with @xmath496 and complex values of the form @xmath497 and @xmath498 with @xmath137 . we are not interested in the zero eigenvalue of @xmath84 , hence we will only consider the @xmath499 matrix formed by the first @xmath500 rows and columns , which we label with the symbol @xmath501 . then , we introduce a matrix @xmath502 defined as ( here @xmath503 ) @xmath504 the corresponding secular equations are @xmath505 the determinant @xmath506 has already been computed in [ app.a1 ] : @xmath507 to determine @xmath508 , we expand the determinant with respect to the first and the last rows . we obtain the following recursion relation : @xmath509 a simpler expression is obtain by eliminating @xmath510 and @xmath511 , and expressing the result in terms of @xmath512 and @xmath513 . this can be done by using the recursion relation satisfied by the determinants @xmath506 , see eq . ( [ recursion - an ] ) with @xmath514 . we obtain finally @xmath515 d_{n-2}. \nonumber \end{aligned}\ ] ] note that the recursion relation ( [ rec - cn - eta1-eta2 ] ) is symmetric with respect to the exchange of @xmath91 and @xmath92 . the eigenvalues of @xmath516 are obtained by requiring @xmath517 . setting @xmath518 and @xmath519 , we obtain eq . ( [ secular - equation ] ) . note that we have simplified @xmath520 in the denominators , hence eq . ( [ secular - equation ] ) has a spurious solution for @xmath143 . a true solution with @xmath141 occurs only when the coefficients of @xmath512 and @xmath513 in eq . ( [ rec - cn - eta1-eta2 ] ) both vanish for @xmath141 . this only occurs when @xmath521 . the low - energy behavior of the model with hamiltonian ( [ isc ] ) can be understood analytically in the limit @xmath391 . for @xmath522 , it is trivial to compute the spectrum of hamiltonian ( [ isc ] ) in the presence of the magnetic boundary term ( [ hb ] ) , since @xmath47 is diagonal in the basis in which @xmath16 is diagonal . assuming @xmath523 and @xmath524 , there are two family of states that control the low - energy spectrum . first , we should consider the magnetized states @xmath525 then , we should also consider the @xmath526 states ( we call them kink states @xcite ) @xmath527 if @xmath528 is the hamiltonian for @xmath529 we have ( @xmath3 ) @xmath530 the value of @xmath198 determines which of the states is the ground state of the system . for @xmath339 , the ground state is doubly degenerate and spanned by @xmath531 and @xmath532 , while for @xmath533 the @xmath526 kink states are the lowest - energy ones . for @xmath534 , the magnetized and the kink states are degenerate . we wish now to determine how this picture changes when the perturbation @xmath535 is added . let us first understand the behavior for @xmath339 . for @xmath522 the ground state is doubly degenerate . such a degeneracy is lifted by perturbation ( [ hg - def ] ) . note , however , that @xmath536 vanishes for any @xmath537 . therefore , the energy gap is proportional to @xmath538 , in agreement with the exact results obtained in sec . [ sec5 ] . let us now consider the corrections to the ground state which are proportional to @xmath303 . at order @xmath303 , @xmath531 mixes with the states generated by @xmath539 , so that we can write @xmath540 where @xmath541 , @xmath542 , @xmath201 are of order @xmath303 and state @xmath543 is defined so that @xmath544 for @xmath545 and @xmath546 . requiring @xmath547 to be normalized , we immediately obtain @xmath548 . the coefficients @xmath541 , @xmath542 , and @xmath201 are fixed by the eigenvalue condition @xmath549 we obtain @xmath550 the analysis for @xmath532 is analogous and leads to @xmath551 . once the degeneracy is lifted , the ground state should be an eigenstate of @xmath552 , hence it should be of the form @xmath553 we are now in the position to compute correlation functions on the ground state . for the local magnetization , we obtain @xmath554 this result shows the the ground state does not show a local magnetization , except at the boundaries . for @xmath555 , the local magnetization @xmath297 vanishes as in the zero - field case . the applied field @xmath198 is too small to destabilize the low - energy state . we can analogously compute the correlation function @xmath298 : @xmath556 where @xmath557 . except on the boundaries , the correlation function is equal to the square of the bulk magnetization @xmath302 , as in the zero - field case . let us now consider the large - field region in which the ground state is @xmath526 degenerate and the relevant states are the kink states ( [ kink - states ] ) . the degeneracy is lifted when @xmath558 is included . the full hamiltonian restricted to the subspace spanned by the kink states has the form ( we write it for @xmath559 ) @xmath560 where @xmath561 . the eigenvalues and eigenvectors of this matrix have been computed in [ app.a1 ] . there are @xmath562 energy states @xmath563 with @xmath564 . the ground state corresponds to @xmath565 . the corresponding eigenfunctions are @xmath566 with @xmath567 . using this expression , we can compute the average of @xmath16 . for the average on state @xmath116 , we obtain @xmath568 where @xmath569 . if we now consider the ground state , in the large-@xmath0 limit , we can rewrite eq . ( [ magnetization ] ) as @xmath570 if we define @xmath383 and @xmath308 we obtain the more symmetric form @xmath571 analogously , we can compute the correlation function . we obtain @xmath572 where @xmath573 . focusing again on the ground state and taking the limit @xmath260 , if @xmath574 , @xmath575 , we obtain @xmath576 we should now discuss the behavior of the system in the intermediate case in which kink states and ferromagnetic states are degenerate . for @xmath522 this case corresponds to @xmath577 . as discussed in sec . [ sec5 ] , when the @xmath303 dependent term is added , the intermediate case corresponds to @xmath578 . therefore , to discuss the intermediate case , we set @xmath579 and consider the model restricted to the @xmath500 subspace spanned by @xmath580 , @xmath581 ( @xmath582 ) , @xmath583 . if @xmath584 is the energy of the kink states , the hamiltonian can be written as ( here @xmath503 ) @xmath585 the eigenvalues can be computed using the results of [ app.a1 ] . indeed , @xmath47 has the same form as matrix @xmath586 defined in eq . ( [ matrix - atilde - def ] ) . since @xmath587 , the secular equation ( [ recursion - tildean ] ) becomes @xmath588 which implies @xmath589 as the energy of each mode is given by @xmath590 , the ground state is obtained by taking @xmath591 . the ground - state energy is @xmath592 and the corresponding eigenvector is simply @xmath593 it is easy to compute the local magnetization and the correlation function . we find @xmath594 which shows that @xmath297 varies linearly between @xmath595 and @xmath596 . if we set , as usual , @xmath383 and @xmath308 we obtain in the large-@xmath0 limit @xmath597 the two - point correlation function is also easily computed @xmath598 let us now study the crossover behavior . we consider again the same basis as in [ app.b3 ] . the hamiltonian becomes ( here @xmath503 ) @xmath599 the secular equation is obtained by using the results of [ app.a1 ] , eq . ( [ recursion - tildean ] ) . we set @xmath600 , @xmath601 , @xmath602 , and @xmath590 . since @xmath603 , we obtain @xmath604 \sin k ( l+1 ) = 0.\ ] ] in the kink phase in which @xmath365 , @xmath159 varies between @xmath605 for @xmath419 and @xmath606 for @xmath370 . hence , we write @xmath607 . expanding the secular equation to order @xmath446 we obtain @xmath608 which coincides with eq . ( [ sec - etac-1 ] ) . the solutions of this equation either solve @xmath609 or @xmath610 . the ground state corresponds to the lowest value of @xmath381 that is a solution of eq . ( [ gap - appb ] ) , hence it satisfies @xmath609 . note the limiting values : @xmath611 we wish now to compute the ground - state eigenfunction , which we express as @xmath612 in the basis @xmath613 , @xmath614 , defined in [ app.b3 ] . if @xmath615 , the coefficients satisfy the relations @xmath616 the solution of eq . ( [ rec - alpha - scaling-2 ] ) is @xmath617 the constants @xmath475 and @xmath476 are fixed by condition ( [ rec - alpha - scaling ] ) and by the normalization condition @xmath618 . in the large-@xmath0 limit these two conditions give @xmath619 we can now compute @xmath297 and @xmath298 . for the local magnetization we have @xmath620 in the large-@xmath0 limit , using @xmath621 , we obtain @xmath622 where @xmath383 , @xmath308 . note that the prefactor guarantees that @xmath297 is equal to @xmath623 at the two boundaries . we can also compute the correlation function , which satisfies the relation @xmath624 correspondingly , we obtain @xmath625 and @xmath626 let us finally consider @xmath627 defined in eq . ( [ defden ] ) . for @xmath377 , @xmath628 , so that @xmath629 . using the definition we obtain @xmath630 substituting the explicit expression of @xmath631 and using @xmath609 , we find @xmath632 the above - reported results apply to the kink phase @xmath352 . for @xmath356 the ground state is a localized state with @xmath633 , where @xmath358 satisfies the equation @xmath634 the solutions of this equation satisfy either @xmath635 or @xmath636 . since @xmath637 , the ground state is obtained by considering the largest positive solution of the equation . it is then easy to prove that it satisfies @xmath635 . the solution for @xmath356 is then obtained by analytic continuation of the results obtained for @xmath365 . it is enough to replace @xmath381 with @xmath638 . we therefore obtain @xmath639 note that @xmath640 for @xmath372 , so that @xmath641 in this limit . 99 g. g. cabrera and r. jullien , universality of finite - size scaling : the role of the boundary conditions , phys . rev . lett . * 57 * , 393 ( 1986 ) ; role of boundary conditions in the finite - size ising model , phys . b * 35 * , 7062 ( 1987 ) . a. dutta , u. divakaran , d. sen , b. k. chakrabarti , t. f. rosenbaum , and g. aeppli , quantum phase transitions in transverse field spin models : from statistical physics to quantum information , arxiv:1012.0653v3 ; a. dutta , g. aeppli , b. k. chakrabarti , u. divakaran , t. f. rosenbaum and d. sen , _ quantum phase transitions in transverse field spin models : from statistical physics to quantum information _ ( cambridge university press , cambridge , 2015 ) .

with the progress in short pulse laser technology @xcite high intensity electromagnetic fields are becoming broadly available . in particular , they make it possible to create strongly correlated quantum plasmas under extreme nonequilibrium conditions which opens a broad range of applications , e.g. @xcite . at the same time , optical techniques for time resolved diagnostics are improving remarkably @xcite . these developments create the need for a quantum kinetic theory of dense nonideal plasmas in intense laser fields . nonequilibrium properties of dense plasmas in which collisions are important are usually studied on the basis of kinetic equations of the boltzmann type . however , in spite of their fundamental character , boltzmann like kinetic equations have a number of shortcomings , in particular in view of their application to dense plasmas in intense laser fields : i ) : : they are valid only for times larger than the correlation ( or collision ) time @xmath0 , ii ) : : they conserve only the mean kinetic energy instead of the sum of kinetic and potential energy , iii ) : : they are valid only in the weak field limit since the corresponding collision integrals are independent of the electromagnetic field , iv ) : : they are not applicable to high frequency processes ( fields ) , where @xmath1 , cf . i ) . obviously , in the case of strong correlations , high frequency electromagnetic fields and/or short time phenomena generalizations are necessary . _ generalized kinetic equations for correlated plasmas _ have been derived already in the 60ies by prigogine @xcite , zwanzig @xcite , kadanoff and baym @xcite , balescu @xcite , silin @xcite , klimontovich @xcite and others . in recent years , the increasing interest in ultrafast processes has revived the theoretical activities , e.g. @xcite , acompanied by progress in numerical solutions , e.g. @xcite ; for textbook overviews , see @xcite . furthermore , kinetic equations for classical plasmas in _ high frequency fields _ have been derived for the first time in papers of silin @xcite . among other problems , he computed the high frequency conductivity of a plasma . we mention that , for the weak field limit , this problem has been studied by many authors , including oberman et al . @xcite and dubois et al . an essential further development of the theory has been given by klimontovich and co workers @xcite . klimontovich used his powerful technique of second quantization in phase space @xcite to investigate the density density and microfield fluctuations in low and high frequency fields . this allowed him to derive collision integrals for classical plasmas in strong fields which take into account dynamical screening and to derive a complete theory of transport processes @xcite . a result of central importance is an expression for the collisional heating rate and the electron ion collision frequency in strong fields in terms of the imaginary part of the inverse dielectric function @xmath2 @xcite . recently , expressions of the same form were derived again @xcite . for a recent overview on the collision frequency in laser plasmas , we refer to mulser et al . @xcite . the above kinetic theories for plasmas in electromagnetic fields were limited to classical plasmas . a first extension to quantum plasmas was given by silin and uryupin @xcite . more recently , a kinetic equation for dense quantum plasmas in strong static fields has been derived @xcite , whereas a systematic quantum kinetic theory for plasmas in strong fields of arbitrary time dependence was presented in ref . @xcite . there , electron electron and electron ion quantum collision integrals were derived within the static born approximation ( landau collision integrals ) . in this paper , we extend this theory to the case of full dynamical screening for dense quantum plasmas under arbitrary nonequilibrium conditions by using the random phase approximation ( polarization approximation ) for the particle particle scattering processes . our approach is based on the nonequilibrium green s functions formalism which allows for the most straightforward derivation and for an explicit solution of the gauge problem . we derive a kinetic equation which is a generalization of klimontovich s classical result @xcite and , on the other hand , generalizes previous quantum results for the case of zero field @xcite and static field @xcite . furthermore , we derive results for the polarization and screening properties and longitudinal field fluctuations in a strong field of arbitrary time dependence . we consider the time evolution of a dense charged particle system under the influence of a strong time dependent electromagnetic field and inter particle correlations . it is instructive to recall first the motion of classical free charges in an external field . from integrating newton s equation , @xmath3 , we obtain the velocity change of a particle with charge @xmath4 and mass @xmath5 in the field @xmath6 during a time interval @xmath7 $ ] , @xmath8 and the field induced displacement @xmath9 in the equations above , we have dropped contributions from the acceleration and velocity at the initial moment @xmath10 since they are not related to the field . a further important quantity is the average kinetic energy @xmath11 where , in case of a periodic field , @xmath12 is the oscillation period . below we will also need the change of the relative velocity of a particle pair @xmath13 gained in the field @xmath14 @xmath15 and the change of the inter particle distance @xmath16 @xmath17 the most important special case is that of a _ harmonic time dependence _ @xmath18 which leads to the following explicit results for the quantities introduced above @xmath19 , \label{va - h } \\ \delta{\bf r}_a(t , t ' ) & = & { \bf r}_a^0 \left[\omega ( t - t')\sin \omega t + \cos \omega t - \cos \omega t'\right ] . \label{va - ra - h}\end{aligned}\ ] ] here , we introduced two important quantities , the `` quiver '' velocity @xmath20 and the so called excursion amplitude @xmath21 furthermore , the cycle averaged ( @xmath22 ) kinetic energy gain of a charged particle ( [ e - kin - av ] ) is the so called ponderomotive energy @xmath23 obviously , the above results may be extended to electromagnetic fields with arbitrary time dependence , for example by expanding the field in terms of harmonic components . nevertheless , it is useful to explicitly consider a second situation frequently encountered in modern applications : pulsed fields , e.g. those produced by femtosecond lasers . we will consider pulses of the following form @xmath24 and @xmath25 otherwise . typically , @xmath26 , although modern femtosecond laser pulses may be as short as a few periods of the main frequency @xmath27 . for the field ( [ e - pulse ] ) , we obtain the velocity change and displacement @xmath28 , \quad { \bf v}_a^{0\pm } \equiv \frac{e_a { \bf e}_0}{m_a \omega^{\pm } } , \quad \omega^{\pm } \equiv \omega_p \pm \omega , \\ \delta{\bf r}_a(t , t ' ) & = & \sum_{s=\pm}{\bf r}_a^{0s } \left[-\omega^s ( t - t')\cos \omega^s t + \sin \omega^s t - \sin \omega^s t'\right ] , \ ] ] with @xmath29 , whereas the change of relative velocity and two particle distance , @xmath30 and @xmath31 follow from @xmath32 and @xmath33 by replacing @xmath34 by @xmath35 and @xmath36 by @xmath37 , respectively . the above results trivially include the case of a time - independent electric field which is recovered by letting in eq . ( [ efeld ] ) @xmath38 . the corresponding results are @xmath39 it is instructive to consider a number of parameters which characterize the state of the plasma , field strength , quantum properties etc : 1 . the field strength can be characterized by the ratio @xmath40 of the amplitude of the oscillation velocity ( quiver velocity ) @xmath41 , eq . ( [ va - h ] ) , to the thermal velocity @xmath42 . the relative importance of the field and of particle particle interaction is characterized by @xmath43 , where @xmath44 is the amplitude of the field induced displacement ( [ va - ra - h ] ) , and @xmath45 is the debye radius ( interaction range ) . 3 . the frequency of the field has to be compared to the eigenfrequencies of the plasma , most importrantly , the electron langmuir ( plasma ) frequency , @xmath46 , which reflects competition between field frequency and plasma density effects . the relevance of collisional processes depends on the ratio @xmath47 , where @xmath48 is the total collision frequency of electrons in the plasma . the photon energy is characterized by its ratio to the thermal energy , @xmath49 . modern lasers easily produce strong fields which satisfy the inequalities @xmath50 and @xmath51 . in high frequency fields and/or plasmas of moderate density , @xmath52 , which allows to treat collisions perturbatively , see below . coulomb interaction between the charged carriers as well as quantum effects , obviously , may drastically modify the free particle behavior . scattering of two particles with charges @xmath53 in quantum states @xmath54 and @xmath55 on the coulomb potential @xmath56 leads to a transfer of momentum @xmath57 between them , so after the collision time @xmath58 the particles are in momentum states @xmath59 and @xmath60 . while conventional kinetic approaches treat collision as instantaneous , @xmath61 , this is not appropriate for correlated plasmas as well as in the presence of rapidly varying fields with @xmath62 not being small . in this case , during the collision time , the scattering partners will be accelerated by the external field which is called _ intra collisional field effect _ , which essentially modifies the scattering process . using a quantum language , during the collision time , the particles may absorb photons of the electromagnetic field which is the familiar _ inverse bremsstrahlung _ , or re emit them ( bremsstrahlung ) . the kinetic treatment of two particle scattering on the coulomb potential ( [ coulomb ] ) leads to the well known divergencies at short and long wavelengths . while the first is naturally cured by a quantum theoretical approach , the origin of the latter is the long range of the coulomb interaction . the familiar solution lies in the replacement of the bare coulomb potential ( [ coulomb ] ) by a screened one @xmath63 where @xmath64 is the retarded dielectric function . this is not only of fundamental interest but has also practical relevance . the dielectric function includes collective plasma oscillations and instabilities which , especially in nonequilibrium situations , may strongly enhance scattering , transport and energy exchange with the electromagnetic field ( anomalous transport ) . the simplest approximation for the dielectric function is the random phase approximation ( rpa ) being the quantum generalization of the vlasov dielectric function @xmath65 where @xmath66 denotes the single particle energy . interestingly , this result was derived by klimontovich and silin @xcite two years before lindhard @xcite . while this approximation is applicable to nonequilibrium situations in which the wigner distributions @xmath67 are weakly time dependent ( when the time scale @xmath68 is much longer than @xmath69 ) , on short times or for fast processes , such as in high frequency fields , generalizations are necessary . such generalizations avoid the assumption of separation of the two time scales and lead to an explicit dependence of the functions @xmath70 and @xmath71 on two times . moreover , the presence of an electromagnetic field may be expected to modify the dielectric and screening properties , leading to a very complex problem of coupled particle , screening and field dynamics . the appropriate theoretical concept to tackle this problem is provided by quantum field theory . numerous concepts have been developed to describe the mentioned above dynamics of particles and fields . among them , the most systematic and powerful is the theory of nonequilibrium green s functions . it is based on the method of relativistic quantum field theory , where charged particles and the longitudinal and transverse electromagnetic field are described on equal footing by field operators @xcite . from the equations of motion for the field operators - the dirac equation and maxwell s equations , one can derive equations of motion for all quantities of interest . among them , the most important are two time correlation functions ( green s functions ) which allow for systematic and far reaching generalizations of traditional kinetic theory . in this paper , we focus on nonrelativistic particle dynamics and start our derivations from the familiar equations of motion for the particle correlation functions @xmath72 and @xmath73 while the electromagnetic field is treated classically . the field theoretical description of plasmas is based on the creation and annihilation operators @xmath74 and @xmath75 @xcite which are defined to guarantee the spin statistics theorem , @xmath76 where @xmath77 has been assumed . the upper ( lower ) sign refers to bosons ( fermions ) , @xmath78 , and @xmath79 labels the particle species . below , we will drop the spin index and assume fermions . the nonequilibrium state of a correlated plasma is described by the two time correlation functions which are statistical averages ( with the initial density operator of the system ) of field operator products @xmath80 where @xmath81 and @xmath82 are , in nonequilibrium , independent from one another . they contain the complete dynamical and statistical information . the latter follows from their elements _ along the time diagonal _ : the one - particle density matrix is immediately obtained from the function @xmath73 according to @xmath83 whereas the dynamical information ( e.g. the single particle spectrum and the correlations ) follows from the function values _ across the diagonal _ in the @xmath84plane , in particular , from the _ spectral function _ @xmath85 , @xmath86 where @xmath87 are the retarded and advanced green s functions , defined below in eq . ( [ gra - def ] ) . in the following , it will often be convenient to use microscopic and macroscopic time and space variables being defined as @xmath88 in particular , in cases where the microscopic variables vary on much smaller scales than the macroscopic ones , it is advantageous to perform a fourier transformation with respect to @xmath89 and/or @xmath90 which leads to the frequency and momentum variables @xmath91 and @xmath92 , respectively . in particular , eq . ( [ f - def ] ) then yields the familiar wigner distribution function @xmath93 the time evolution of the correlation functions in an electromagnetic field is determined by the kadanoff baym equations @xcite @xmath94g_a^{\gtrless}(1 , 1 ' ) - \int d{\bar { \bf r}}_1 \ , \sigma_a^{\rm hf}(1,{\bar { \bf r}}_1t_1 ) g_a^{\gtrless}({\bar{\bf r}}_1t_1,1 ' ) \nonumber\\ & & = \int_{t_0}^{t_1 } d{\bar 1 } \ , \left[\sigma_a^>(1,{\bar 1})-\sigma_a^<(1,{\bar 1})\right ] \ , g_a^{\gtrless}({\bar 1 } , 1 ' ) - \int_{t_0}^{t'_1 } d{\bar 1 } \ , \sigma_a^{\gtrless } ( 1,{\bar 1})\,\left[g_a^>({\bar 1 } , 1')- g_a^<({\bar 1 } , 1')\right ] , \label{kb_eq}\end{aligned}\ ] ] which have to be fulfilled together with the adjoint equations . here , @xmath95 denotes the initial time where the system is assumed to be uncorrelated ( otherwise , the equations have to be supplemented with an initial correlation contribution to @xmath96 , cf . @xmath97 is the hartree fock selfenergy ( mean field energy with exchange ) , @xmath98 and @xmath99 are the correlation selfenergies which will be discussed below . for the following derivations , it is useful to introduce , in addition , the retarded and advanced green s functions @xmath100\left\{g_a^>(1,1')-g_a^<(1,1')\right\ } , \label{gra - def}\ ] ] which obey the simpler equations @xmath101 g_a^{r / a}(1,1 ' ) \nonumber\\ & & -\int d2 \,\sigma_a^{r / a}(1,2)g_a^{r / a}(2,1')\,\ , = \,\ , \delta(1 - 1 ' ) . \label{gra_eq}\end{aligned}\ ] ] in equations ( [ kb_eq ] ) and ( [ gra_eq ] ) , the electromagnetic field is given by the vector and scalar potentials @xmath102 and @xmath103 and will be treated classically . @xmath102 is the full vector potential ( external plus induced ) which obeys maxwell s equations , whereas @xmath103 is understood as to be due to external sources only , the induced longitudinal field is fully accounted for in the screened coulomb potential @xmath71 which enters the selfenergies @xmath104 and @xmath105 , see below . although one can directly analyze and solve the two time kadanoff baym equations ( [ kb_eq ] ) , e.g. @xcite , for an overview , see @xcite , it is easier to consider the kinetic equation for the wigner distribution function ( [ f - def2 ] ) , which we will be concerned with here . this equation is immediately obtained from the equal time limit , @xmath106 , of eq . ( [ kb_eq ] ) plus its adjoint . introducing further the variables @xmath107 and @xmath90 , eqs . ( [ r - r - def ] ) and ( [ t - tau - def ] ) , we obtain , after fourier transformation with respect to @xmath90 , for the spatially homogeneous case , @xmath108 this is an exact equation and , therefore , well suited for deriving generalized kinetic equations . however , this equation is not closed yet since it contains under the collision integral functions depending on two times . therefore , to obtain explicit expressions for the collision integral , one has : 1 . to find appropriate approximations for the self energy . for this , the green s functions approach provides powerful approximation schemes based on feynman diagrams which allow for a very systematic development of the theory . here , we are interested in the plasma dynamics with screening effects properly included . therefore , the appropriate choice for the selfenergy will be the random phase approximation ( rpa ) ; 2 . to express the correlation functions @xmath109 as functionals of the wigner functions @xmath67 ( reconstruction problem ) . this problem can be solved approximately on the basis of the generalized kadanoff baym ansatz ( gkba ) of lipavsk et al . @xcite , see sec . [ gkba_ss ] . it is well known that the electromagnetic field can be introduced in various ways ( gauges ) which may lead to essentially different explicit forms of the resulting kinetic equations . although alternative derivations are successfully applied too , gauge invariance becomes a particular problem if the resulting kinetic equations are treated by means of approximations , such as retardation or gradient expansions . a critical issue is that the result of these approximations maybe essentially different in different gauges , see e.g. @xcite for examples . to avoid these difficulties , we will formulate the theory in terms of correlation functions which are made explicitly gauge invariant . while the main results have been presented in ref . @xcite , here we provide some additional details . in this section , we use a co - variant 4-vector notation as it makes the following transformations more compact and symmetric . the corresponding definitions are @xmath110 and the convention @xmath111 is being used . one readily proofs that the kadanoff baym equations ( [ kb_eq ] ) remain covariant under gauge transformations , i.e. , under the following transformations of the potentials and field operators @xmath112 the corresponding gauge transform of the green s functions leads to @xmath113 } g_a(x , x)\ , . \ ] ] following an idea of fujita @xcite , we now introduce a gauge invariant green s function @xmath114 which is given by the modified fourier transform @xmath115\right\ } g_a(x , x ) , \label{gauge - ft}\end{aligned}\ ] ] where use has been made of the identity @xmath116 indeed , one readily confirms that under any gauge transform ( [ gauge - a ] ) , the phase factors cancel , and @xmath117 , @xcite . in the following , we focus on spatially homogeneous electric fields and use the vector potential gauge @xmath118 in this case , relation ( [ gauge - ft ] ) simplifies to @xmath119 g_a({\bf r},\tau;{\bf r},t ) , \label{gauge - ftv}\end{aligned}\ ] ] what means that the gauge invariant green s function @xmath120 follows from the wigner transformed function @xmath121 by replacing the canonical momentum @xmath92 by the gauge invariant kinematic momentum @xmath122 according to @xmath123 let us illustrate this for the examples studied in sec . [ sec_bas ] . for a _ harmonic electric field _ given by eq . ( [ efeld ] ) , the vector potential and the momentum relation become , according to eq . ( [ v - gauge ] ) , @xmath124 similarly , for a _ pulsed field _ of the form eq . ( [ e - pulse ] ) , the result for the vector potential and the momenta is @xmath125 finally , for a _ static field _ we obtain , @xmath126 for the derivations below , we will need the gauge invariant fourier transform of the convolution of two functions which , in the homogeneous case , is given by @xmath127 after straightforward manipulations which involve the back transform of ( [ gauge - ftv ] ) , we arrive at @xmath128 \nonumber \\ & & \qquad \times\,\ , c \left[{\bf k}+ \frac{e_a}{c}\int_{t'_1}^{t_1 } dt''\ , \frac{{\bf a } ( t'')}{t_1-t'_1 } - \frac{e_a}{c } \int^{\bar{t}}_{t'_1 } dt''\ , \frac{{\bf a}(t '' ) } { \bar{t}-t'_1};\,\bar{t},t'_1\right ] . \quad \label{conv - ft}\end{aligned}\ ] ] in particular , the derivation of the collision integral in the kinetic equation for the wigner function , requires the equal time limit of this expression , @xmath106 , @xmath129 \nonumber \\ & & \qquad \times\,\ , c \left[{\bf k}+\frac{e_a}{c}{\bf a}(t ) - \frac{e_a}{c}\int_{\bar t}^{t } dt '' \ , \frac{{\bf a}(t'')}{t-{\bar t } } ; \,\bar{t},t\right ] , \quad \label{conv - ftt}\end{aligned}\ ] ] where @xmath130 and @xmath131 will be replaced by @xmath132 and @xmath133 . notice that in this case the momentum arguments of @xmath130 and @xmath131 are equal . to simplify the notation below , we introduce the field induced momentum shift @xmath134 which has the important property @xmath135 where @xmath136 is nothing but minus the momentum gain of a free particle in the field , @xmath137 , cf . ( [ va ] ) . another important relation follows from multiplication by the time interval : @xmath138 where @xmath139 is just the field induced displacement of a free particle , eq . ( [ ra ] ) , i.e. @xmath140 . definition ( [ ka ] ) allows us to rewrite eq . ( [ conv - ftt ] ) as @xmath141 \ , c \left[{\bf k}+{\bf k}^a_a(t,{\bar t});\ , t,{\bar t}\right ] . \label{conv - ftt2}\end{aligned}\ ] ] as noted in sec . [ kbe_ss ] , for the derivation of the collision integral in the kinetic equation , we need to express the functions @xmath72 and @xmath73 in terms of the wigner function . in addition , such a reconstruction ansatz involves the retarded and advanced green s functions @xmath142 for which suitable expressions have to be found . we first determine these quantities for free particles in an electromagnetic field which allows to simplify eq . ( [ gra_eq ] ) to @xmath143 g_{a}^{r / a}({\bf p};t_1,t'_1)=\delta ( t_1-t'_1),\ ] ] which is solved immediately by @xmath144 ^ 2/2m_a\right ] , \label{gra_free}\end{aligned}\ ] ] and @xmath145 is obtained from the symmetry relation @xmath146^*$ ] . from this result , we can calculate the spectral function @xmath147 , eq . ( [ a - def ] ) , @xmath148 ^ 2/2m_a\right ] . \label{a_free}\end{aligned}\ ] ] obviously , the results ( [ gra_free ] ) and ( [ a_free ] ) are gauge - dependent since @xmath149 and @xmath150 are functions of the canonic momentum @xmath92 . but one can easily obtain the corresponding gauge invariant results by applying the transform ( [ gauge - ft ] ) , with the result @xmath151 } , \label{gra_fg } \\ \nonumber\\ \mbox{where}\quad s_a({\bf a};\tau , t ) & = & \frac{e_a^2}{2 m_a c^2 } \left [ \int_{t-\frac{\tau}{2}}^{t+\frac{\tau}{2 } } dt ' a^2(t ' ) + \frac{1}{\tau } \left(\int_{t-\frac{\tau}{2}}^{t+\frac{\tau}{2 } } dt ' { \bf a}(t')\right)^2\right]\,.\end{aligned}\ ] ] this result has a simple physical interpretation . for a free particle without field , the spectral function shows free undamped oscillations along @xmath89 ( i.e. perpendicular to the time diagonal ) with the one particle energy @xmath152 , and its fourier transform is @xmath153 . \label{afree}\ ] ] this clearly underlines the meaning of the functions @xmath87 and , in particular , the spectral function @xmath79 they contain the full information on the single particle energy spectrum . furthermore , in a correlated system , the single particle spectrum is affected by interactions with other particles . this leads to a shift of the oscillation frequency and to damping of the oscillations , i.e. to finite life time effects , and it is reasonable to call the corresponding single particle excitations _ quasi particles_. on the other hand , the result ( [ gra_fg ] ) reflects the influence of an electromagnetic field on the particle spectrum , while correlation effects have been neglected . ( [ gra_fg ] ) shows that the field causes a time dependent shift of the single particle energy which , obviously , reflects the well known fact that the proper eigenstates of the system contain the electromagnetic field and are given by volkov states @xcite . the spectrum may even contain additional peaks which becomes particularly transparent in the limiting case of a harmonic time dependence : for the field ( [ efeld ] ) , the time integrations in @xmath154 can be performed , and simple trigonometric relations lead to @xcite @xmath155 , \label{gra_fgh}\end{aligned}\ ] ] where @xmath156 is the ponderomotive potential which was introduced in eq . ( [ e - pond - h ] ) . the first term in the brackets leads to a shift of the single - particle energy , the average kinetic energy of the particles increased by @xmath156 . the remaining terms modify the spectrum qualitatively giving rise to additional peaks which are related to _ photon sidebands _ @xcite . now we turn to the solution of the reconstruction problem . the simplest solution is the common _ kadanoff baym ansatz _ , @xmath157 where @xmath158 and @xmath159 , and the upper ( lower ) sign refers to @xmath72 ( @xmath73 ) . indeed , the two time functions @xmath160 and @xmath161 are now expressed in terms of one time wigner distribution functions and a known spectral function . however , due to the expected retardation effects , this ansatz is not applicable here . as mentioned above , a more general solution which properly takes into account retardation ( memory ) effects is the _ generalized kadanoff baym ansatz _ proposed by lipavsk et al . @xcite which reads @xmath162 where for the functions on the time diagonal @xmath163 , cf . ( [ f - def2 ] ) . within the quasiparticle approximation and with static selfenergies , eq . ( [ gkba ] ) is exact . in more complex situations , it is an approximation to the exact reconstruction solution , which has prooved extremely successful in many applications . in particular , it has been used for more general selfenergies and also with more general propagators @xmath164 , e.g. @xcite . we , therefore will use this ansatz below . ( [ gkba ] ) is written in terms of the momentum @xmath92 and is , therefore , gauge dependent . to transform this relation into a gauge invariant form , we use its coordinate representation , @xmath165 and apply the gauge invariant fourier transform ( [ gauge - ftv ] ) together with the back transforms of @xmath164 and @xmath166 which leads to the _ gauge invariant generalization of the gkba _ @xmath167 - f_a^{\gtrless}\left[{\bf k}-{\bf k}^a_a(t , t ' ) ; t_1\right ] \ , g_a^a({\bf k};t_1,t'_1 ) , \label{gkba_gi}\end{aligned}\ ] ] where the definition ( [ ka ] ) for @xmath168 has been used . as in the field free case , the first term is nonzero only for @xmath169 and the second in the opposite case . notice the difference of the time arguments in the two distribution functions . we now come back to the time diagonal limit of the kadanoff baym equations , cf . ( [ kb_eq ] ) and ( [ f_eq ] ) , and derive the quantum kinetic equation for a plasma in a laser field thereby fully taking into account dynamical screening . again , it is advantageous to derive this equation for the gauge invariant wigner distribution . to this end , we take the fourier transform ( [ gauge - ftv ] ) of the time diagonal kadanoff baym equation @xmath170 where the full arguments are , according to the convolution relation ( [ conv - ftt2 ] ) , given by @xmath171 \ , g_a^{\lessgtr}\left[{\bf k}_a+{\bf k}_a^a(t,{\bar t});{\bar t},t\right ] . \nonumber\end{aligned}\ ] ] this expression is valid for arbitrary approximations for the selfenergies @xmath172 and @xmath173 . in our previous paper @xcite , we used the simple static born approximation . here , we are interested in a fully selfconsistent _ inclusion of dynamical screening _ , so the appropriate choice is the random phase approximation ( rpa ) . starting from the familiar expression in coordinate representation , application of the gauge invariant fourier transform ( [ gauge - ftv ] ) straightforwardly leads to the following gauge invariant result @xmath174 which transforms the collision integral of eq.([f_eqgi ] ) into @xmath175 \nonumber\\ & \times & i\hbar v_{aa}^{s\,>}({\bf q};t,{\bar t})\ , g_a^<\left[{\bf k}_a+{\bf k}_a^a(t,{\bar t});{\bar t},t\right]\ , \quad - \quad [ \;>\ ; \longleftrightarrow \;<\ ; ] \bigg\}\end{aligned}\ ] ] in eq . ( [ sigma - rpa ] ) we introduced the correlation functions of the screened potential ( plasmon correlation functions ) @xmath176 which contain the whole screening problem and are directly related to the correlation function of the _ longitudinal field fluctuations _ ( microfield fluctuations ) @xcite @xmath177 . \label{dede}\end{aligned}\ ] ] while in the classical case , the contributions from @xmath178 and @xmath179 are equal , in the quantum case a symmetrization is useful which is indicated by the bar over the fluctuation term . @xmath180 and @xmath181 can be related to the retarded and advanced screened potentials via the optical theorem @xmath182 where @xmath183 and @xmath184 obey the following equation of motion ( dyson equation ) @xmath185 and @xmath186 are related to the nonequilibrium inverse dielectric function according to @xmath187^{-1}. \label{idf}\end{aligned}\ ] ] in the above equations , @xmath188 is the bare coulomb potential ( [ coulomb ] ) and @xmath189 the retarded and advanced longitudinal polarization functions ( plasmon selfenergies ) . to close this system of equations , the polarization functions have to be expressed in terms of the particle correlation functions for which the simplest approximation is provided by the rpa , @xmath190 and @xmath191 follows from the relation @xmath192^*$ ] . this set of equations completely defines the non markovian polarization approximation ( rpa ) for a quantum plasma in a strong transverse field . what is left to obtain a closed expression for the collision integral in equation ( [ f_eqgi ] ) is to apply the gauge invariant gkba ( [ gkba_gi ] ) together with the free particle approximation ( [ gra_fg ] ) to all two time functions . this leads to the following results for the optical theorem : @xmath193 } \nonumber\\ & & \times f_b^{\gtrless}\left[{\bf k}+{\bf q}+{\bf q}_b(t_3,t_4);t_4\right]\ , f_b^{\lessgtr}\left[{\bf k}+{\bf q}_b(t_3,t_4);t_4\right ] \nonumber\\ & & + \theta(t_4-t_3)\,e^{-\frac{i}{\hbar}\left [ \left(\epsilon^b_{{\bf k}+{\bf q}}-\epsilon^b_{{\bf k}}\right)(t_3-t_4 ) + { \bf q}{\bf r}_b(t_4,t_3 ) \right ] } \nonumber\\ & & \times f_b^{\gtrless}\left[{\bf k}+{\bf q}+{\bf q}_b(t_4,t_3);t_3\right]\ , f_b^{\lessgtr}\left[{\bf k}+{\bf q}_b(t_4,t_3);t_3\right ] \bigg\}. \label{vs - gkba}\end{aligned}\ ] ] similarly , the result for @xmath194 can be transformed to @xmath195- f_a\left[{\bf k+q}+{\bf q}_a(t , t');t'\right ] \right\ } , \label{pira - gkba}\end{aligned}\ ] ] where the momentum shift @xmath136 and field induced displacement @xmath139 were defined above in eqs . ( [ ka - dif ] ) and eqs . ( [ ka - dt ] ) , respectively . in the absence of the electromagnetic field , ( @xmath196 ) , eq . ( [ pira - gkba ] ) reduces to the well known nonequilibrium rpa polarization function . the effect of the field is two fold : first , it introduces an additional retardation @xmath136 in the distributions ( intra collisional field effect ) and second , it leads to a modification of the one particle energies in the exponent given by @xmath139 which we discussed in detail in ref . @xcite . we now can transform the collision integral eq . ( [ coli1 ] ) by applying the gkba ( [ gkba_gi ] ) to @xmath197 , and using the result for @xmath198 and @xmath199 , eq . ( [ vs - gkba ] ) . after straightforward calculations , we obtain the final result @xmath200}\ , f_a^>\left[{\bar { \bf k}}_a+{\bf q}_a(t,{\bar t});{\bar t}\right ] f_a^<\left[{\bf k}_a+{\bf q}_a(t,{\bar t});{\bar t}\right]\ , \nonumber\\ & \times&\int_{t_0}^t d t_3 \;\int_{t_0}^{\bar t } d t_4 v^{s\,r}({\bf k}_a-{\bar { \bf k}}_a;t_1,t_3)\ , v^{s\,a}({\bf k}_a-{\bar { \bf k}}_a;t_4,t_2)\,\times \nonumber\\ & & \bigg [ \theta(t_3-t_4)\,e^{-\frac{i}{\hbar}\left [ \left({\bar \epsilon}_b-\epsilon_b\right)(t_3-t_4 ) + ( { \bf k}_b-{\bar { \bf k}}_b){\bf r}_b(t_3,t_4 ) \right ] } \nonumber\\ & & \times f_b^>\left[{\bar { \bf k}}_b+{\bf q}_b(t_3,t_4);t_4\right]\ , f_b^<\left[{\bf k}_b+{\bf q}_b(t_3,t_4);t_4\right ] \nonumber\\ & & + \theta(t_4-t_3)\,e^{-\frac{i}{\hbar}\left [ \left({\bar \epsilon}_b-\epsilon_b\right)(t_3-t_4 ) -({\bf k}_b-{\bar { \bf k}}_b){\bf r}_b(t_4,t_3 ) \right ] } \nonumber\\ & & \times f_b^>\left[{\bar { \bf k}}_b+{\bf q}_b(t_4,t_3);t_3\right]\ , f_b^<\left[{\bf k}_b+{\bf q}_b(t_4,t_3);t_3\right ] \bigg ] \quad - \quad [ \;>\ ; \longleftrightarrow \;<\ ; ] \bigg\}.\end{aligned}\ ] ] this is the general non markovian collision integral for a homogeneous weakly coupled _ dynamically screened plasma in an electromagnetic field_. it is a generalization of numerous results which were previously obtained by various authors , including our static screening result for strong time dependent fields @xcite and the result of silin and uryupin @xcite , the rpa result for a static field of morawetz @xcite and the field - free rpa results of kuznetsov @xcite and haug and ell @xcite . furthermore , it generalizes previous results obtained for classical plasmas by silin , oberman et al . , klimontovich and others @xcite . in particular , as we will see below , the classical dynamical screening result of klimontovich and puchkov @xcite is straightforwardly recovered from the collision integral ( [ coli3 ] ) . this collision integral is the basis for computing electron ion collision frequencies , plasma heating inverse bremsstrahlung effects etc . , thereby fully taking into account dynamical screening , plasma instabilities and anomalous transport despite the complicated structure of the integral ( [ coli3 ] ) , a direct numerical integration of the kinetic equation ( [ f_eqgi ] ) appears to be within reach , as recently solutions of non markovian rpa - type equations ( without longitudinal field ) for semiconductors have been reported @xcite . on the other hand , to gain deeper insight in the physical processes contained in the collision term ( [ coli3 ] ) , it is useful to consider analytical simplifications . if the collision frequency is low compared to the oscillation frequency of the field , i.e. if the parameter @xmath52 , one may follow an idea of silin @xcite and solve the kinetic equation ( [ f_eqgi ] ) with a perturbation ansatz @xmath201 where @xmath202 obeys the collisionless equation @xmath203 , \label{silin_an}\end{aligned}\ ] ] where @xmath204 is an arbitrary function depending on the initial conditions . the equation for @xmath205 reads @xmath206 \right\}. \label{silin_an1}\end{aligned}\ ] ] with this scheme , there follow essential simplifications of the above results because it effectively eliminates the time retardation of the distribution functions in the collision integrals @xcite . indeed , one easily verifies that the arguments of the distributions which appear in the formulas above , now become @xmath207 \longrightarrow f_{a0}\left[{\bf k}+\frac{e_a}{c}{\bf a}(t)\right ] , \label{farg - silin}\end{aligned}\ ] ] and do not depend on the time @xmath10 anymore . this simplification allows to compute the transport , screening and fluctuation properties quite efficiently . we first consider the modification of the longitudinal polarization @xmath208 . straightforward transformations of eq . ( [ pira - gkba ] ) including a change of the momentum integration variable lead to @xmath209 using this result and the adiabatic approximation ( neglecting the ion contribution to the polarization ) , we obtain from the dyson equation , eq . ( [ vdyson ] ) , for the retarded screened potential @xmath210 where again , @xmath211 denotes the screened potential in the zero field limit and without retardation in the distribution functions . simplifications are possible also for the collision integrals . in particular , we obtain for the electron ion scattering term , @xmath212\tau } \nonumber\\ & & \ ; \times \int_{0}^{\infty}d\tau_1\int_{-\infty}^{0}d\tau_2\frac{1 } { \epsilon_{0}^{r}({\bf q},\tau_1)}\frac{1}{\epsilon_{0}^{a}({\bf q},\tau_2 ) } \ , e^{-\frac{i}{\hbar}\frac{e_e}{m_e c}{\bf q } \int\limits_{t-\tau-\tau_2}^{t-\tau_1 } d{\bar t}\,{\bf a}({\bar t } ) } \left\{f_{e0}({\bf k - q})-f_{e0}({\bf k } ) \right\ } , \quad \label{iei - silin}\end{aligned}\ ] ] where @xmath213 are the field free inverse dielectric functions . ( [ iei - silin ] ) is the generalization of klimontovich s result @xcite to quantum plasmas and . finally , we consider the simplifications introduced by the ansatz ( [ silin_an1 ] ) to the field fluctuations . the quantity of central importance is the correlation function of the temporal and spatial microfield fluctuations @xmath214 , eq . ( [ dede ] ) . a lengthy but straightforward calculation leads to the following result @xmath215(t_1-t_2 ) } } { |\epsilon_{0}({\bf q},\epsilon^{e}_{{\bf k+q}}-\epsilon^{e}_{{\bf k}})|^2 } \left\{f_{e0}({\bf k+q})[1-f_{e0}({\bf k})]+ f_{e0}({\bf k})[1-f_{e0}({\bf k+q})]\right\ } \nonumber\\ & & \qquad + \;e_i^2 \ , e^{-\frac{i}{\hbar}[\epsilon^{i}_{{\bf k+q}}-\epsilon^{i}_{{\bf k}}](t_1-t_2 ) } \int_{0}^{\infty } d\tau_1 \ , \frac{e^{-\frac{i}{\hbar } [ \epsilon^{i}_{{\bf k+q}}-\epsilon^{i}_{{\bf k}}]\tau_1 } } { \epsilon_{0}^{r}({\bf q},\tau_1 ) } \nonumber\\ & & \qquad\quad\times \int^{0}_{-\infty } d\tau_2 \ , \frac{e^{\frac{i}{\hbar } [ \epsilon^{i}_{{\bf k+q}}-\epsilon^{i}_{{\bf k}}]\tau_2 } } { \epsilon_{0}^{a}({\bf q},\tau_2 ) } \ ; e^{-\frac{i}{\hbar}\left(\frac{e_e}{m_e}-\frac{e_i}{m_i}\right){\bf q } \int\limits_{t_2-\tau_2}^{t_1-\tau_1 } d{\bar t}\,\frac{{\bf a}}{c}({\bar t } ) } \left[f_{i0}({\bf k+q})+f_{i0}({\bf k})\right ] \bigg\ } , \nonumber \label{de - silin}\end{aligned}\ ] ] where the first term in parantheses ( second line ) is the electron contribution , and the second ( third and fourth line ) results from the ions . again , this is a generalization of klimontovich s remarkable result @xcite who considered the classical limit and the equal time fluctuations , @xmath77 . our result fully includes the two time fluctuations which are directly measurable quantities . from the above fluctuation spectrum , all major observables of dense quantum plasmas in a strong laser field can be computed . the corresponding analysis will be presented in a forthcoming paper . in this paper , we have presented a gauge invariant derivation of the quantum kinetic equation for dense plasmas in a laser field . our main result , eq . ( [ coli3 ] ) , generalizes previous work to quantum systems . this equation can be used to calculate the transport properties of a dense plasma in a laser field on arbitrary time scales , i.e. , over the whole frequency range . the use of the random phase approximation allows for a highly consistent treatment of the combined effect of internal longitudinal fields ( dynamical screening ) and transverse electromagnetic fields , including intense laser pulses . in particular , it allows to investigate the influence of the electromagnetic field on the two particle scattering process and the screening properties of the plasma and on the screening buildup in the presence of a strong field . besides , the presented gauge invariant approach is completely general and can be extended straightforwardly to more complex situations , including strong coupling effects , bound states , impact and field ionization . moreover , it can be directly generalized to relativistic systems und ultra intense fields . it is a great pleasure to dedicate this paper to yuri lvovich klimontovich on his seventyfith birthday . all of us have , at different times , come in contact with the remarkable papers and monographs of klimontovich . we have very much enjoyed countless discussions with him as well as rewarding scientific collaborations on problems of kinetic theory of nonideal gases and plasmas in general , in particular on dielectric properties @xcite , kinetic equations with bound states @xcite , foundation of reaction diffusion equations for nonideal plasmas @xcite , selforganization and structure formation @xcite and many others . the authors acknowledge stimulating discussions with g. kalman , p. mulser and r. sauerbrey . this work is supported by the deutsche forschungsgemeinschaft ( schwerpunkt `` wechselwirkung intensiver laserfelder mit materie '' ) , and by the european commission through the tmr network silasi . it is interesting to note that klimontovich s phase space densities @xmath216 @xcite are essentially the classical limit of the field operators . therefore , his kinetic theory is of similar generality as modern field theory , cf . his text books @xcite .

the storage and retrieval properties , as well as the dynamics , of large attractor neural networks have been studied over some time and numerous results are now available . the presence of an exponentially large number of unwanted spin - glass - like states in fully connected networks may limit severely the storage capacity , the information content and the performance , in particular the retrieval quality of previously stored patterns which act as attractors in the dynamics of the network . also , the size of the corresponding basins of attraction may be considerably reduced . other tasks , as the categorization or generalization ability of a network , are also impaired due to the presence of such states and , except for a low storage ratio , that is for a reduced level of stochastic noise , a performing network is very likely to be trapped in one of those states preventing the occurence of finite overlaps with the patterns of interest . the study of the equilibrium behavior of the hopfield model with binary neurons and extremely low _ symmetric _ connectivity @xmath3 suggested that @xcite , except for zero storage ratio @xmath4 or in the absence of synaptic noise @xmath5 , the properties of the model may be quite different from those of the extremely diluted network with asymmetric synapses , which has a trivial dynamics @xcite . in particular , the equations for the order parameters are equivalent to those of the sherrington and kirkpatrick ( sk ) spin - glass ( sg ) model @xcite . it has been shown , nevertheless , that strong symmetric dilution of synapses in the hopfield model @xcite may considerably reduce the stability of spin - glass states . indeed , in the limit @xmath6 , the retrieval states are globally stable in their whole domain of existence of the @xmath7 phase diagram , thus leaving only locally stable sg states @xcite , in contrast to the situation in the fully connected hopfield model @xcite . it has also been shown that already a gradual dilution reduces the stability of the sg states in the region where they compete with the retrieval states and that the storage capacity of the network is increased @xcite . these are important results on the performance of a network which also suggest that the basins of attraction of the memory ( retrieval ) states should be increased in symmetrically diluted networks . networks of q - ising neurons in various architectures have been studied over some time , in particular , fully connected , layered feed - forward and extremely dilute networks with asymmetric connectivity @xmath8 $ ] . they are of interest for possible hardware applications and an eventual biological modeling , because of the increased flexibility of the states of the network to account for complex neural behavior . it has been suggested that an attractor neural network model with multi - state neurons may describe the short term memorization performance of the @xmath9 region of the hyppocampus in both the brain of primates and in the human brain , and results of numerical simulations for the selective performance of a network with asymmetric synapses and small connectivity ( @xmath10 ) are now available @xcite . however , complete phase diagrams that give a global picture of the performance of a network and a physical explanation of why biological networks seem to prefer a low connectivity are still missing . attractor network models with a hebbian type of learning rule may also serve to account for long - term memory in the brain @xcite . both the parallel dynamics and the equilibrium properties of extremely diluted networks with symmetric connectivity between multi - state q - ising neurons have been studied in recent works @xcite . the symmetry of the synaptic connections allows for a detailed balance assumption in the dynamics and enables one to perform full analytic calculations of the equilibrium properties of a network . this is particularly convenient in the search for phase diagrams . so far , there are no results , to our knowledge , concerning the memorization performance of symmetrically diluted q - ising networks with small but finite connectivity and low - activity patterns , which characterize a biological network . the purpose of the present work is to consider the retrieval performance of an attractor neural network modeled on some of the features of a biological network . we study here the equilibrium behavior of a symmetrically diluted hopfield model with finite connectivity and low - activity units in q - ising states . multi - state networks are known to have complex properties and our main aim here is to find out how that behavior is changed by a finite dilution of the synapses and the low activity of the embedded patterns . we consider explicitly a @xmath0 , @xmath1 and a continuous @xmath2-state network . the outline of the paper is the following . in sec . ii we present the model and show how the hamiltonian becomes the sum of an effective hopfield - model hamiltonian and that of a sherrington - kirkpatrick - like spin - glass model . the replica - symmetric mean - field theory for the model and the relevant order parameters are derived in sec . iii and explicit expressions are presented in the appendix . the results for the phase diagrams and the retrieval performance are discussed in sec . iv , and our conclusions with a summary of the results can be found in sec . v. attractor neural networks are dynamical systems . consider a network of @xmath11 neurons , @xmath12 . at the time step @xmath13 , the state of neuron @xmath14 is described by the variable @xmath15 , that can be in any one of the @xmath16 ising states @xmath17 in the interval [ -1,1 ] , for @xmath18 . a macroscopic set of @xmath19 independent and identically distributed random patterns @xmath20 , with @xmath21 , is embedded in the network by means of a hebbian - like learning rule , specified below . here , @xmath22 is the storage ratio per connected site and @xmath3 is the connectivity of the network . every bit of each pattern can be in any one of @xmath16 equally spaced states , also in the interval [ -1,1 ] , which are assumed to have zero mean and variance @xmath23 this is a measure of the size of the patterns , describing their mean activity , that accounts for bits that either are turned off or depressed . the first task to be performed by the network is to attain a finite storage ratio @xmath4 and this ratio can be optimized for patterns of a given size by means of an appropriate tuning of the states of the network , discussed below . the second aim is to reach a sufficiently small hamming distance @xmath24 between the network state @xmath25 and a given pattern @xmath26 , as a measure of the retrieval performance of the network . this depends on both the overlap @xmath27 and the dynamical activity of the network , @xmath28 . we consider next the dynamics of the partially connected model . for a given configuration @xmath29 of the network , the local field acting on neuron @xmath14 is given by @xmath30 where @xmath31 is the synaptic coupling between neurons @xmath14 and @xmath32 . we assume that the synapses of the network are symmetrically diluted being left with a _ finite _ connectivity @xmath3 ( the fraction of connected neurons ) which , eventually , may become vanishingly small . specifically , the synapses are of the hebbian - like form , @xmath33 for our non - sparse network , in which @xmath34 . here , @xmath35 is a set of independent identically distributed random variables , such that @xmath36 with probability @xmath3 and zero with probability @xmath37 , while @xmath38 . thus , the symmetric dilution introduces an additional randomness both into the dynamics of the network , which becomes non - trivial even in the extremely diluted case due to feedback loops @xcite , as well as in the thermodynamics . we are interested , in the following , in the case of a large connected network , in which @xmath39 is very large . the state of each neuron is updated asynchronously according to a glauber ( single spin - flip ) dynamics @xcite in which the transition probabilities are given by @xmath40 } { \sum_{l=1}^q\exp[\beta\epsilon_j(\sigma_l|h_j(\{s_i(t)\}))]}\ ; , \label{7}\ ] ] where @xmath41 is the inverse synaptic noise ( temperature ) and the single site energy , @xmath42 , is given by @xmath43 here , @xmath44 is a tuning parameter favoring local states of small dynamical activity . in the absence of stochastic noise , the deterministic evolution of the system follows the updating rule @xmath45 where @xmath46 is the non - decreasing step function , for finite @xmath16 , @xmath47 \label{10}\ ] ] with @xmath48 and @xmath49 , in which @xmath50 , if @xmath51 , and zero otherwise . clearly , @xmath44 is a threshold parameter since the state of neuron @xmath32 assumes the value @xmath52 given by eq . ( [ 1 ] ) if the local field @xmath53 is bound by @xmath54 . the width of the intermediate states with constant @xmath52 for @xmath55 ( that is , excluding the limiting values of @xmath56 ) , is given by @xmath57 . thus , the width of the zero state for the three - state network studied below is @xmath58 . in the limit @xmath59 , the input - output function becomes the piecewise linear function @xmath60 where @xmath61 means the minimum between @xmath62 and @xmath63 . the slope of the linear part , @xmath64 , is the gain parameter of the continuous network . both , the locally stable states of the dynamics , as well as the globally stable states that characterize the equilibrium thermodynamic properties of the diluted network , that follow from the above dynamics , are described by the hamiltonian @xmath65 where the first sum is over all distinct pairs of neurons @xmath66 . eventually , a field - dependent term , @xmath67 , may be added to generate the overlap with a specific pattern , and this will be implicitly assumed below . we adapt next the procedure of viana and bray @xcite for diluted spin glasses in order to deal with the random dilution . in distinction to their case , which is that of a strongly interacting spin glass , we have here a diluted network with randomness in the patterns and weak interactions between units . the latter allows for an exact truncation to the relevant order in @xmath68 that is sufficient for the mean - field calculation of the following section . consider the disorder dependent part of the hamiltonian in the exponentiated form @xmath69 given a fixed set @xmath70 of patterns embedded in the network , we first have to build up a finite connectivity between units . that is to say , we have to find a network such that the mean of @xmath71 is precisely @xmath3 and , to this end , we have to perform first an average over these random variables . the average over the random patterns , that is necessary to evaluate the performance of the network , comes at a later stage . the configurational average over the set @xmath72 of the @xmath73 times replicated function @xmath74 becomes @xmath75 where @xmath76 denotes the value of @xmath77 for @xmath36 . we are interested in dense networks , that is , @xmath78 and large @xmath39 , letting eventually @xmath79 after the thermodynamic limit . for finite @xmath80 , the weak @xmath81 may be used to expand the logarithm to next - to - leading order in @xmath76 . using the law of large numbers to deal with @xmath82 we obtain @xmath83 in which the effective coupling @xmath84 is given by @xmath85 the first term is the hebbian synapsis of the fully connected network with multi - state patterns of activity @xmath86 and @xmath87 is a gaussian random variable with zero mean and a pattern - independent variance @xmath88 ; the brackets on the right - hand side of eq.(15 ) denote an average over this random variable . this generalizes an earlier result by sompolinsky @xcite showing that dilution appears as a synaptic noise . thus , the symmetric dilution introduces an effective hamiltonian , in the large @xmath39 limit , @xmath89 where the first term is a sum of a hopfield - model hamiltonian and a kind of sk - model term . this is used in the following section to derive the mean - field theory for the model . we consider now the mean - field theory for finite @xmath4 and for any connectivity @xmath78 . the averaged free energy per connected site is given by @xmath90 with the averages first over the gaussian noise in the synaptic interactions and then over the pattern distribution . the effective partition function is then given by @xmath91 and this is used in the replica method to perform the random averages by means of @xmath92 the essential point is then the calculation of the averaged replicated partition function . assuming , as usual , that a finite number of patterns @xmath93 is condensed , with finite overlaps with the state of the network , we perform the random noise average and sum over the uncondensed patterns to obtain @xmath94\ } \right\rangle_{\{\xi^{\mu}_i\}}\ , . \label{21}\end{aligned}\ ] ] where , the sum over @xmath95 counts the configurations for all possible replicas @xmath4 . the dilution appears only in the first factor and @xmath96 is the average over the uncondensed patterns . for both , the calculation of the function @xmath97 and the linearization of the quadratic form in the dilution term , we introduce the replica matrix elements @xmath98 of the sg order parameter and @xmath99 which is the dynamical activity @xmath100 of the network . the latter is one only in the case of binary units and , in general , @xmath101 . in the thermodynamic limit @xmath102 , we obtain @xcite @xmath103 where @xmath104 is the matrix of elements @xmath105 and @xmath106 . introducing as usual the overlap parameter @xmath107 associated to the correlation between the overlaps of the patterns that do not condense , we restrict ourselves to replica symmetry , in which @xmath108 the free - energy per connected site , in the thermodynamic @xmath102 limit , then follows as @xmath109\nonumber\\ & + & \alpha(1-c)(\frac{\chi^2}{4\beta}+\frac{q\chi}{2 } ) -\frac{1}{\beta}\left\langle \int{\cal d}z\,\ln\sum_{\{s\}}\exp(\beta{\cal h}_{eff})\right\rangle_{\{\xi^{\nu}\}}\ ; , \label{27}\end{aligned}\ ] ] where @xmath110 is a gaussian measure , and @xmath111 , given by @xmath112 , is the susceptibility of the network . the new , site independent effective hamiltonian @xmath113 , is given by @xmath114 in terms of which the thermal averages are defined as @xmath115 note , incidentally , that the explicit term in @xmath116 in the free energy comes from the variance of the gaussian random noise due to the dilution . the order parameters that describe the performance of the network are given by the saddle - point equations @xmath117 @xmath118 and the susceptibility becomes @xmath119 in which @xmath120 , for a given gaussian noise @xmath121 is given by eq.(32 ) . noting that the effective hamiltonian @xmath113 is formally the same as either that of the fully connected or the extremely diluted network , with a different stochastic gaussian noise and different effective threshold @xmath122 , for each connectivity @xmath3 , our equations for @xmath123 , @xmath124 and @xmath125 will be formally similar to the equations for both these networks , and explicit expressions are given in the appendix . the parameter @xmath126 follows from the algebraic saddle - point equation @xmath127 and @xmath128 is the effective width of the intermediate states . eventually , depending on the state of the network specified by the dynamical activity @xmath100 and the sg order parameter @xmath124 , @xmath129 may become negative , favoring an order with large absolute values for @xmath130 . whenever this is the case the network acts , at zero temperature , as a binary network with zero threshold , as will be seen below . the combination @xmath131 , in which the second term is the gaussian noise due to the macroscopic number of uncondensed patterns , can be viewed as an effective random field that will influence the network performance through a competition with the effective threshold @xmath129 . having performed the thermodynamic limit , one may now allow _ any _ value for the connectivity within @xmath132 , including the limit @xmath6 . it can easily be seen that one recovers for @xmath133 both the mean - field free energy for the fully connected network when @xmath134 @xcite and that for the extremely dilute case , when @xmath79 @xcite . in this limit , @xmath135 and @xmath136 , whereas for @xmath134 we have , @xmath137 and @xmath138 , also in agreement with the known results . these relations are valid for any temperature parameter @xmath139 . thus , we have the complete form for the replica - symmetric free energy for arbitrary connectivity . note , incidentally , that @xmath135 not only in the above limit but also for @xmath140 , in the @xmath141 limit , for any connectivity @xmath3 , ( that is , for any architecture ) and for all @xmath16 , and this is based on @xmath142 in that limit . otherwise , the susceptibility remains finite , even at zero temperature , since @xmath143 , when @xmath144 , while at finite temperature we have , in general , @xmath145 . the limit of stability of the replica - symmetric solution comes from the study of quadratic fluctuations of the free - energy in the vicinity of the symmetric saddle - point . following the de almeida and thouless ( at ) analysis @xcite , we obtain @xmath146 ^ 2 \right\rangle_{\{\xi^{\nu}\ } } \leq 1\ , , \label{35}\ ] ] as the stability condition for the replica - symmetric solution . this equation is to be solved together with the saddle - point equations for the order parameters . the formal results obtained so far are valid for any finite number of condensed patterns with finite overlaps with the state of the network . we are mainly interested in this work in the retrieval performance with a single condensed pattern , and this will be discussed in the following section . for the retrieval of a single condensed pattern , say @xmath147 , we have @xmath148 , and omit the index @xmath149 from now on . we consider separately the results for @xmath0 , @xmath1 and @xmath2 that follow by solving the saddle - point equations and restrict ourselves to the simplest case of uniformly distributed patterns . in the @xmath0 case , the patterns take the values @xmath150 with probability @xmath151 and @xmath152 with probability @xmath153 , in which @xmath154 in the case of uniformly distributed patterns . the effective transfer function @xmath155 that follows from the average in eq.(32 ) is given by @xmath156 which becomes @xmath157 in the zero temperature limit . clearly , when @xmath158 , the network acts as a binary network at @xmath159 . in the cases of the fully connected @xcite or the extreme symmetrically diluted network @xcite , explicit closed form expressions that signal the appearance of either a retrieval or a spin - glass phase , at zero temperature , have been obtained in the @xmath160 limit when @xmath161 . these are particularly useful to understand the low - threshold behavior of the phase diagrams . in the present case , of a network with finite and less than complete connectivity , closed form expressions for the onset of the ordered phases can not be obtained and one has to resort fully to numerical solutions . nevertheless , the zero - temperature behavior of the thermodynamic transition can easily be analyzed as in previous works , to demonstrate that the retrieval state corresponds to the most stable phase , despite the presence of a spin - glass and a paramagnetic phase ( which is a frozen zero - spin state ) . indeed , the physical free energy , @xmath162 at zero temperature becomes @xmath163 since the susceptibility vanishes in the @xmath164 limit when @xmath159 for any of the three phases , and at the same time @xmath165 , the retrieval free energy is the minimum whenever @xmath166 . note that the susceptibility of the paramagnetic phase , @xmath167 also vanishes in the zero temperature limit for finite @xmath4 and @xmath122 converging to @xmath44 , ensuring a minimum retrieval free energy for small @xmath4 . although much emphasis is often made on the thermodynamic transition to globally stable retrieval states , which have the lowest free energy , it is worth keeping in mind that neural networks are dynamical systems with accessible _ locally _ stable retrieval states , in particular in the presence of some amount of noise . as far as the performance of the network is concerned , these are the most interesting states and they usually appear for higher values of @xmath4 @xcite . we are interested here in the characteristic features of the phase diagrams and the specific performance of the network . to see the effects of a gradual change in the connectivity , we show in fig . 1 the @xmath168 phase diagrams for @xmath159 . the full lines represent the phase boundaries where the locally stable retrieval states appear at the critical storage ratio @xmath169 , whereas the long - dashed lines indicate the thermodynamic transitions to the globally stable retrieval phase . the sg phase appears to the left of the short - dashed lines representing the boundary to the paramagnetic state . to distinguish in what follows the transitions involving locally stable states from the thermodynamic transitions , we refer to the former simply as retrieval transitions . as long as the connectivity remains finite , all the transitions are discontinuous and , as usual in connected networks , the sg state is globally stable only above the thermodynamic transition @xcite . consider first the case of the fully connected network , with @xmath134 , which has been redone and completed here for the purpose of comparison . there are two retrieval regions , i and ii for small @xmath4 , separated by a sharp phase boundary , and this is the case both below and above the thermodynamic transition . the first is a region of non - optimal performance characterized by a moderately large hamming distance which decreases with increasing @xmath44 , whereas in region ii the hamming distance is small , dropping discontinuously at the phase boundary between the two regions with optimal network performance along the dotted line . the situation should change with decreasing connectivity , even at zero temperature . due to the synaptic noise produced by the dilution , given by the variance @xmath88 , one now expects an end to the discontinuous transition between the retrieval regions i and ii at a @xmath3-dependent critical point for finite @xmath4 , below @xmath169 . this starts to appear for @xmath170 , with @xmath171 and @xmath172 , as shown in the inset of the figure for the retrieval transition . a critical point also appears at the thermodynamic transition and in fig . 1 we show the results for @xmath173 and @xmath174 . there can be now a continuous transition with increasing @xmath44 induced by stochastic noise between the non - optimal and optimal performance retrieval states above the critical point suggesting the presence of increasingly larger regions of continuous changes . the interesting point is that this is a feature that already starts to appear for finite and _ intermediate _ connectivity , between that of the fully connected network and the @xmath6 limit . to understand the role of the threshold one may use as a guide the case of vanishing stochastic noise , with @xmath164 . when @xmath44 is small , the state of a unit will be essentially @xmath175 , with an overlap @xmath176 in region i for uniformly distributed three - state patterns @xmath177 , that take the values @xmath178 , @xmath152 and @xmath179 . since @xmath180 , the hamming distance in this region will be @xmath181 . on the other hand , as @xmath44 becomes larger , the state @xmath182 becomes increasingly important and , despite the fact that @xmath183 also yields an overlap @xmath184 , the sg order parameter @xmath124 is now reduced to the activity @xmath185 . thus , the resulting hamming distance @xmath186 will be vanishingly small . note that these results do not depend on the connectivity and they are , therefore , independent of the architecture of the network . this also follows from the zero - temperature saddle - point equations as can easily be checked with the appendix . in the case of a finite non - zero stochastic noise , instead , the performance of the network becomes explicitly dependent on the connectivity , but the overall qualitative dependence on @xmath44 below the critical phase boundaries is expected to follow that at @xmath140 . the critical storage capacity now increases with decreasing connectivity and the presence of two comparable maxima for @xmath169 is only a feature of intermediate @xmath3 . indeed , as the synapses are further diluted a single maximum is left , albeit with a shift to higher values of @xmath44 . finally , it is also worth noting that , in the @xmath6 limit , the retrieval state is the globally stable phase everywhere below the critical @xmath169 line and to the left of the globally stable paramagnetic phase , despite the relatively large stochastic noise due to the presence of spin - glass states in most of this region . the zero - temperature results presented so far are not stable to replica - symmetry breaking perturbations but it is expected that most of the features described here will be present at already a small but finite temperature above the at line , shown as dash - dotted in fig . 2 , for @xmath187 . the full lines again represent the transitions to the locally stable retrieval states and the thermodynamic transitions are not shown . note that , even for small connectivity , there is only a low synaptic - noise region in which the network is not stable to replica - symmetry breaking perturbations . for a larger @xmath188 we expect a similar behavior for the transition to the locally stable retrieval state with decreasing connectivity as that found before for the fully connected network , but still quite different from the behavior for lower @xmath44 @xcite . in the case of @xmath1 the phase diagrams are more involved . the patterns are assumed to take the values @xmath150 with probability @xmath189 and @xmath190 with probability @xmath191 , in which @xmath192 , where @xmath193 for uniformly distributed patterns . we consider in the following the zero - temperature behavior of the network and begin with the fully connected case as a guide . we recover precisely the retrieval phase boundaries found before @xcite . similar results with exclusively sharp phase boundaries , between now enlarged ordered regions , are found for somewhat lower connectivity , as shown in fig . 3(a ) for @xmath173 . for low @xmath4 , we find three different ordered retrieval ferromagnetic phases @xmath194 , @xmath195 and @xmath196 , in a previous notation @xcite and characterized below , separated by discontinuous phase boundaries and in which @xmath197 for uniformly distributed patterns . as @xmath4 increases , these phases disappear discontinuously at the critical phase boundaries @xmath169 into the @xmath198 phase . the three phases correspond to possible locally stable states which become globally stable at the thermodynamic transition for lower @xmath4 , not shown in the figure for simplicity . which of the locally stable states is actually reached in the dynamic evolution of the network will depend , as usual , on their basins of attraction and the choice of the initial state . one may consider four regimes in the @xmath199 limit . first , when @xmath44 is small , that is within the phase @xmath194 , mainly the high activity states are favored with @xmath200 with an overemphasized overlap @xmath201 ; the spin - glass order parameter and the hamming distance become @xmath180 and @xmath202 , respectively . for larger and intermediate @xmath44 , there should be a phase , called @xmath195 , characterized by states of the network that follow essentially the patterns , with @xmath203 and an overlap given by @xmath204 , the spin - glass order parameter @xmath205 and a vanishing hamming distance . between the small and the intermediate @xmath44 regimes there could be a coexistence region of the phase @xmath194 with the phase @xmath195 where the states of the network start to recognize the full structure of the patterns . finally , there should be a phase characteristic of the large @xmath44 regime in which mainly the intermediate states with @xmath206 are activated leading to a performance with @xmath207 , @xmath208 and again @xmath202 . this is the phase @xmath196 , which should have an overlap with the phase @xmath195 at intermediate @xmath44 . these expectations have been confirmed by means of the solutions to the saddle - point equations in the @xmath199 limit and some of the results can be found in earlier work @xcite . the four regimes are given by @xmath209 , @xmath210 , @xmath211 and @xmath212 . they do not depend on the connectivity and are therefore independent of the architecture of the network , in accordance with earlier results either on the fully connected or the extreme symmetrically dilute network @xcite . on the other hand , we confirm the symmetry of the limiting @xmath169 for @xmath213 and for @xmath214 , in accordance with earlier results @xcite . we also find that the optimal performance line appears within the @xmath195 phase and that the network has a relatively high performance with a small hamming distance in that phase , with an overlap at the critical phase boundary that is @xmath215 of that at @xmath140 . when the connectivity is reduced to @xmath216 , the distinction between the phases @xmath194 and @xmath195 starts to disappear , as shown by the enlarged gap in fig . 3(b ) , allowing for a continuous change into the high performance phase for intermediate @xmath4 . note that there is still a discontinuous phase boundary between the phases @xmath194 and @xmath196 and that the presence of this phase boundary is important in order to inhibit the transition to the low performance phase @xmath196 . furthermore , we still find four regimes for low @xmath4 and that the three main retrieval phases , @xmath194 , @xmath195 and @xmath196 , end discontinuously at the critical phase boundary @xmath169 . the optimal performance line is still purely within the @xmath195 phase , as in the previous case , and the network has a high performance up to @xmath169 , with an overlap close to one on the phase boundary , for @xmath44 around @xmath217 . as the connectivity is further decreased to @xmath218 we find the phase diagram shown in fig . 3(c ) with the three main retrieval phases that disappear discontinuously at @xmath169 and the four low-@xmath4 regimes discussed above . the distinction between the phases @xmath194 and @xmath195 disappears now at lower @xmath4 and the optimal network performance in the central phase @xmath195 can be reached continuously within a considerable range of @xmath4 from the @xmath194 phase . as in the previous cases , there is a coexistence region between the phases @xmath194 and @xmath195 , now only for small @xmath4 . moreover , there is no need now for a specific choice of threshold parameter @xmath44 in order to access most of the high performance domain of the network . incidentally , note that the continuous retrieval phase boundary for the common phase @xmath194 and @xmath195 is similar to that found for the fully connected network with pattern activity @xmath219 @xcite . we have no further insight , at present , of this feature . finally , in order to check the overall simplification of the phase diagrams that appears with decreasing connectivity , we also present results for @xmath220 that are shown in fig . the phase boundaries are still lines of discontinuous transitions and the distinction between the four regimes is restricted to even lower values of @xmath4 . there is now a considerably larger region of continuous changeover from the phase @xmath194 to the phase @xmath195 , with access to optimal performance , without the need of a fine adjustment in @xmath44 . in the case of @xmath2 we again consider uniformly distributed patterns between @xmath178 and @xmath149 , implying that @xmath221 , and restrict the results to the zero temperature case . the discontinuous transitions to the ordered phase are shown in full lines in the @xmath222 phase diagram for decreasing connectivity in fig . 4 , where we omit again the thermodynamic transitions and the long - dashed lines indicate now the onset of the binary - network behavior . note that the disappearance of the ordered phase takes place at @xmath223 for any finite connectivity , as in the case of the fully connected network and in networks of different architecture , like the extremely asymmetric diluted and the @xmath16-ising layered network @xcite . in the case of the extremely diluted networks , the retrieval phase boundaries have a reentrance for @xmath224 @xcite . this seems to be a feature of the @xmath6 limit , as one can see from our further results for the connectivity dependence of the maximum storage capacity @xmath225 and the corresponding @xmath226 , both shown in fig . 5 . as in the case of both the fully connected and the symmetrical extremely diluted network , and in contrast with the @xmath0 state network , we find that even at zero temperature most of the retrieval regions for different @xmath3 are stable to replica - symmetry - breaking perturbations , that is for @xmath44 above the at lines . this includes the maximum storage capacity and it follows from a positive replicon eigenvalue for this case obtained from eq.(38 ) , @xmath227 where @xmath125 is the susceptibility for the continuous network presented in the appendix . the at line is given by @xmath228 and , again , in both the @xmath134 and the @xmath6 limit , in which @xmath135 , this result coincides with that for the fully connected and the symmetrical extremely diluted network @xcite . we derived in this work the replica symmetric mean - field theory for @xmath16-ising attractor networks with low - activity patterns and arbitrary symmetric dilution of the synaptic connections . we extended earlier studies on the retrieval behavior and thermodynamic properties of either fully connected or symmetrical extremely diluted @xmath16-ising neural networks with low activity patterns , in order to study the effects of a gradual dilution of the synaptic connections guided by the motivation that neurons in biological networks of associative memory are neither fully connected nor very sparsely linked to other neurons . we are mainly interested in the nature of the phase transitions to locally stable retrieval states and in the role that synaptic dilution has in either reducing or destroying sharp transitions motivated by the plasticity of biological networks . in the context of the networks studied here , we focus mainly on the dependence of the retrieval properties on the threshold or gain parameter @xmath44 for decreasing connectivity @xmath3 and ask to which extent can a network go from a given locally stable retrieval state to a nearby high - performance state without crossing phase boundaries of discontinuous transitions . in order to answer that question , one has to look for appropriate phase diagrams that were obtained here in replica - symmetric mean - field theory . since deviations from that theory are very small and appear only in a small region near @xmath159 , we may still draw relevant conclusions from those diagrams . the explicit phase diagrams obtained in this work apply to uniformly distributed patterns and to networks with arbitrary symmetric dilution . the behavior of both the fully connected and the extremely diluted network are recovered when @xmath134 and @xmath6 , respectively , in that our general saddle - point equations become identical to those for either case that have been obtained before @xcite . we find that common features of the @xmath199 limiting behavior in either fully connected or symmetrical extremely diluted networks also appear for arbitrary finite connectivity @xmath3 . this confirms the expectations of earlier works that pointed out the architecture independent nature of some properties @xcite . among these is the particular @xmath188 where the thermodynamic transition ends in the @xmath0 state network and where the optimal performance takes place for low @xmath4 , both for @xmath159 . the common limiting @xmath5 in the @xmath229 phase diagram for varying @xmath3 is a further property of this kind , as well as the four distinct domains in the @xmath1 state network and the @xmath188 limiting threshold for the @xmath2 network at @xmath140 . the main dependence of the behavior of the network on the connectivity arises for finite @xmath4 . for both odd and even @xmath16 , we find that a common feature that appears with an increase in the dilution of the synaptic connections is to suppress selected sharp phase boundaries of discontinuous transitions that make the optimal performance domain readily accessible to a wide region of low - threshold locally stable retrieval states . note that , on the other hand , the sharp boundary for the @xmath1 state network between the low performance phases @xmath194 and @xmath196 survives synaptic dilution , at least to quite an extent . these features of the network for small but finite connectivity appear long before the extremely diluted limit and they should be of considerable interest . concentrating , for simplicity , on the @xmath0 state network we also found that the boundaries between thermodynamic transitions are suppressed by an increase of the synaptic dilution , and expect a similar behavior for the @xmath1 state network . the results of our work may be used to infer the behavior of other networks . since the fully connected network is strongly sensitive to pattern activity , one may consider other than uniformly distributed patterns @xcite . there are , essentially , two kinds of phase diagrams in that case about which we can make definite predictions . one is the type of phase diagram for patterns of relatively large activity that has mostly a decreasing phase boundary with increasing @xmath44 and an optimal performance line that appears only at small @xmath44 . the other type , that appears for medium or small pattern activity , has an optimal performance only at intermediate @xmath44 , like the cases shown in fig . 3 , and both types appear for @xmath0 and @xmath1 , while only the first type seems to appear for @xmath2 . we expect that the main effect of a finite synaptic dilution on the first type of phase diagram is simply to shift the retrieval phase boundaries upwards towards a larger @xmath169 . in the second type of phase diagrams , however , we expect also a disappearance of the discontinuous phase boundary between the @xmath194 and @xmath195 phases , in essentially the same way we found in the present work , allowing for a smooth changeover from states of non - optimal to those of optimal performance . the main result of this work , that partially connected multi - state hebbian networks can attain near - optimal performance without a fine tuning of neuron activity may be a simplified statistical mechanics explanation of why biological memory networks seem to prefer low - activity patterns between partially connected neurons . of course , biological networks have asymmetric synaptic connections which may lead through a dynamic evolution to different stationary states , the search of which is certainly an interesting issue that deserves a separate investigation , currently in progress . the study of the effects of symmetric synaptic dilution may be extended to other problems that deal with associative memory , like the categorization problem as a classification task in @xmath16-ising networks @xcite . this has been done recently for @xmath230 @xcite and there is work in progress for general @xmath16 @xcite . * acknowledgments * we thank d. boll and g. m. shim for clarifying a point on a previous version of the manuscript . this work was financially supported in part by cnpq ( conselho nacional de desenvolvimento cientfico e tecnolgico ) brazil . we present here , for completeness , the explicit expressions for the saddle - point equations obtained for the symmetrically diluted network with arbitrary connectivity @xmath3 and uniformly distributed patterns , for @xmath0 , @xmath1 and @xmath2 . for @xmath0 taking patterns @xmath150 with probability @xmath151 and @xmath152 with probability @xmath153 we have @xmath231 @xmath232\ , , \label{a.2}\ ] ] @xmath233\ , , \label{a.3}\ ] ] with @xmath234 given by eq.(39 ) . these equations reduce to the equations for the extreme symmetrically diluted network , as well as for the fully connected network , in the @xmath6 limit and @xmath134 , respectively . the same applies for the cases of @xmath1 and @xmath2 , presented below . in the zero - temperature limit ( @xmath144 ) , the integrations over the gaussian variable @xmath121 can be done explicitly . in the @xmath0 case we obtain , for @xmath235 , @xmath236\ , , \label{a.4}\ ] ] @xmath237\ , , \label{a.5}\ ] ] @xmath238\ , , \label{a.6}\ ] ] with the relation between @xmath126 , @xmath124 and @xmath125 given by eq.(36 ) . for the @xmath1 state model with states and uniformly distributed patterns that take the values @xmath178 , @xmath239 , @xmath240 and @xmath179 , @xmath193 and we obtain @xmath241\ , , \label{a.7}\ ] ] @xmath242\ , , \label{a.8}\ ] ] @xmath243\ , . \label{a.9}\ ] ] where @xmath244 in the zero - temperature limit @xmath245\ , , \label{a.11}\ ] ] and the above equations yield , for positive @xmath122 , @xmath246\nonumber\\ & + & \frac{1}{10}\left[{\rm erf}(\frac{m+4\tilde\theta } { 3\sqrt{2\alpha rc } } ) + { \rm erf}(\frac{m-4\tilde\theta}{3\sqrt{2\alpha rc } } ) + { \rm erf}(\frac{m}{3\sqrt{2\alpha rc}})\right]\ , , \label{a.12}\end{aligned}\ ] ] @xmath247\ , , \label{a.13}\ ] ] @xmath248\ , . \label{a.14}\end{aligned}\ ] ] finally , in the zero - temperature limit for @xmath2 and uniformly distributed patterns between @xmath178 and @xmath149 , implying @xmath221 , we obtain for @xmath235 , @xmath249\ , , \label{a.11}\ ] ] @xmath250\ , , \label{a.12}\end{aligned}\ ] ] @xmath251 where @xmath252 and this integrations can be performed directly . watkin and d. sherrington , europhys . lett . * 14 * , 791 ( 1991 ) . b. derrida , e. gardner and a. zippelius , europhys . lett . * 4 * , 167 ( 1987 ) . d. sherrington and s. kirkpatrick , phys . lett . * 35 * , 1792 ( 1975 ) . hopfield , proc . usa * 79 * , 2554 ( 1982 ) . j. hertz , a. krogh and r. palmer , _ introduction to the theory of neural computation _ ( reading , ma . addison - wesley , 1991 ) . a. canning and j .- p . naef , j. phys . i * 2 * , 1791 ( 1992 ) . amit , h. gutfreund and h. sompolinsky , ann . ( ny ) * 173 * , 30 ( 1987 ) . yedidia , j. phys . a * 22 * , 2265 ( 1989 ) . h. rieger , j. phys . a * 23 * , l1273 ( 1990 ) . c. meunier , d. hansel and a. verga , j. stat . phys . * 55 * , 859 ( 1989 ) . d. boll , p. dupont and b. vinck , j. phys . a. * 25 * , 2859 ( 1992 ) . d. boll , b. vinck and v.a . zagrebnov , j. stat . phys . * 70 * , 1099 , ( 1993 ) . d. boll , g.m . shim , b. vinck and v.a . zagrebnov , j. stat . phys . * 74 * , 565 ( 1994 ) . d. boll , g.m . shim and b. vinck , j. stat . phys . * 74 * , 583 ( 1994 ) . d. boll , h. rieger and g.m . shim , j. phys . a * 27 * , 3411 ( 1994 ) rolls , a. treves , d. foster and c. perez - vicente , neural networks * 10 * , 1559 ( 1995 ) . arbib , editor ( m.i.t . press , 1995 ) . d. boll , g. jongen and g.m . shim , j. stat . phys . * 96 * , 861 ( 1999 ) . d. boll , d. m. carlucci and g.m . shim , j. phys . a * 33 * , 6481 ( 2000 ) . l. viana and a.j . bray , j. phys . c * 18 * , 3037 ( 1985 ) . h. sompolinsky , phys . rev . * 34 * , 2571 ( 1986 ) . m.a . p. idiart and a. theumann , j. phys . a * 25 * , 779 ( 1992 ) . j.r . almeida and d. thouless , a * 11 * , 983 ( 1978 ) . r. erichsen jr . , w.k . theumann and d.r.c . dominguez , phys rev . e * 60 * , 7321 ( 1999 ) . p.r . krebs and w.k . theumann , phys . e * 60 * , 4580 ( 1999 ) . r. erichsen jr . and w.k . theumann , unpublished . + * figure captions * + fig . phase diagram @xmath168 for the @xmath0 state network with uniformly distributed patterns and connectivity @xmath3 as shown , at @xmath159 . the full and the long - dashed lines represent the retrieval and the thermodynamic transition , respectively . the latter ends on the right at the ( dotted ) optimal performance line and the spin - glass phase appears at the left of the paramagnetic phase boundary indicated by the short - dashed lines . the two retrieval regions i and ii are discussed in the text and the inset corresponds to @xmath253 where the distinction betweeen these phases starts to disappear . stable phase diagram to the right of the de almeida - thouless ( dot - dashed ) line in the @xmath229 plane for the @xmath0 state network with uniformly distributed patterns , @xmath187 and connectivity @xmath3 as shown . the full lines represent the retrieval transition . phase diagram @xmath168 for the @xmath1 state network with uniformly distributed patterns and connectivity ( a)@xmath173 , ( b)@xmath254 , ( c)@xmath218 and ( d)@xmath220 , at @xmath159 . the full lines represent retrieval transitions and the optimal performance is indicated in dotted lines . the central region is the best performance phase @xmath195 and there is a low-@xmath4 coexistence region between phases @xmath194 and @xmath195 . phase diagram @xmath168 for the @xmath2 state network with uniformly distributed patterns and connectivity @xmath3 as shown , at @xmath159 . the full lines represent retrieval transitions , the optimal performance is indicated in dotted lines and the onset of the binary network is shown by long - dashed lines . the phases are stable to the right of the de almeida - thouless line ( dot - dashed ) . fig . 5 : connectivity dependence for the maximum storage capacity @xmath225 and the corresponding @xmath226 in the @xmath2 state network with uniformly distributed patterns at @xmath159 .

stability and interactions in biological and soft - matter systems often depends on the underlying properties of electrostatic interactions @xcite . charges on macromolecular surfaces in aqueous environments , as in the case of membranes , self - assembled micelles , globular proteins and fibrous polysaccharides , affect a wealth of functional , structural and dynamical properties @xcite . the traditional approach to charged ( bio)colloidal systems has been the mean - field poisson - boltzmann ( pb ) formalism applicable at weak surface charges , low counter - ion valency and high temperature @xcite . the limitations of this approach become practically important in highly - charged systems where counterion - mediated interactions between charged bodies start to deviate substantially from the mean - field accepted wisdom @xcite . one of the fundamental recent advances in this field has been the systematization of these non - pb effects based on the notions of _ weak _ and _ strong _ coupling approximations . the latter approach has been pioneered by rouzina and bloomfield @xcite , elaborated later by shklovskii _ et al . _ @xcite , levin _ et al . _ @xcite , and brought into final form by netz _ these two approximations allow for an explicit and exact treatment of charged systems at two disjoint _ limiting conditions _ whereas the parameter space in between can be analyzed only approximately @xcite and is mostly accessible solely _ via _ computer simulations @xcite . in the absence of a general approach that would cover thoroughly all the regions of the parameter space one has to take recourse to various partial formulations that take into account only this or that facet of the problem . in this respect the counterion - only or the one - component coulomb fluid model system has proved to be of substantial value @xcite . heuristically as well as numerically . a proper understanding of the behavior of charged systems would thus start with the analysis of counter - ion distribution around charged macromolecular surfaces , neglecting completely the effects of salt . both the weak and the strong coupling approximations are based on a functional integral or field - theoretic representation @xcite of the grand canonical partition function of a system composed of fixed surface charges with intervening mobile counterions , and depend on the value of a single dimensionless coupling parameter @xmath0 @xcite . the distance at which two unit charges interact with thermal energy @xmath1 is known as the bjerrum length @xmath2 ( in water at room temperature , one has @xmath3 nm ) . if the charge valency of the counterions is @xmath4 then the aforementioned distance scales as @xmath5 . similarly , the distance at which a counterion interacts with a macromolecular surface ( of surface charge density @xmath6 ) with an energy equal to @xmath1 is called the gouy - chapman length , defined as @xmath7 . a competition between ion - ion and ion - surface interactions can be quantitatively measured with a ratio of these characteristic lengths , that is @xmath8 , which is known as the ( netz - moreira ) electrostatic coupling parameter @xcite . the weak coupling ( wc ) regime @xmath9 ( appropriate for low valency counterions and/or weakly charged surfaces ) , is characterized by the fact that the width of the counterion layer @xmath10 is much larger than the separation between two neighboring counterions in solution and thus the counterion layer behaves basically as a three - dimensional gas . each counterion in this case interacts with many others and the collective mean - field approach of the poisson - boltzmann ( pb ) type is completely justified . on the other hand in the strong coupling ( sc ) regime @xmath11 ( appropriate for high valency counterions and/or highly charged surfaces ) , the mean distance between counterions , @xmath12 , is much larger than the layer width ( _ i.e. _ , @xmath13 ) , indicating that the counterions are highly localized laterally and form a strongly correlated quasi - two - dimensional layer next to a charged surface . in this case , the weak - coupling approach breaks down due to strong counterion - surface and counterion - counterion correlations . since counterions can move almost independently from the others along the direction perpendicular to the surface , the collective many - body effects that enable a mean - field description are absent , necessitating a complementary sc description @xcite . the range of validity of both limiting theories at intermediate values of the coupling parameter has been explored thoroughly in the literature @xcite . formally the weak coupling limit can be straightforwardly identified with the saddle - point approximation of the field theoretic representation of the grand canonical partition function , and is reduced to the mean - field pb theory in the lowest order for @xmath14 . the quadratic fluctuations around the mean field provide a second - order correction to the mean - field solution for small finite @xmath15 @xcite . the strong coupling approximation has no pb - like correlates @xcite since it is formally equivalent to a single particle description obtained from a systematic @xmath16 expansion in the limit @xmath17 , and corresponds to two lowest order terms in the virial expansion of the grand canonical partition function . the consequences and the formalism of these two limits of the coulomb fluid description have been explored widely and in detail ( for reviews , see refs . @xcite ) . considering the inhomogeneity of charged surfaces in various biological contexts it has always been of interest to investigate not just electrostatic interactions between symmetrical charged surfaces , _ i.e. _ those bearing equal charges of the same sign , but also interactions between surfaces bearing unequal charges or even charges of opposite sign @xcite . this problem has a venerable history starting from the seminal work of parsegian and gingell @xcite who formulated a linearized pb theory of the interactions in the presence of salt . the linearization _ ansatz _ was later generalized in the work of lau and pincus @xcite and ben - yaakov _ et al . _ @xcite who formulated the appropriate non - linear mean - field theory of non - symmetric electrostatic interactions . it is thus our goal in this contribution to show how and to what extent the asymmetry in the distribution of charges on two apposed planar surfaces affects the interactions between macromolecular surfaces carrying them . below we shall present a complete analysis of the asymmetric case in the weak coupling limit , _ i.e. _ the mean - field poisson - boltzmann theory supplemented with a quadratic - fluctuations analysis , as well as in the strong coupling limit _ via _ the asymptotic strong - coupling theory and evaluate how these analytical results compare with extensive numerical simulations . we will show that in their respective regimes of validity ( _ i.e. _ small / large couplings ) both approximations present a very accurate quantitative statistical description of the system . in our model system we consider uniform surface charge distributions on two plane - parallel surfaces ( located at @xmath18 ) given by the surface charge density of the form @xmath19 we may interchangeably use the half - separation @xmath20 , or @xmath21 to identify the surface - surface distance . we assume furthermore that the charge of both bounding surfaces is compensated by mobile counterions of charge valency @xmath4 immersed in an aqueous medium of dielectric constant @xmath22 and distributed in between the two surfaces ( see fig . [ fig : geometry ] ) . we thus neglect all coions . this approximation is relevant for low salt concentrations where the debye screening length is much larger than the scales of interest @xcite . we consider the surfaces as impenetrable to counterions and neglect the dielectric discontinuity across the bounding surfaces which was addressed at various levels of approximation in @xcite . without loss of generality we can assume here that @xmath23 and @xmath24 it will be helpful for our later developments to introduce an _ asymmetry parameter _ @xmath25 that will allow us to quantify the dissimilarity between the two bounding surfaces as @xmath26 furthermore , by suitably normalizing the results one can concentrate exclusively on the interval @xmath27 . all other cases can be mapped onto this interval with appropriate rescaling of the parameters . the values @xmath28 and @xmath29 represent exceptional points in the parameter space : @xmath28 is the standard symmetric case ( @xmath30 ) already amply treated in the literature , and @xmath29 represents the antisymmetric case ( @xmath31 ) with no counterions between surfaces that reduces to the trivial case of a planar capacitor . these two well - understood limiting cases will be thus omitted from our discussion . counterions between the surfaces satisfy electroneutrality condition that can be written in the form @xmath32 where @xmath33 denotes the ( infinite ) area of each surface . because of the asymmetry present in the system , we have two length scales describing the interaction of the counterions with each of the bounding surfaces . these two length scales are given by the corresponding gouy - chapman lengths associated with the two surfaces as @xmath34 for the same reason we can thus define two different coupling parameters @xmath35 each one being defined by the ratio between the bjerrum length and the corresponding gouy - chapman length . in what follows , we rescale the surface separation as @xmath36 ( or the rescaled half - distance as @xmath37 ) with respect to plate 1 . with an appropriate rescaling one could also equivalently define all dimensionless lengths with respect to plate 2 . thus the minimal set of dimensionless parameters that fully characterize the system in the thermodynamic limit is given by @xmath38 . other physical quantities such as the mean electrostatic potential @xmath39 , number density of counterions , @xmath40 , and the pressure , @xmath41 , acting on each surface can be rescaled as well . we shall use the standard rescaled electrostatic potential @xmath42as well as the rescaled density and pressure @xmath43 where @xmath44 , and all the other quantities have been defined above . in the weak coupling regime , the leading contribution to the partition function comes from the saddle - point configuration of the local fluctuating electrostatic potential , @xmath45 @xcite . the saddle - point configuration can be straightforwardly translated into a solution of the pb equation and corresponds to an exact asymptotic result in the limit @xmath14 @xcite . for a system containing only counterions , the pb equation for the dimensionless potential , @xmath46 , can be written in the standard form @xcite @xmath47 with boundary conditions @xmath48 integration of the pb equation gives rise to the first integral of the system of the form @xmath49 where the constant @xmath50 is nothing but the mean - field pb pressure acting between the bounding surfaces @xcite and @xmath51 is the pb number density profile of counterions between the surfaces . the normalization factor @xmath52 follows from the electroneutrality condition ( [ eq : electroneutral ] ) as @xmath53 the nature of the solution @xmath54 obviously crucially depends on the sign of the pressure @xmath50 @xcite . different forms are obtained for positive and negative pressures , corresponding to repulsion and attraction between the bounding surfaces respectively . we review these different cases separately . in the case of repulsive pressure the appropriate solution of eq . ( [ eq : pbeq ] ) can be written as @xmath56 where the constants @xmath57 and @xmath58 are obtained from the boundary conditions ( [ bc - rep ] ) and satisfy the set of two equations @xmath59 eliminating @xmath57 we obtain an equation for @xmath60 of the form @xmath61 the solution of this equation provides the final result for the repulsive pb pressure @xmath50 . note in particular that in rescaled units and by definition one has @xmath62 where @xmath63 . once @xmath60 is known , the parameter @xmath57 can be simply obtained from eqs . ( [ z01 ] ) or ( [ z02 ] ) and thus the potential @xmath64 , eq . ( [ psi0 ] ) , is fully determined . the density profile of counterions then follows from @xmath65 a positive ( repulsive ) solution for the pressure as considered in this section is always possible for any given asymmetry parameter @xmath25 ( excluding the trivial case of @xmath66 ) . in particular , it easily follows that within the mean - field theory two surfaces of equal sign ( @xmath67 ) always repel , that is at all separation distances @xmath68 . when the surfaces bear charges of opposite sign @xmath69 , they attract at large separations ( see below ) and a repulsion emerges only at sufficiently small separations . at small separations @xmath70 , we can obtain the limiting solution for @xmath60 and thus the limiting small - distance pressure @xmath50 as @xmath71 for arbitrary @xmath72 as noted in scetion [ sec : geo ] . this is of course nothing but the ideal - gas osmotic pressure of counterion confined between the two plates ( _ i.e. _ , @xmath73 in actual units ) , which dominates over the energetic contributions at small separations . at large separations @xmath74 and for @xmath67 , we obtain the asymptotic expansion @xmath75 , \label{rep - limit - zeta}\ ] ] which is valid for @xmath76 . thus in the limit @xmath77 , the pressure behaves as @xmath78 which agrees with the asymptotic pressure between two equally charged surfaces ( @xmath79 ) . for smaller @xmath25 than determined above , _ i.e. _ for @xmath80 , one needs to invoke a different asymptotic expansion and specifically for @xmath81 ( one surface being neutral ) , one obtains @xmath82 this latter asymptotic result may be obtained from the one in eq . ( [ rep - limit ] ) by redefining @xmath83 . this may be understood simply by noting that because of symmetry a system with @xmath67 may be decomposed into two halves each with an effective asymmetry parameter @xmath84 . an attractive pressure on the mean - field level is possible only if the surfaces are oppositely charged @xmath69 . the appropriate solution in this case is given by @xmath86 where the constants @xmath57 and @xmath87 can again be obtained from boundary conditions , this time in the form @xmath88 eliminating @xmath57 we obtain an equation for @xmath60 as @xmath89 in this case we have in rescaled units @xmath90 and for the density profile of counterions @xmath91 the asymptotic form of the attractive pressure at large separations @xmath92 can be derived as @xmath93 where @xmath69 as noted above . for infinite separations , this pressure does _ not _ vanish and exponentially approaches @xmath94 since for @xmath95 the system behaves partially as a simple capacitor . in the case of charged surfaces with opposite sign ( @xmath69 ) , the large - distance attraction regime and the short - distance repulsion regime merge at the point of zero pressure , @xmath97 , where the surfaces are at equilibrium . in this case , the pb solution for the potential reads @xmath98 and the density profile of counterions is given by @xmath99 where @xmath57 is found from the boundary conditions as @xmath100 , and the _ bound - state separation _ , @xmath101 , follows in rescaled units as @xmath102 the surfaces attract for @xmath103 and repel for @xmath104 . in the vicinity of @xmath105 , that is for @xmath106 , the pressure behaves as @xmath107 this concludes the calculation of the inter - surface pressure on the mean - field pb level strictly valid for @xmath14 . the preceding results may be summarized in a phase diagram shown in fig . [ fig : bound - sc ] in terms of @xmath101 and the asymmetry parameter @xmath25 displaying the mean - field attraction and repulsion regimes separated by the boundary line ( [ eq : pb_a * ] ) . the forms of the pressure here are completely consistent with those derived by lau and pincus @xcite _ via _ a different route . the first non - zero correction to the saddle point is second order in the fluctuations of the local electrostatic potential around the mean - field pb solution , @xmath108 . our goal here is to calculate the corrections in pressure , @xmath109 , stemming from these quadratic fluctuations , which leads then to the total wc pressure @xmath110 this approach has correlates in many diverse areas of physics where fluctuations around a mean - field solution are important @xcite and goes under different names , though the physics is always the same . we may conventionally refer to the mean - field pb term , @xmath111 , and the fluctuations contribution , @xmath109 , as the _ zeroth - order _ and the _ second - order _ correction terms on the wc level , respectively . this procedure formally also corresponds to a series expansion in powers of @xmath0 ( loop expansion ) around the asymptotic mean - field solution ( @xmath14 ) @xcite and is thus expected to be valid for sufficiently small coupling parameters as will be determined later . note also that in this latter sense the second - order pressure turns out to be proportional to @xmath0 , that is @xmath112 , and thus corresponds to a first - loop correction @xcite . in order to proceed , one needs to evaluate the appropriate hessian of the field action in the partition function and study its fluctuation spectrum ( see refs . @xcite for more details ) . the hessian of the field action can be derived in the form @xmath113 where @xmath114 is the inverse coulomb operator and @xmath115 is the zeroth - order pb density as derived in the previous section . hence , @xmath116 the corresponding correction , @xmath117 , to the free energy of the system is then given by the trace - log of the hessian . it can be written equivalently in the following form @xcite @xmath118 this form can be derived rather straightforwardly by using the argument principle @xcite and converting the discrete sum of eigenvalues of the hessian operator into an integral over the transverse wave - vector @xmath119 , with density of modes @xmath120 of the logarithm of the secular determinant @xmath121 of the same operator . the index @xmath122 in the secular determinant refers to the eigenvalue equation that can be derived in the form @xmath123 by simply writing @xmath124 for the quotient @xmath125 , and noting that the secular determinant depends explicitly also on the value of the inter - surface spacing , @xmath20 , the free energy contribution from the quadratic fluctuations can be equivalently expressed exactly in a dimensionless form as @xmath126 where @xmath127 , @xmath128 and the rescaled area @xmath129 . here the secular determinant of the hessian for homogeneous transverse modes has been written as a function of dimensionless quantities @xmath130 , @xmath128 . this determinant has to be standardly regularized so that all irrelevant constants , _ i.e. _ all the terms not depending on the separation between the bounding surfaces , are dropped , amounting to a rescaling @xmath131 this corresponds to a subtraction of the part of the free energy for two separate interfaces at infinite separation from the total free energy . in the next step one has to calculate the secular determinant @xmath132 for each of the pressure regimes separately , since the appropriate eigenfunctions of the hessian depend on the mean - field solution that in its turn depends on the sign of the interaction pressure , see eq . ( [ sign - eigen ] ) . in what follows we shall follow closely the derivations in refs . @xcite . the total pressure in the weak - coupling limit is thus the sum of the pb pressure and the quadratic fluctuations correction and can be written as @xmath133 in this regime the secular determinant of the hessian , eq . ( [ eigf ] ) , can be obtained by solving @xmath134 with appropriate boundary conditions implying continuity of the solution and its derivative across the bounding surfaces at @xmath18 . the general solution of eq . ( [ sec - rep ] ) for various regions in the perpendicular direction can be written in the form @xmath135 where @xmath136,\\ & & y_{2}=\cfrac{q^2}{q^2+\alpha^2}\>e^{-qz}\left[1- \cfrac{\alpha}{q}\,\tg\,\alpha ( z - z_0)\right].\end{aligned}\ ] ] taking into account the continuity of the solution and its derivatives we get a set of four homogeneous equations for the coefficients @xmath137 and @xmath138 . the solution exists only if the ( secular ) determinant of this system equals zero . thus we derive the secular determinant of the hessian operator in this case in the rescaled form @xmath139 where @xmath140 the regularized form of the secular determinant is obtained by taking the quotient as indicated in eq . ( [ renorm ] ) . while doing this , it is important to realize that @xmath141 also depends on the inter - surface distance . in fact from eq . ( [ rep - limit ] ) it follows that the appropriate limit of @xmath141 is @xmath142 in the regularization of the secular determinant this limiting behavior should be consistently taken into account . finally , the dimensionless quadratic fluctuations free energy @xmath143 can be calculated numerically _ via _ eqs . ( [ second - f ] ) and ( [ eq : seculard_final ] ) . the fluctuations contribution to the pressure , @xmath144 , then follows from eq . ( [ secondord ] ) . the asymptotic form of the second - order dimensionless pressure can be obtained analytically . note that at large separations @xmath145 , a repulsive mean - field pressure @xmath55 , as considered in this section , is possible only for non - negative @xmath25 . for not too small @xmath146 , _ i.e. _ when @xmath147 , we find @xmath148 while for @xmath81 , _ i.e. _ when @xmath80 , we get @xmath149 again the difference in the two cases above is due to the symmetry of the problem in the latter case , that can be described by redefining @xmath83 and discarding the sub - dominant terms . the second - order pressure is obviously _ attractive _ and in this regime leads to a reduction of the total pressure from the mean - field value @xmath50 . this clearly shows that electrostatic correlations favor attraction between two repelling asymmetrically charged plates . however , the total pressure never becomes negative as the fluctuations are assumed to be small within the second - order weak - coupling analysis . in this case the secular determinant of the hessian , eq . ( [ eigf ] ) , is obtained by solving @xmath150 the general solutions for particular regions in the @xmath151 direction are @xmath152 where @xmath153,\\ & & y_{2}=\cfrac{q^2}{q^2-\alpha^2}\,e^{-qz}\left[1+\cfrac{\alpha}{q}\,\cth\,\alpha ( z - z_0)\right].\end{aligned}\ ] ] again the solution exists only if the determinant of the system of equations , which connect coefficients @xmath154 and stems from the application of the boundary conditions at @xmath155 , is identically zero . this again defines the secular determinant @xmath124 appropriate for this case . it is easy to show that the secular determinant @xmath124 can be obtained from the @xmath55 result , eq . ( [ eq : seculard_final ] ) , simply by substituting @xmath156 and so we can write @xmath157 where @xmath158 here we can again regularize the secular determinant to discard divergences , eq . ( [ renorm ] ) . again one has to be careful by taking the correct limit for @xmath141 in the above regularization scheme . in this regime , the appropriate limit is given by @xmath159 as follows straightforwardly from eq . ( [ att - limit ] ) . the fluctuations contribution to the pressure , @xmath144 , can then be evaluated numerically from eq . ( [ secondord ] ) . the asymptotic form of @xmath144 for @xmath74 can be derived analytically as @xmath160 which is applicable only for charged surfaces of opposite sign , @xmath161 , which can attract ( @xmath85 ) at large separations . the function @xmath162 is defined as @xmath163 for @xmath164 , and @xmath165 for @xmath166 . the second - order pressure thus asymptotically decays exponentially and can be only attractive . for not too small values of @xmath25 , it is thus qualitatively very different from the case @xmath167 . the total weak - coupling pressure , @xmath168 , is shown in fig . [ fig : wc_p ] for a few different asymmetry parameters and a relatively small value of the coupling parameter @xmath0 . as noted above the foregoing weak - coupling analysis is valid as long as the quadratic corrections are sufficiently small so that the series expansion around the mean - field solution does not diverge @xcite . as an approximate measure for the validity regime of this scheme , one can require that the second - order correction term is smaller than the leading order term , _ i.e. _ @xmath169 this leads to a useful criterion identifying the regime of coupling parameters and distances in which the weak - coupling theory is applicable . for @xmath55 and by employing the closed - form expressions obtained for large separations @xmath92 , we find the validity condition @xmath170 this indicates that at a given non - vanishing @xmath0 , the weak - coupling scheme becomes increasingly more accurate at larger separations , while as the surfaces get closer a smaller coupling parameter needs to be chosen . on the other hand , for @xmath85 ( which occurs for @xmath161 ) and at large separations @xmath92 , we obtain @xmath171 the right hand side here is exponentially large meaning that for charged surfaces of opposite sign , the weak - coupling analysis performs far better at finite coupling parameters and smaller inter - surface separations than for the surfaces of equal sign ( @xmath67 ) . finally , note that for @xmath96 that corresponds to the equilibrium phase boundary line in fig . [ fig : bound - sc ] for @xmath161 , we deal with a situation where the leading order term is zero and the fluctuations are _ dominant _ at any finite value of @xmath0 . the convergence of the loop expansion has to be determined in this case by evaluating the higher order terms which we shall not consider in this paper . the strong - coupling approximation coincides with the lowest order non - trivial expansion of the partition function in terms of the fugacities of the counterions . this expansion may be expressed as a @xmath16 series expansion @xcite , whose leading order term ( @xmath17 ) corresponds to the so - called sc theory . we will not delve into the strong - coupling expansion in more detail since it has been exhaustively reviewed in the literature @xcite . on the leading order , the free energy is obtained as @xmath172 where @xmath173 is electrostatic interaction energy of charged surfaces @xmath174 with @xmath33 representing surface area of each plate , and @xmath175 and @xmath176 are electrostatic interaction energies between a single counterion and individual charged surfaces , _ i.e. _ @xmath177 since in the strong - coupling regime the free energy is given _ via _ simple quadratures , it is much simpler to evaluate it than on the weak - coupling level . defining the rescaled free energy @xmath178 we obtain @xmath179.\ ] ] differentiating the free energy with respect to the surface - surface distance @xmath180 we get the corresponding pressure acting between the bounding surfaces @xmath181 . \label{eq : p_sc}\ ] ] the dependence of this dimensionless pressure on the separation for different values of @xmath25 is presented in fig . [ fig : sc_p ] . note that the sc pressure can become attractive for _ both _ like - charged and oppositely charged surfaces which contrasts with the mean - field theory that does not allow attraction between like - charged surfaces . this is because of the strong electrostatic correlations mediated by counterions between the charged surfaces for @xmath182 and has been investigated throughly before for equally charged surfaces @xcite . our results show that a similar attraction mechanism holds for asymmetrically charged surfaces in the sc limit . the pressure exhibits two well - defined limiting laws obtainable for small and large inter - surface separations . for small separations @xmath70 , we have @xmath183 for arbitrary @xmath72 as noted in section [ sec : geo ] . for large separations @xmath74 , we obtain @xmath184 note that in the limit @xmath185 the sc pressure coincides with the pb result on the leading order ( compare eqs . ( [ rep - limit1 ] ) and ( [ psc_limit ] ) ) and represents the ideal - gas osmotic pressure of counterions which dominates over the electrostatic contributions . the pb and sc forms for @xmath161 coincide also in the limit of @xmath74 ( compare eqs . ( [ att - limit ] ) and ( [ newlimit ] ) ) and reduce to the pressure in the capacitor . the dependence of the pressure on the inter - surface separation points to the existence of a bound state defined _ via _ @xmath186 . the sc bound - state separation @xmath101 can be expressed analytically as @xmath187 and is presented in fig . [ fig : bound - sc ] as a function of @xmath25 . obviously both like - charged and oppositely charged surfaces can form bound states at small surface - surface separations . the bound - state separation approaches infinity and the surfaces _ unbind _ asymptotically as @xmath188 , that is when one plate becomes electroneutral . finally one can also derive the explicit form of the counterion density as a function of the normal coordinate @xmath151 . this can be read off simply from the integrand in eq . ( [ deff ] ) , that is @xmath189 , where @xmath190 is a normalization factor @xcite . according to the electroneutrality condition , we normalize the density as @xmath191 , or in rescaled units @xmath192 where we have defined @xmath193 with @xmath194 being the gouy - chapman length with respect to plate 1 . from here , the density profile is obtained as @xmath195 } , \label{denssc}\ ] ] as a function of @xmath151 and @xmath25 . the regime of applicability of the leading order sc theory follows from a simple criterion that has been discussed and confirmed previously in the case of equally charged surfaces by both mc simulations and higher - order calculations @xcite . the generalization to asymmetrically charged surfaces is straightforward . for large couplings , counterions are strongly attracted to an oppositely charged surface as the counterion - surface interaction becomes large and equivalently , the gouy - chapman length , @xmath10 , and the thickness of the counterionic layer at the surface become small . the layer thickness has to be compared with the typical lateral spacing between counterions @xmath196 . for counterions sandwiched between two asymmetrically charged surfaces , this latter quantity follows from the local electroneutrality condition as @xmath197 up to a factor of the order unity and assuming that the surfaces are sufficiently close so that they may be strongly coupled _ via _ the counterions as will be determined consistently here . in rescaled units , one gets @xmath198 where @xmath199 . obviously , @xmath200 becomes large relative to the layer thickness as @xmath0 grows . note that on the other hand the lateral coulomb repulsion between counterions in this quasi - two - dimensional layer becomes much larger than the thermal energy , _ i.e. _ @xmath201 , indicating that counterions form a strongly correlated liquid in which they are highly localized within correlation holes of lateral size @xmath202 @xcite . thus for surface - surface separations , @xmath68 , smaller than the correlation hole size , _ i.e. _ @xmath203 counterions can move almost independently from each other in the direction normal to the surface and one can safely assume that the effective surface - surface interaction as well as the counterionic density profile follow only from the interactions of _ individual _ counterions with the bounding charged surfaces . the counterion - counterion interactions contribute on the sub - leading order and matter at larger separations . this picture is of course confirmed on a systematic level by the sc expansion analysis @xcite , and the above equation sets a criterion for the validity regime of the single - particle leading order sc theory ( @xmath204 ) , when applied to finite coupling parameters . we performed monte - carlo ( mc ) simulations in order to study the system of two asymmetrically charged surfaces beyond the analytical limits of weak and strong coupling discussed above . all simulations were performed in the canonical ensemble ( nvt ) using the standard metropolis algorithm @xcite . the mobile counterions were modeled as point charges @xcite enclosed in a simulation box bounded in @xmath151 direction by two charged surfaces of distance @xmath138 and surface charge densities @xmath205 and @xmath206 ( compare fig . [ fig : geometry ] ) . periodic boundary conditions were applied in the lateral directions parallel to the bounding surfaces . the lateral size of the charged surfaces , @xmath207 , which is set equal to the lateral size of the simulation box , was held fixed throughout all the simulations . the number of counterions , @xmath208 , was varied between 60 and 1800 , depending on the system parameters in order to fulfill the electroneutrality condition . the counterions interact through the coulombic potential @xmath209 with @xmath210 being the charge of the @xmath211-th counterion and @xmath212 the separation distance between the @xmath211-th and the @xmath213-th counterions ( with @xmath214 ) . the interaction energy of the @xmath211-th counterion with the charged surfaces in the simulation box are given by @xmath215 ^ 2}\bigg ) + \frac{\pi}{2}\bigg\}\bigg ] , \label{coulomb_wallterm}\end{aligned}\ ] ] where @xmath216 and @xmath206 is surface charge density of the @xmath217-th surface ( with @xmath218 ) and @xmath219 is the normal distance between the @xmath211-th counterion and the @xmath217-th surface . the long - ranged coulomb interactions in this system were accounted for _ via _ a charged sheet scheme similar to that proposed by torrie and valleau @xcite . this scheme makes use of the counterion profile in the simulation box in order to calculate an external field , stemming from the long - ranged interactions . this external field is iteratively updated and self - consistency is achieved normally in a few iterations . in the course of simulations , new configurations were created by trial displacements of the counterions and equilibration was accomplished by running through @xmath220 configurations . the @xmath221 following configurations were then used for the production runs . the pressure @xmath41 was calculated in the production runs according to the contact - value theorem as @xmath222 where @xmath223 is the density of counterions at contact with the @xmath217-th surface . in thermodynamic equilibrium , the pressure does not depend on which surface ( @xmath224 or @xmath225 ) is chosen in order to calculate the pressure from the above equation , and the contact condition at both surfaces leads to precisely the same value for the pressure . all simulations were conducted at fixed temperature @xmath226 k , lateral simulation box size @xmath227 and dielectric constant @xmath228 , which is assumed to be the same throughout the system . the simulations were performed at different values of the coupling parameter @xmath0 and the asymmetry parameter @xmath25 . we explored the @xmath0 parameter space extensively by using @xmath229 , @xmath230 and @xmath231 in order to cover exhaustively both the weak coupling and the strong coupling regimes . the concurrent values of the asymmetry parameter were always taken as @xmath232 at each value of the coupling parameter . the results are plotted in the form of rescaled density and pressure as previously defined in this paper ( section [ sec : rescal ] ) . in order to asses the validity of the weak and strong coupling results presented above for asymmetrically charged surfaces , we performed extensive mc simulations and compared them to analytical results in both limits . it transpires from this comparison that the simulation results corresponding to an exact evaluation of the partition function are always bracketed by the wc and the sc limiting forms , smoothly approaching them in the appropriate limits of the coupling parameter @xmath0 . first we compare the density profiles of simulations with theoretical results given by eqs . ( [ denswc1 ] ) and ( [ denswc2 ] ) for the pb limit ( solid lines ) and eq . ( [ denssc ] ) for the sc limit ( dashed lines ) . as seen from fig . [ dens ] the theoretical pb and sc rescaled density profiles represent two extremal cases and all mc simulations results with finite values of @xmath0 are located consistently between these two limits . the mc results for small @xmath0 are almost exactly spot on the pb prediction , while larger discrepancies are observed as @xmath0 grows . for large enough @xmath233 , the mc results slowly converge to the sc result . this is especially clear for surfaces with charges of equal sign , @xmath234 , whereas for surfaces with opposite sign , @xmath235 , there is no big difference between pb and sc profiles . next we consider the inter - surface pressure as obtained from the simulations ( symbols in fig . [ fig : sim_pressure ] ) as well as the pb theory , eqs . ( [ pres - pb - rep ] ) and ( [ pres - pb - att ] ) , and the sc theory , eq . ( [ eq : p_sc ] ) ( solid and dashed lines , respectively ) . the pb result is expected to be valid for separations @xmath236 @xcite ( see also section [ sec : wc_crit ] ) . therefore , for @xmath237 the pb line expectedly agrees nicely with the simulation data ( open squares ) in the whole range of separations shown in the figure . upon closer inspection , however , we find small deviations from the pb result as shown in the insets in fig . [ fig : sim_pressure ] for all three values of the asymmetry parameter @xmath25 . in this case , the fluctuation correction to the mean - field pressure accurately compensates for these deviations and adding the second - order correction to the pb pressure leads to a total wc pressure , eq . ( [ secondord ] ) ( shown as a dotted line in the inset ) that matches the simulation data perfectly . note that the pressure changes can be drastic even on the wc level as @xmath25 assumes different values . for @xmath238 and 0 the wc pressure is strictly repulsive , while it turns attractive and leads to a bound state ( zero pressure point ) for @xmath239 . in the intermediate regime of coupling parameters , the simulation results for the pressure are clearly bracketed by the two limiting analytical forms , given by the pb plus the second order correction and the sc expressions of the interaction pressure . the sc prediction is expected to be valid for separations @xmath240 as discussed in section [ sec : sc_crit ] . consistently , the interaction pressure starts off close to the strong coupling limit at small separations and then smoothly converges to the weak coupling limit for larger separations . this is strictly true for @xmath238 and 0 . in the case of @xmath241 the difference between the strong and weak coupling results for the rescaled interaction pressure is marginal and the simulation data and the analytical results nearly coincide for all rescaled separations @xmath242 . we emphasize that the pressures and the density profiles are plotted here in _ rescaled representation _ ; in actual units , fig . [ fig : sim_pressure ] corresponds to different ranges of separation , @xmath138 , for the wc and sc regimes as the gouy - chapman length , @xmath10 , is typically very different between the two limits ( small at high couplings and large at small couplings as may be realized , _ e.g. _ , by changing the counterion valency at fixed surface charge densities and bjerrum length ) . for large values of the coupling parameter the simulation results for the interaction pressure expectedly follow very closely the strong coupling prediction for a wider range of inter - surface separations . the correspondence between the sc theory and simulations is better for @xmath241 than for @xmath238 , which can be again traced back to the fact that the strong and the weak coupling results are very close to one another for charged surfaces of opposite sign in the whole range of rescaled separations , whereas they differ significantly in the case of surfaces of equal sign . for intermediate and large couplings , we have not attempted to compare our data with the second - order wc approximation as this approximation breaks down at the range of distances shown in the figures ( section [ sec : wc_crit ] ) . note also that in all cases considered here the theoretical and the simulated values of the interaction pressure converge for very small surface - surface separations . in fact , the sc and pb results coincide in the leading order as the rescaled distance , @xmath68 , tends to zero , eqs . ( [ rep - limit1 ] ) and ( [ psc_limit ] ) , since both are dominated by the osmotic pressure of counterions . the sub - leading corrections for very small @xmath68 are different in the pb and sc limits and on this level the simulation data with finite @xmath0 are generally expected to agree better with the sc prediction at small separations @xcite . we now consider the simulated bound - state separation , @xmath101 , as a function of the coupling parameter in fig . [ fig : bound_xi ] ( symbols ) . for @xmath234 , the pb theory ( @xmath14 ) gives only repulsion and predicts no bound state . while the sc theory predicts a closely packed bound state with @xmath243 given by eq . ( [ eq : d*_sc ] ) , explicitly here @xmath244 . as seen , by increasing the coupling parameter the simulation data for @xmath101 ( filled squares ) decrease monotonically and rather slowly converge to the sc prediction ( dotted line ) . the comparison is again worse for the @xmath238 than for the @xmath241 case ( dashed line and filled circles ) . for @xmath234 , even at @xmath245 the difference between the simulations and the analytical result is still close to 10% . the deviations are quite pronounced for smaller values of the coupling parameter . the opposite is true for @xmath241 . here , the simulation results are close to the strong - coupling analytical limit in the whole range of @xmath0 values . for @xmath246 we can discern , see inset in fig . [ fig : bound_xi ] , weak coupling behavior that starts off with the pb - predicted value @xmath247 in the limit of @xmath248 ( eq . ( [ eq : pb_a * ] ) , solid line ) , that later follows the pb plus second - order corrections line ( dot - dashed line ) and then rapidly approaches the strong - coupling result @xmath249 ( dashed line ) . as already noted the differences between strong and weak coupling in this case are marginal in the rescaled representation . the dependence of @xmath105 on @xmath25 in fig . [ fig : bound_zeta ] complements the above observations . the bound - state separation diverges for @xmath250 both in simulations as well as in the analytical limits . here we reproduce the simulation results at @xmath234 and @xmath251 , which again clearly show the convergence to the sc result for @xmath234 and the crossover from the pb result ( solid line ) to the sc result ( dashed line ) for @xmath235 upon increasing the coupling parameter . the quantitative agreement between the simulations and the analytical results in the two limiting cases of pb and sc is excellent . note here again that the second - order wc expansion ( dotted line in the inset ) begins to fail at small separations as it deviates from the limiting pb results as well as from the simulation data . since the pb pressure is zero on the pb phase boundary line , the validity of the second order wc correction can be assessed analytically only by performing a two - loop calculation which goes beyond the scope of this paper . to summarize , we have derived theoretical forms for the interaction pressure as well as the counterionic density profiles of asymmetrically charged planar surfaces with neutralizing counterions in between . based on the field - theoretical methods we analyzed two different regimes of weak and strong coupling as defined by the electrostatic coupling parameter @xmath0 . the crossover between these two regimes is studied _ via _ monte - carlo simulations . for small values of @xmath0 , the system is described very well by the _ weak coupling _ ( wc ) theory , that in the lowest ( zeroth ) order coincides with the mean - field poisson - boltzmann ( pb ) result . the second order of wc corresponds to a first - order loop expansion and represents the contribution from correlated quadratic fluctuations around the mean - field or saddle - point solution . this second - order correction , which is proportional to @xmath0 , always lowers the interaction pressure between the surfaces and thus leads to an attractive contribution to the total interaction pressure . since it corresponds to an expansion of the partition function around the mean - field saddle point it has to be smaller , in absolute terms , than the mean - field result . net attraction given by second - order fluctuations term ( for interacting surfaces of equal sign ) is therefore inconsistent with the nature of the wc approximation . for large values of the coupling parameter @xmath0 , the weak coupling approach breaks down and the virial expansion amounting to the _ strong coupling _ ( sc ) approximation must be used . the sc approach is effectively a one - particle theory and takes properly into account the strong correlation and interaction of the counterions with external surface fields on the leading order @xcite . following standard procedures , we derived an analytical expression for interaction pressure in the sc limit . the interaction pressure in this case is always lower than the pb result and can be negative ( corresponding to a net attractive force ) even for charged surfaces of equal sign . we compared both our theories , _ i.e. _ wc with second - order corrections and sc , with monte - carlo simulations . we found very good agreement for both theories in their expected regime of validity . as expected , the wc approach is valid for separations @xmath252 . the second - order correction improves the small discrepancies between pb and mc results at large separations but it tends to fail for smaller distances when discrepancies get more pronounced . on the other hand , the sc theory describes the behavior perfectly at small separations @xmath253 . for small enough coupling parameter @xmath0 , the validity of the pb approximation spreads to smaller separations , @xmath68 , where pb and sc results nearly coincide ( in the rescaled representation ) . therefore , we may conclude that for sufficiently small @xmath0 , the pb result is valid on the whole interval @xmath68 . note that the second - order wc correction term consistently diverges ( toward large negative values ) for small inter - surface separations , @xmath254 , irrespective of @xmath25 , which makes it in general inapplicable in this limit . the reason for this is simple . for small inter - surface separations the mean - field solution becomes more and more homogeneous , almost a constant , and the interaction free energy approaches its standard zero - frequency van der waals form that diverges for small separations . in the case of surface charges with equal sign , @xmath67 , the wc theory predicts no attraction and hence no bound state . the attraction and the corresponding bound state appear only for coupling parameters @xmath0 that are large enough , as predicted by the sc theory . in the case of charged surfaces of opposite sign , @xmath161 , the attraction appears also in wc limit above a threshold value @xmath255 that represents the equilibrium surface - surface separation . it is notable that for charged surfaces of opposite sign , the wc analysis in general performs much better than for the surfaces of equal sign and that the sc and the wc results are very close to one another for charged surfaces of opposite sign in the whole range of rescaled separations , whereas they differ significantly in the case of surfaces of equal sign . there is also only a marginal difference between the pb and sc counterion density profiles in the rescaled representation for surfaces of opposite sign . a reasonable explanation for this would be in our opinion that for oppositely charged surfaces the counterions mostly feel the effect of the strong uniform external field provided by the surface charges , which acts similarly in the strong as well as the weak coupling limit . thus the mean - field and the strong - coupling approaches should converge . in the case of similarly charged surfaces , the mean - field theory depends more on the local counterion density whereas the strongly coupled counterions still feel mostly the external field . thus the difference between the wc and the sc frameworks in the @xmath161 and @xmath67 cases . our results support an emerging new paradigm , according to which the wc and the sc limit bracket the exact results for the interaction pressure between charged surfaces neutralized by mobile counterions . they indeed provide quantitatively correct results for the interaction pressure in the limit of small and large inter - surface separations , while at intermeditae separations the exact results are always located between the two limits . it thus seems advisable that in analyzing the electrostatic interactions in colloidal systems one always calculates both analytic limits , the wc as well as the sc , in order to get a good handle on the range of values that the interaction can assume for any value of the electrostatic coupling parameter . in future we intend to study the same system in the presence of added salt and dielectric discontinuities . m.k . and r.p . would like to acknowledge the financial support by the agency for research and development of slovenia ( grants p1 - 0055(c ) , z1 - 7171 , l2 - 7080 ) . this study was supported by the intramural research program of the nih , national institute of child health and human development . this research was supported in part by the national science foundation under grant no . phy05 - 51164 . in our model excluded - volume counterion - counterion repulsions are neglected . these two - body interactions will be irrelevant in the leading order within the strong - coupling limit @xcite . however , the counterion size ( or the counterion - wall exluded - volume interactions ) is incorporated exactly by taking @xmath256 as the closest - approach surface - surface distance leaving out the counterion diameter from the actual distance . on a more fundamental level it is indeed possible to incorporate excluded- volume effects within a modifed poisson - boltzmann formulation as in _ e.g. _ : i. borukhov , d. andelman , h. orland , phys . lett . * 79 * 435 ( 1997 ) ; electrochimica acta * 46 * 221 ( 2000 ) . r. r. netz , eur . j. e * 3 * 131 ( 2000 ) ; r. menes , p. pincus , b. stein , phys . e * 62 * 2981 ( 2000 ) ; j. schmit , r. menes , p. a. pincus , phys e * 66 * 061502 ( 2002 ) ; m. m. hatlo , r. a. curtis , l. lue , j. chem . phys . * 128 * 164717 ( 2008 ) .

recent photometric surveys such as the acs survey of galactic globular clusters @xcite have substantially increased the quality and volume of data available for the galactic globular cluster system . despite the fact that this survey focused on metal poor stellar populations it was found that there was a need for stellar evolution models with @xmath0 } } > $ ] 0 @xcite . many other photometric surveys such as the two micron all sky survey ( 2mass ) and the sloan digital sky survey ( sdss ) and extensions such as segue underscore the need for stellar evolution libraries that span a wide range of compositions and photometric systems . @xcite introduced a set of stellar evolution tracks , isochrones , and computer programs to accompany the acs survey of galactic globular clusters . the set includes tracks with masses between 0.1 and 1.8 @xmath2 , isochrones with ages ranging from 2 to 15 gyr , [ fe / h ] from 2.5 to 0 with a variety of options in @xmath1}}$ ] and initial he content ( y@xmath5 ) . in this paper , that set of models is expanded to include : ( i ) tracks and isochrones with [ fe / h]= + 0.15 , + 0.3 , and + 0.5 ; ( ii ) higher mass tracks and isochrones with ages as low as 250 myr for all compositions ; ( iii ) a wider selection of photometric systems ; and ( iv ) a web interface to three of the computer programs so that visitors can easily generate specific models without downloading the entire database and running the programs locally . collectively , these tracks , isochrones , and supporting computer programs constitute the dartmouth stellar evolution database . the stellar evolution code used to generate these models is reviewed in @xmath62 . details of the compositions and model grids are outlined in @xmath63 and the color@xmath7 transformations in @xmath64 . in @xmath65 , the isochrones are compared to those of comparable isochrone libraries and photometry of galactic open clusters and one globular cluster . the results are summarized and discussed in @xmath66 . the dartmouth stellar evolution program ( dsep ) has been described previously by @xcite and @xcite . the following is a brief description of the physics employed by dsep . partially inhibited atomic diffusion and gravitational settling of both he ( y ) and metals ( z ) are included as described by @xcite . diffusion is calculated based on the formalism of @xcite but inhibited in the outer 0.01 @xmath2 where the diffusion coefficients are set to zero in the outermost 0.005 @xmath2 and ramped from zero to full diffusion in the inner 0.05 @xmath2 . this method produces results that are consistent with recent observations of diffusion in ngc 6397 @xcite . dsep uses high temperature opacities from opal @xcite and low temperature opacities based on @xcite . the opal tables were obtained for each of the different @xmath1}}$ ] mixtures from the opal web interface and the low temperature tables were calculated by one of us ( ferguson ) . the low temperature opacity tables have been made available via the world wide web . conductive opacities are taken from @xcite . the basic equation of state ( eos ) in dsep is a general ideal gas eos with the debye - hckel correction @xcite and is used for tracks with m @xmath8 . for lower mass tracks ( m@xmath9 ) , the freeeos equation of state @xcite is used in the eos4 configuration . ccccccccccc 2.5 & 1.6 & 1.6 & 1.6 & 1.5 & 1.3 & 1.3 & 1.5 & 1.5 & 1.5 & 1.5 + 2.0 & 1.4 & 1.4 & 1.4 & 1.4 & 1.3 & 1.3 & 1.5 & 1.5 & 1.5 & 1.5 + 1.5 & 1.4 & 1.4 & 1.4 & 1.4 & 1.3 & 1.3 & 1.4 & 1.4 & 1.4 & 1.4 + 1.0 & 1.4 & 1.4 & 1.4 & 1.4 & 1.2 & 1.2 & 1.4 & 1.4 & 1.4 & 1.4 + 0.5 & 1.2 & 1.2 & 1.2 & 1.2 & 1.2 & 1.0 & 1.2 & 1.2 & 1.2 & 1.2 + 0.0 & 1.2 & 1.2 & 1.1 & 1.1 & 1.0 & 0.9 & 1.2 & 1.1 & 1.2 & 1.1 + + 0.15 & & 1.1 & & & & & & & & + + 0.3 & 1.1 & 1.1 & 1.1 & & & & & & & + + 0.5 & 1.1 & 1.1 & 1.0 & & & & & & & + dsep treats convection with the standard mixing length theory and a solar - calibrated mixing length , @xmath10=1.938 . the models include convective core overshoot according to the scheme developed by @xcite . the underlying idea is that the amount of overshoot grows with the size of the convective core . the amount of core overshoot is parametrized as a multiple of the pressure scale height ( @xmath11 ) and is a function of stellar mass and composition . the minimum stellar mass receives 0.05 @xmath11 of overshooting , models with masses greater than the minimum by 0.1 @xmath2 receive 0.1 @xmath11 of overshooting , and models with masses greater than or equal to the minimum plus 0.2 @xmath2 receive 0.2 @xmath11 of overshooting . for details on the minimum mass as a function of composition see table [ conv ] . convective envelope undershooting is not considered in these models . the surface boundary condition ( bc ) was derived from a grid of ` phoenix ` model atmospheres @xcite covering @xmath7=2,000 to 10,000 k and log g=0.5 to 5.5 . the solar abundance scale used in these calculations was that of ( * ? ? ? * hereafter gs98 ) . ` phoenix ` model atmosphere grids were computed for the full range @xmath0}}$]=2.5 to + 0.5 and @xmath1}}$]=0.2 to + 0.8 covered by the stellar evolution calculations . a small number of model atmospheres with enhanced he were also computed to gauge the importance of enhanced he to the physical structure and the spectrum . the gas pressure ( and thus the surface bc ) experiences a small change with enhanced he and this is accounted for in the surface bc . @xcite atmospheres were used for @xmath12 10,000 k ; @xcite models were computed only for @xmath1}}$]=0 and + 0.4 . stellar evolution models with @xmath1}}\leq$ ] + 0.2 made use of surface bc derived from @xmath1}}$ ] = 0 model atmospheres for @xmath12 10,000 k. for stellar evolution models with @xmath1}}\geq$ ] + 0.4 the @xmath1}}$ ] = + 0.4 boundary conditions were used for @xmath12 10,000 k. the transition between the ` phoenix ` and @xcite were always smooth regardless of the composition . the smooth transition is due to the stellar evolution models being relatively insensitive to the surface bc at higher @xmath7 coupled with the reduced sensitivity of the model atmospheres to abundance variations in the transitional @xmath7 . the surface bc was fit at the point in the atmospheric structure where t=@xmath7 for a given model . the gas pressure was extracted at t=@xmath7 and added to the radiation pressure to satisfy the surface bc . @xcite found that , at least in the vicinity of 1 @xmath2 , the choice of model atmosphere boundary condition fitting point ( either at t=@xmath7 or at an optical depth of @xmath13=100 ) has little influence on the resulting evolution . the ` phoenix ` model atmosphere code employs its own set of physics that is not identical to the physics used in dsep , though there is some overlap . in particular , the low temperature opacities of @xcite are calculated with ` phoenix ` and this implies a level of consistency between the two codes . ` phoenix ` employs a standard mixing length theory of convection with mixing length @xmath14 = 2.0 ( whereas dsep uses @xmath14 = 1.938 ) . perhaps the largest difference between ` phoenix ` and dsep in terms of stellar structure is the eos : ` phoenix ` uses an ideal eos that includes the contributions of a large number of elements and molecules while freeeos ( used by dsep ) includes several non - ideal effects but accounts for fewer elements and only molecules involving combinations of h and he . the nuclear reaction rates are taken from @xcite for most reactions . the exceptions are the @xmath15n(p,@xmath16)@xmath17o reaction @xcite , the triple-@xmath18 reaction from the nacre compilation @xcite , and the @xmath19c(@xmath18,@xmath16)@xmath20o reaction @xcite . neutrino cooling rates were taken from @xcite . stellar evolution models were computed using one of six different heavy element mixtures with the distinguishing characteristic of each mixture being the level of @xmath18-element enhancement . opacities were generated for each of @xmath1}}$]= 0.2 , 0 , 0.2 , 0.4 , 0.6 , and 0.8 . the heavy element mass fractions of these mixtures are given in table [ mix ] . the mixtures were calculated starting with the solar mixture of gs98 by adding the value of @xmath1}}$ ] to the log(n+12 ) abundance of each of the following @xmath18-elements : o , ne , mg , si , s , ca , and ti ; the abundance of ar was not altered from its gs98 value in any of the mixtures . overall , models were computed for 67 different x , z pairs ; the initial compositions are listed in table [ init ] . for @xmath0}}>$ ] 0 the maximum level of @xmath18-enhancement was chosen to be @xmath1}}$]=+0.2 in order to keep the total z value below 0.1 since the opacity tables do not extend above z=0.1 . the model atmosphere grid included only provided scaled - solar ( @xmath1}}$]=0 ) models at @xmath0}}$]=+0.15 . the he content of the models begins with y=0.245 at @xmath0}}$ ] = 2.5 , consistent with @xcite , and a he enrichment relation of @xmath21y/@xmath21z=1.54 . for @xmath0}}\leq$]0 , additional models were computed with y=0.33 and 0.40 ( see table [ init ] ) . stellar evolution tracks were computed for stellar masses ranging from 0.1 to 4 @xmath2 in increments of 0.05 @xmath2 ( between 0.1 and 1.8 @xmath2 ) , 0.1 @xmath2 ( between 1.8 and 3 @xmath2 ) , and 0.2 @xmath2 ( between 3 and 4 @xmath2 ) . the tracks cover the evolution of the models from the fully convective pre - main sequence ( pre - ms ) to either the core he flash for m @xmath22 2 @xmath2 or either 100 gyr or the top of the white dwarf cooling curve ( whichever came first ) for m @xmath22 0.5 @xmath2 . for models that experience the core he flash , separate he burning tracks were evolved from the zero age horizontal branch ( zahb ) using the surface compositions and he core masses listed in tables [ zahb ] and [ core ] , respectively . for models that either transition smoothly to core he burning ( m @xmath23 2 @xmath2 ) or were evolved from the zahb , the he burning evolution was followed until the onset of thermal pulsations on the asymptotic giant branch ( agb ) . isochrones were generated from these tracks with ages from 250 myr to 15 gyr in increments of : 50 myr between 250 myr and 1 gyr , 250 myr between 1 and 5 gyr , and 500 myr between 5 and 15 gyr . the isochrones extend from the lower ms to the tip of the red giant branch ( rgb ) or , for ages where the tracks permit it ( ms mass above @xmath242 @xmath2 ) , at or near the onset of the thermal pulsations on the agb . the low mass limit of the isochrones depends on the age : the isochrones only include masses for which the models have reached the ms . the lowest mass stars considered in this paper ( 0.1 @xmath2 ) may require @xmath241 gyr before reaching the ms and therefore will not be present in isochrones for younger ages . for example , the @xmath0}}$]=0 isochrones have the following minimum masses ( @xmath2 ) : 0.43 at 250 myr , 0.32 at 500 myr , 0.22 at 750 myr , and 0.11 at 1 gyr . evolutionary tracks are also provided and these include the pre - ms evolution for all masses . all tracks and isochrones were transformed to the color - magnitude diagram ( cmd ) using each of the color-@xmath7 transformations described in @xmath64 . the track files consist of lines indicating age , log @xmath7 , log g , log l/@xmath25 , followed by the absolute magnitude in each of the filters for a given photometric system . a theoretical track file is also available , this file contains age , log @xmath7 , log g , log l/@xmath25 , log r/@xmath26 , core he mass fraction ( y@xmath27 ) , core z mass fraction ( z@xmath27 ) , the surface z / x ratio , h burning luminosity ( @xmath25 ) , he burning luminosity ( @xmath25 ) , he core mass , and c / o core mass . the isochrones list stellar mass , log @xmath7 , log g , log l/@xmath25 , and absolute magnitude in each of the filters for a given photometric system . the collection of fortran 77 programs introduced by @xcite have been updated to deal with the expanded compositions and photometric systems presented in this paper but are otherwise largely unchanged . see the appendices of @xcite for a detailed description of these programs and their output . ccccccc c & 2.296e-01 & 1.721e-01 & 1.233e-01 & 8.490e-02 & 5.685e-02 & 3.731e-02 + n & 6.726e-02 & 5.041e-02 & 3.611e-02 & 2.487e-02 & 1.665e-02 & 1.093e-02 + o & 3.940e-01 & 4.680e-01 & 5.314e-01 & 5.800e-01 & 6.155e-01 & 6.403e-01 + ne & 9.044e-02 & 1.050e-01 & 1.165e-01 & 1.271e-01 & 1.349e-01 & 1.403e-01 + na & 2.773e-03 & 2.078e-03 & 1.489e-03 & 1.025e-03 & 6.866e-04 & 4.506e-04 + mg & 3.366e-02 & 3.998e-02 & 4.540e-02 & 4.955e-02 & 5.258e-02 & 5.470e-02 + al & 4.814e-03 & 3.608e-03 & 2.585e-03 & 1.780e-03 & 1.192e-03 & 7.822e-04 + si & 3.715e-02 & 4.412e-02 & 5.010e-02 & 5.468e-02 & 5.803e-02 & 6.036e-02 + p & 6.493e-04 & 4.866e-04 & 3.486e-04 & 2.401e-04 & 1.607e-04 & 1.055e-04 + s & 1.851e-02 & 2.199e-02 & 2.497e-02 & 2.725e-02 & 2.892e-02 & 3.008e-02 + cl & 3.900e-04 & 2.923e-04 & 2.094e-04 & 1.442e-04 & 9.656e-05 & 6.337e-05 + ar & 5.793e-03 & 4.342e-03 & 3.111e-03 & 2.142e-03 & 1.434e-03 & 9.414e-04 + k & 3.045e-04 & 2.282e-04 & 1.635e-04 & 1.126e-04 & 7.539e-05 & 4.948e-05 + ca & 3.268e-03 & 3.882e-03 & 4.408e-03 & 4.811e-03 & 5.106e-03 & 5.311e-03 + ti & 1.519e-04 & 1.804e-04 & 2.048e-04 & 2.236e-04 & 2.372e-04 & 2.468e-04 + cr & 1.470e-03 & 1.102e-03 & 7.894e-04 & 5.436e-04 & 3.640e-04 & 2.389e-04 + mn & 1.075e-03 & 8.054e-04 & 5.771e-04 & 3.974e-04 & 2.661e-04 & 1.746e-04 + fe & 1.020e-01 & 7.641e-02 & 5.474e-02 & 3.770e-02 & 2.524e-02 & 1.657e-02 + ni & 6.026e-03 & 4.516e-03 & 3.235e-03 & 2.228e-03 & 1.492e-03 & 9.791e-04 cllllllllll 2.5 & x=0.7549 & 0.7548 & 0.7548 & 0.7547 & 0.7546 & 0.7543 & 0.6699 & 0.6699 & 0.6000 & 0.5999 + & z=4.09e-5 & 5.48e-5 & 7.63e-5 & 1.11e-4 & 1.65e-4 & 2.52e-4 & 4.85e-5 & 9.85e-5 & 4.34e-5 & 8.82e-5 + 2.0 & x=0.7547 & 0.7545 & 0.7544 & 0.7541 & 0.7536 & 0.7529 & 0.6698 & 0.6697 & 0.5999 & 0.5997 + & z=1.30e-4 & 1.72e-4 & 2.41e-4 & 3.50e-4 & 5.24e-4 & 7.97e-4 & 1.53e-4 & 3.50e-4 & 1.37e-4 & 2.79e-4 + 1.5 & x=0.7539 & 0.7536 & 0.7530 & 0.7521 & 0.7507 & 0.7484 & 0.6695 & 0.6690 & 0.5996 & 0.5991 + & z=4.10e-4 & 5.47e-4 & 7.62e-4 & 1.11e-3 & 1.65e-3 & 2.50e-3 & 4.85e-4 & 9.83e-4 & 4.34e-4 & 8.81e-4 + 1.0 & x=0.7516 & 0.7505 & 0.7487 & 0.7459 & 0.7415 & 0.7346 & 0.6685 & 0.6669 & 0.5986 & 0.5972 + & z=1.29e-3 & 1.72e-3 & 2.40e-3 & 3.46e-3 & 5.15e-3 & 7.77e-3 & 1.53e-3 & 3.46e-3 & 1.37e-3 & 2.78e-3 + 0.5 & x=0.7444 & 0.7441 & 0.7355 & 0.7270 & 0.7139 & 0.6942 & 0.6652 & 0.6603 & 0.5957 & 0.5913 + & z=4.05e-3 & 5.37e-3 & 7.44e-3 & 1.07e-2 & 1.57e-2 & 2.32e-2 & 4.82e-3 & 9.70e-3 & 4.31e-3 & 8.69e-3 + 0.0 & x=0.7174 & 0.7071 & 0.6882 & 0.6628 & 0.6247 & 0.5710 & 0.6550 & 0.6402 & 0.5866 & 0.5734 + & z=1.43e-2 & 1.89e-2 & 2.55e-2 & 3.52e-2 & 4.97e-2 & 7.02e-2 & 1.50e-2 & 2.98e-2 & 1.34e-2 & 2.66e-2 + + 0.15 & x= & 0.6880 & & & & & & & & + & z= & 2.56e-2 & & & & & & & & + + 0.3 & x=0.6842 & 0.6638 & 0.6334 & & & & & & & + & z=2.70e-2 & 3.48e-2 & 4.64e-2 & & & & & & & + + 0.5 & x=0.6491 & 0.6185 & 0.5792 & & & & & & & + & z=4.04e-2 & 5.21e-2 & 6.71e-2 & & & & & & & + cllllllllll 2.5 & x=0.7564 & 0.7565 & 0.7561 & 0.7549 & 0.7549 & 0.7546 & 0.6835 & 0.6828 & 0.5967 & 0.6212 + & z=3.97e-5 & 5.31e-5 & 7.40e-4 & 1.60e-4 & 1.08e-4 & 2.44e-4 & 4.63e-5 & 9.40e-5 & 8.40e-3 & 8.33e-5 + 2.0 & x=0.7547 & 0.7548 & 0.7551 & 0.7541 & 0.7524 & 0.7510 & 0.6820 & 0.6810 & 0.6208 & 0.6193 + & z=1.26e-4 & 1.67e-4 & 2.33e-4 & 3.41e-4 & 5.06e-4 & 7.69e-4 & 1.46e-4 & 2.97e-4 & 1.30e-4 & 2.64e-4 + 1.5 & x=0.7529 & 0.7517 & 0.7509 & 0.7491 & 0.7464 & 0.7427 & 0.6785 & 0.6763 & 0.6180 & 0.6156 + & z=3.96e-4 & 5.29e-4 & 7.37e-4 & 1.07e-3 & 1.60e-3 & 2.43e-3 & 4.64e-4 & 9.44e-4 & 4.12e-4 & 8.37e-4 + 1.0 & x=0.7472 & 0.7453 & 0.7423 & 0.7383 & 0.7325 & 0.7238 & 0.6732 & 0.6698 & 0.6125 & 0.6076 + & z=1.25e-3 & 1.67e-3 & 2.33e-3 & 3.37e-3 & 5.02e-3 & 7.59e-3 & 1.47e-3 & 3.00e-3 & 1.31e-3 & 2.67e-3 + 0.5 & x=0.7350 & 0.7309 & 0.7242 & 0.71485 & 0.7010 & 0.6810 & 0.6641 & 0.6600 & 0.6034 & 0.5967 + & z=3.94e-3 & 5.24e-3 & 7.27e-3 & 1.05e-2 & 1.54e-3 & 2.28e-2 & 4.67e-3 & 9.48e-3 & 4.16e-3 & 8.41e-3 + 0.0 & x=0.7166 & 0.6928 & 0.6741 & 0.6497 & 0.6131 & 0.5640 & 0.6477 & 0.6303 & 0.5877 & 0.5854 + & z=1.41e-2 & 1.85e-2 & 2.50e-2 & 3.46e-2 & 4.88e-2 & 6.87e-2 & 1.47e-2 & 2.91e-2 & 1.31e-2 & 2.68e-2 + + 0.15 & x= & 0.6737 & & & & & & & & + & z= & 2.51e-2 & & & & & & & & + + 0.3 & x=0.6696 & 0.6505 & 0.6204 & & & & & & & + & z=2.66e-2 & 3.42e-2 & 4.57e-2 & & & & & & & + + 0.5 & x=0.6360 & 0.6075 & 0.5734 & & & & & & & + & z=3.98e-2 & 5.12e-2 & 6.57e-2 & & & & & & & + ccccccccccc 2.5 & 0.506 & 0.505 & 0.504 & 0.502 & 0.500 & 0.497 & 0.489 & 0.485 & 0.474 & 0.473 + 2.0 & 0.500 & 0.499 & 0.498 & 0.494 & 0.492 & 0.491 & 0.483 & 0.479 & 0.469 & 0.468 + 1.5 & 0.495 & 0.493 & 0.491 & 0.489 & 0.488 & 0.486 & 0.478 & 0.474 & 0.464 & 0.463 + 1.0 & 0.489 & 0.488 & 0.486 & 0.484 & 0.481 & 0.478 & 0.473 & 0.469 & 0.460 & 0.458 + 0.5 & 0.483 & 0.482 & 0.480 & 0.477 & 0.473 & 0.469 & 0.469 & 0.465 & 0.457 & 0.455 + 0.0 & 0.475 & 0.473 & 0.469 & 0.464 & 0.458 & 0.447 & 0.465 & 0.460 & 0.454 & 0.451 + + 0.15 & & 0.468 & & & & & & & & + + 0.3 & 0.468 & 0.464 & 0.459 & & & & & & & + + 0.5 & 0.461 & 0.456 & 0.449 & & & & & & & + synthetic color-@xmath7 transformations were derived from the ` phoenix ` model atmosphere grid ( also used to generate surface boundary conditions in the stellar evolution calculations ) for @xmath28 10,000 k and for all @xmath0}}$ ] and @xmath1}}$ ] combinations . the ` phoenix ` grid was supplemented by @xcite models for higher @xmath7 . in order to insure a smooth transition , the @xcite colors were adjusted to line up with the ` phoenix ` colors at 10,000 k and the two sets of tables were ramped between ` phoenix ` and @xcite in the temperature range of 9,000 k to 10,000 k. as mentioned in @xmath62 , the ` phoenix ` model atmospheres used to generate synthetic color@xmath7 transformations were computed for each of the @xmath0}}$ ] , @xmath1}}$ ] pairs but not for all of the enhanced y values . the influence of enhanced he on broad - band synthetic photometry is negligible for the temperatures considered here , as discussed briefly by @xcite and in more detail by @xcite . the @xcite model atmospheres were only computed for @xmath1}}$]=0 and + 0.4 . in constructing the color transformations the same approach was used as described in @xmath62 for the surface bc . for evolutionary tracks and isochrones with @xmath1}}\leq + 0.2 $ ] , colors based on @xcite synthetic spectra with @xmath1}}= 0 $ ] were applied and for tracks and isochrones with @xmath1}}\geq + 0.4 $ ] , colors based on @xcite synthetic spectra with @xmath1}}= + 0.4 $ ] were applied . while this is not an ideal situation , the influence of @xmath18-enhancement , or indeed of total z , on broad - band synthetic colors is small for @xmath29 5,000 k. the only exceptions are in the bluest bands ( equivalent to u ) and at @xmath29 35,000 k. bands equivalent to u are the most uncertain in terms of the synthetic photometry and very few of the evolutionary tracks ( only the lowest mass , metal poor he - burning tracks ) presented in this paper extend beyond @xmath7 = 35,000 k. the synthetic colors have been tested and perform well in bandpasses equivalent to v or redder ( with central wavelengths longer than @xmath30 ) but the blue and ultraviolet bands ( with central wavelengths shorter than @xmath30 ) suffer from inaccuracy of the synthetic fluxes at shorter wavelengths . in cases where analysis in the bluer bands is important , empirical color transformations are strongly recommended . @xcite explored the dependence of broadband synthetic colors on @xmath1}}$ ] for @xmath1}}$]=0 and + 0.4 . their findings indicate that u b and b v are the colors most strongly effected by @xmath18-enhancement and that the influence of @xmath18-enhancement on broadband colors increases with increasing metallicity and decreasing @xmath7 . the color transformations derived from the ` phoenix ` model atmosphere grid have a greater span of @xmath7 and @xmath1}}$ ] , including @xmath1}}$ ] = 0.2 for the first time , and better resolution ( 0.2 dex ) in @xmath1}}$ ] . the comparisons performed with the ` phoenix ` synthetic colors qualitatively agree with the findings of @xcite though it is important to recognize that @xcite compared colors at constant z while in this paper comparisons are made at constant @xmath0}}$ ] . hence an increase in @xmath1}}$ ] represents an increase in total z and thus one expects the colors to become redder with increasing @xmath1}}$ ] as a general rule . figure [ alpha ] displays the changes to cmd morphology based solely on the level of @xmath18-enhancement at @xmath0}}$]=0 using a 10 gyr scaled - solar isochrone . the figure shows @xmath1}}$ ] = 0 , + 0.4 , and + 0.8 colors . the comparisons were assembled using a scaled - solar isochrone in order to demonstrate the effect of @xmath18-enhancement on the colors only . two colors , v i and j@xmath4 are shown in figure [ alpha ] . the figure shows that v i increases with @xmath1}}$ ] and while the difference is small near the msto , it grows as @xmath7 decreases , becoming largest near m@xmath31 = 10 . the situation in j@xmath4 is reversed : the highest level of @xmath1}}$ ] produces the bluest j@xmath4 color while the lowest level produces the reddest color . the j@xmath4 result is , at first glance , counter - intuitive . however , the j band is particularly sensitive to absorption from water and , in the presence of enhanced oxygen ( from @xmath1}}$ ] ) , substantially more water is able to form . this pushes flux away from the j band and causes the j@xmath4 color to become bluer with increasing @xmath1}}$ ] . as reported by @xcite , the color differences due to variations in @xmath1}}$ ] decrease as total metallicity decreases however , isochrones presented in this paper suggest that the separation along the lower ms in j@xmath4 remains noticeable even at @xmath0}}$]=2 and this feature may prove a useful diagnostic for near infrared studies of metal poor systems of sufficient depth to see the lower ms . the remainder of this section give a brief description of the photometric systems into which the stellar evolution tracks and isochrones have been transformed . @xcite defined filter transmission curves for the johnson cousins ubv(ri)@xmath3 photometric system . these filter curves were combined with 2mass jh@xmath4 filters from @xcite to create a set of isochrones suitable for use with ground based photometry from the ultraviolet to the near infrared . the ubvri magnitudes were calibrated to vega by following appendix a of @xcite . the 2mass magnitudes were calibrated by setting the magnitudes of vega to zero in each filter @xcite . @xcite defined a method for generating synthetic photometry based on the detailed filter transmission curves for the hubble space telescope ( hst ) advanced camera for surveys ( acs ) and wide field planetary camera 2 ( wfpc2 ) from @xcite . the synthetic photometry presented here are all on the vegamag system . the following filters are included for the acs - wfc : f435w , f475w , f555w , f606w , f625w , f775w , f814w , and f850lp . the following filters are included for the wfpc2 : f336w , f439w , f555w , f606w , f791w , f814w , and f850lp . the performance of the synthetic color transformation in the acs f606w - f814w cmd has been documented by @xcite . sdss produced ( and its ongoing extensions continue to produce ) vast quantities of stellar photometry and spectroscopy with important implications for our knowledge of galactic structure and evolution . the sdss ugriz photometric system is based on the ab magnitude system but each of the five filters has an additional offset from ab of at most a few hundredths of a magnitude . the offsets , in the sense of mag(sdss)mag(ab ) are : 0.04 in u , 0.01 in g , 0 in r , 0.015 in i , and 0.03 in z. the filter transmission curves and procedure for transforming synthetic fluxes to sdss magnitudes magnitudes is outlined on the sdss photometric flux calibration web page . typically , the synthetic colors provide good fits to sdss data in g , r , and i. synthetic photometry in the u band is subject to the same shortcomings as other bands at bluer wavelengths while problems with the underlying calibration of the sdss photometric systems persist in both u and z. for general use , and comparison with the synthetic color transformations , the tracks and isochrones are presented in colors based on the semi empirical transformations of ( * ? ? ? * hereafter vc03 ) in bv(ri)@xmath3 . vc03 compiled a set of @xmath7color transformations and bolometric corrections covering a wide range of metallicity , @xmath7 , and log g. the vc03 colors are based on marcs model atmospheres @xcite for lower temperatures and kurucz models ( consistent with @xcite ) for higher temperatures . the color transformations are semi empirical in the sense that vc03 began with a collection of synthetic colors and then applied corrections to the colors , primarily for cooler ms stars approaching and above @xmath0}}$]=0 , based on observational constraints from open clusters , globular clusters , and field stars . @xcite presented color transformations for the strmgren uvby photometric system . @xcite used a similar approach to vc03 by adopting marcs models @xcite for @xmath32 8,000 k and colors based on kurucz models @xcite for higher temperatures . the uvby color transformations of @xcite provide much better fits to observations than any synthetic color tables ( see that paper for a detailed discussion ) and thus no synthetic strmgren photometry has been provided . this section will compare the dsep models to comparable models and observational data from the literature . @xcite presented comparisons with old , metal poor isochrones ( 12 gyr , z=0.0001 ) from several groups and compared the dsep isochrones to acs photometry of two globular clusters : m 92 and 47 tuc from the acs galactic globular cluster survey @xcite . as a result , this section will focus on model comparisons at @xmath0}}\geq 0 $ ] , along with younger , more metal rich open clusters . the one exception is a comparison of sdss photometry from m 13 and ngc 2420 to the ugriz isochrones in @xmath65.2.6 . following @xcite , comparison isochrones were drawn from ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * hereafter basti ) , ( * ? ? ? * hereafter padova ) , ( * ? ? ? * hereafter victoria - regina ( vr ) ) , and ( * ? ? ? * ; * ? ? ? * hereafter yale - yonsei ( y@xmath33 ) ) . in this section , the comparison is extended to solar metallicity and above as well as to younger ages . comparisons are shown in both the log @xmath7-log l/@xmath25 plane and the v vs. v i color - magnitude diagram . in each case , the dsep isochrone is shown with the synthetic color transformation . a comparison of different color transformations applied to the dsep isochrones appears later in this section . figures [ 600myr ] and [ 600msto ] show isochrones at @xmath0}}\sim$]+0.15 and an age of 625 - 630 myr ( compatible with the the hyades , see @xmath65.2.2 ) . the isochrones range in initial z=0.025 to 0.03 and y=0.28 to 0.3 . in figure [ 600myr ] , the padova , basti , and dsep isochrones all show the agb though the point at which the agb terminates differs : dsep and basti stop at or before the onset of thermal pulsations though the basti library has been updated to include extended agb evolution @xcite . fainter than log l/@xmath25=1 , the large differences are due to different mass ranges , the eos , and low temperature opacities used by each group . the basti isochrone begins at 0.5 @xmath2 , vr and y@xmath33 isochrones begin at 0.4 @xmath2 , the padova isochrone begins at 0.15 @xmath2 , and the dsep isochrone begins at 0.3 @xmath2 at this age . in figure [ 600msto ] the agreement is quite good among dsep , basti , and y@xmath33 isochrones . the padova isochrone is less luminous in the region of the ms and subgiant branch ( sgb ) while the vr isochrons is more luminous in the same region . these differences in luminosity are primarily due to different adopted treatment of convective core overshoot . figure [ 600myr ] indicates that the y@xmath33 isochrone extends to cooler temperatures than any of the others despite the fact that it terminates at 0.4 @xmath2 while dsep and padova terminate at considerably lower mass . this effect is almost certainly due to the eos used in the y@xmath33 calculations . figures [ 4gyr ] and [ 4msto ] show isochrones at solar metallicity and 4 gyr ( compatible with m 67 , see @xmath65.2.4 ) . each isochrone has an initial z=0.018 - 0.020 and y=0.27 - 0.28 ; the adopted physics ( especially the eos and convective core overshoot prescription ) and solar heavy element mixture differ from one group to another . in the left panels of figure [ 4gyr ] and [ 4msto ] , the isochrones are in reasonable agreement for points brighter than log l/@xmath25=1 ( padova and basti isochrones include he burning stars while the others do not ) . in figure [ 4msto ] , the basti , vr , and dsep isochrones all agree well on the temperature and luminosity of the ms turn off ( to ) and sgb . in the right panels of figure [ 4gyr ] and [ 4msto ] , all isochrones agree well at magnitudes brighter than m@xmath31=6 except for padova that is consistently redder and brighter on the upper rgb . in figure [ 4gyr ] the dsep isochrone includes the complete lower mass range ( down to 0.1 @xmath2 ) and the fact that the y@xmath33 isochrone is cooler and almost as faint in m@xmath31 is significant . conversely , the padova isochrone is hotter and bluer than the others for points fainter that log l/@xmath25=1 ( m@xmath31=7 ) although the mass - luminosity relation is similar to the dsep isochrone . figures [ 600myr ] , [ 600msto ] , [ 4gyr ] and [ 4msto ] show general agreement amongst the different sets of isochrones . the main exceptions are the padova isochrones that appear cooler and redder near the msto and on the rgb but hotter and bluer on the lower main sequence . amongst the other models the largest differences occur near the msto where the adopted core overshoot treatments differ and on the lower main sequence where the adopted eos and minimum masses differ . the dsep models use the same eos and opacities as the vr and basti models and it is reassuring that these models all lie close together aside from the noted exceptions due to the included mass range and treatment of convective core overshoot . several papers have demonstrated the performance of the dsep isochrones when applied to hst / acs data in two of the most popular bands for globular cluster photometry ( f606w and f814w ) including : @xcite , @xcite , @xcite , and @xcite . this section will focus on the ground - based photometric systems presented in this paper and also compares the synthetic color transformation with the vc03 transformation in b v and v i . ccccccc m 37 & 11.55 & 0.27 & 0.09 & 0.5 & b v j @xmath4 & 1,2 + hyades & & & 0.13 & 0.65 & b v i @xmath4 & 3,4 + ngc 2420 & 12.1 & 0.05 & 0.27 & 2 & b v i @xmath4 & 1 + m 67 & 9.65 & 0.04 & 0 & 4 & b v i @xmath4 & 1,5,6 + ngc 6791 & 13.35 & 0.1 & 0.4 & 9 & b v i @xmath4 & 7,8,9 + ( * ? ? ? * hereafter gs03 ) combined optical photometry of six galactic open clusters with near - infrared data from 2mass . similar compilations for the hyades @xcite and ngc 6791 @xcite are presented along with three clusters from gs03 to span a range of ages and metallicities in open clusters . these combined optical and near - infrared data provide an excellent source of comparisons with the dsep isochrones . table [ clusters ] lists the relevant parameters of the clusters featured in this section . the adopted reddening values are listed as e(b v ) . the extinction curve of @xcite was used to determine the reddening in other filter combinations : e(v i ) = 1.4 e(b v ) , e(v j ) = 2.4 e(b v ) and e(v-@xmath4 ) = 2.8 e(b these comparisons are not meant to determine the ages of the clusters but rather to demonstrate the performance of the isochrones . the ages were chosen to be consistent with other published results but if the age shown is considerably different from the true age of the cluster it will influence the quality of the fit . figure [ m37 ] shows the cmd of m 37 in b v @xcite and v j and v@xmath4 ( gs03 ) along with a @xmath0}}$]=0 , 500 myr isochrone . the isochrone is plotted using the synthetic ( solid line ) and vc03 ( dashed line ) colors . in b v ( left panel of figure [ m37 ] ) , the synthetic color isochrone is redder than the data by @xmath240.1 mag throughout the cmd while the vc03 isochrone provides a reasonable fit to the data from the msto to v=16 before drifting to the red of the data . the he burning portion of the isochrone is redder than the data by 0.1 mag in both color transformations . in v j and v@xmath4 , the synthetic color isochrone provides an excellent fit to the data for the entire extent of the ms ( the ms data cuts off at v@xmath2420 in v j and v@xmath4 ) . the isochrone is redder than the data for the he burning stars in both v j and v@xmath4 . we note in passing that the vc03 b v isochrone shown in the left panel of figure [ m37 ] does not match the shape of the cmd for v @xmath34 17 . although m 37 and m 67 are both roughly @xmath0}}$ ] = 0 clusters their b v cmds from @xcite and m 67 @xcite , respectively , adjusted for distance and reddening , appear to disagree somewhat for m@xmath35 6 when viewed together . since the vc03 colors are based in part on the cmd of m 67 , it is impossible for them to fit both clusters simultaneously . figure [ hya ] shows the hipparcos b v ( left panel ) and v i ( center ) cmds of the hyades @xcite and the combined hipparcos and 2mass v@xmath4 cmd compiled by @xcite . along with the data , a @xmath0}}$]=+0.15 , 625 myr isochrone is shown in the synthetic ( solid line ) and vc03 ( dashed line ) color transformations . the choice of a 625 myr isochrone may be an underestimate of the age which is typically quoted as between 650 and 700 myr ( e.g. vc03 ) . the he abundance assumed for the @xmath0}}$]=+0.15 models may also be a cause for disagreement since the adopted @xmath21y/@xmath21z yields y@xmath5=0.2864 whereas the hyades y value inferred from model comparisons is lower by @xmath240.02 ( see vc03 and references therein for a discussion ) . in b v , the synthetic color isochrone is redder than the data by 0.050.1 mag down to m@xmath31=7 where the slope changes and the isochrone becomes bluer than the data by m@xmath31=8 . the vc03 color isochrone , on the other hand , is a good fit for the entirety of the ms . both isochrones appear cooler than the two he burning stars in the b v cmd . in v i , the color transformations are quite similar and both perform well down to m@xmath31=6 . at this point , the isochrones become redder than the data before matching the data again at m@xmath31=8 . in v@xmath4 , the synthetic color isochrone fits the data for stars with m@xmath36 3 but becomes bluer than the data just below the msto . the discrepancy just below the msto would be avoided by plotting a slightly older isochrone . figure [ n2420 ] shows the b i , and v@xmath4 cmds of ngc 2420 from gs03 . plotted along with the data is an isochrone interpolated to @xmath0}}$]=0.25 and 2.15 gyr . given the adopted distance and reddening in table [ clusters ] , the best fit age plotted in figure [ n2420 ] is older than the age quoted by gs03 . this is not surprising since ultimately the gs03 age estimate for ngc 2420 was based on padova isochrones and these were shown to be fainter and redder near the msto than the dsep isochrones ( see figures [ 600myr ] and [ 4gyr ] and the discussion in @xmath65.1 ) . as a result , the best fit padova isochrone will be younger than the best fit dsep isochrone . in b v the synthetic isochrone is redder than the data , particularly on the lower ms and the rgb , though it matches well near the msto . the vc03 color isochrone fits the entire cmd well . the he burning track lies fainter than the red clump for both color transformations with the color predicted correctly by vc03 but too red by the synthetic color . if ngc 2420 is younger than 2.1 gyr , then the stars in the red clump will be more massive and more luminous . in v i , both synthetic and vc03 colors fit the cmd well except for the synthetic color on the rgb that is slightly cooler than the data . the color of the red clump is correctly predicted by vc03 but slightly too red in the synthetic color . in v@xmath4 , the synthetic isochrone fits the sgb and rgb and correctly predicts the color of the red clump . however , the v@xmath4 color appears to be too blue on the ms . figure [ m67 ] shows the b i , and v@xmath4 cmds of m 67 from @xcite and gs03 along with a @xmath0}}$]=0 , 4 gyr isochrone . the isochrone has been transformed using the synthetic ( solid line ) and vc03 ( dashed line ) colors . a he burning track with @xmath0}}$]=0 and m=1.25 @xmath2 is shown to indicate the predicted location of the red clump . the fit of the synthetic color isochrone to the data in b v ( left panel of figure [ m67 ] ) is poor , with the isochrone redder than the data by about 0.05 - 0.1 mag throughout the cmd while the vc03 color isochrone provides an excellent fit . the he burning model has the correct v magnitude but is slightly redder than the observed red clump in b v for both color transformations . in v i ( middle panel of figure [ m67 ] ) , both synthetic and vc03 colors match the data well over most of the cmd . the only exception is that , for stars fainter than v=17 , the vc03 colors are redder than the data while the synthetic colors follow the data closely down to v=20 . the he burning model correctly predicts the color of the red clump in v i for both color transformations . in v@xmath4 only the synthetic color isochrone is plotted in the right panel of figure [ m67 ] . as with v i , the synthetic v@xmath4 color is an excellent fit to the data for the lower ms , the msto , the rgb , and the location of the red clump . @xcite combined jhk photometry of ngc 6791 with optical data from @xcite . the resulting cmds in b v , v i , and v k are shown in figure [ n6791 ] . k cmd has been adjusted by + 0.024 mag to convert from cit k to 2mass @xmath4 @xcite . the isochrones shown in figure [ n6791 ] have @xmath0}}_{init}$ ] = + 0.36 and an age of 9 gyr . the b v cmd continues to be problematic for the synthetic colors where , using the adopted e(b v ) , the synthetic isochrone is @xmath240.1 mag redder than the data . the isochrone with vc03 colors is a substantially better fit to the data . this is not surprising since the vc03 colors were calibrated at high metallicity with the @xcite data . i cmd is equally well fit by both synthetic and vc03 colors . the synthetic color isochrone in v k is an excellent fit to the shape of the cmd as well . for the isochrone used in figure [ n6791 ] , neglecting mass loss , the mass at the tip of the rgb is 1.12 @xmath2 . a 1 @xmath2 he burning track is plotted to show the predicted level of the red clump in figure [ n6791 ] . the he burning track provides a good fit to the v mag of the red clump but lies a few hundredths of a magnitude redder than the red clump in b v for both color transformations . in v i , the color of the red clump is in excellent agreement with the model . in v k , the model is bluer than the red clump by a few hundredths of a magnitude . it should be noted that while the adopted reddening value is e(b v ) = 0.1 , @xcite found e(b v ) = 0.14 at @xmath0}}$]=+0.4 and the @xcite dust maps yield e(b v ) = 0.155 for ngc 6791 . these values are both higher than the adopted value of e(b v ) and would alter the isochrone fits shown in figure [ n6791 ] . if instead e(b v ) = 0.15 is adopted as the reddening then the isochrone fits require either younger age or lower metallcity ( or both ) . using the same isochrones ( with @xmath0}}_{init}$]=+0.36 ) but with the larger reddening , a reasonably good fit is achieved at 8 gyr but with an increased distance modulus of @xmath37 = 13.45 . reducing @xmath0}}$ ] to + 0.2 introduces a substantial mismatch between the observed cmd and the models along the ms with the isochrones becoming bluer than the data at 2 mag below the msto . given that the adopted @xmath0}}$ ] value provides a good fit to the entire cmd and agrees with recent spectroscopic estimates by , for example , @xcite and @xcite , the isochrones show a clear preference for the lower reddening value of e(b v)=0.1 used in figure [ n6791 ] . constraining the distance , reddening , composition , and age of ngc 6791 remains a thorny problem ( see @xcite for a detailed discussion ) . because of it high metallicity , a conclusive age determination of ngc 6791 will most likely require that the models follow the detailed abundance patterns found by spectroscopy and not simply match the observed @xmath0}}$ ] . for example , @xcite found that @xmath1}}\sim$ ] 0 but [ c / fe]=0.35 in ngc 6791 . @xcite present spectroscopically selected cluster members for stars in a sample of galactic globular and open clusters for purposes of validating the segue stellar parameter pipeline . the selected stars from m 13 and ngc 2420 are used here because they represent the cleanest available data sets of clusters on the sdss ugriz system . though these sequences are quite narrow and well defined , they are not as deep as allowed by sdss , but are suitable for the purpose of demonstrating the performance of the isochrones . figure [ sdss ] shows the g vs. ( g i)@xmath38 cmds of m 13 ( left ) and ngc 2420 ( right ) along with appropriate isochrones . the photometry has been de - reddened by @xcite using the @xcite dust maps . plotted on top of the m 13 data in figure [ sdss ] are isochrones with @xmath0}}$]=-1.5 , @xmath1}}$]=+0.2 at 11 and 13 gyr assuming a distance modulus of 14.4 in agreement with @xcite . the isochrones follow the locus of stars on the ms and rgb well . the only real discrepancy is the slope of the sgb but this is a common problem with synthetic color isochrones compared to metal poor systems . plotted on top of the ngc 2420 data in figure [ sdss ] is the same isochrone as in @xmath65.2.3 but now transformed to g vs. g i plane . the agreement between the data and the isochrone is excellent on the ms but the observational constraints of the sample do not permit comparison with stars in the convective hook or past the msto . bluer synthetic colors , as represented by b v in this section , continue to require adjustments in order to match observed cmds as figures [ m37 ] through [ n6791 ] indicate . the semi - empirical color transformation of vc03 are an improvement over the synthetic b v color but seem unnecessary for v i when used in conjuction with the dsep isochrones . while the synthetic colors and magnitudes are suitable analysis of observational data in bandpasses with central wavelengths longer than @xmath30 , empirically adjusted colors ought to be used at shorter wavelengths . in these redder bands , the synthetic colors perform well throughout the cmd , including on the lower main sequence where they perform better than the empirical colors . the optical to near - infrared comparisons in v j ( m 37 only ) and v k of the open clusters are favorable for entire extent of the ms and the rgb in the older clusters but larger reddening than suggested by the adopted reddening curve is needed in some cases . the location of he burning stars in the open clusters is matched by the models in the older clusters ( m 67 and ngc 6791 ) but for the younger clusters the models predict the red clump to be fainter ( ngc 2420 ) and/or redder ( m 37 and the hyades ) than the data suggest . the general lack of evolved stars in the cmds of both m 37 and the hyades makes it difficult to speculate as to whether the models are intrinsically too cool or whether the colors are too red for a given @xmath7 . the dartmouth stellar evolution program has been used to compute a large grid of stellar evolution tracks . a subset of these tracks and isochrones were presented by @xcite and are extended in this paper to include younger ages , higher metallicities , and transformations to multiple photometric systems . these models take advantage of the latest advances in radiative opacities , nuclear reactions rate , and the equation of state . the dartmouth stellar evolution database comprises a set of stellar evolution tracks and isochrones transformed to several photometric systems along with a set of computer programs that allow for interpolation and construction of luminosity functions and synthetic horizontal branch models . the database spans a wide range of compositions from @xmath0}}$]=2.5 to + 0.5 , @xmath1}}$]=0.2 to + 0.8 ( at @xmath0}}\leq$]0 ) or + 0.2 ( at @xmath0}}>$]0 ) , and additional models with enhanced he for @xmath0}}$]=2.5 to 0 . the models were transformed to the observational plane using a synthetic color transformation based on ` phoenix ` model atmospheres for @xmath2810,000 k and @xcite for hotter temperatures . the synthetic color transformations are compared to the vc03 semi - empirical transformations in b v and v i . the synthetic colors perform well in v i but are consistently too red in b v . as a general rule , the synthetic colors should only be trusted in bandpasses equivalent to v or redder , i.e. those with central wavelengths longer than @xmath30 . in these bands , the synthetic colors were found to perform well and better than the empirical colors on the lower main sequence . synthetic photometry in the bluer bandpasses ( those with central wavelengths shorter than @xmath30 ) does not compare favorably with observations . empirically adjusted magnitudes and colors should be favored for the bluer bandpasses whenever possible . the isochrones were compared to isochrones from other groups at @xmath0}}$]=0 and + 0.15 as well as photometry from open clusters and one globular cluster . the isochrones extend to lower mass ( 0.1 @xmath2 ) than those of other major isochrone libraries and provide reasonable fits to the ms of the clusters shown in @xmath65 . the dartmouth stellar evolution database should prove useful in studies of resolved stellar populations in the local universe and is suitable for application to population synthesis and integrated light models . the database is available through a dedicated website . ad wishes to thank alan irwin for his work on freeeos and for making it freely available , arnold boothroyd for sharing his subroutines to manage the opal type 2 opacity tables , kevin covey for helpful discussions regarding the sdss photometric system , and harvey richer for advice regarding the hst photometric system . thanks also to aaron grocholski , ata sarajedini , and jason kalirai for sharing their open cluster photometry . the work of dj and eb was supported in part by by nasa grants nag5 - 3505 and nng04gb36 g , nsf grants ast-0307323 and ast-0707704 , and us doe grant de - fg02 - 07er41517 . this research used resources of the national energy research scientific computing center ( nersc ) , which is supported by the office of science of the u.s . department of energy under contract no . de - ac03 - 76sf00098 ; and the hchstleistungs rechenzentrum nord ( hlrn ) . we thank both these institutions for a generous allocation of computer time . jf acknowledges support from nsf grant ast-0239590 and grant no . eia-0216178 , grant no . eps-0236913 with matching support from the state of kansas and the wichita state university high performance computing center . adelberger , e. c. et al . 1998 , rev . phys . , 70 , 1265 angulo , c. , et al . 1999 , , 656 , 3 bedin , l. r. , cassisi , s. , castelli , f. , piotto , g. , anderson , j. , salaris , m. , momany , y. , & pietrinferni , a. 2005 , , 357 , 1038 bell , r. a. & gustafsson , b. 1978 , , 34 , 229 bell , r. a. & gustafsson , b. 1989 , , 236 , 653 bessell , m. s. 1990 , , 102 , 1181 bessell , m. s. , & brett , j. m. 1988 , , 100 , 1134 bessell , m. s. , castelli , f. , & plez , b. 1998 , , 333 , 231 bjork , s. r. & chaboyer , b. 2006 , , 641 , 1102 boesgaard , a. m. , & friel , e. d. 1990 , , 351 , 467 carney , b. w. , lee , j .- w . , & dodson , b. 2005 , , 129 , 656 carpenter , j. m. 2001 , , 121 , 2851 carraro , g. , villanova , s. , demarque , p. , mcswain , m. v. , piotto , g. , & bedin , l. r. 2006 , , 643 , 1151 cassisi , s. , salaris , m. , castelli , f. , & pietrinferni , a. 2004 , , 616 , 498 castelli , f. 1999 , , 346 , 564 castelli , f. , gratton , r. g. , & kurucz , r. l. 1997 , , 318 , 841 castelli , f. , & kurucz , r. l. 2003 , iau symp . n. piskunov , w. w. weiss , & d. f. gray ( san francisco : asp ) , a20 chaboyer , b. , fenton , w. h. , nelan , j. e. , patnaude , d. j. , & simon , f. e. 2001 , , 562 , 521 chaboyer , b. , greene , e. m. , & liebert , j. 1999 , , 117 , 1360 chaboyer , b. & kim , y .- c . 1995 , , 454 , 767 clem , j. l. , vandenberg , d. a. , grundahl , f. , & bell , r. a. 2004 , , 127 , 1227 cohen , m. , wheaton , w. a. , & megeath , s. t. 2003 , , 126 , 1090 cordier , d. , pietrinferni , a. , cassisi , s. , & salaris , m. 2007 , , 133 , 468 de bruijne , j. h. j. , hoogerwerf , r. , & de zeeuw , p. t. 2001 , , 367 , 111 demarque , p. , woo , j .- h . , kim , y .- c . , & yi , s. k. 2004 , , 155 , 667 dotter , a. , chaboyer , b. , jevremovi , baron , e. , ferguson , j. w. , sarajedini , a. , & anderson , j. 2007 , , 134 , 376 ferguson , j. w. , alexander , d. r. , allard , f. , barman , t. , bodnarik , j. g. , hauschildt , p. h. , heffner - wong , a. , & tamanai , a. 2005 , 623 , 585 girardi , l. , bressan , a. , bertelli , g. , & chiosi , c. 2000 , , 141 , 371 girardi , l. , castelli , f. , bertelli , g. , nasi , e. 2007 , , 468 , 657 grevesse , n. & sauval , a. j. 1998 , , 85 , 161 ( gs98 ) grocholski , a. j. & sarajedini , a. 2003 , , 345 , 1015 haft , m. , raffelt , g. , & weiss , a. 1994 , , 425 , 222 harris , w.e . 1996 , , 112 , 1487 hauschildt , p. h. , allard , f. , & baron , e. 1999a , , 512 , 377 hauschildt , p. h. , allard , f. , ferguson , j. , baron , e. , & alexander , d. 1999b , , 525 , 871 houdashelt , m. l. , bell , r. a. , & sweigart , a. v. 2000 , , 119 , 1448 hubbard , w. b. & lampe , m. 1969 , , 18 , 297 iglesias , c.a . & rogers , f.j . 1996 , , 464 , 943 imbriani , g. et al . 2004 , , 420 , 625 irwin , a. 2004 , technical report ( http://freeeos.sourceforge.net ) johnson , h. l. 1965 , , 141 , 923 kalirai , j. s. , ventura , p. , richer , h. b. , fahlman , g. g. , durrell , p. r. , dantona , f. , & marconi , g. 2001 , , 122 , 3239 korn , a.j . , grundahl , f. , richard , o. , barklem , p.s . , mashonkina , l. , collet , r. , piskunov , n. , & gustafsson , b. , 2006 , the messenger , 125 , 6 kunz , r. , fey , m. , jaeger , m. , mayer , a. , & hammer , j. w. 2002 , , 567 , 643 lee , y .- w . , demarque , p. , & zinn , r. 1990 , , 350 , 155 lee , y. s. et al . 2007 , submitted to ( http://arxiv.org/abs/0710.5778 ) montgomery , k. a. , marschall , l. a. , & janes , k. a. 1993 , , 106 , 181 paulson , d. b. , sneden , c. , & cochran , w. d. 2003 , , 125 , 3185 ( psc ) peterson , r. c. & greene , e. m. 1998 , , 502 , 39 origlia , l. , valenti , e. , rich , r. m. , & ferraro , f. r. 2006 , , 646 , 499 pietrinferni , a. , cassisi s. , salaris m. & castelli f. 2004 , , 612 , 168 pietrinferni , a. , cassisi , s. , salaris , m. , & castelli , f. 2006 , , 642 , 797 pinsonneault , m. h. , terndrup , d. m. , hanson , r. b. , stauffer , j. r. 2004 , , 600 , 946 richer , h. b. , dotter , a. , hurley , j. , anderson , j. , king . i. , davis , s. , fahlman , g. g. , hansen , b. m. s. , kalirai , j. , paust , n. , rich , r. m. , & shara , m. m. 2007 , accepted ; http://arxiv.org/0708.4030 sarajedini , a. , bedin , l. r. , chaboyer , b. , dotter , a. , siegel , m. , anderson , j. , aparicio , a. , king , i. , majewski , s. , mar ' in - franch , a. , piotto , g. , reid , i. n. , & rosenberg , a. 2007 , , 133 , 1658 schlegel , d. j. , finkbeiner , d. p. , & davis , m. 1998 , , 500 , 525 siegel , m. h. , dotter , a. , majewski , s. r. , sarajedini , a. , chaboyer , b. , nidever , d. l. , anderson , j. , marn - franch , a. , rosenberg , a. , bedin , l. r. , aparicio , a. , king , i. , piotto , g. , reid , i. n. 2007 , , 667 , 57 sirianni , m. et al . 2005 , , 117 , 1049 spergel , d. n. , et al . 2003 , , 148 , 175 stetson , p. b. , bruntt , h. , & grundahl , f. 2003 , , 115 , 413 taylor , b. j. 2007 , , 133 , 370 thoul , a. a. , bahcall , j. n. , & loeb , a. 1994 , , 421 , 828 vandenberg , d. a. & bell , r. a. 1985 , , 58 , 561 vandenberg , d. a. , bergbusch , p. a. , & dowler , p. d. 2006 , , 162 , 375 vandenberg , d. a. & clem , j. l. 2003 , , 126 , 778 ( vc03 ) vandenberg , d. a. , edvardsson , b. , eriksson , k. , & gustafsson , b. 2007 , http://arxiv.org/0708.1188 vandenberg , d. a. & stetson , p. b. 2004 , , 116 , 997 yi , s. k. , demarque , p. , & kim , y .- c . 2004 , , 291 , 261 yi , s. k. , kim , y .- c . , demarque , p. 2003 , , 144 , 259

the drell - yan process @xcite , the production of a lepton pair in hadron - hadron collisions , has played an important role in establishing the parton picture underlying the description of hard interactions in qcd . in current experiments , studies of the drell - yan cross section as a function of the invariant mass of the lepton pair are used to search for new heavy particles such as a hypothetical @xmath5 boson , while the differential distributions provide detailed information about the parton distribution functions ( pdfs ) , including in particular the sea - quark distributions . drell - yan production also serves as a prototype for other collider processes , such as higgs production or the production of new particles . a lot of effort has been put in obtaining accurate theoretical predictions for the drell - yan process in perturbative qcd . the calculation of the cross section and rapidity distribution at next - to - leading order ( nlo ) in @xmath6 was accomplished in the pioneering work @xcite . the first next - to - next - to - leading order ( nnlo ) result for the cross section was presented in @xcite and confirmed much later in @xcite . the rapidity distribution at nnlo was derived in @xcite , while the fully differential cross section was obtained recently in @xcite . as the invariant mass @xmath7 of the lepton pair approaches the center - of - mass energy of the collision , there is less and less phase space available for the emission of qcd radiation . after the cancellation of virtual and real soft divergences large sudakov logarithms remain , because the scale associated with the soft radiation is much smaller than @xmath7 . these threshold logarithms " threaten the convergence of the perturbative expansion and need to be resummed to all orders . for the inclusive cross section @xmath8 this was accomplished in the seminal papers @xcite based on the solution of certain evolution equations in mellin moment space ( see also @xcite ) . the generalization of this method to the drell - yan rapidity distribution was obtained in @xcite . recent analyses of threshold resummation for the rapidity distribution in drell - yan or electroweak gauge boson production can be found in @xcite in next - to - leading logarithmic ( nll ) approximation , while in @xcite the resummation is extended to the next - to - next - to - next - to - leading logarithmic ( n@xmath9ll ) order . because the pdfs are strongly suppressed in the endpoint region @xmath10 , the cross section @xmath8 is a steeply falling function as @xmath7 approaches the kinematical endpoint @xmath11 . in fact , in a typical experiment it will not be possible to observe drell - yan pairs with masses exceeding about one half of the center - of - mass energy . for instance , at the lhc one does not expect to discover new heavy particles with masses in the 10tev range . in practice , one is therefore never in a region where the ratio @xmath12 approaches 1 . since threshold resummation deals with logarithms of the form @xmath13 , it is then not obvious why such terms should be treated on different footing than other higher - order terms . in view of this , it is surprising that large resummation effects were recently claimed to be important for the drell - yan rapidity distribution as measured by the e866/nusea collaboration @xcite . specifically , ref . @xcite claims that for @xmath14gev and @xmath15gev ( corresponding to @xmath16 ) the resummation of threshold logarithms at nnll order would lower the cross section by about 30% compared with the fixed - order nlo result , whereas the fixed - order nnlo corrections increase it by a small amount @xcite . however , significantly smaller resummation effects were found by other authors for lhc kinematics ( @xmath17gev and @xmath18tev ) @xcite and for @xmath19-boson production at rhic ( @xmath20gev and @xmath21gev ) @xcite . an argument why threshold resummation effects could be important even if @xmath22 has been given in refs . the idea is that the sharp fall - off of the parton luminosity at large @xmath0 dynamically enhances the contribution of the _ partonic _ threshold region @xmath23 , i.e. , the region where the center - of - mass energy @xmath24 of the initial - state partons is just sufficiently large to produce the drell - yan pair . it could then be important to resum logarithms of the form @xmath25 in the hard partonic cross section . however , since @xmath26 is not related to a small ratio of external physical scales , it is not obvious how to give a formal justification of this argument . to study this question quantitatively and to assess the importance of resummation effects for the drell - yan rapidity distribution were the main motivations for the present work . to do so , we use a recent approach to sudakov resummation based on effective field theory @xcite . contrary to the standard treatment in mellin moment space , this framework completely separates the effects associated with different scales in the problem , thereby avoiding the landau - pole ambiguities inherent in the standard approach . it then uses renormalization - group ( rg ) evolution to resum logarithms of scale ratios . the resummation is performed directly in momentum space , which makes it simpler to compare to and match onto fixed - order calculations . our framework is particularly well suited to study the resummed rapidity distribution , for which we derive an exact analytic expression as a one - dimensional integral over pdfs . using the convergence properties of the perturbative expansion after scale separation as the primary criterion , we study in detail how and under which circumstances an effective physical scale @xmath27 emerges , which is associated with the soft emission in the process . our approach resums logarithms of the ratio @xmath28 to all orders in perturbation theory . our main findings can be summarized as follows : * in the true endpoint region @xmath29 , the effective soft scale @xmath30 is an order of magnitude smaller than the naive guess @xmath31 . for pdfs behaving like @xmath32 near @xmath10 , we find @xmath33 with @xmath34 . this result provides a formal justification to the argument of a dynamical enhancement of the partonic threshold region due to the fall - off of parton densities . * the dynamical enhancement of the threshold contributions remains effective down to moderate values @xmath35 , while at very small @xmath36 values the parameter @xmath37 decreases to about 2 . this reflects the fact that for small @xmath0 values the fall - off of the pdfs is much weaker than for large @xmath0 . * even far away from the true threshold the drell - yan cross section receives its dominant contributions from those terms in the hard partonic cross section that are leading in the limit @xmath38 . assuming this is true for other processes as well , the evaluation of virtual corrections plus soft emissions provides a simple and efficient way to obtain useful approximations for higher - order perturbative corrections . * with the appropriate choice of the effective soft scale @xmath30 , the convergence of the perturbative expansion is greatly improved by the resummation . however , for small drell - yan masses the terms beyond @xmath39 in the resummed expression for the cross section are numerically unimportant . we thus do not confirm the large impact of threshold resummation on the drell - yan rapidity distribution reported in @xcite . for larger masses the effects can be significant . for instance , the experiment e866/nusea has reported data up to @xmath40gev ( corresponding to @xmath41 ) @xcite . we find that at @xmath42gev resummation effects enhance the fixed - order predictions for the cross section by about 25% at nlo , and 7% at nnlo . * for the case of the integrated cross - section @xmath8 , we perform a detailed comparison with the traditional resummation approach in moment space . similar to the case of deep - inelastic scattering ( dis ) in the region @xmath10 , we find that the two approaches are equivalent up to power corrections , which turn out to be numerically small . an important conceptual difference is that in the effective - theory approach the running coupling is evaluated at physical short - distance scales depending only on the external variables @xmath43 and @xmath7 ( and perhaps the rapidity @xmath44 ) . in this way , the landau - pole ambiguities inherent in the standard approach are avoided . we begin our analysis discussing the structure of the hard - scattering kernels relevant for the drell - yan rapidity distribution in fixed - order perturbative qcd . in section [ sec : scet ] we use the framework of soft - collinear effective theory ( scet ) @xcite to derive the standard factorization formula for the partonic cross section in the limit @xmath38 in terms of hard and soft functions , which we define in terms of wilson coefficients of operators in the effective theory . the solutions to the rg equations obeyed by these coefficients are derived in section [ sec : resummation ] . with these results at hand , we present exact analytic expressions for the resummed drell - yan cross section and rapidity distribution . a detailed numerical analysis of our results is presented in section [ sec : numerics ] . after choosing the hard and soft matching scales in the effective theory by analyzing the perturbative expansions of the wilson coefficient functions , we investigate the stability of the results under scale variations and discuss the impact of the resummation . before concluding , we discuss the connection with the conventional moment - space approach we consider the production of a lepton pair with invariant mass @xmath7 in hadron - hadron collisions at center - of - mass energy @xmath45 ( drell - yan process ) , focusing for simplicity on the reaction @xmath46 followed by @xmath47 . our goal is to calculate the double differential cross section in the variables @xmath48 and @xmath49 , where @xmath44 denotes the rapidity of the lepton pair in the center - of - mass frame . up to power corrections this cross section can be calculated in perturbative qcd and expressed in terms of convolutions of short - distance partonic cross sections with pdfs : @xmath50 here @xmath51 is the probability of finding a parton @xmath52 with longitudinal momentum fraction @xmath0 inside the hadron @xmath53 , and @xmath54 is the factorization scale . the hard - scattering kernels @xmath55 have an expansion in powers of the strong coupling @xmath6 . the sum extends over all possible partonic channels contributing at a given order in this expansion . at leading order ( @xmath56 ) only the channels @xmath57 contribute , while at nlo ( @xmath58 ) one must include @xmath59 in the sum . it will be useful for our purposes to introduce the ratios @xmath60 where @xmath61 is the center - of - mass energy squared of the partonic subprocess that creates the lepton pair . this determines the maximum energy transferred to the leptons and the maximum invariant mass @xmath7 they can have . in @xcite the coefficient functions @xmath55 are expressed in terms of the variable @xmath62 and a second quantity @xmath63 these variables take values on the intervals @xmath64 and @xmath65 subject to the condition that the parton momentum fractions @xmath66 do not exceed 1 . the allowed range for the rapidity is such that @xmath67 . we then define new kernels via @xmath68[1-(1-y)(1-z ) ] } \,.\ ] ] at nlo the explicit results for these functions can be written in the form ( with @xmath69 and @xmath70 denoting the electric charges of the quarks in units of @xmath71 ) @xcite @xmath72 \nonumber\\ & & \mbox{}+ \frac{c_f\alpha_s}{\pi}\,\bigg\ { \frac{\delta(y ) + \delta(1-y)}{2}\,\bigg [ ( 1+z^2 ) \left [ \frac{1}{1-z } \ln\frac{m^2(1-z)^2}{\muf^2 z } \right]_+ + 1 - z \bigg ] \nonumber\\ & & \quad \mbox{}+ \frac12 \left [ 1 + \frac{(1-z)^2}{z}\,y(1-y ) \right ] \left [ \frac{1+z^2}{1-z } \left ( \left [ \frac{1}{y } \right]_+ + \left [ \frac{1}{1-y } \right]_+ \right ) - 2(1-z ) \right ] \bigg\ } \ , , \nonumber\\ \frac{c_{qg}}{e_q^2 } & = & \frac{t_f\alpha_s}{2\pi}\,\bigg\ { \delta(y ) \left [ \left ( z^2 + ( 1-z)^2 \right ) \ln\frac{m^2(1-z)^2}{\muf^2 z } + 2z(1-z ) \right ] \nonumber\\ & & \quad \mbox{}+ \left [ 1 + \frac{(1-z)^2}{z}\,y(1-y ) \right ] \left [ \left ( z^2 + ( 1-z)^2 \right ) \left [ \frac{1}{y } \right]_+ + 2z(1-z ) + ( 1-z)^2 y \right ] \bigg\ } \ , . \quad\end{aligned}\ ] ] the @xmath54-dependent terms can be derived from the fact that the cross section in ( [ sig ] ) is scale independent , while the pdfs obey the dglap evolution equations @xcite . the remaining functions follow from the symmetry relations @xmath73 the hard - scattering kernels at nnlo have been calculated in @xcite and are available in the form of a computer program @xcite . the explicit expressions for the coefficient functions given above contain terms that are singular in the partonic threshold region " @xmath38 , in which the center - of - mass energy of the parton subprocess is just large enough to create a lepton pair with invariant mass @xmath7 . indeed , the arguments of the logarithms in ( [ cijres ] ) suggest the relevance of two mass scales : a hard " scale @xmath74 , and a soft " scale @xmath75 . physically , the hard scale is set by the invariant mass of the lepton pair , while the soft scale is of the order of the energy of the remnant jet @xmath76 produced in the collision . in the region of parton kinematics where @xmath38 these scales are separated , @xmath77 , in which case the coefficient functions contain large logarithms irrespective of the choice of the factorization scale @xmath54 . threshold resummation for the drell - yan cross section @xcite aims at resumming these logarithms to all orders in perturbation theory . let us return to the structure of the relations ( [ cijres ] ) and identify the leading singular terms in the partonic threshold region . they are contained in @xmath78 and up to nlo multiply @xmath79-functions in the variable @xmath80 . beyond nlo some of the leading singular terms in the expressions obtained in @xcite multiply nontrivial functions of @xmath80 , but since the @xmath80-dependence of the parton variables @xmath81 and @xmath82 in ( [ xi ] ) is subleading in the @xmath38 limit one can always rearrange the expressions in such a way that the leading singular terms multiply @xmath79-functions in @xmath80 . explicitly , we then obtain @xmath83 where @xmath84_+ \right\ } \nonumber\\ & & \mbox{}+ c_f \left ( \frac{\alpha_s}{\pi } \right)^2 \big [ c_f p_f(z ) + c_a p_a(z ) + t_f n_f p_f(z ) \big ] \,,\end{aligned}\ ] ] and we have defined @xmath85 and @xmath86 $ ] . the terms in the first line can be readily read off from ( [ cijres ] ) . the two - loop coefficients @xmath87 are given in appendix [ app : a ] . note that the factor @xmath62 in the argument of the logarithm @xmath88 could be set to 1 at leading order , but it is correctly reproduced by our resummation formula below and so we will keep it . the goal of this paper is to derive a formalism that resums these leading terms to all orders in perturbation theory . upon performing the integration over @xmath80 , the leading singular terms in ( [ llterms ] ) give rise to the following contribution to the cross section : @xmath89 . \nonumber\end{aligned}\ ] ] the lower limit of the @xmath62 integral is @xmath90 , as appropriate for the two terms . at tree level , we recover the parton - model result @xmath91 .\ ] ] we also note that integrating over rapidity , we obtain for the leading singular terms in the single - differential cross section @xmath92 , \ ] ] where @xmath93 , and the integration is restricted to the region where @xmath94 . this result can be rewritten in the more convenient form @xmath95 where @xmath96\ ] ] denotes the mellin convolution of the pdfs . in the following section we will derive a factorization formula for the coefficient @xmath97 using methods of effective field theory . the result is @xmath98 where @xmath99 and @xmath100 will be referred to as hard and soft functions , respectively , and will be defined in terms of wilson coefficients of operators in scet . the calculation of the components @xmath99 and @xmath100 at any order in perturbation theory is much simpler than the calculation of the drell - yan cross section at the same order . ( [ factform ] ) thus provides an approximation to the cross section that requires a minimal amount of calculational work . the all - order resummation of the partonic threshold logarithms is then achieved by solving rg equations . it must be emphasized at this point that the variable @xmath62 is not set by external kinematics , but instead is integrated over the interval between @xmath12 and 1 . it is therefore necessary to specify under which conditions the partonic threshold region requires special attention . for instance , also in inclusive processes such as @xmath101 there are kinematical situations where scales much smaller than the center - of - mass energy are important , such as the emission of a soft gluon into the final state . it is a well - known fact , however , that upon integration over the entire phase space the perturbation series is insensitive to scales lower than @xmath11 at leading power . there are two limits in which the threshold contributions are parametrically enhanced . first , if the invariant mass of the lepton pair is near the kinematic limit set by the total center - of - mass energy of the hadron - hadron collision , then @xmath102 and hence @xmath103 is always near 1 . threshold resummation is necessary in this case and proceeds in close analogy to the resummation of threshold logarithms for the dis structure functions for @xmath10 @xcite . in practice , the region @xmath102 is irrelevant for phenomenology , since the strong suppression of the pdfs near the endpoint implies a very low parton luminosity in this case . [ ] [ ] @xmath80 [ b]@xmath104 fall - off of the parton luminosity function @xmath104 for @xmath105gev . the dashed lines show the asymptotic behavior for small and large @xmath80.,title="fig:",scaledwidth=50.0% ] a second way in which the threshold contributions can be enhanced arises dynamically , if the weight function multiplying the hard - scattering kernel under the @xmath62-integral is steeply falling with @xmath26 @xcite . in this case threshold resummation can be justified even if @xmath36 is much less than 1 . in practice , such a behavior has to result from the fall - off of the parton densities with increasing @xmath0 . consider for simplicity the total cross section @xmath8 , for which the relevant combination of pdfs is given by the function @xmath104 in ( [ ffdef ] ) . figure [ fig : ffdrop ] shows that this function is indeed very steeply falling with @xmath80 . taking @xmath105gev , one finds that @xmath106 for @xmath107 and @xmath108 for @xmath109 , where @xmath110 and @xmath111 . the figure shows that the first form reasonably well describes the behavior for @xmath112 , while the second form holds for @xmath113 . using these asymptotic forms for the parton luminosity function , we find that for @xmath114 @xmath115 while for @xmath116 @xmath117 in the first case the cross section is given by a low - order moment of the hard - scattering kernel @xmath118 , in which case the partonic threshold region @xmath38 is not parametrically enhanced . in the second case , on the other hand , in the limit where we treat the exponent @xmath119 as a large parameter , the @xmath62 integral receives important contributions only from the region where @xmath120 . even for @xmath36 values not near 1 there is thus a parametric enhancement of the partonic threshold region , which turns the threshold logarithms into logarithms of the exponent @xmath119 . the intermediate range @xmath121 is a transition region , in which the dynamical enhancement of the threshold region ceases to be effective as @xmath36 is lowered . cc [ ] [ ] @xmath44 [ b]@xmath122[pb / gev ] comparison of the complete fixed - order results ( solid lines ) and the contributions arising from the leading singular terms ( dashed lines ) to the drell - yan rapidity distribution at different orders in perturbation theory . on the left we use the pdf sets mrst01lo , mrst04nlo , and mrst04nnlo , as appropriate to the order of the calculation ; in the right plot mrst04nnlo is used throughout.,title="fig:",scaledwidth=45.0% ] & [ ] @xmath44 [ b]@xmath122[pb / gev ] [ ] @xmath123 q & \!=\ ! & 38.76\,\mbox{gev } \nonumber\\[-0.25 cm ] m & \!=\ ! & 8\,\mbox{gev } \nonumber\\[-0.25 cm ] \muf & \!=\ ! & m \nonumber \end{aligned}\ ] ] comparison of the complete fixed - order results ( solid lines ) and the contributions arising from the leading singular terms ( dashed lines ) to the drell - yan rapidity distribution at different orders in perturbation theory . on the left we use the pdf sets mrst01lo , mrst04nlo , and mrst04nnlo , as appropriate to the order of the calculation ; in the right plot mrst04nnlo is used throughout.,scaledwidth=45.0% ] while , as we have just discussed , the dominance of the partonic threshold region can not be justified parametrically for @xmath114 , it nevertheless appears that even in this case the drell - yan cross section and rapidity distribution are well approximated by keeping only the leading singular terms in the hard - scattering kernel ( [ cdecomp ] ) . the reason is an inherent property of the hard - scattering kernel , which appears to receive the largest radiative corrections from the region of phase space corresponding to born kinematics . in other words , the effects of hard real emissions appear to be suppressed compared with virtual corrections and soft emissions . to illustrate this point , we show in figure [ fig : fixedvsresummed ] the full fixed - order predictions for the drell - yan rapidity distribution at @xmath15gev and @xmath14gev and compare them to the results obtained by keeping only the threshold terms in the hard - scattering kernel . we use the mrst sets of pdfs compiled in @xcite . throughout our work we use the three - loop running coupling @xmath124 normalized to @xmath125 , and we take @xmath126 for the number of light quark flavors . in the right plot we use the same nnlo parton densities for all curves , so that the size of the different perturbative corrections to the hard - scattering kernel can be read off directly . since our main focus in this work is on the behavior of the perturbative expansion of the hard - scattering kernel , we will from now on always use the set mrst04nnlo for the pdfs . even though @xmath16 is very small in this example , we observe that at central rapidity about 80% of the nlo correction and 63% of the nnlo correction arise from the leading singular terms . for the total cross section @xmath8 about 93% of the nlo correction and 97% of the nnlo correction are accounted for by the leading singular terms . if we lower the factorization scale to @xmath127 , then the leading singular terms come even closer to reproducing the fixed - order results . the factorization theorem ( [ factform ] ) for the leading singular terms in the hard - scattering kernel @xmath118 has been established a long time ago @xcite . in particular , it has been understood that the soft function @xmath100 can be represented as the vacuum expectation value of a certain wilson loop of soft gluon fields . nevertheless we find it useful to rederive this formula using methods of effective field theory . the advantage of this approach is that we will relate the hard and soft functions , @xmath99 and @xmath100 , to wilson coefficients of operators in the effective theory , which obey certain rg equations . solving these equations we accomplish threshold resummation directly in momentum space . in the framework of scet the standard factorization formula ( [ sig ] ) for the drell - yan cross section has been discussed in @xcite . however , the issue of glauber gluons @xcite has not yet been addressed in the effective theory . glauber - gluon interactions have presented an important complication in the factorization analysis of the drell - yan process @xcite , and it would be worthwhile to rederive their cancellation using scet . as discussed below , glauber gluons do not affect the factorization of the hard - scattering kernel @xmath128 in the threshold region . the threshold resummation for the total cross section @xmath8 has been performed in @xcite using an effective - theory variant of the traditional mellin moment - space approach . in the present paper we go beyond these works by presenting a derivation of the factorization formula ( [ factform ] ) in scet and performing the threshold resummation directly in @xmath62 space . this will provide us with the tools to discuss the relevance of resummation away from the true threshold , i.e. , for @xmath129 much smaller than @xmath43 . we begin with the standard formula for the drell - yan cross section , @xmath130 where @xmath131 is the electromagnetic current . to derive the factorization theorem , we match the product of currents onto operators in scet . the matching proceeds in two steps , and the corresponding wilson coefficients are the hard function @xmath99 and the soft function @xmath100 , respectively . the remaining effective - theory matrix element can be identified with the pdfs . a similar two - step matching procedure has been used in many scet applications . in particular , the factorization theorem for dis in the region @xmath10 was derived in scet by proceeding this way @xcite . in the present paper we are solely interested in the factorization of the hard - scattering coefficient @xmath128 . for this purpose , we can simplify the effective - theory analysis by considering ( [ textbook ] ) with partonic instead of hadronic matrix elements . these partonic matrix elements correspond directly to the hard - scattering coefficients , and the hadronic cross sections are then obtained after convoluting the results with the pdfs . an effective - theory analysis of the hadronic matrix elements would be more complicated because of their sensitivity to nonperturbative physics governed by the scale @xmath132 , which does not enter the hard - scattering kernels . note that the product of current operators in the matrix element in ( [ textbook ] ) is not time - ordered . this is in contrast to the case of dis , where the decay rate can be obtained from the discontinuity of a time - ordered product of currents . it is not possible to rewrite ( [ textbook ] ) in the same way , because the imaginary part of the time - ordered product also gets contributions from virtual corrections to @xmath133 , where @xmath76 is a purely hadronic final state . a path integral framework to analyze operator products that are not time ordered is the keldysh formalism @xcite , which we review in appendix [ sec : keldysh ] . in our discussion below , the operator ordering does not lead to any complications , but it implies that the soft function is not given by feynman diagrams , but appropriate cuts of such diagrams , as explained in @xcite . before entering the technicalities of the discussion of factorization in scet , it is useful to understand the physics behind the factorization theorem ( [ factform ] ) by means of a simple , intuitive argument . consider the special kinematics of drell - yan production in the ( partonic ) threshold region , as illustrated in figure [ fig : cartoon ] . because the partonic center - of - mass energy is just above the invariant mass of the produced drell - yan pair , only soft emissions from the initial - state partons are allowed . as is well known , these soft emissions are described by eikonal interactions and exponentiate into wilson lines . furthermore , to leading power they leave the incoming partons on the mass shell , so that the production of the drell - yan pair is described by an on - shell quark form factor . [ b]@xmath134 [ b]@xmath135 [ b]@xmath136 soft gluon emissions from the initial - state partons in drell - yan production.,title="fig:",scaledwidth=32.0% ] we will now formalize this argument using the language of effective field theory . the effective theory is constructed by introducing fields for the momentum regions that contribute to the matrix elements in the given kinematics . as stressed above , for our purposes it is sufficient to analyze partonic matrix elements . to distinguish the different momentum modes , we introduce the light - cone decomposition @xmath137 where @xmath138 and @xmath139 are two light - like reference vectors ( @xmath140 ) in the directions of the colliding partons . we denote the small expansion parameter by @xmath141 and quote the components @xmath142 of parton momenta . the relevant momentum regions are @xmath143 where the large momentum scale in the process is set by the partonic center - of - mass energy @xmath24 . in the first matching step , the contributions from the hard region are absorbed into wilson coefficients , while the remaining contributions are represented by fields in the effective theory . the version of the effective theory with fields scaling in the above way is often called scet@xmath144 . note that , while in many applications the soft scale is associated with nonperturbative physics , in our case @xmath145 is of order the mass of the hadronic final - state jet and assumed to be in the perturbative domain . as explained above , the kinematics of the drell - yan process near threshold is such that the hadronic final state is made up of soft partons ; the virtuality of the hard - collinear partons is parametrically larger than the jet mass : @xmath146 . the potential contribution from glauber gluons @xcite presents a challenge when deriving the standard factorization theorem of the drell - yan cross section into pdfs convoluted with hard partonic cross sections @xcite . however , the glauber region does not present an additional difficulty for the factorization of the hard - scattering coefficient @xmath147 in the threshold region , which we will derive below . the hard - scattering coefficient is given by the on - shell @xmath148 cross section , so that for its factorization only glauber exchanges involving active quarks need to be studied . the momenta of the glauber gluons relevant for the threshold region scale as @xmath149 , and these gluons are thus not allowed in the hadronic final state , which is assumed to have @xmath150 . writing down the loop integrals for glauber exchanges between the two inital - state quarks , one finds that they are scaleless and vanish in dimensional regularization . in more physical terms , this means that the glauber contribution is accounted for already by the standard momentum regions introduced above . in order not to clutter the notation , we consider for the moment a single quark flavor with charge @xmath151 . we thus match the current @xmath152 onto an effective current operator in scet containing a hard - collinear quark and an anti - hard - collinear anti - quark @xcite . at leading power in @xmath153 , this yields @xmath154 where @xmath155 and @xmath156 are the usual hard - collinear ( along the @xmath157-direction ) and anti - hard - collinear ( along the @xmath158-direction ) wilson lines of scet . note that the matching relation for the ( transverse ) vector current does not include a two - gluon operator at leading power in the effective theory . such an operator would arise , however , in the corresponding matching relation for a scalar current relevant for higgs production . mixed quark - gluon operators corresponding to the @xmath159 , @xmath160 , @xmath161 , and @xmath162 scattering channels are power suppressed . accordingly , the corresponding hard - scattering kernels @xmath163 do not contain terms that are singular in the limit @xmath38 , as is explicitly seen in ( [ cijres ] ) . the fourier transform of the position - space wilson coefficient @xmath164 is a function of the product of the large light - cone momentum components carried by the quark fields . in our case , @xmath165 is equal to the hard scale set by the mass of the drell - yan pair . as a result the wilson coefficient is evaluated at the time - like momentum transfer carried by the current , @xmath166 . the @xmath167 prescription is required since this function has a branch cut along the positive real @xmath168 axis . note that the same wilson coefficient @xmath169 appears in dis , but evaluated at space - like momentum transfer @xcite . the coefficient @xmath169 can be determined by on - shell matching ; indeed , it is simply given by the on - shell massless form factor in qcd @xcite . the drell - yan cross section involves the current squared , so that the hard function in ( [ factform ] ) is given by @xmath170 the expression for the coefficient @xmath169 up to @xmath39 was derived in @xcite and can be found in appendix [ appendix : rgfunctions ] . at leading power in @xmath153 , only the @xmath171 component of the soft gluon field couples to the hard - collinear fields . these eikonal interactions can be represented by wilson lines . in the effective theory this is achieved by redefining the hard - collinear fields as @xcite @xmath172 which implies @xmath173 . here @xmath174 is a soft wilson line along the @xmath158 light - cone . the same redefinition , but with @xmath158 and @xmath157 interchanged , decouples the soft gluon field also from the anti - hard - collinear fields . as a result , the current operator @xmath175 splits into three parts , which no longer interact with each other . in the same way , the matrix element for the drell - yan process factorizes in the form @xmath176 to obtain this expression , we have fierz rearranged the fermion fields and have averaged over the color of the external states . we have simplified the dirac algebra making use of the projection properties @xmath177 and @xmath178 of the hard - collinear fermion fields ( and likewise @xmath179 and @xmath180 for the anti - hard - collinear fields ) . also , we have neglected the power - suppressed dependence of the hard - collinear matrix element on @xmath181 and @xmath182 ( @xmath183 and @xmath182 for the anti - hard - collinear matrix element ) , using the fact that up to power corrections the incoming partons fly along the beam axis . in more technical terms , we have multi - pole expanded the corresponding fields to leading power @xcite . the soft matrix element @xmath184 ( not to be confused with the hard - collinear wilson lines ) is a closed wilson loop formed from the product of the soft wilson lines in the two currents , @xmath185\ , { \bf t } \big [ s_{\bar n}^\dagger(0 ) s_n(0 ) \big]|0 \rangle \,,\ ] ] where the trace is over color indices , and the operator @xmath186 makes explicit that the wilson lines in the complex conjugate current @xmath187 are anti - time - ordered . a detailed discussion of the operator ordering is given in appendix [ sec : keldysh ] . it will become evident in section [ sec : kin ] that the function @xmath184 is closely related to the soft function for the drell - yan process . in the second matching step we lower the renormalization scale below the soft scale and integrate out the soft fields . because the soft scale is in the short - distance domain , this simply amounts to a perturbative calculation of the wilson loop ( [ softmatrixelement ] ) . to make contact with the standard treatment , where the factorization theorem is derived within perturbative qcd , we note that after the decoupling transformation ( [ decoupl ] ) the leading - power lagrangian in each of the three sectors of the effective theory is completely equivalent to the qcd lagrangian . we can thus equally well evaluate the soft function with qcd wilson lines instead of soft wilson lines . the same is true for the hard - collinear matrix elements , where one can replace @xmath188\ , \psi(0)|n_1\rangle \,,\ ] ] where @xmath189 is the qcd quark field and @xmath190 $ ] a qcd wilson line of @xmath191 gluons extending from @xmath192 to @xmath183 . we have simplified the effective - theory treatment by restricting ourselves to on - shell partonic instead of hadronic matrix elements , which has the advantage that we do not need to discuss momentum regions that scale with powers of @xmath132 . , a full factorization analysis ( including messenger fields @xcite ) was performed in @xcite . ] these momentum regions can not enter the hard or the soft functions , and given that the drell - yan cross section is known to factorize for @xmath193 , we are guaranteed that their presence will not spoil factorization . the restriction to partonic matrix elements can thus safely be dropped . the hadronic matrix elements of the form ( [ collinearmatrixelemt ] ) are the usual pdfs @xcite @xmath194\,\psi(0)\,|n(p)\rangle \,.\ ] ] inserting the factorized form of the matrix element ( [ eq : factor ] ) into expression ( [ textbook ] ) for the cross section , we obtain @xmath195 \nonumber\\ & & \times \int\frac{d^3\vec{q}}{(2\pi)^3 2q^0}\ , \frac{1}{2\pi } \int d^4x\ , e^{i(x_1 p_{1-}+x_2 p_{2+}-q)\cdot x}\,\hat w_{\rm dy}(x ) \,.\end{aligned}\ ] ] from now on we will drop the @xmath167 prescription in the argument of @xmath169 . to proceed , we need to study the kinematics in the partonic threshold region , where @xmath38 . with the expansion parameter @xmath141 , it is easy to show that in the parton center - of - mass system @xmath196 it follows that @xmath197 is parametrically larger than the spatial components of @xmath198 . this observation implies that the term @xmath197 in the denominator in ( [ sigma1 ] ) is independent of @xmath199 at leading power , in which case performing the integral over @xmath200 yields @xmath201 . it also follows that the rapidity of the drell - yan pair in the parton center - of - mass system vanishes at leading power , @xmath202 which explains why the rapidity in the laboratory frame , @xmath203 , is determined by the ratio @xmath204 at leading power in @xmath26 , as can be seen from ( [ xi ] ) . finally , we need that @xmath205 using these results , we obtain at leading power in the @xmath38 limit the expression @xmath206 \nonumber\\ & & \times \frac{1}{2\sqrt{\hat s } } \int\frac{dx^0}{2\pi}\,e^{i\sqrt{\hat s}(1-z)x^0/2}\ , \hat w_{\rm dy}(x^0,\vec{x}=0,\muf ) \,,\end{aligned}\ ] ] where the integration region over the parton momentum fractions is such that @xmath94 , so that @xmath207 . in the final step we introduce the fourier transform of the position - space wilson loop at time - like separation via @xmath208 this wilson loop plays the role of the jet function in dis in the sense that it describes the properties of the hadronic final state . it has been introduced previously in @xcite . the drell - yan cross section now takes the form @xmath209 \nonumber\\ & & \times \sqrt{\hat{s}}\,w_{\rm dy}(\sqrt{\hat s}\,(1-z),\muf ) \,,\end{aligned}\ ] ] and from comparison with ( [ dsigdm2 ] ) and ( [ factform ] ) we identify the soft function as @xmath210 we should mention that at leading power in @xmath26 the argument of the soft function could be simplified as @xmath211 ; however , since the exact expressions ( [ cijres ] ) for the hard - scattering kernels at nlo contain the logarithm @xmath212 , we prefer to keep the argument of the soft function in the form written above . at this point we have identified the two components in the factorized expression ( [ factform ] ) for the hard - scattering coefficient @xmath128 with field - theoretic objects defined in terms of operator matrix elements . the resummation of threshold logarithms arising in the @xmath38 region can now be accomplished by solving the rg evolution equations obeyed by these quantities . the evolution equation for the hard matching coefficient @xmath169 evaluated at time - like momentum transfer and its solution can be obtained from the corresponding results valid for space - like momentum transfer @xcite by analytic continuation . this leads to @xmath213 c_v(-m^2-i\epsilon,\mu ) \,.\ ] ] we have reinserted the @xmath167 regulator , which determines the sign of the imaginary part of the anomalous dimension . the appearance of the logarithm and its coefficient , the cusp anomalous dimension @xmath214 @xcite , can be explained using arguments presented in @xcite . this term in the evolution equation is associated with sudakov double logarithms . the remaining term , @xmath215 , accounts for single - logarithmic evolution . the exact solution to ( [ gammav ] ) is @xmath216 \left ( \frac{m^2}{\muh^2 } \right)^{-a_\gamma(\muh,\muf)}\ , c_v(-m^2,\muh ) \,,\ ] ] where @xmath74 is a hard matching scale , at which the value of @xmath169 is calculated using fixed - order perturbation theory . note that the wilson coefficient at time - like momentum transfer is a complex quantity . the sudakov exponent @xmath100 and the exponents @xmath217 are given by @xcite @xmath218 and similarly for the function @xmath219 . the perturbative expansions of the anomalous dimensions and the resulting expressions for the evolution functions valid at nnlo in rg - improved perturbation theory are collected in appendix [ appendix : rgfunctions ] . in order to derive the evolution equation for the soft function it is important to have a consistent definition of the threshold region . to this end we consider the limit @xmath220 in ( [ sigma3 ] ) , which implies @xmath38 , since @xmath103 . the condition @xmath94 then implies that @xmath221 , in which case the dglap evolution for the pdfs can be written in the simplified form @xmath222_+ } + \dots \,,\ ] ] which is obtained by expanding the altarelli - parisi splitting function as @xmath223_+ } + 2\gamma^\phi(\alpha_s)\,\delta(1-z ) + \dots \,.\ ] ] this asymptotic form of the splitting function holds to all orders in perturbation theory @xcite . the cusp anomalous dimension and the coefficient @xmath224 have been calculated at three - loop order @xcite . they are given in appendix [ appendix : rgfunctions ] . requiring that the drell - yan cross section in the threshold region be independent of the arbitrary factorization scale @xmath54 , we find that the momentum - space wilson loop obeys the integro - differential evolution equation @xmath225 w_{\rm dy}(\omega,\mu ) \nonumber\\ & & \mbox{}- 4\gamma_{\rm cusp}(\alpha_s ) \int_0^{\omega}\!d\omega'\ , \frac{w_{\rm dy}(\omega',\mu)-w_{\rm dy}(\omega,\mu ) } { \omega-\omega ' } \,,\end{aligned}\ ] ] where @xmath226 curiously , this quantity starts only at two - loop order . the analogous relation in the case of dis is @xmath227 , where the anomalous dimension @xmath228 of the dis jet function is defined in analogy with @xmath229 in ( [ jrge ] ) . combining the two relations we obtain @xmath230 , which is an exact relation that links the anomalous dimensions of the soft function in drell - yan production and the dis jet function to the anomalous dimension of the vector current . the exact solution to the evolution equation ( [ jrge ] ) can be written in the form @xcite @xmath231 \widetilde s_{\rm dy}(\partial_\eta,\mui)\ , \frac{1}{\omega } \left ( \frac{\omega}{\mui } \right)^{2\eta}\ , \frac{e^{-2\gamma_e\eta}}{\gamma(2\eta ) } \,,\ ] ] where @xmath232 denotes a derivative with respect to an auxiliary parameter @xmath233 , which is then identified with @xmath234 . this result is well defined for @xmath235 . the solution for negative @xmath233 is obtained by analytic continuation . for instance , to obtain the result for @xmath236 we use the identity @xmath237 where @xmath238 is a smooth test function . for @xmath239 additional subtractions are required . the function @xmath240 is obtained from the momentum - space wilson loop by the laplace transformation @xcite @xmath241 the position - space wilson loop on the right - hand side of ( [ wdydef ] ) can be shown to have the functional form @xmath242 @xcite , and a straightforward calculation shows that the function @xmath243 can be expressed in terms of @xmath244 as @xmath245 we can then use the explicit two - loop expression for the position - space wilson loop obtained in @xcite to compute the function @xmath240 at two - loop order . the result is presented in appendix [ appendix : rgfunctions ] . we are now in a position to derive the rg - resummed expression for the hard - scattering coefficient @xmath128 in ( [ factform ] ) , which is given by the product of the solutions ( [ cvsol ] ) and ( [ sonice ] ) for the hard and soft functions . the result can be simplified by eliminating the anomalous dimension @xmath229 using ( [ gw ] ) , and combining the sudakov exponents using the relation @xmath246 we find @xmath247 where @xmath234 , and we have defined the evolution function @xmath248 .\ ] ] as before , equation ( [ cfinal ] ) is valid for @xmath235 ( @xmath249 ) . for negative @xmath233 ( @xmath250 ) , integrals of @xmath251 with test functions @xmath252 must be defined using a subtraction at @xmath253 and analytic continuation in @xmath233 . we emphasize that the result ( [ cfinal ] ) is formally independent of the scales @xmath254 and @xmath30 , at which the matching conditions for the hard and soft functions are evaluated . on the other hand , the hard - scattering kernel @xmath118 does depend on the factorization scale @xmath54 , at which the pdfs are renormalized . in practice , a residual dependence on the matching scales arises when the perturbative expansions of the matching coefficients and anomalous dimensions are truncated , and this dependence can be used to estimate the remaining perturbative uncertainties . setting the three scales @xmath254 , @xmath30 , and @xmath54 equal to each other in the resummed expression ( [ cfinal ] ) , one can readily reproduce the leading singular terms for @xmath38 in the fixed - order perturbative qcd expression for the hard - scattering kernel . in this way we have obtained the two - loop corrections in ( [ llterms ] ) . ccrccc rg - impr . pt & log . approx . & accuracy @xmath255 & @xmath214 & @xmath215 , @xmath224 & @xmath169 , @xmath240 + + & ll & @xmath256 & 1-loop & tree - level & tree - level + lo & nll & @xmath257 & 2-loop & 1-loop & tree - level + nlo & nnll & @xmath258 & 3-loop & 2-loop & 1-loop + nnlo & nnnll & @xmath259 & 4-loop & 3-loop & 2-loop + the final expression ( [ cfinal ] ) for the hard - scattering kernel can be evaluated at any desired order in resummed perturbation theory . table [ tab : counting ] shows what is required to obtain different levels of accuracy . in this work we adopt the counting scheme of rg - improved perturbation theory , where at lo one includes all @xmath260 terms , at nlo one includes all @xmath261 terms , etc . the large logarithm @xmath262 is counted like @xmath263 . in the literature on threshold resummation the alternative notation n@xmath264ll is often used instead of n@xmath265lo . the leading logarithmic ( ll ) approximation is listed only for completeness , as it misses some @xmath260 terms . in the following section we will perform a detailed numerical analysis of the drell - yan cross section and rapidity distribution . in most cases of phenomenological relevance the invariant mass of the drell - yan pair will be small compared with the center - of - mass energy , i.e. @xmath266 . nevertheless , it is interesting to briefly consider the limit @xmath29 , in which the need for threshold resummation is justified parametrically ( see the discussion in section [ sec : pqcd ] ) . in this case the convolution integrals in formula ( [ dsigdm2 ] ) for the drell - yan cross section can be performed analytically if a reasonably simple model for the pdfs near the endpoint is adopted . we parameterize the behavior near @xmath267 as @xmath268 \,,\ ] ] and similarly for the anti - quark distribution . it then follows that at leading power in @xmath269 the parton luminosity function defined in ( [ ffdef ] ) is given by @xmath270 furthermore , the sum over flavors is dominated by the terms with the smallest exponent @xmath271 . for example , in @xmath1 collisions at one finds @xmath272 , @xmath273 , and @xmath274 at @xmath275gev @xcite . these exponents increase by a flavor - independent amount @xmath276 when @xmath54 is raised to larger values , e.g. , by about 0.4 at @xmath105gev @xcite . it follows that the leading behavior near the endpoint is due to the down - quark contribution . introducing the drell - yan @xmath277-factor as the ratio @xmath278 we then obtain at leading power @xmath279 this may be compared with the @xmath277-factor for dis at large bjorken @xmath0 and momentum transfer @xmath280 , which has been derived in @xcite and reads @xmath281 \nonumber\\ & & \times ( 1-x)^\eta\,\ , \widetilde j_{\rm dis}\big ( \ln\frac{q^2(1-x)}{\mui^2 } + \partial_\eta,\mui \big)\ , \frac{e^{-\gamma_e\eta}\,\gamma(1+b_u)}{\gamma(1+b_u+\eta ) } \,,\end{aligned}\ ] ] where in this case the dominant contribution comes from the valence up - quark , which has the smallest @xmath282 parameter , @xmath283 at @xmath284gev @xcite . obviously the structure of threshold logarithms is very similar in the two cases once we consider equal hard scales ( @xmath285 ) and compare the small parameter @xmath286 in dis near the endpoint with the small parameter @xmath287 in drell - yan production . differences arise from the following facts : ( i ) drell - yan production has time - like kinematics , whereas dis probes the nucleon at space - like momentum transfer . this gives rise to a difference in the hard matching coefficients @xmath288 starting at one - loop order . ( ii ) the soft / jet functions in the two cases are different starting from one - loop order . ( iii ) the drell - yan cross section involves a convolution with two pdfs , whereas in dis a single parton density appears . this explains the different coefficients in front of @xmath289 and @xmath233 in the expressions for the @xmath277-factors . ( iv ) for the same reason , the resulting convolution integrals over the pdfs give rise to different expressions involving the @xmath282 and @xmath290 exponents . there is one more important piece of information that we can extract from the result ( [ kdy ] ) for the drell - yan @xmath277-factor . as we have seen , the exponents @xmath282 and @xmath290 take rather large values . therefore , the arguments of the @xmath291-functions in ( [ kdy ] ) contain the large quantity @xmath292 . it is straightforward to show that the derivative with respect to @xmath233 in the argument of the soft function has the effect of changing the argument of the logarithm as follows : @xmath293 up to @xmath260 factors . it follows that a proper choice for the soft matching scale near @xmath29 is @xmath294 which is an order of magnitude less than the naive choice @xmath31 . the fact that the fall - off of the parton densities strongly favors the large-@xmath62 region leads to a strong additional suppression of the effective soft scale . in this section we perform a detailed numerical analysis of our results . one of our goals is to study to what extent threshold resummation is important ( or even justified ) in processes where the invariant mass of the drell - yan pair is not very close to the center - of - mass energy . we have seen in section [ sec : pqcd ] that for very small values of the ratio @xmath12 the threshold contributions are not parametrically enhanced . even though empirically these terms still give rise to the dominant contributions to the cross section , there is no need to perform a resummation of the threshold logarithms . on the other hand , the result ( [ magicmui ] ) derived in the previous section shows that near the true endpoint the threshold logarithms are enhanced by two effects : the kinematic restriction of the @xmath62-integral to the small interval between @xmath36 and 1 , as well as the dynamical enhancement from the strong fall - off of the parton densities . as a result , the appropriate value of the soft matching scale is an order of magnitude smaller than the naive choice @xmath31 . from our discussion so far it is not obvious how to interpolate between these two extreme cases . we now perform a detailed numerical study to assess the importance of resummation at intermediate values of @xmath36 . we begin with a discussion of the proper choice of the matching scales @xmath254 and @xmath30 , using the convergence of the perturbative expansions of the matching coefficients @xmath169 and @xmath240 in the resummed hard - scattering kernel ( [ final ] ) as the primary guiding principle . while it is obvious that the hard matching scale should be chosen of order @xmath7 , the choice of the soft scale is more problematic . naively , based on the structure of the result ( [ cfinal ] ) one would expect that @xmath295 should in some sense be identified with the average " value of @xmath296 $ ] . unfortunately , however , the distribution in the variable @xmath62 is both singular at @xmath253 and not positive definite , so that it does not lend itself to a probabilistic interpretation . a simple way to avoid large logarithmic contributions would be to make the scale choice @xmath297 $ ] inside the @xmath62-integral . however , this would lead to landau - pole singularities in the integrand and hence would upset the proper scale separation that is at the heart of our approach . [ ] @xmath298 [ ] @xmath299 [ ] @xmath300 [ ] @xmath301 dependence of the first two expansion coefficients in the perturbative series for the hard function on the matching scale @xmath254.,title="fig:",scaledwidth=50.0% ] our approach to threshold resummation based on effective field - theory methods applied directly in momentum space provides a natural way of resolving this question . the matching scales @xmath254 and @xmath30 should be chosen such that the perturbative expansions of the wilson coefficient functions @xmath169 and @xmath240 in ( [ cfinal ] ) are well behaved . for the case of the soft function , this criterion should be applied _ after _ the integration over @xmath62 in ( [ final ] ) has been performed . this is the essence of effective field theory : one performs matching calculations at scales where these calculations can be done in fixed - order perturbation theory , and use the renormalization group to perform the evolution ( running " ) between the different matching scales . [ ] @xmath30[gev ] [ b]@xmath6 correction relative contributions to the drell - yan cross section @xmath8 at @xmath302gev arising from the one - loop ( left ) and two - loop ( right ) corrections to the soft function @xmath240 , as a function of the soft matching scale @xmath30 . the curves are labeled by the corresponding values of @xmath12.,title="fig:",scaledwidth=45.0% ] [ ] @xmath303 correction relative contributions to the drell - yan cross section @xmath8 at @xmath302gev arising from the one - loop ( left ) and two - loop ( right ) corrections to the soft function @xmath240 , as a function of the soft matching scale @xmath30 . the curves are labeled by the corresponding values of @xmath12.,title="fig:",scaledwidth=46.0% ] we begin by applying this criterion to the hard function @xmath304^n .\ ] ] figure [ fig : muhchoice ] shows the dependence of the expansion coefficients @xmath300 and @xmath301 on the ratio @xmath298 . the one - loop coefficient @xmath300 vanishes for @xmath305 and @xmath306 . the second solution is in a region of very small @xmath254 , where the expansion coefficients vary strongly and where the @xmath307 and @xmath308 terms have opposite sign . we will thus discard it . in the region around the first solution the two - loop coefficient @xmath301 is stable and positive . in our numerical analysis we will vary @xmath254 between @xmath7 and @xmath309 , taking @xmath310 as the default choice . the matching scale @xmath30 must be determined separately for each process , since it is sensitive to the integration range of the @xmath62 variable ( which depends on @xmath36 and @xmath44 ) and to the shape of the pdfs . in figure [ fig : muichoice1 ] , we plot the relative contributions to the cross section @xmath8 ( normalized to the total cross section ) arising from the one- and two - loop terms in the soft function @xmath240 as a function of @xmath30 . we choose @xmath302gev and consider different values of @xmath12 between 0.01 and 0.7 . the plots have been obtained by setting the factorization scale equal to @xmath7 and using mrst2004nnlo parton densities @xcite . we have checked that virtually indistinguishable results are obtained when @xmath54 is varied by a factor of 2 . notice that with increasing @xmath36 values the regions where the one- and two - loop contributions are of modest size shift toward lower @xmath30 values . to be specific , we consider two criteria for a good convergence of the perturbative expansion ( see the left plot in the figure ) : * starting from a high scale , we determine the value of @xmath30 at which the one - loop correction drops below 15% . * we choose the value of @xmath30 for which the one - loop contribution is minimal . note that with either choice the two - loop corrections at the corresponding @xmath30 values are very small , indicating that the first two terms in the perturbation series for the soft function are well behaved for the same choice of scale . the same analysis can be repeated for different masses of the drell - yan pair . [ ] @xmath12 [ b]@xmath311 scale - setting results for the soft matching scale @xmath30 for different values of @xmath12 and drell - yan masses @xmath312 , 50 , 20 , 10gev . the lighter curves correspond to lower scales . the upper set of curves corresponds to convergence criterion i , the lower one to criterion ii.,title="fig:",scaledwidth=50.0% ] the resulting values for the soft scale @xmath30 determined using these convergence criteria are shown in figure [ fig : scaling ] , where we consider the choices @xmath313 , 20 , 50 , 100gev . to a good approximation the curves for the ratio @xmath311 as a function of @xmath36 exhibit scaling , i.e. , they are almost independent of @xmath7 . we thus obtain a relation of the form @xmath314 . small scaling violations arise from the scale dependence of the pdfs and of the running coupling in the perturbative expansion of the soft function . our numerical results are well reproduced by the empirical functions @xmath315 for the two criteria . their form should not be taken too seriously except to note that for @xmath29 both functions approach @xmath316 , as required by the resummation formula ( [ kdy ] ) for the @xmath277-factor valid near the true endpoint . indeed , the smallness of the constants in the two cases ( 0.125 and 0.080 , respectively ) is in good agreement with our estimate in ( [ magicmui ] ) . it results from the dynamical suppression provided by the pdfs . in the opposite limit @xmath317 , both forms yield @xmath318 with an @xmath260 constant , as required by the fact that there is only a single physical scale in this case . below we will vary the soft scale between @xmath319 and @xmath320 and , somewhat arbitrarily , use the average of the two scales as the default choice . the same analysis can be carried out for the double differential decay rate @xmath321 , and it leads to similar results . in general , one finds that at central rapidity ( @xmath322 ) the resulting values for the soft matching scale are higher by about 20% than in the case of @xmath8 , while at larger rapidity they are lower , so that on average , after integration over @xmath44 , the results for the total cross section are reproduced . we now proceed to study the stability of our resummed expression for the drell - yan cross section , using scale dependence as an estimator of yet unknown higher - order perturbative effects . we vary the matching scales @xmath254 and @xmath30 about the default values @xmath323 and @xmath324 determined in the previous section . to simplify comparisons with the literature we adopt the conventional choice @xmath325 for the factorization scale , at which the pdfs are renormalized . from the point of view of effective field theory it would be more natural to choose a lower value for @xmath54 , given that the cross section is sensitive to physics at scales much below the hard scale @xmath7 . we will see , however , that our results are very stable with respect to variations of the factorization scale , so that the choice of the default value is not particularly important . since our focus is on the behavior of different perturbative approximations to the hard - scattering kernel , we use the same set of pdfs ( mrst04nnlo ) throughout the analysis . our results are shown in the first three plots in figure [ fig : scales ] , in which we study the drell - yan @xmath277-factor defined in ( [ kdef ] ) for @xmath302gev and various values of @xmath12 . in the calculation of the @xmath277-factor we keep the factorization scale @xmath325 fixed in the leading - order expression in the denominator , even when @xmath54 is varied in the numerator . for the time being we only include the leading terms in the @xmath38 limit , corresponding to the result ( [ cfinal ] ) . adding the small power - suppressed corrections would not change any of our conclusions . we observe an excellent convergence of the perturbative expansion for the cross section after resummation . the bands corresponding to the lo , nlo , and nnlo approximations overlap , and the dependence on the matching scales @xmath254 and @xmath30 becomes negligible beyond lo , indicating that the residual perturbative uncertainty is very small . [ l ] @xmath326 [ b]@xmath327 [ ] [ cols="^,^ " , ] the fact that the moment - space resummation approach differs from the momentum - space approach by first - order corrections in @xmath26 leads to visible numerical effects . in order to illustrate this we take the inverse mellin transformation of ( [ sign ] ) to compute the drell - yan @xmath277-factor . for simplicity , we fit the polynomial form @xmath328 to the parton luminosity function defined in ( [ ffdef ] ) , so that the product of the moments of the pdfs in ( [ sign ] ) can be evaluated in closed form . the first plot in figure [ fig : compmoment ] shows that for the same choices of matching scales the moment - space approach leads to somewhat larger results for the @xmath277-factor than the momentum - space approach , and that the convergence is slightly better in the second case . in the traditional moment - space approach based on expression ( [ gn ] ) one implicitly makes the scale choice @xmath329 , which leads to a landau pole at @xmath330 in the mellin inversion . to see how the results obtained with this choice compare with our default results we adopt the so - called minimal prescription @xcite , which amounts to choosing an inversion contour in the @xmath53-plane that does not include this pole . when this is done , we find that the moment - space formula ( [ momentspacecn ] ) evaluated with @xmath331 and @xmath329 gives numerical results similar to those found using our default scale choices once we go beyond leading order . this is illustrated in the right plot in figure [ fig : compmoment ] . we have presented a detailed analysis of threshold resummation for drell - yan production . instead of the conventional mellin moment formalism , we have performed the resummation directly in momentum space , using rg evolution . we have obtained analytic expressions for the resummed hard - scattering kernels that are free of the landau - pole ambiguities inherent in the traditional approach . after deriving the necessary matching coefficients and anomalous dimensions from known perturbative results in the literature , we have performed the resummation for the cross section and rapidity distribution at nnlo in rg - improved perturbation theory , corresponding to n@xmath9ll accuracy . the main goal of this work was to address the question to what extent resummation can be phenomenologically relevant given that in all practical applications the true threshold region is experimentally not accessible , because the parton distributions fall off very steeply at large @xmath0 . it has been argued in the literature that precisely this fall - off forces the drell - yan process to the partonic threshold region , in which large logarithmic corrections arise . our analysis confirms the existence of this effect and quantifies its importance . in the true endpoint region , we find that the scale of soft radiation is an order of magnitude smaller than the naive estimate . the extra suppression factor is given by the sum of the exponents of the fall - off of the quark and anti - quark distributions near @xmath267 . on the other hand , for very small lepton - pair masses this effect becomes inoperative because the fall - off of the parton distributions is not very steep at low @xmath0 . an analysis of the convergence of perturbation theory in the intermediate region shows that resummation effects become relevant for @xmath7 larger than about @xmath332 . at the largest measured drell - yan masses resummation effects can thus be significant . in the case of the experiment e866/nusea , a fixed - target experiment with @xmath1 collisions at @xmath14gev , these effects increase the nnlo ( nlo ) cross section by 7% ( 25% ) at @xmath42gev . we do not confirm the recent claim of large negative resummation effects for the rapidity distribution measured in the e866/nusea experiment at @xmath15gev @xcite . we find that the fixed - order nnlo threshold contribution is positive and a good approximation to the full nnlo correction . higher - order terms beyond nnlo turn out to be negligible , so that it is unnecessary to resum them . to check whether the large effect seen in @xcite could be an artifact of the moment - space formalism , we have applied our approach also in moment space and showed that it encompasses the traditional moment - space approach . we find that the two methods give compatible results , and that in both schemes resummation increases the cross section . the drell - yan cross section is dominated by the contribution of the threshold region even at small masses of the lepton pair . for example , for @xmath15gev the threshold terms give rise to @xmath333 of the nlo correction to the total cross section and @xmath334 of the nnlo contribution . if this is a more generic feature of hard cross sections , then threshold resummation techniques provide an efficient way to obtain the dominant part of the higher - order corrections . perhaps the most important outcome of our analysis is a quantitative understanding of the emergence of an effective physical scale characterizing the soft radiation in the drell - yan process . this scale is generated through an intricate interplay of dynamical and kinematical effects . it will be interesting to explore how the same mechanism affects other collider processes such as higgs production at the tevatron and lhc . furthermore , it would be useful to analyze resummation for @xmath335-production , which is phenomenologically relevant , since the tevatron now produces top - quark pairs with quite high invariant masses . there was some controversy on how to best perform the resummmation in this case @xcite . it would be interesting to revisit the issue using effective field - theory methods . we are grateful to lance dixon , keith ellis , frank petriello , and werner vogelsang for useful discussions . m.n . likes to thank the fermilab theory group for hospitality and support as a frontier fellow . the research of t.b . was supported by the department of energy under grant de - ac02 - 76ch03000 . the research of g.x . was supported by the national science foundation under grant phy-0355005 . fermilab is operated by the fermi research alliance under contract with the u.s . department of energy . the leading singular terms in the fixed - order perturbative expressions for the hard - scattering kernels @xmath163 can be derived by using the perturbative expansions of the hard and soft functions in ( [ factform ] ) . in this way , we obtain for the coefficients @xmath87 in ( [ llterms ] ) @xmath336 \nonumber\\ & & \mbox{}+ 16\zeta_3 \left [ \frac{1}{1-z } \right]_+ + \left ( - l^2 + 3l - 8 - \frac{2\pi^2}{3 } \right ) \left [ \frac{l_z}{1-z } \right]_+ + \left [ \frac{l_z^3}{1-z } \right]_+ \ , , \nonumber\\ p_a(z ) & = & \delta(1-z ) \left [ - \frac{11}{16}\,l^2 + \left ( \frac{193}{48 } - \frac{11\pi^2}{36 } - \frac{3\zeta_3}{2 } \right ) l - \frac{1535}{192 } + \frac{47\pi^2}{36 } - \frac{23\pi^4}{720 } + \frac{43\zeta_3}{12 } \right ] \nonumber\\ & & \mbox{}+ \left ( - \frac{101}{27 } + \frac{11\pi^2}{18 } + \frac{7\zeta_3}{2 } \right ) \left [ \frac{1}{1-z } \right]_+ + \left ( \frac{67}{18 } - \frac{\pi^2}{6 } \right ) \left [ \frac{l_z}{1-z } \right]_+ - \frac{11}{12 } \left [ \frac{l_z^2}{1-z } \right]_+ \ , , \nonumber\\ p_f(z ) & = & \delta(1-z ) \left [ \frac14\,l^2 + \left ( - \frac{17}{12 } + \frac{\pi^2}{9 } \right ) l + \frac{127}{48 } - \frac{4\pi^2}{9 } + \frac{\zeta_3}{3 } \right ] \nonumber\\ & & \mbox{}+ \left ( \frac{28}{27 } - \frac{2\pi^2}{9 } \right ) \left [ \frac{1}{1-z } \right]_+ - \frac{10}{9 } \left [ \frac{l_z}{1-z } \right]_+ + \frac13 \left [ \frac{l_z^2}{1-z } \right]_+ \,.\end{aligned}\ ] ] as before , we use the abbreviations @xmath85 and @xmath86 $ ] . if desired , the plus distributions involving powers of @xmath88 can be reduced to distributions of the form @xmath337_+$ ] using the identity @xmath338_+ = f(z ) \left [ g(z ) \right]_+ - \delta(1-z ) \int_0 ^ 1\!dz'\,f(z ' ) \left [ g(z ' ) \right]_+ \,.\ ] ] for completeness we list the perturbative expansions of the various matching coefficients and anomalous dimensions required to evaluate our rg - improved result ( [ cfinal ] ) at nnlo . the matching conditions for the wilson coefficient @xmath169 evaluated at time - like momentum transfer can be obtained by analytic continuation from the corresponding expression valid at space - like momentum transfer . using the known two - loop result for the on - shell qcd form factor @xcite , we find @xcite @xmath339 , \ ] ] where @xmath340 and @xmath341 . this result agrees with the corresponding expression given in @xcite . the matching condition for the function @xmath240 can be derived most easily using relation ( [ wonderful ] ) , which relates it to the perturbative expansion of the position - space wilson loop @xmath342 at time - like separation . we can then use the explicit two - loop expression for the position - space wilson loop obtained in @xcite . to this end , we have obtained the fourth - order coefficient in the expansion of a certain appell hypergeometric function , @xmath343 2+\epsilon,1 - 2\epsilon \end{array}\,\big|\,1,1 \big ) = - \frac{1+\epsilon}{2\epsilon } \left ( 1 - \frac{\pi^2}{3}\,\epsilon^2 - 14\zeta_3\epsilon^3 - \frac{5\pi^4}{18}\,\epsilon^4 + \dots \right ) .\ ] ] the resulting two - loop expression for the soft function reads @xmath344 , \ ] ] where @xmath345 note that the @xmath346 terms exponentiate , which is a consequence of the non - abelian exponentiation theorem for wilson loops @xcite . our result for the function @xmath240 agrees with a corresponding expression entering in the moment - space resummation approach studied in @xcite . here we list expressions for the anomalous dimensions and the qcd @xmath347-function , quoting all results in the @xmath348 renormalization scheme . we define the expansion coefficients of the anomalous dimensions and the qcd @xmath347-function as @xmath349 , \end{aligned}\ ] ] and similarly for the other anomalous dimensions . the expansion of the cusp anomalous dimension @xmath214 to two - loop order was obtained some time ago @xcite , while recently the three - loop coefficient has been calculated in @xcite . for the four - loop coefficient @xmath350 we use the pad approximant @xmath351 . the results are @xmath352 , \nonumber\\ \gamma_2 & = & 4 c_f \bigg [ c_a^2 \left ( \frac{245}{6 } - \frac{134\pi^2}{27 } + \frac{11\pi^4}{45 } + \frac{22}{3}\,\zeta_3 \right ) + c_a t_f n_f \left ( - \frac{418}{27 } + \frac{40\pi^2}{27 } - \frac{56}{3}\,\zeta_3 \right ) \nonumber\\ & & \mbox{}+ c_f t_f n_f \left ( - \frac{55}{3 } + 16\zeta_3 \right ) - \frac{16}{27}\,t_f^2 n_f^2 \bigg ] \ , , \nonumber\\ \gamma_3 & \approx & 7849,~ 4313 , ~ 1553 \quad \mbox{for } \quad n_f = 3,\,4,\,5 \,.\end{aligned}\ ] ] the anomalous dimension @xmath215 can be determined up to three - loop order from the partial three - loop expression for the on - shell quark form factor in qcd , which has recently been obtained in @xcite . we find @xmath353 the anomalous dimension @xmath224 is know to three - loop order from the nnlo calculation of the altarelli - parisi splitting functions @xcite . the expansion coefficients are @xmath354 using these results , one can compute the expansion coefficients for the anomalous dimension @xmath229 of the drell - yan soft function from the relation @xmath355 . this yields @xmath356 we do not list the three - loop coefficient . finally , the expansion coefficients for the qcd @xmath347-function to four - loop order are @xmath357 the value of @xmath358 is taken from @xcite and corresponds to @xmath359 and @xmath360 . we now give the perturbative expansions of the functions @xmath100 and @xmath361 defined in ( [ rgesols ] ) , working consistently at nnlo in rg - improved perturbation theory . at this order we need to keep terms through @xmath39 in the final expressions . the resulting expression for @xmath361 is given by @xmath362 \frac{\alpha_s^2(\mu ) - \alpha_s^2(\nu)}{32\pi^2 } + \dots \bigg\ } \,.\end{aligned}\ ] ] similar expressions with the @xmath363 replaced by the coefficients @xmath364 or @xmath365 hold for the functions @xmath219 and @xmath289 , respectively . the nnlo expression for the sudakov exponent @xmath100 is more complicated . it reads @xcite @xmath366 \nonumber\\ & & \mbox{}+ \left ( \frac{\alpha_s(\nu)}{4\pi } \right)^2 \bigg [ \left ( \frac{\beta_1\beta_2}{\beta_0 ^ 2 } - \frac{\beta_1 ^ 3}{2\beta_0 ^ 3 } - \frac{\beta_3}{2\beta_0 } + \frac{\beta_1}{\beta_0 } \left ( \frac{\gamma_2}{\gamma_0 } - \frac{\beta_2}{\beta_0 } + \frac{\beta_1 ^ 2}{\beta_0 ^ 2 } - \frac{\beta_1\gamma_1}{\beta_0\gamma_0 } \right ) \frac{r^2}{2 } \right ) \ln r \nonumber\\ & & \hspace{1.0 cm } \mbox{}+ \left ( \frac{\gamma_3}{\gamma_0 } - \frac{\beta_3}{\beta_0 } + \frac{2\beta_1\beta_2}{\beta_0 ^ 2 } + \frac{\beta_1 ^ 2}{\beta_0 ^ 2 } \left ( \frac{\gamma_1}{\gamma_0 } - \frac{\beta_1}{\beta_0 } \right ) - \frac{\beta_2\gamma_1}{\beta_0\gamma_0 } - \frac{\beta_1\gamma_2}{\beta_0\gamma_0 } \right ) \frac{(1-r)^3}{3 } \nonumber\\ & & \hspace{1.0 cm } \mbox{}+ \left ( \frac{3\beta_3}{4\beta_0 } - \frac{\gamma_3}{2\gamma_0 } + \frac{\beta_1 ^ 3}{\beta_0 ^ 3 } - \frac{3\beta_1 ^ 2\gamma_1}{4\beta_0 ^ 2\gamma_0 } + \frac{\beta_2\gamma_1}{\beta_0\gamma_0 } + \frac{\beta_1\gamma_2}{4\beta_0\gamma_0 } - \frac{7\beta_1\beta_2}{4\beta_0 ^ 2 } \right ) ( 1-r)^2 \nonumber\\ & & \hspace{1.0 cm } \mbox{}+ \left ( \frac{\beta_1\beta_2}{\beta_0 ^ 2 } - \frac{\beta_3}{\beta_0 } - \frac{\beta_1 ^ 2\gamma_1}{\beta_0 ^ 2\gamma_0 } + \frac{\beta_1\gamma_2}{\beta_0\gamma_0 } \right ) \frac{1-r}{2 } \bigg ] + \dots \bigg\ } \,,\end{aligned}\ ] ] where @xmath367 . whereas the three - loop anomalous dimensions and @xmath347-function are required in ( [ asol ] ) , the expression for @xmath100 also involves the four - loop coefficients @xmath350 and @xmath358 . to perform the effective theory analysis of the process , we would like to have a path - integral definition of the hadronic quantity of interest . in the case of dis ( and also for inclusive @xmath368-decays ) , one considers the discontinuity of forward matrix elements of time - ordered products of the electroweak currents and studies their factorization properties . the drell - yan cross section can not be written in a similar form . the reason is that not all cuts of the relevant feynman graphs correspond to the same physical process . for example , in addition to the contribution from the cut indicated in figure [ fig : cutdiagram ] , the discontinuity of the same diagram also gets a contribution from a cut through the triangle loop on the left , which describes @xmath369 with a virtual lepton pair . a path - integral method for the direct evaluation of cut diagrams is the keldysh ( or time - loop ) formalism @xcite . let us illustrate the method for a scalar field theory with field @xmath372 . instead of the usual action , where one integrates over time from @xmath373 , one considers a path integral with an action @xmath374 where the contour @xmath375 first runs from @xmath376 and then back , with a negative imaginary part , along the contour shown in figure [ fig : contour ] . when evaluating expectation values of fields , one needs to specify on which part of the contour the fields reside . we denote the fields living on the first half of the contour by @xmath377 and the ones on the second half by @xmath378 . the path - integral expectation values of fields correspond to path - ordered vacuum expectation values . since the fields @xmath379 are positioned at earlier points along the contour , they are to the right of all fields @xmath380 . furthermore path - ordering translates into anti - time - ordering on the second part of the contour , so that we find @xmath381 here @xmath382 denotes anti - time - ordering . note that under a field redefinition @xmath383 , the plus and minus fields transform separately @xmath384 this is relevant when one uses a field redefinition to decouple the soft gluon fields from the quark fields in the current operator . the keldysh formalism is useful , because it gives a path integral formulation of squared amplitudes , e.g. @xmath385 which is then rewritten as an expectation value of @xmath386 fields using ( [ expectation ] ) . for our perturbative analysis , we need the feynman rules to calculate matrix elements of the form ( [ expectation ] ) . the rules are simple : those for the field @xmath379 are the usual feynman rules and those for @xmath380 are the complex conjugate of the usual rules . since all interactions are local , there are no vertices involving both @xmath379 and @xmath380 fields . the only connection is the cut propagator @xmath387 this expression is familiar from the cutkosky rules @xcite used to extract the contribution of a given cut to an ordinary feynman diagram . [ t]@xmath388 -plane along which the lagrangian is integrated to obtain the action in the keldysh formalism . ( to dampen the oscillatory behavior of the path integrand , the entire contour should further be rotated clockwise by a small angle . ) [ fig : contour],title="fig:",scaledwidth=40.0% ] s. d. drell and t. m. yan , phys . lett . * 25 * , 316 ( 1970 ) [ erratum - ibid . * 25 * , 902 ( 1970 ) ] . g. altarelli , r. k. ellis and g. martinelli , nucl . b * 157 * , 461 ( 1979 ) . r. hamberg , w. l. van neerven and t. matsuura , nucl . b * 359 * , 343 ( 1991 ) [ erratum - ibid . b * 644 * , 403 ( 2002 ) ] . r. v. harlander and w. b. kilgore , phys . lett . * 88 * , 201801 ( 2002 ) [ hep - ph/0201206 ] . c. anastasiou , l. j. dixon , k. melnikov and f. petriello , phys . lett . * 91 * , 182002 ( 2003 ) [ hep - ph/0306192 ] . c. anastasiou , l. j. dixon , k. melnikov and f. petriello , phys . d * 69 * , 094008 ( 2004 ) [ hep - ph/0312266 ] . k. melnikov and f. petriello , phys . lett . * 96 * , 231803 ( 2006 ) [ hep - ph/0603182 ] . k. melnikov and f. petriello , phys . d * 74 * , 114017 ( 2006 ) [ hep - ph/0609070 ] . g. sterman , nucl . b * 281 * , 310 ( 1987 ) . s. catani and l. trentadue , nucl . b * 327 * , 323 ( 1989 ) . l. magnea , nucl . b * 349 * , 703 ( 1991 ) . g. p. korchemsky and g. marchesini , phys . b * 313 * , 433 ( 1993 ) . g. sterman and w. vogelsang , jhep * 0102 * , 016 ( 2001 ) [ hep - ph/0011289 ] . a. mukherjee and w. vogelsang , phys . d * 73 * , 074005 ( 2006 ) [ hep - ph/0601162 ] . p. bolzoni , phys . b * 643 * , 325 ( 2006 ) [ hep - ph/0609073 ] . v. ravindran , j. smith and w. l. van neerven , hep - ph/0608308 . v. ravindran and j. smith , arxiv:0708.1689 [ hep - ph ] . j. c. webb _ et al . _ [ nusea collaboration ] , hep - ex/0302019 . d. appell , g. sterman and p. b. mackenzie , nucl . b * 309 * , 259 ( 1988 ) . s. catani , m. l. mangano and p. nason , jhep * 9807 * , 024 ( 1998 ) [ hep - ph/9806484 ] . t. becher and m. neubert , phys . lett . * 97 * , 082001 ( 2006 ) [ hep - ph/0605050 ] . t. becher , m. neubert and b. d. pecjak , jhep * 0701 * , 076 ( 2007 ) [ hep - ph/0607228 ] . c. w. bauer , s. fleming , d. pirjol and i. w. stewart , phys . d * 63 * , 114020 ( 2001 ) [ hep - ph/0011336 ] . c. w. bauer , d. pirjol and i. w. stewart , phys . d * 65 * , 054022 ( 2002 ) [ hep - ph/0109045 ] . m. beneke , a. p. chapovsky , m. diehl and t. feldmann , nucl . b * 643 * , 431 ( 2002 ) [ hep - ph/0206152 ] . v. n. gribov and l. n. lipatov , sov . j. nucl . phys . * 15 * , 438 and 675 ( 1972 ) [ yad . 15 * , 781 and 1218 ( 1972 ) ] . y. l. dokshitzer , sov . jetp * 46 * , 641 ( 1977 ) [ zh . eksp . 73 * , 1216 ( 1977 ) ] . g. altarelli and g. parisi , nucl . b * 126 * , 298 ( 1977 ) . the code dyrap we used is available at http://www.slac.stanford.edu/~lance / vrap/. a. d. martin , r. g. roberts , w. j. stirling and r. s. thorne , phys . b * 604 * , 61 ( 2004 ) [ hep - ph/0410230 ] . c. w. bauer , s. fleming , d. pirjol , i. z. rothstein and i. w. stewart , phys . d * 66 * , 014017 ( 2002 ) [ hep - ph/0202088 ] . g. t. bodwin , s. j. brodsky and g. p. lepage , phys . lett . * 47 * , 1799 ( 1981 ) . j. c. collins , d. e. soper and g. sterman , phys . b * 109 * , 388 ( 1982 ) . j. c. collins , d. e. soper and g. sterman , nucl . b * 261 * , 104 ( 1985 ) . j. c. collins , d. e. soper and g. sterman , nucl . b * 308 * , 833 ( 1988 ) . a. idilbi , x. d. ji and f. yuan , nucl . b * 753 * , 42 ( 2006 ) [ hep - ph/0605068 ] . j. s. schwinger , j. math . * 2 * , 407 ( 1961 ) . l. v. keldysh , zh . eksp . 47 * , 1515 ( 1964 ) [ sov . jetp * 20 * , 1018 ( 1965 ) ] . a. v. belitsky , phys . b * 442 * , 307 ( 1998 ) [ hep - ph/9808389 ] . t. becher , r. j. hill , b. o. lange and m. neubert , phys . d * 69 * , 034013 ( 2004 ) [ hep - ph/0309227 ] . a. v. manohar , phys . d * 68 * , 114019 ( 2003 ) [ hep - ph/0309176 ] . t. becher , r. j. hill and m. neubert , phys . d * 69 * , 054017 ( 2004 ) [ hep - ph/0308122 ] . m. beneke and t. feldmann , phys . b * 553 * , 267 ( 2003 ) [ hep - ph/0211358 ] . j. c. collins and d. e. soper , nucl . b * 194 * , 445 ( 1982 ) . g. p. korchemsky and a. v. radyushkin , nucl . b * 283 * , 342 ( 1987 ) . i. a. korchemskaya and g. p. korchemsky , phys . b * 287 * , 169 ( 1992 ) . m. neubert , eur . j. c * 40 * , 165 ( 2005 ) [ hep - ph/0408179 ] . g. p. korchemsky and g. marchesini , nucl . b * 406 * , 225 ( 1993 ) [ hep - ph/9210281 ] . s. moch , j. a. m. vermaseren and a. vogt , nucl . b * 688 * , 101 ( 2004 ) [ hep - ph/0403192 ] . m. neubert , phys . d * 72 * , 074025 ( 2005 ) [ hep - ph/0506245 ] . s. alekhin , jetp lett . * 82 * , 628 ( 2005 ) [ pisma zh . * 82 * , 710 ( 2005 ) ] [ hep - ph/0508248 ] . a. d. martin , r. g. roberts , w. j. stirling and r. s. thorne , phys . b * 531 * , 216 ( 2002 ) [ hep - ph/0201127 ] . w. k. tung , h. l. lai , a. belyaev , j. pumplin , d. stump and c. p. yuan , jhep * 0702 * , 053 ( 2007 ) [ hep - ph/0611254 ] . t. becher and m. neubert , phys . b * 637 * , 251 ( 2006 ) [ hep - ph/0603140 ] . s. moch and a. vogt , phys . b * 631 * , 48 ( 2005 ) [ hep - ph/0508265 ] . e. laenen and l. magnea , phys . b * 632 * , 270 ( 2006 ) [ hep - ph/0508284 ] . s. moch , j. a. m. vermaseren and a. vogt , nucl . b * 726 * , 317 ( 2005 ) [ hep - ph/0506288 ] . v. ravindran , nucl . b * 746 * , 58 ( 2006 ) [ hep - ph/0512249 ] . v. ravindran , nucl . b * 752 * , 173 ( 2006 ) [ hep - ph/0603041 ] . m. beneke and v. m. braun , nucl . b * 454 * , 253 ( 1995 ) [ hep - ph/9506452 ] . s. catani , m. l. mangano , p. nason and l. trentadue , nucl . b * 478 * , 273 ( 1996 ) [ hep - ph/9604351 ] . t. matsuura and w. l. van neerven , z. phys . c * 38 * , 623 ( 1988 ) . t. matsuura , s. c. van der marck and w. l. van neerven , nucl . b * 319 * , 570 ( 1989 ) . t. gehrmann , t. huber and d. maitre , phys . b * 622 * , 295 ( 2005 ) [ hep - ph/0507061 ] . s. moch , j. a. m. vermaseren and a. vogt , jhep * 0508 * , 049 ( 2005 ) [ hep - ph/0507039 ] . j. g. m. gatheral , phys . b * 133 * , 90 ( 1983 ) . j. frenkel and j. c. taylor , nucl . b * 246 * , 231 ( 1984 ) . t. van ritbergen , j. a. m. vermaseren and s. a. larin , phys . b * 400 * , 379 ( 1997 ) [ hep - ph/9701390 ] . r. e. cutkosky , j. math . * 1 * , 429 ( 1960 ) . s. catani , m. l. mangano , p. nason and l. trentadue , phys . lett . b * 378 * , 329 ( 1996 ) [ hep - ph/9602208 ] . e. l. berger and h. contopanagos , phys . d * 54 * , 3085 ( 1996 ) [ hep - ph/9603326 ] . e. l. berger and h. contopanagos , phys . d * 57 * , 253 ( 1998 ) [ hep - ph/9706206 ] .

electron scattering off magnetic impurities , unlike the potential scattering , substantially modifies the properties of _ s_-wave superconductors . it was realized in the seminal work by abrikosov and gorkov@xcite ( ag ) that the presence of magnetic impurities reduces the superconducting gap below its bcs value and eventually may lead to the phenomenon of gapless superconductivity . the gap suppression was investigated by means of tunneling between normal and superconducting electrodes.@xcite beyond the ag theory , it was realized that a single magnetic impurity creates a localized state within the bcs gap.@xcite a small concentration of magnetic impurities results in the formation of an impurity band@xcite which merges with the continuum as the concentration exceeds a certain value ( see ref . for a review ) . ( the original ag theory is valid above that value . ) the impurity band was investigated by tunneling experiments with alloys ( such as pbmn)@xcite and normal metal superconductor bilayers,@xcite and also by means of thermal transport in superconducting films.@xcite the observation of discrete levels associated with a separate magnetic impurity came later in the scanning tunneling microscopy experiments.@xcite the effect of magnetic impurities came under scrutiny recently in connection with limits on performance of superconducting elements in qubits@xcite and photon detectors.@xcite in these applications , superconductors should perform in the ac regime . there is also some experimental effort directly aimed at observation of the effect of magnetic impurities on the surface impedance of superconducting multilayers@xcite and unconventional superconductors.@xcite theoretically , the results on impedance@xcite were analyzed in the framework of a simple two - fluid model . the two - fluid model assumes a simplified frequency - independent drude - type dissipative conductivity @xmath0 , while the influence of magnetic impurities on the electron spectrum should lead to a nontrivial @xmath1 dependence on the frequency @xmath2 . a theoretical approach to studying the surface impedance in superconductors was developed in the celebrated work of mattis and bardeen,@xcite which treated the anomalous skin effect ( the skin depth is smaller than the mean free path ) taking nonmagnetic disorder into account . independently , the clean case was considered by abrikosov , gorkov , and khalatnikov@xcite in the green - function language . later , nam@xcite employed the same technique and demonstrated that the extreme anomalous limit is equivalent ( up to the expressions for the normal - metallic conductivities ) to the local limit , realized in the dirty regime . the complex conductivity of a superconductor containing both potential and magnetic impurities was addressed by skalski _ _ et al.__@xcite their paper does contain general results applicable in the case of weak magnetic scattering ; however , similarly to the ag theory , scattering off magnetic impurities was considered in the born approximation , which did not allow to treat the effect of localized states on the impedance . in ref . , larkin and ovchinnikov considered superconductors with inhomogeneous pairing , which under certain assumptions is formally equivalent to the presence of magnetic impurities in the ag regime . some of their results can therefore be applied to this problem . however , a convenient formulation of the theory , suitable for the experimentally relevant dirty limit and making it possible to treat both the ag regime and the impurity band , as well as detailed quantitative investigation of the effect of magnetic impurities on the surface impedance in different regimes over temperature , frequency , and electron scattering rate off magnetic impurities is still lacking , to the best of our knowledge . in this paper , we study the influence of magnetic impurities on the complex conductivity and surface impedance of a superconductor . we are especially interested in the limit of low frequency of the impinging radiation and low temperature , and in the effect of the impurity band on the dissipation . quantitative results are presented for the experimentally relevant dirty limit ( with respect to the nonmagnetic disorder ) . the paper is organized as follows . first , in sec . [ sec : clean ] , we consider the limit of weak scattering off magnetic impurities and disregard nonmagnetic disorder ( `` clean '' limit ) . in that limit , we find the localized states bound to magnetic impurities and discuss the formation of the impurity band at finite but small concentration of magnetic impurities and the formation of the ag density of states at higher concentrations . we also present there a qualitative consideration of the dissipative conductivity . the main features of the developed simple picture carry over to the realistic case of a `` dirty '' superconductor doped with magnetic impurities . in sec . [ sec : dirty ] , we consider the dirty limit for the nonmagnetic disorder in the framework of the quasiclassical formalism of the usadel equation . we first obtain the spectrum of quasiparticle excitations in the presence of magnetic impurities ( treated beyond the born approximation ) . then we address the effect of magnetic impurities on the low - temperature dissipative conductivity . in addition to the low - frequency limit , we also consider in detail the frequency dependence of the dissipation near the thresholds determined by the gaps in the excitation spectrum . our results are summarized in sec . [ sec : concl ] . some technical details are presented in the appendixes . to highlight the main features of the problem related to the impurity states , we consider here a simplified case of a superconductor with magnetic impurities only . in addition , considering the hybridization of single - impurity states into an impurity band we first assume magnetic impurities to be polarized ( magnetic moments pointing up the @xmath3 axis ) , while the impurity positions are random . furthermore , we assume the exchange constant @xmath4 of interaction of localized classical spins @xmath5 with the spins of itinerant electrons to be small . introducing the dimensionless parameter @xmath6 describing the `` strength '' of a single magnetic impurity , we can write the latter requirement as @xmath7 here @xmath8 is the electron density of states ( dos ) per spin projection at the fermi level in the normal state , with @xmath9 and @xmath10 being the fermi wave vector and the fermi velocity , respectively . weak exchange interaction will modify substantially the excitation spectrum only close to the gap . concentrating on that energy region , we simplify the hamiltonian of the system in the following way . first , we express the operator of exchange interaction of electrons with an impurity ( placed at origin , @xmath11 ) in terms of the bogoliubov quasiparticles ( @xmath12 and @xmath13 are the operators of electrons and bogoliubov quasiparticles , respectively ) : @xmath14,\end{gathered}\ ] ] where @xmath15 are momenta and @xmath0 is the spin index ( @xmath16 for @xmath17 , @xmath18 for @xmath19 ) . at the same time , the standard bcs part of the hamiltonian becomes diagonal after the bogoliubov transformation , with the excitation energies @xmath20 in the superconducting state . here @xmath21 is the bcs gap and @xmath22 are the excitation energies in the normal state . close to the gap ( @xmath23 ) , we can approximate @xmath24 , while the bogoliubov coherence factors are simply @xmath25 , so finally the total hamiltonian takes the form @xmath26 where @xmath27 and we measure the quasiparticle energies from the edge of the continuous spectrum @xmath21 : @xmath28 it is easy to see that hamiltonian ( [ ham ] ) has a bound state for quasiparticles with spin @xmath0 satisfying the condition @xmath29 . for definiteness , hereafter we set @xmath30 ; then the bound state exists for spin down . we denote the wave function of the bound state in the momentum representation as @xmath31 ; the schrdinger equation for it , which follows from the form of hamiltonian ( [ ham ] ) , reads @xmath32 clearly , the wave function must have the form @xmath33 . substituting it into eq . ( [ eqphi ] ) , we find the equation for the energy of the bound state ( @xmath34 ) , @xmath35 yielding @xmath36 at @xmath37 . the power - law dependence , @xmath38 , is a direct consequence of the excitations spectrum @xmath39 . the convergence of the integral in eq . ( [ sch_e ] ) at @xmath40 validates the use of the simplified form of @xmath41 . we emphasize that in order to obtain the bound state , it was essential to go beyond the born approximation.@xcite the wave function of the bound state in the coordinate representation has the form @xmath42 where @xmath43 is the coherence length of a clean superconductor . the wave function of the localized state decays exponentially on the spatial scale which exceeds the coherence length by the factor @xmath44 . clearly , the effect of multiple magnetic impurities on the quasiparticle spectrum depends crucially on the characteristic distance between them compared to @xmath45 . at lowest concentrations @xmath46 , separate magnetic impurities create a @xmath47-peak in the quasiparticle dos . with the increase of the concentration the peak broadens . as long as the average inter - impurity distance is large , the broadening is due to rare occurrences of closely located impurity pairs producing split levels due to their hybridization.@xcite however , at higher concentrations , @xmath48 , the bound state on a given impurity overlaps with a large number of other localized states and forms a well - defined impurity band . this large parameter allows one to treat the dos in that band self - consistently . we start with an estimate of the impurity band width @xmath49 . a quasiparticle initially localized on a given impurity may hop on @xmath50 `` nearest neighbors '' . because of their large number , the return of the quasiparticle is improbable . therefore , we may introduce the escape rate @xmath51 for a given impurity . it will determine the level width , and we will associate it with the width of the band , @xmath52 . on the other hand , the escape rate from a given site is proportional to the dos on the `` receiving '' ones . the latter is inversely proportional to @xmath49 . this way , the formula for the escape rate turns into a self - consistent equation for @xmath49 ( this is the essence of the self - consistent born approximation for degenerate states , analyzed in detail in the context of the broadening of high landau levels@xcite ) . to implement the scheme , we denote the hopping matrix element between two impurity states @xmath53 and @xmath54 , at positions @xmath55 and @xmath56 , as @xmath57 . the said self - consistency equation reads @xmath58 ( upon summing over positions @xmath54 , the dependence on @xmath53 disappears ) . the estimate of @xmath57 should be obtained from a consideration of the two - impurity problem . tunneling between two impurity sites splits the degenerate level in two . the resulting energy splitting depends on the inter - impurity distance @xmath59 . similarly to the conventional tight - binding problem , it is this energy splitting that should be identified with @xmath60 . this way , we find @xmath61 . to estimate the proportionality coefficient here , we note that the energy splitting reaches a value of the order of @xmath62 if @xmath63 . therefore , @xmath64 . this estimate is confirmed by a solution of the energy splitting problem , which is easily obtained by the generalization of eqs . ( [ ham])([sch_e ] ) to the case of two impurities.@xcite summation over the random positions of the impurities @xmath54 results in @xmath65 . remarkably , we used here only the value of the wave function @xmath66 close to the impurity and the fact that @xmath66 is normalized . this is why one may expect that the obtained estimate of @xmath67 remains valid in the presence of nonmagnetic disorder , as long as the corresponding mean free path is large compared to the electron fermi wave length . ( this expectation is indeed confirmed by the rigorous consideration of sec . [ sec : dirty ] . ) substituting the estimate of @xmath67 into eq . ( [ w ] ) and replacing @xmath68 , we find expression for @xmath49 up to a numerical factor . together with this factor ( which will be found later ) , the bandwidth in terms of measurable quantities @xmath46 and @xmath69 , reads @xmath70 this expression is valid when the band is narrow , @xmath71 . the next question is the dos @xmath72 for the bogoliubov quasiparticles ( hence the @xmath73 subscript ) inside the impurity band which can be found from the momentum - integrated green function @xmath74 as @xmath75 the green function @xmath76 of the bogoliubov quasiparticles is determined by the hamiltonian , which now , in contrast to hamiltonian ( [ ham ] ) , contains magnetic impurities of finite concentration : @xmath77 the green function can be written in terms of the self - energy @xmath78 and we treat the self - energy within the self - consistent @xmath79-matrix approximation,@xcite which diagrammatically amounts to the geometric series of fig . [ fig : scba ] : @xmath80 ( the sign in front of @xmath4 in the self - energy corresponds to quasiparticle spin down , which is of interest for us ) . integrating eq . ( [ g ] ) over momentum , we obtain the self - consistency condition for @xmath81 : @xmath82 the impurity band corresponds to nonzero dos , that is , to the nonzero imaginary part of @xmath83 , in some domain of negative @xmath84 . analyzing eq . ( [ g ] ) in the limit @xmath85 ( the limit of narrow band ) , we find a semicircle dos inside the impurity band : @xmath86 which is centered at @xmath69 , found earlier in eq . ( [ bs ] ) , and has the width given by eq . ( [ wfin ] ) [ note that it is the calculation leading to eq . ( [ nuom ] ) that establishes the numerical factor in eq . ( [ wfin ] ) ] . the total dos inside the impurity band is equal to @xmath46 , demonstrating that each impurity brings one state into the system.@xcite the same eq . ( [ g ] ) can be employed to demonstrate that the lower edge of the continuous spectrum shifts from zero to @xmath87 at finite impurity concentrations , and the behavior of the dos near the edge is @xmath88 ( the positive value of @xmath89 is a manifestation of the level repulsion between the states of the continuum and the impurity states . ) corresponding to the self - consistent @xmath79-matrix approximation . the scattering on a single impurity ( dashed lines ) is treated in all orders of the perturbation series . the self - consistency is achieved by inserting the bold lines for the green functions , which means that they are calculated with the help of the presented self - energy . ] as we mentioned before , a well - defined impurity band exists [ and can be described by the self - consistent @xmath79-matrix method leading to dos ( [ nuom ] ) ] if @xmath48 , which translates into @xmath90 , where @xmath91 is the fermi energy . with the increase of the impurity concentration , the impurity band broadens , @xmath92 . at @xmath93 , the width of the band becomes of the same order as the distance from the center of the band to the continuum edge , resulting in the merger of the impurity band with the continuum of quasiparticle states . [ quantitatively , the analysis of eq . ( [ g ] ) shows that the energy gap between the impurity band and continuum exists if @xmath94 . ] the two conditions for the existence of the impurity band , @xmath48 and @xmath95 , are consistent since @xmath96 . . ] in the above discussion , we did not treat the self - consistency equation explicitly , assuming @xmath21 that we used to be the self - consistent order parameter corresponding to the given impurity parameters . actually , @xmath21 is suppressed by magnetic impurities . a gapless state is realized at a critical concentration @xmath97 . at that concentration the average zeeman energy associated with the magnetic impurities , @xmath98 is comparable to @xmath21.@xcite figure [ fig : axis_ns ] illustrates the results of the above discussion . the above consideration of spin - polarized magnetic impurities can be easily generalized to the case of randomly oriented spins ( see appendix [ app : relation ] ) . it yields eq . ( [ app : nuom3 ] ) for the dos now summed over the quasiparticle spin directions inside an impurity band of width @xmath99 [ see eq . ( [ app : wfin3 ] ) ] . comparing eqs . ( [ app : nuom3 ] ) and ( [ app : wfin3 ] ) with eqs . ( [ nuom ] ) and ( [ wfin ] ) , we notice the relation @xmath100 between the widths of the band . indeed , randomness of impurity spin orientations leads to an additional @xmath101 factor in the hopping matrix element @xmath57 between two impurity states [ see eq . ( [ w ] ) ] , and hence to the @xmath102 factor in the average @xmath103 , finally changing the width of the band by the @xmath104 factor . ( here @xmath105 is the angle between the spins of the two impurities . ) another difference of the random - spin case from the spin - polarized case is the absence of the average zeeman energy . upon increasing @xmath46 , the gapless state is then reached due to the electron spin - flip scattering off magnetic impurities.@xcite the latter is characterized by a scattering rate @xmath106 which , in the born approximation , is @xmath107 the transition occurs when @xmath108 , that is , at a concentration of magnetic impurities @xmath109 . at finite but low temperatures @xmath79 , quasiparticles first populate the impurity band . this results in a finite dissipative part @xmath110 of conductivity at arbitrary low frequencies . we estimate @xmath110 at @xmath111 using a suitable version@xcite of the mattis - bardeen theory.@xcite we disregard possible spin selection rules@xcite by assuming that the spins of magnetic impurities are randomly oriented . we may relate @xmath112 to the absorption power per unit volume , @xmath113 , where @xmath114 is the amplitude of the applied ac electric field . the absorption power can be evaluated as the product of the photon energy quantum @xmath2 and the rate of creation of quasiparticles . the latter is @xmath115 and is evaluated with the help of the fermi golden rule . the perturbation leading to the excitation of the system is @xmath116 , where @xmath117 is the operator of velocity of an electron and @xmath118 is the vector potential of the ac field ; its amplitude is @xmath119 . the absorption power associated with the photon - induced transitions of quasiparticles within the impurity band equals@xcite @xmath120 \\ \times \nu_b ( e_1 ) \nu_b ( e_2 ) ( u_1 u_2+v_1 v_2)^2 \delta ( e_1-e_2-\omega).\end{gathered}\ ] ] writing eq . ( [ abspow2 ] ) , we accounted for the fact that the states in the impurity band do not correspond to a defined momentum , and consequently the latter is not conserved in the transitions . to estimate the average over the states @xmath121 , we employ the arguments of the work of mattis and bardeen,@xcite which allow one to express it in terms of an analog of the normal conductivity for the impurity band . that leads us to the estimate @xmath122 , where @xmath123 is the relaxation time and @xmath124 is a characteristic value of the dos in the impurity band . substituting @xmath125 and @xmath126 [ see eq . ( [ app : nuom3 ] ) ] , we find @xmath127 in the case of `` shallow '' levels ( @xmath37 ) we may also replace the combination of the coherence factors , @xmath128 . at lowest frequencies , @xmath129 , we may expand the difference of the quasiparticle distribution functions in eq . ( [ abspow2 ] ) as @xmath130 . as a result , we find frequency - independent asymptotic forms for the dissipative conductivity at @xmath131 and @xmath132 , which we express in terms of the density @xmath133 of quasiparticles in the impurity band [ hence the @xmath134 superscript ] : @xmath135 in equilibrium , the density of quasiparticles is @xmath136 in the opposite limit of photon energies high compared to temperature , we may neglect @xmath137 in eq . ( [ abspow2 ] ) , finally finding @xmath138 in an ideal superconductor at zero temperature , photon absorption is possible only for photons with energy @xmath2 exceeding the threshold @xmath139 ( each photon creates a pair of quasiparticles ) . the presence of states within the gap @xmath21 reduces the threshold frequency and modifies the dependence of the dissipative part of conductivity @xmath112 on frequency for @xmath2 exceeding the threshold . similarly to the above consideration , one may relate @xmath110 to the absorption power ( cf . ref . ) , @xmath140 as follows from eq . ( [ app : nuom3 ] ) , absorption is possible only if the frequency exceeds the threshold , @xmath141 to evaluate the absorption , however , the accepted approximation of shallow levels is insufficient , because it yields @xmath142 . as follows from the detailed calculation presented in sec . [ sec : dissip_zerot ] this factor @xmath143 ; the dissipative conductivity scales as @xmath144 near the threshold . we turn now to the practically important case of short elastic mean free time @xmath145 caused by potential scattering of electrons and consider the dirty limit , @xmath146 . we will see that potential disorder does not affect the electron spectrum.@xcite however , its role is essential for transport properties ( complex conductivity and impedance ) . in the dirty limit , magnetic impurities of small concentration can be described in the framework of the usadel equation.@xcite the form of the usadel equation , widely used in literature,@xcite corresponds to the born limit for magnetic scattering ( with random positions and orientations of magnetic impurities ) . at the same time , the generalization to arbitrary strength of magnetic scattering can also be obtained , as shown by marchetti and simons.@xcite employing the standard @xmath147-parametrization of the normal and anomalous quasiclassical green functions , @xmath148 we can write the generalized form as@xcite @xmath149 where @xmath150 is the diffusion constant and @xmath151 is the electron energy measured from the fermi level . we are interested in the homogeneous solution . following ref . , we define a `` renormalized '' energy and order parameter , @xmath152 with these quantities , the homogeneous usadel equation becomes simple , @xmath153 and can be formally solved as @xmath154 where @xmath155 . at the same time , this is not a solution yet , but an alternative parametrization of the green functions in terms of @xmath156 . the usadel equation actually determines the green functions dependence on energy @xmath151 ; hence , in eq . ( [ tetd ] ) @xmath156 itself depends on @xmath151 . the latter dependence is determined by the equation , directly following from eqs . ( [ tildee ] ) and ( [ tilded ] ) : @xmath157 where @xmath158 here @xmath159 is the exact scattering rate due to magnetic impurities , normalized by @xmath21 , and @xmath160 can be shown to determine the center of the impurity band , also in units of @xmath21 . since we focus on the case when each impurity is weak , @xmath37 , we simplify the above expressions as @xmath161 once found , @xmath156 can be immediately applied for analyzing the spectrum . the dos per spin is given by @xmath162 it turns out that the potential scattering does not influence the energy spectrum , since eqs . ( [ u_eq ] ) , ( [ gammaepsilon ] ) , and ( [ dos_u ] ) coincide@xcite with those obtained by shiba@xcite and rusinov@xcite in the clean limit ( see also appendix [ app : relation ] ) . finite dos corresponds to complex solutions of eq . ( [ u_eq ] ) for @xmath156 . analyzing the behavior of the function in the left - hand side ( l.h.s . ) of eq . ( [ u_eq ] ) at @xmath163 , shiba@xcite found the domain of energies [ the right - hand side ( r.h.s . ) of eq . ( [ u_eq ] ) ] that can not be matched by _ real _ values of @xmath156 . at positive energy , this domain is the impurity band centered around @xmath164 . calculating @xmath156 inside the narrow band , we find @xmath165 for the dos and @xmath166 for the width . the dos at @xmath167 is the mirror image of eq . ( [ nuom3 ] ) . the total number of states in the two impurity bands equals @xmath46 . the dos ( [ nuom3 ] ) coincides with the result derived in the clean limit [ see eq . ( [ app : nuom3 ] ) ] up to a factor of 4 . this factor accounts for the difference in the definitions of dos . [ the dos ( [ app : nuom3 ] ) , defined for bogoliubov quasiparticles existing only at @xmath168 , corresponds to eq . ( [ nuom3 ] ) ( electron component ) folded together with its mirror image ( hole component ) and , additionally , summed over the spin projections . ] as the concentration of magnetic impurities ( and hence @xmath159 ) grows , the upper edge of the band merges with the continuum at @xmath169 [ this result can be obtained from eq . ( [ u_eq ] ) and differs only numerically from the condition obtained for the case of polarized magnetic impurities in sec . [ subsec : imp - band ] ] . at higher concentrations the ag regime is realized , with the gap width reduced due to the scattering off magnetic impurities . as follows from eq . ( [ u_eq ] ) , the quasiparticle continuum starts at energy lower than the single - impurity bound - state energy @xmath170 , and the gap @xmath171 is reduced progressively with the increase of @xmath159 . at @xmath172 , one may replace @xmath173 with @xmath174 in eq . ( [ u_eq ] ) . then it simplifies to the form @xmath175 exactly as considered by abrikosov and gorkov.@xcite note that @xmath159 itself can still be small , and we mainly focus on that case : @xmath176 the ag regime , corresponding to a single gap in the spectrum , has two characteristic features , essential for the low - frequency dissipation : first , the gap is suppressed ( the relative scale is determined by @xmath177 ) , and second , the square - root bcs singularity of the dos at @xmath178 is smeared . the gap @xmath171 in units of @xmath21 is@xcite @xmath179 solving the usadel equation at energies close to the actual gap , @xmath180 , we obtain@xcite @xmath181 the square - root behavior of the dos [ see eq . ( [ dos_u ] ) ] in the vicinity of the gap was found in refs . . at low concentration , magnetic impurities substantially modify the spectrum in a relatively narrow domain of energies , @xmath182 . we take @xmath178 as a reference point within the domain , and find @xmath183 and its derivative with respect to energy : @xmath184 the maximum of the dos is achieved slightly above @xmath178 . at the same time , considering the derivative @xmath185 and the bcs solution that makes it possible to approach the maximum from the side of higher energies , we conclude that the maximal dos is of the same order as the value determined by eq . ( [ at1a ] ) . this result was found in ref . . the results for the dos are summarized in fig . [ fig : dos ] . note that the square - root behavior of the dos near the gap edge [ following from eq . ( [ neargap ] ) ] , being applied at @xmath178 ( which is already beyond its applicability range ) , yields @xmath186 , which only slightly differs from the accurate result @xmath187 following from eq . ( [ at1a ] ) . in the evaluation of dissipation with the help of eq . ( [ dissip ] ) , a good approximation for the dos is given by @xmath188 at the matching points , the above expressions match by the order of magnitude , being different only by numerical factors close to 1 . therefore , this is a rather accurate approximation for calculating the dissipation up to order - of - one numerical factors ( at the same time , in the limiting cases that we consider below , the numerical factors will be asymptotically exact ) . at small @xmath189 and @xmath159 , magnetic impurities significantly affect the electron spectrum only at energies close to @xmath21 . in that energy range [ corresponding to eq . ( [ nuom3 ] ) and to the first two lines in eq . ( [ dos ] ) ] , we have @xmath190 and hence @xmath191 [ as follows from eq . ( [ tetd ] ) ] , so that finally @xmath192 . outside that domain , the functions @xmath193 and @xmath183 in the leading order have the conventional bcs form [ with @xmath194 instead of the last line in eq . ( [ dos ] ) ] . note that we have discussed @xmath195 and @xmath196 at @xmath168 . at the same time , @xmath197 is even while @xmath198 is odd with respect to @xmath151 . having found the green functions , we are ready to calculate the dissipation . the expression for the response kernel of the current to the external electromagnetic field is derived in appendix [ sec : app : kernel ] . from the result for the kernel , we extract the expression for the dissipative part of conductivity : @xmath199,\end{gathered}\ ] ] where @xmath200 and @xmath201 is the drude conductivity . this formula is the dirty - limit counterpart of eqs . ( [ abspow2 ] ) and ( [ abspow3 ] ) for the absorption power in the clean limit . the combination of the green functions @xmath183 and @xmath193 here replaces the product of the quasiparticle densities of states and the combinations of the coherence factors in eqs . ( [ abspow2 ] ) and ( [ abspow3 ] ) . the analysis of the present section is be based on eq . ( [ dissip ] ) . at the same time , the results can be easily converted into the language of the surface impedance , as discussed in sec . [ sec : app : impedance ] . small but finite dissipative conductivity implies a small but finite real part of the surface impedance . at zero temperature the dissipation is absent below a certain threshold frequency . in the absence of magnetic impurities , @xmath2 must exceed @xmath139 so that a cooper pair can be broken and two quasiparticles are created above the gap . at frequencies slightly above the threshold , @xmath203 , the dissipative part of conductivity is@xcite @xmath204 in the presence of magnetic impurities , the bcs singularity of the green functions is smeared out and hence the threshold behavior of the dissipation changes . the threshold frequency in the ag regime is @xmath205 , and the nonzero contribution to the integral in eq . ( [ dissip ] ) comes only from the region of energies @xmath206 , where the arguments of the green functions are close to @xmath207 so that , at first sight , we can employ the square - root form for @xmath197 and @xmath198 [ the first line in eq . ( [ dos ] ) ] and take into account that @xmath195 is even while @xmath196 is odd . however , approximating @xmath208 at @xmath168 we find that the integrand in eq . ( [ dissip ] ) vanishes in the main order , so we should take into account a small difference between @xmath197 and @xmath198 . this is easily done on the basis of the normalization condition @xmath209 . note that it is the large absolute value of @xmath183 near @xmath171 [ see eq . ( [ neargap ] ) ] that leads to @xmath191 and hence to @xmath192 . at the same time , the normalization condition immediately gives also a correction to this result : @xmath210 then we find nonzero dissipation which is quadratic in deviation from the threshold at @xmath211 : @xmath212 this result was obtain in ref . . the quadratic threshold behavior of @xmath112 is a consequence of the square - root threshold behavior of the dos , and experimental results of ref . seem to confirm this @xmath110 dependence . at lowest concentrations of magnetic impurities , the impurity band exists inside the gap , so that the threshold frequency for absorption is determined by transitions in which a cooper pair is broken and two quasiparticles are created inside the impurity band . the threshold frequency for this process is @xmath213 , which is eq . ( [ omth ] ) written in different notation . just above the threshold , at @xmath214 , only the edge behavior ( near an edge of the impurity band ) of @xmath197 and @xmath198 enters eq . ( [ dissip ] ) . according to eq . ( [ nuom3 ] ) , both functions near an edge grow as a square root . we again find a cancellation in the main order and have to employ eq . ( [ imfreg ] ) . the functional form is thus the same as in the ag regime , and the result differs from eq . ( [ sgma ] ) only by coefficients : @xmath215 next we consider the case of finite ( but low ) temperatures , when the threshold for absorption is absent since equilibrium thermally excited quasiparticles can absorb photons of arbitrarily small energy . we start with the consideration of the ag regime in the limit of small @xmath159 , so that quasiparticle states exist only above the gap @xmath171 , which is weakly suppressed compared to @xmath21 , as described by eqs . ( [ gammas ] ) and ( [ epsilong ] ) . most of the results of this subsection were obtained in previous publications , although sometimes in the framework of physically different problems and with the help of different techniques . our consideration makes it possible to incorporate them in a unified manner . in the limit that we consider , @xmath195 and @xmath196 entering eq . ( [ dissip ] ) are nonzero only above @xmath171 which is close to @xmath21 , so that at low frequencies and temperatures , @xmath216 two simplifications are possible : ( a ) we can substitute the @xmath217 s with their asymptotic exponential forms , and ( b ) since the exponentials limit the integration to the region of @xmath218 , where @xmath196 is approximately equal to @xmath195 , the contribution of the @xmath219 term is the same as the contribution from the @xmath220 term . therefore , we can simplify eq . ( [ dissip ] ) as @xmath221 at the same time , at @xmath222 the external electromagnetic field can not excite quasiparticles across the gap ; therefore , the dissipation will be due to quasiparticles that are already excited due to low but finite temperature . it is thus natural to express the results for the dissipation in terms of the equilibrium quasiparticle density , @xmath223 for which under the same assumptions ( @xmath224 , @xmath225 ) we obtain two limiting cases : @xmath226 below we consider various relations between the three energy scales @xmath2 , @xmath79 , and @xmath227 ( while all of them are much smaller than @xmath21 ) . starting with the case of relatively strong magnetic scattering , @xmath228 , we note that , due to the exponentials , the main contribution to the integral in eq . ( [ sigmagg ] ) comes from the narrow region above @xmath171 , where the square - root asymptote for the dos is valid [ see the first line in eq . ( [ dos ] ) ] . as a result , @xmath229 where @xmath230 is the macdonald function ( the modified bessel function of the second kind ) . asymptotic forms of the macdonald function at large and small values of the argument describe the two limiting cases for the @xmath231 ratio , so that finally at @xmath232 we reproduce the result of maki,@xcite @xmath233 while at @xmath234 we find @xmath235 the result ( [ k1 ] ) was actually obtained by larkin and ovchinnikov in a physically different but formally equivalent problem.@xcite they considered a superconductor with inhomogeneous effective interaction between electrons . for small - scale inhomogeneities the problem turns out to be mathematically equivalent to the case of homogeneous effective interaction in the presence of magnetic impurities . the results of ref . thus apply to the problem considered in the present paper . in the limit of weak magnetic scattering , @xmath236 , the main contribution to the integral in eq . ( [ sigmagg ] ) comes from the region of energies above the smeared bcs singularity ; in that region we can apply the unperturbed result for the dos [ see the last line in eq . ( [ dos ] ) ] . we thus reproduce the results obtained in the absence of magnetic impurities@xcite ( we have to take into account the correspondence noted by nam@xcite between the extreme anomalous limit in the clean case@xcite and the dirty case to which our calculations refer ) : @xmath237 asymptotic forms of the macdonald function @xmath238 at large and small values of the argument describe the two limiting cases for the @xmath231 ratio , so that finally at @xmath239 we find @xmath240 ( where @xmath241 is euler s constant ) , while at @xmath242 the result is @xmath243 the logarithmic divergence in eq . ( [ dis4 ] ) at @xmath244 is due to the overlap of two square - root bcs singularities of the dos . similarly to the case of eq . ( [ dis1 ] ) , at @xmath245 the second exponential in eq . ( [ sigmagg ] ) can be neglected while the @xmath2-shifted dos is nearly constant , @xmath246 , in the essential region of integration . as a result , the remaining integral in both cases is exactly the same as the one determining the quasiparticle density ( [ nqp ] ) at @xmath247 . therefore , we obtain exactly the same result ( [ dis1 ] ) in terms of @xmath248 , although the expressions for @xmath248 itself are different at @xmath242 and @xmath245 [ eqs . ( [ nqp1 ] ) and ( [ nqp2 ] ) , respectively ] . this result actually follows from ref . . finally , at @xmath249 , relatively large temperature allows rather wide region of energy integration in eq . ( [ sigmagg ] ) , while the shift of the two dos can be neglected . similarly to the case of eq . ( [ dis4 ] ) , a logarithmic singularity appears in the integral ; however , now it is cut off not due to the finite shift @xmath2 [ as it was the case in eq . ( [ dis4 ] ) ] but due to smearing of the square - root bcs singularity of the dos by magnetic impurities . as a result , with the logarithmic accuracy we find @xmath250 the results for @xmath110 at @xmath222 are valid as long as the quasiparticle population may be described by the boltzmann distribution with an effective temperature . the quasiparticle chemical potential should not necessarily be @xmath251 and the temperature @xmath79 may deviate from equilibrium as long as @xmath252 . at equilibrium , the results for the dissipation @xmath112 calculated in sec . [ sec : ag ] for the ag regime in the limit of low temperatures always contain an exponentially small factor @xmath253 , since the absorption of the electromagnetic field is due to the quasiparticles thermally excited above the gap . at the same time , as we discussed above , at low concentrations of magnetic impurities a band of quasiparticles states appears below the continuum . this means that at temperatures much smaller than the distance between the band and the continuum , the dominant contribution to the dissipation will be due to quasiparticles residing inside the impurity band . below we consider this case , which is the main focus of our interest . our aim now is to calculate the dissipation ( [ dissip ] ) for the case when it is due to transitions inside the impurity band ; hence , @xmath254 . at the same time , we assume the band to be narrow , so that its width , which can be written as @xmath255 , is much smaller than the distance from the center of the band @xmath170 to the bottom of the continuum @xmath21 : since @xmath256 , the condition can be written as @xmath257 the contribution of the impurity band is dominant at @xmath258 , when the continuum contribution is exponentially smaller due to the @xmath259 factor . since we consider energies of the order of @xmath170 , that is , of the order of @xmath21 , and @xmath260 inside the narrow band ( as discussed in sec . [ sec : dos ] ) , we can make the same simplifications that lead us to eq . ( [ sigmagg ] ) . the only difference is that now the region of the energy integration is confined to within the impurity band . taking into account the explicit form of the dos inside the band [ eq . ( [ nuom3 ] ) ] and making the variable of integration dimensionless , we obtain @xmath261 the integral contains the product of the two semicircle dos ( [ nuom3 ] ) shifted by the frequency @xmath2 . the overlap exists and produces nonzero dissipation only at @xmath262 . similarly to sec . [ sec : ag ] , we express the final results in terms of the equilibrium quasiparticle density ( within the band ) . calculating eq . ( [ nqp ] ) for the impurity band , we find @xmath263 where @xmath264 is the modified bessel function of the first kind . in the two limiting cases for the @xmath265 ratio , we obtain @xmath266 below we consider various relations between the three energy scales , @xmath2 , @xmath79 , and @xmath99 ( while all of them are much smaller than @xmath21 ) . starting with the case of relatively large band width , @xmath267 , we note that , due to the exponentials , the main contribution to the integral in eq . ( [ sigmaggb ] ) comes from the narrow region above the lower limit of integration where the simplified square - root result for the dos is valid : @xmath268 . this situation is very similar to the case of eq . ( [ k1 ] ) ( cf . the discussion above that formula ) . as a result , we obtain the same functional form [ with a different prefactor and with @xmath269 instead of @xmath171 in the exponent ] : @xmath270 asymptotic forms of the macdonald function at large and small values of the argument describe the two limiting cases for the @xmath231 ratio , so that finally at @xmath271 we find @xmath272 while at @xmath273 the result is @xmath274 at @xmath275 ( while @xmath99 and @xmath2 are not necessarily close to each other ) , the exponential in eq . ( [ sigmaggb ] ) can be replaced by unity in the region of integration , and the remaining integral can be written as @xmath276.\end{gathered}\ ] ] here @xmath277 and @xmath151 are the complete elliptic integrals of the first and second kind , respectively , and their argument is @xmath278 in the limit of small frequency and high temperature , @xmath279 , with the help of eq . ( [ nqp1b ] ) , expression ( [ ek ] ) can be simplified as @xmath280 there is an upper threshold for impurity band absorption at @xmath281 . close to that threshold , eq . ( [ ek ] ) gives @xmath282 , where @xmath147 is the heaviside step function . comparing eqs . ( [ eq : sigmab - dirty1 ] ) , ( [ eq : sigmab - dirty3 ] ) , and ( [ eq : sigmab - dirty2 ] ) in the dirty case with eqs . ( [ eq : sigmab - clean1 ] ) and ( [ eq : sigmab - clean2 ] ) in the clean case , we notice that the dependences of the dissipative part of conductivity on @xmath2 , @xmath79 , and @xmath99 are the same . at the same time , the overall coefficients in the two cases are different due to different limits with respect to potential scattering . at low temperatures , the density of equilibrium quasiparticles becomes negligible , and in reality @xmath248 is dominated by extraneously produced quasiparticles.@xcite in quantum devices based on josephson junctions , the adverse effect of quasiparticles on the coherence may be mitigated by the inclusion of traps sections of the superconducting leads with reduced value of the gap . we may pose a similar question with respect to the impedance : can magnetic impurities reduce the dissipation caused by nonequilibrium quasiparticles ? such a possibility is most intriguing in the range of frequencies @xmath283 [ see eq . ( [ bs ] ) ] and quasiparticle densities @xmath284 ( see fig . [ fig : axis_ns ] ) . in the following , we use the results of the two previous subsections to elucidate the dependence of the normalized dissipative conductivity @xmath285 on @xmath46 . for estimates , we assume the density of quasiparticles @xmath248 fixed , and their effective temperature low , @xmath286 , so that thermal ionization to the continuum does not prevent magnetic impurities from trapping the quasiparticles . in the absence of magnetic impurities , the dissipative response is defined by eq . ( [ dis1 ] ) . at small density , impurities reduce by @xmath46 the density of quasiparticles at the bottom of the continuous spectrum , yielding @xmath287 . in the interval of densities @xmath288 , the impurity band may accommodate all the quasiparticles , while being too narrow to allow absorption of the field of frequency @xmath2 . this low - absorption regime stretches up to the densities @xmath289 . at higher @xmath46 , the impurity band is broad enough to allow for intraband absorption ; compared to the benchmark @xmath290 dissipation level , eq . ( [ eq : sigmab - dirty3 ] ) yields @xmath285 smaller by a factor @xmath291 . upon further increase of @xmath46 , the impurity band merges with the continuum . deep in the ag regime , @xmath292 , one may use eq . ( [ dis3 ] ) for the estimate of dissipation ; compared to the @xmath290 dissipation level , eq . ( [ dis3 ] ) yields @xmath285 smaller by a factor @xmath293 . to summarize the above estimates , in the selected region of frequencies , @xmath283 , and nonequilibrium quasiparticle densities , @xmath294 , magnetic impurities indeed reduce the dissipative part of conductivity due to the reduction of the dos available for the quasiparticle transitions . finally , we reformulate the above results for the complex conductivity @xmath1 in terms of another important experimentally measurable quantity , the surface impedance:@xcite @xmath295 the impedance can be written as @xmath296 , where @xmath297 and @xmath298 are the surface resistance and reactance , respectively . the complex conductivity @xmath299 is given by eqs . ( [ sigma ] ) and ( [ qkeld ] ) . since we always consider the case of weak dissipation , @xmath300 , we also obtain @xmath301 . the response of the superconductor is then almost purely reactive with @xmath302 under assumptions of sec . [ sec : dissip ] , we can calculate @xmath303 in the simplest zero - temperature bcs model without magnetic scattering ( since corrections due to low temperature and weak magnetic scattering , @xmath304 , are negligible ) . at the same time , the new qualitative feature arising due to finite @xmath112 ( calculated in sec . [ sec : dissip ] ) , is the dissipation , i.e. , small but nonzero @xmath297 : @xmath305 in the near - threshold situation of sec . [ sec : dissip_zerot ] , we are interested in frequencies @xmath306 , then@xcite @xmath307 , and we find @xmath308 in the small - frequency limit considered in secs . [ sec : ag ] and [ sec : impband ] , at @xmath309 , we find@xcite @xmath310 and finally obtain @xmath311 thus , with the help of eqs . ( [ rth ] ) and ( [ rlow ] ) , the results for @xmath110 ( the dissipative part of conductivity ) calculated in sec . [ sec : dissip ] are directly translated into the surface resistance . we have employed the quasiclassical approach to study dissipation in superconductors due to small concentration of magnetic impurities . the superconductor was assumed to be in the dirty limit with respect to the potential ( nonmagnetic ) scattering . the concentration of magnetic impurities was assumed to be small enough so that the gap suppression is weak . employing the extension ( proposed by marchetti and simons@xcite ) of the usadel equation beyond the born approximation for magnetic scattering , we have considered in a unified manner both the ag regime ( with a continuum of states above a gap ) and the limit of lowest magnetic impurities concentrations where the impurity band due to overlap of localized impurity states is formed below the edge of the continuum . the deviation from the born limit was assumed to be finite but small , so that the subgap impurity states lie close to the continuum edge . our main results refer to the limit where the temperature @xmath79 , the frequency @xmath2 , and the gap suppression @xmath227 are all much smaller than the bcs gap @xmath21 in the absence of magnetic impurities . at the same time , the relation between @xmath79 , @xmath2 , and @xmath227 was assumed to be arbitrary , and we have obtained explicit analytical expressions in various limiting cases . our results can be expressed in terms of quantities describing the response of superconductor to external electromagnetic field : the dissipative conductivity @xmath112 ( the real part of the complex conductivity @xmath0 ) and the surface resistance @xmath297 ( the real part of the surface impedance @xmath312 ) . in the limit of small temperatures and small frequencies that we considered , at equilibrium , the dissipation is always proportional to the density of thermally excited quasiparticles and thus exponentially suppressed due to gapped character of the spectrum . on the other hand , fluctuations in positions of magnetic impurities can lead to finite dos below the mean - field gap.@xcite investigating the effect of the `` tail '' states on the dissipation is beyond the scope of our work . we only note that since the number of tail states is small ( compared to the number of the `` mean - field '' states considered in this paper ) , there is a wide region of applicability for our results at subgap temperatures @xmath313 , excluding only ultra - low temperatures . _ note added . _ during preparation of this paper , we became aware of a preprint by kharitonov _ _ et al.__@xcite addressing the same topic with a complementary approach . in that work the surface impedance of superconductors with magnetic impurities is evaluated numerically , with emphasis on the superconducting gapless regime having low energy states in the electron spectrum . we are grateful to m. h. devoret , r. j. schoelkopf , m. v. feigelman , and j. s. meyer for helpful discussions . ya.v.f . was supported by the rfbr ( grant no . 11 - 02 - 00077-a ) , the russian federal agency of education and science ( contract no . p799 ) , and the program `` quantum physics of condensed matter '' of the ras . this work was also supported by doe contract no . de - fg02 - 08er46482 ( yale ) and the nanosciences foundation at grenoble , france . thanks yale university for the hospitality . his visit was made possible by the gift of victor and marina vekselberg to yale university . in this appendix we demonstrate that the electron spectrum in the presence of magnetic impurities , regardless of their polarization , is insensitive to the nonmagnetic disorder . to see this we consider separately two different models . in sec . [ a : clean ] we derive explicitly the equation determining the dos in clean superconductors with randomly oriented spins and we establish the direct relation with the result obtained in sec . [ sec : dirty ] ( dirty superconductors , randomly oriented spins ) . in sec . [ a : dirty ] we formulate the usadel equation ( dirty limit ) for superconductors with polarized magnetic impurities and find that the dos corresponds to the one obtained in sec . [ sec : clean ] ( clean limit , polarized magnetic impurities ) . in the clean limit , the hamiltonian describing quasiparticles near the edge of the spectrum and interacting with randomly oriented weak magnetic impurities is @xmath314 where @xmath315 is the vector of the pauli matrices acting in the spin space and @xmath316 is the ( random ) orientation of the spin of impurity @xmath54 . [ the hamiltonian ( [ ham1 ] ) for polarized magnetic impurities corresponds to all @xmath317 . ] the @xmath79-matrix approximation for this model yields the self - consistency equation @xmath318 for the momentum - integrated green function @xmath74 , which is now diagonal in spin space . at low concentration of weak magnetic impurities ( @xmath37 and @xmath319 ) and for energies close to the edge of the continuous spectrum ( i.e. , @xmath320 ) , eq . ( [ grandom ] ) agrees with eq . ( [ u_eq ] ) that was derived in the dirty limit provided one identifies @xmath321 [ see also eq . ( [ tetd ] ) ] . in particular , the analysis of eq . ( [ grandom ] ) in the limit @xmath322 yields a narrow impurity band with @xmath323 for the total dos ( summed over quasiparticle spins ) and @xmath324 for the width [ which coincides with eq . ( [ wfin3 ] ) ] . following the derivation of eq . ( [ usadel_mod ] ) by marchetti and simons,@xcite we find that the usadel equation for the case of polarized magnetic impurities is @xmath325 here @xmath326 parametrizes the quasiclassical green functions that are different for electrons with spins up and spins down . at low concentration of weak magnetic impurities ( @xmath37 and @xmath319 ) and for energies close to the edge of the continuous spectrum , eq . ( [ usadel_modup ] ) in a uniform superconducting state can be rewritten as @xmath327 in terms of the spin - resolved momentum - integrated green functions @xmath328 . the equation for down spin is the same as eq . ( [ g ] ) derived in the clean limit , while the one for spin up does not produce an impurity band . the usadel equation in a superconductor reads @xmath329 -i[\check\sigma_s , \check g]=0 , \label{eq : usadel}\\ \hat\delta = \begin{pmatrix } 0 & \delta \\ \delta^ * & 0 \end{pmatrix},\end{gathered}\ ] ] where the quasiclassical green s function @xmath330 is a matrix in the nambu ( pauli matrices @xmath331 ) and keldysh spaces ( pauli matrices @xmath332 ) , @xmath333 and the self - energy @xmath334 describes magnetic scattering.@xcite the covariant derivative @xmath335 $ ] depends on the vector potential @xmath336 ( we choose the gauge without the scalar potential ) . time convolution is implicit in the matrix product . the superconducting gap @xmath337 enters the matrix ( in the nambu space ) @xmath338 and solves the self - consistency equation @xmath339 where @xmath340 is the bcs pairing constant . in the next order , @xmath344,\ ] ] so that @xmath345\right)\ ] ] in the same order . if we choose the london gauge with @xmath346 ( accompanied by the requirement of @xmath347 at the surface ) , then the usadel equation in the first order reads @xmath348.\ ] ] the l.h.s . of this equation is a linear operator @xmath349 acting on @xmath350 , while the source term in the r.h.s . is proportional to @xmath351 . taking the first order also in the self - consistency equation ( [ eq : self ] ) , we see that our choice of the gauge leads to @xmath352 and @xmath353 . ( it is actually well known that @xmath354 in the london gauge.@xcite ) the current flowing in the superconductor reads @xmath355 ( where @xmath201 is the drude conductivity ) , which in the lowest order in @xmath118 reduces to @xmath356 making the fourier transform , we obtain @xmath357 with the response kernel [ we omit the @xmath358 superscript of the green function for brevity ] the keldysh component of the green function can be expressed in terms of the retarded and advanced ones , while the advanced component can be written in terms of the retarded one , @xmath363 so that finally only the retarded component enters . in the case without the superconducting phase , it has the form @xmath364 [ @xmath183 and @xmath193 are the functions introduced in sec . [ sec : dirty ] , see eq . ( [ gftheta ] ) ] and we obtain @xmath365 - \notag \\ - & \tanh\frac{e_+}{2 t } \bigl [ g^*(e_- ) { \mathop\mathrm{re}\nolimits}g(e_+ ) + i f^*(e_- ) { \mathop\mathrm{im}\nolimits}f(e_+ ) \bigr ] \biggr\ } , \label{qkeld}\end{aligned}\ ] ] where @xmath366 [ note that the integrand in eq . ( [ qkeld ] ) is even with respect to @xmath151 ] . the difference of this expression for the kernel from the results of refs . is that our expression is written in terms of the _ quasiclassical _ green functions ( in the dirty limit ) that can be directly found from the usadel equation . a similar expression was employed in ref . , although it was formulated in terms of contour integrals in the plane of complex @xmath151 and the cuts in the plane were made from @xmath207 , the values of the ag gap ; thus , the ag regime with a single gap was explicitly assumed . our expression ( [ qkeld ] ) is more general , applicable also in the case when the impurity band is present and in the case of gapless superconductivity . defining the complex conductivity@xcite @xmath368 that determines the ( local ) response of current to electric field , @xmath369 we see that @xmath370 dissipation is determined by the real part of conductivity , @xmath371 , for which , according to eq . ( [ qkeld ] ) , we obtain eq . ( [ dissip ] ) given in the beginning of sec . [ sec : dissip ] . the exponential in @xmath59 factor of @xmath372 may be obtained by the standard means relating @xmath57 to the product of @xmath66 and its derivative at the midpoint between the impurities . however , obtaining this way the preexponential factor is problematic because of the oscillatory nature of wave function ( [ wf ] ) . the short - range nature of the potential forming @xmath66 allows us a full solution of the two - impurity problem . the main difference of the problem considered here from the corresponding one for a single quantum particle in the field of two short - range potentials@xcite stems from the structure of wave function ( [ wf ] ) which stays finite at @xmath373 . the usadel equation must be complemented by the self - consistency equation for the order parameter @xmath21 . as a result , @xmath21 is suppressed by magnetic impurities . we do not treat the self - consistency equation explicitly , assuming @xmath21 that we use to be the self - consistent order parameter corresponding to the given impurity parameters . alternatively , eq . ( [ u_ag ] ) can be rewritten as a fourth - order polynomial equation , and then an analytical solution is available , according to ref . ( see also the discussion of this solution in ref . ) . however , this formal solution is bulky and nontransparent , therefore rather than simplifying it in the corresponding limiting cases , we choose to consider those cases directly . in sec . va2a of ref . , maki considers the `` low - temperature and low - frequency '' limit , which probably means that @xmath79 and @xmath2 are much smaller than both @xmath21 and @xmath227 . the condition @xmath374 written above eq . ( 102 ) in ref . is probably a misprint for @xmath375 . this would mean that eq . ( 102 ) from ref . is obtained at @xmath376 , and in the limit when the last inequality becomes strong , we obtain agreement with our eq . ( [ dis6 ] ) . the self energy for magnetic impurities with randomly oriented spins reads@xcite @xmath377 for polarized magnetic impurities , the spin structure of the quasiclassical green function @xmath378 needs to be taken into account ( pauli matrices @xmath379 ) and the self energy reads @xmath380

the commonly accepted paradigm for gamma - ray bursts ( grbs ) is that they are formed in ultrarelativistic , collimated jets . typical terminal lorentz factors for these jets , such that the photons can freely escape , are @xmath1 a few hundreds , even above on thousand ( e.g. @xcite ) . opening angles @xmath2 are inferred from achromatic breaks in the afterglow light curves , although these are not clearly detected in several bursts ( e.g. , @xcite ) . since the lorentz factor decreases during the afterglow phase , and the break in the light curves occurs when the beaming angle equals @xmath2 , the product @xmath3 should be larger than one at the start of this phase , typically of the order of a few tens . the long / soft class of grbs are thought to be connected with the death of massive stars , since some of them are associated with type ic supernovae and are observed in star - forming regions of the host galaxies ( see e.g. , @xcite and references therein ) . during this process a compact central object and accretion disk are formed , and the jet is powered by either the neutrino annihilation , or by magnetic fields , tapping the rotational energy of the central object or disk . it is not clear which of the two mechanisms ( or both ) operates , with the detection or not of the thermal photospheric emission being a key factor @xcite . thermal ( fireball ) acceleration is in general a fast and efficient process , with the lorentz factor increasing linearly with the cylindrical distance from the symmetry axis . magnetic acceleration also works provided that the jet is supported externally by an environment whose pressure does not drop faster than the inverse square of the distance from the origin , as was analytically shown in @xcite . the interior of the progenitor star could very well play this role . relativistic magnetohydrodynamic ( mhd ) simulations by @xcite show that the efficiency of the magnetic acceleration is @xmath4 or more . the model faces two problems though : ( 1 ) the jet looses its external support when it exits the progenitor star , its motion becomes ballistic and its acceleration is practically halted . ( 2 ) the magnetic acceleration requires that the flow is expanded in a way such that the separation between neighboring streamlines increases faster than the cylindrical radius . this is achieved through stronger collimation of the inner part of the outflow relative to the outer part , and for this reason the mechanism was dubbed _ collimation - acceleration _ by @xcite . the resulting jets are very narrow with @xmath5 , and the product @xmath6 is close to unity before the start of the afterglow phase , making the breaks unlikely to happen . a solution to both problems can be given by carefully studying the dynamics at the regime where the jet comes out from the star , and its external pressure drops to practically zero . @xcite simulated this transition and found that it is accompanied by a spurt of acceleration . @xcite confirmed their finding numerically and interpreted it as _ rarefaction acceleration_. the loss of external support induces a sideways expansion of the jet , and a strong rarefaction wave that is driven into the flow and accelerates it . in fact this is a powerful mechanism seen in other numerical simulations of both , hydrodynamic and mhd flows with contact discontinuity and flow along it , see @xcite . + a similar mechanism was studied in @xcite , for the problem of an initially static magnetized plasma allowing to move into an environment . here we are interested for non - static cases and their sideways expansion . however , the analysis in all these works were based on time dependent simple waves , while for the grb problem under consideration it is more appropriate to use steady - state simple waves . in the present paper we develop a model for the steady - state , relativistic , magnetized , rarefaction wave . the work is a generalization of the classical steady - state , hydrodynamic rarefaction analyzed in @xcite , and its relativistic ( but again unmagnetized ) counterpart by @xcite . in section [ seceq ] we review the steady - state , special relativistic , mhd equations in planar geometry . in section [ secrss ] we develop the model , in section [ results ] we present and discuss the results and their application to grbs . finally , in section [ discuss ] we give a summary . the system of equations of special relativistic , ideal mhd , consist of the ohm s law @xmath7 the maxwell equations @xmath8 the mass @xmath9 momentum @xmath10 and entropy @xmath11 conservation equations ( e.g. , @xcite ) . + here @xmath12 is the velocity of the outflow , @xmath13 is the associated lorentz factor satisfying @xmath14 @xmath15 the electromagnetic field as measured in the central object s frame , @xmath16 the rest mass density , @xmath17 the gas pressure , and @xmath18 the specific enthalpy ( over @xmath19 ) , whose expression is , for an ideal gas with polytropic index @xmath20 , @xmath21 the polytropic index takes the values @xmath22 or @xmath23 in the limit of ultrarelativistic or nonrelativistic temperatures , respectively . ( any other value would imply a nonadiabatic evolution and hence requires the incorporation of heating / cooling terms into the entropy and momentum equations . see also @xcite for intermediate temperatures . ) by assuming steady state ( @xmath24 ) and a planar symmetric flow , i.e. , @xmath25 in a system of cartesian spatial coordinates @xmath26 , we can carry out a partial integration of the above equations ( [ ohm][enthalpy ] ) . it is also sufficient to simplify the analysis by assuming that the flow lies on the poloidal plane @xmath27 , @xmath28 , and the magnetic field in the transverse direction , @xmath29 . as discussed in section [ results ] , these are reasonable assumptions when the model is applied to grb outflows . it is possible to generalize the analysis to planar symmetric magnetized flows with nonzero @xmath30 and poloidal magnetic field ; this will be presented in a future paper . the continuity equation ( [ continuity ] ) for steady flows on the @xmath27 plane yields @xmath31 $ ] , and thus the flow velocity can be expressed as @xmath32 the stream function @xmath33 is constant along each streamline ( since @xmath34 ) , and can be used as its `` label '' . using ohm s equation ( [ ohm ] ) we express the electric field as @xmath35 . substituting in faraday s equation ( [ maxwell ] ) we find @xmath36 , which means that the quantity @xmath37 is a streamline constant , @xmath38 thus , the electric field can be written as @xmath39 . this expression shows the relation of the function @xmath40 with the scalar electric potential , and also that the streamlines are equipotentials . the component of the momentum equation ( [ momentum ] ) along the flow gives after some manipulation for @xmath41 in the left - hand side of the momentum equation ( [ momentum ] ) and then , by dotting with @xmath12 , we get @xmath42 /4\pi $ ] . the left - hand side , by replacing @xmath43 and using equation ( [ entropy ] ) , becomes @xmath44 . the right - hand side , using @xmath45 and equation ( [ phipsi ] ) , can be written as @xmath46 . since @xmath47 , this is equal to @xmath48 , and the equation of the two sides gives @xmath49=0 $ ] . thus , the quantity inside the brackets is a streamline constant . ] @xmath50 this integral represents the total energy - to - mass flux ratio ( over @xmath19 ) , since the mass flux ( times @xmath19 ) is @xmath51 , the poynting flux @xmath52 , and the matter energy flux ( including thermal , bulk kinetic and rest energy ) is @xmath53 . the entropy conservation equation ( [ entropy ] ) , for an ideal gas whose enthalpy is given by equation ( [ enthalpy ] ) , simplifies to @xmath54 , meaning that the quantity @xmath55 which is related to the specific entropy is a streamline constant @xmath56 the previous partial integrations greatly simplify the original system of equations ( [ ohm])([enthalpy ] ) , yielding several streamline constants , which can be determined at the boundary of the flow . three equations remain to be integrated : the component of the momentum equation ( [ momentum ] ) normal to the flow velocity , and equations ( [ lorentz ] ) , ( [ enthalpy ] ) . there are correspondingly three unknown functions , which we choose to be the stream function @xmath33 , the specific enthalpy @xmath57 , and the ratio of poynting - to - matter energy flux @xmath58 we may write the physical quantities in terms of these variables : @xmath59 @xmath60 @xmath61 knowing the streamline constants ( @xmath40 , @xmath62 , @xmath63 ) we can find the remaining unknowns @xmath64 by solving the following system of equations , with the first coming from equation ( [ enthalpy ] ) , the second from equation ( [ lorentz ] ) , and the third from the component of the momentum equation ( [ momentum ] ) normal to the flow velocity : @xmath65 @xmath66 @xmath67 \nonumber \\ -\frac{1}{2 } \nabla \left ( \frac{4\pi\mu c}{{\phi_\psi}}\frac{\sigma}{1+\sigma } \right)^2 \cdot \frac{\nabla { \psi } } { \mid \nabla { \psi}\mid^2 } + \frac{\mid \nabla { \psi}\mid^2}{2 } \frac{d{\phi_\psi}^2}{d{\psi } } \nonumber \\ -\frac{{\gamma}-1}{{\gamma } } \nabla \left [ 16 \pi^2 c^2 \frac{{h}({h}-1 ) \sigma } { { \phi_\psi}^2 } \right ] \cdot \frac{\nabla { \psi } } { \mid \nabla { \psi}\mid^2 } = 0\,.\end{aligned}\ ] ] the problem under consideration is basically a prandtl - meyer flow around a corner , and thus the appropriate coordinates are polar on the plane @xmath27 with the corner at the origin , defined through @xmath68 and @xmath69 , see fig . [ sketch ] . a self - similar flow is described with a stream function of the form @xmath70 , with constant @xmath71 . it is more convenient to replace the function @xmath72 in terms of the function @xmath73 , defined through @xmath74 , with constant @xmath75 , @xmath76 , in which case the definition of @xmath33 yields @xmath77 the function @xmath73 gives the radial distance from the corner , modulo a scale factor which is different in each streamline . this clearly shows that all streamlines are similar to each - other , hence the term `` self - similarity '' . the derivative of @xmath73 controls @xmath78 , and is thus related to the flow direction . rewriting equation ( [ a22-v ] ) as @xmath79 , where @xmath80 and @xmath81 the unit vectors of the polar coordinates , and defining the angle @xmath82 between the flow velocity and the @xmath83 axis ( see fig . [ sketch ] ) , the @xmath84 yields @xmath85 our goal is to separate the variables @xmath0 and @xmath86 in the system of equations ( [ x - m - v])([transfield - v ] ) , and reduce them to equations with respect to the polar angle @xmath86 alone . from inspection of equations ( [ x - m - v ] ) , ( [ bernoulli - v ] ) we require @xmath87 the last term of equation ( [ bernoulli - v ] ) should be a function of @xmath86 alone , and this gives the form of the streamline constant @xmath88 the so - called bernoulli equation ( [ bernoulli - v ] ) can then be written as @xmath89 or in differential form @xmath90 } { \sigma \left(1+\sigma\right ) \left[\mu^2 - { h}^2 \left(1+\sigma\right)^2\right ] } \frac{d\sigma}{d\theta } \ , , \label{odevartheta } \end{aligned}\ ] ] where we used the differential form of equation ( [ x - m - v ] ) @xmath91 @xmath92 is the square of the proper sound speed ( over @xmath19 ) @xmath93 the transfield equation ( [ transfield - v ] ) , after some manipulation using the previous two equations , gives @xmath94 after solving the system of equations ( [ x - m - v ] ) , ( [ ode10 ] ) , ( [ odethe0 ] ) , ( [ ode2 ] ) for the functions @xmath73 , @xmath95 , @xmath96 , @xmath97 , the physical quantities can be recovered using @xmath98 note that , using the previous expressions , the numerator and denominator of the differential equation ( [ ode2 ] ) can be written as @xmath99 @xmath100 is always positive , while @xmath101 can be written as @xmath102 ( using expression [ omegacoproper ] of appendix [ appendixa ] ) . near the corner the flow properties are expected to depend mostly on the polar angle @xmath86 ; their dependence on the coordinate @xmath0 is only weak . this requires the parameter @xmath71 to be @xmath104 , see equations ( [ densss])([ess ] ) ( the density is proportional to @xmath105 and the other quantities depend on @xmath0 through the density ) . the case @xmath106 corresponds to the classical rarefaction wave , a steady - state simple wave . it is the relativistic mhd generalization of the hydrodynamic steady - state rarefaction wave analyzed by @xcite in the nonrelativistic regime , and @xcite in the relativistic case . as in these studies , the assumption that the flow depends only on the polar angle @xmath86 leads to two possibilities : the first corresponds to a uniform flow , and the second to a rarefaction wave . by inspection of equation ( [ ode2 ] ) for @xmath106 one directly concludes that @xmath107 . the case with constant @xmath108 is the trivial one of a uniform flow , equation ( [ x - m - v ] ) implies that @xmath109 is also constant , equation ( [ odethe0 ] ) yields that @xmath110 , and the combination of the latter with equation ( [ ode10 ] ) gives that @xmath82 is also constant . ] while @xmath111 corresponds to the rarefaction wave . . the @xmath112 component of the flow proper velocity equals @xmath113 ( radius of cycle ) . ] a more robust perspective is to notice that @xmath111 implies that the @xmath81 component of the flow proper velocity ( @xmath114 ) is equal to the comoving proper phase velocity of a magnetosonic wave @xmath113 . equivalently , the lines @xmath115 constant intersect the streamlines at every point at the mach angle @xmath116 , i.e. , @xmath117 , see fig . [ sketch2 ] . this can be seen by noting that @xmath118 and @xmath119 ( see equation [ mach_opening ] of appendix [ appendixa ] ) . + the two sides of the mach cone are the two characteristics , with equations @xmath120 for @xmath117 we again conclude that the minus characteristics are the cones @xmath115 constant . suppose we are interested to model an outflow approaching a corner , see fig . [ sketch ] . the flow is initially uniform , in pressure equilibrium with its environment ( the @xmath121 regime in fig . [ sketch ] ) , and superfast - magnetosonic ( the bulk velocity is higher than the fast - magnetosonic wave speed ) . in such a flow the information is propagating in a mach cone around the flow speed , formed by the plus and minus characteristics . the effect of the corner is propagated only downstream of the minus characteristic that leaves the corner , and corresponds to the head of the rarefaction shown in fig . [ sketch ] . each fluid parcel keeps moving with constant speed till it crosses this line . from this point on the streamlines start to bent , the flow expands and its density , thermal / magnetic energy flux decline . as a result of the energy conservation the flow is accelerated . this mechanism of converting thermal / magnetic energy into kinetic energy of bulk motion is dubbed rarefaction acceleration by @xcite . the bending of streamlines and the acceleration of the flow continues till the angle @xmath122 , the tail of the rarefaction , where the flow becomes ballistic and pressureless ( in equilibrium with the vacuum ) . mathematically , in the initial uniform superfast - magnetosonic part of the outflow @xmath123 , @xmath101 is positive . the same is true in the first portion of the @xmath124 regime . however , as the flow moves in that part ( see fig . [ sketch ] ) the @xmath125 component of the flow velocity decreases leading to a decreasing @xmath101 ( see equation [ expd ] ) . eventually @xmath101 becomes zero at an angle @xmath126 corresponding to the head of the rarefaction , and remains zero in the whole rarefaction phase ( for @xmath127 ) . the system of equations ( [ x - m - v ] ) , ( [ ode10 ] ) , ( [ odethe0 ] ) , ( [ ode2 ] ) gives @xmath128 , @xmath82 , @xmath109 , @xmath108 at each @xmath129 $ ] . these expressions , together with their simplified versions in the limits of cold and unmagnetized flows , are given in appendix [ appendixb ] . the @xmath106 is the most important case ( and the only one with finite density at the corner ) , but we kept the analysis more general including @xmath130 cases ( for @xmath131 the density becomes infinity at the origin ) . in these cases the flow is nonuniform initially , with the density increasing with the distance from the corner . as a result , denser parts tend to move towards the less dense regions , and the resulting flow expansion provides an additional acceleration mechanism on top of the rarefaction acceleration which is still present . + the numerical procedure is to give the model parameters @xmath71 , @xmath20 , the initial quantities @xmath132 , @xmath133 , @xmath134 , and @xmath135 at some initial angle @xmath136 , and find @xmath62 and @xmath137 using equations ( [ a22-v ] ) , ( [ odethe0 ] ) . then solve the system of the two algebraic equations ( [ x - m - v ] ) , ( [ odethe0 ] ) together with the two differential equations ( [ ode10 ] ) , ( [ ode2 ] ) for the functions @xmath73 , @xmath95 , @xmath96 , @xmath97 . since we are interested to apply the model to grb outflows we set the energy - to - mass flux ratio ( @xmath62 ) , which equals the maximum possible bulk lorentz factor if all the energy is transferred to kinetic , a few hundreds . in particular , we choose a value @xmath138 in the numerical results . a jet associated with a long / soft grb is thought to be formed inside the progenitor star , and its first acceleration phase takes place before it crosses the stellar surface . we take as a reference value for the resulting bulk lorentz factor , which is the initial value for the rarefaction acceleration phase that we examine , @xmath139 . for a cold flow ( @xmath140 ) the magnetization is @xmath141 such that equation ( [ a22-v ] ) is satisfied . since the details of the acceleration phase inside the star are not known in general , where @xmath142 the distance from the origin and @xmath143 is related to the flow shape , see @xcite . for example , for @xmath144 we get @xmath145 , where @xmath146 is the stellar radius , while for @xmath147 we get @xmath148 . ] we also examine a model with @xmath149 ( and @xmath150 ) . if the jet is magnetically driven , it is superfast - magnetosonic when it crosses the stellar surface . it is well known from the mhd theory that in this regime the magnetic field is predominantly azimuthal , justifying our choice for ignoring the @xmath151 and @xmath152 components in the model . in @xcite ) . for typical values of a cylindrical distance @xmath153 and @xmath154 cm this ratio is @xmath155 . ] by adopting a planar geometry we ignore the tension of the azimuthal magnetic field . this is reasonable , since the fast variations induced by the rarefaction wave give a much larger magnetic pressure gradient in the radial ( @xmath156 ) direction . we also include a purely hydrodynamic model with @xmath157 and @xmath158 ( from equation [ [ a22-v ] ] ) , and an intermediate case with @xmath159 and @xmath160 . in all cases we started the integration from @xmath161 , with a flow parallel to the @xmath83 axis , @xmath162 . the results of the numerical integration for the rarefaction case @xmath163 are shown in fig . [ raref_gammacont ] for various sets of the initial quantities @xmath132 , @xmath133 and @xmath164 . the first column corresponds to the cold / magnetized case , the third to the hydrodynamic / unmagnetized , and the middle to the intermediate case . in each column the top panels show the variation of the three parts of the energy - to - mass flux ratio ( whose sum is the constant @xmath62 ) : the bulk kinetic ( including the rest mass energy ) which is the lorentz factor , the poynting which is written through the magnetization as @xmath165 , and the enthalpy @xmath166 . during the rarefaction phase the bulk acceleration to its full completion ( @xmath167 ) is clearly seen . the bottom panels show the geometry of the flow ( the solid lines are streamlines ) , together with the lorentz factor ( color ) . in agreement with the discussion in section [ sec_raref ] , three distinct regimes can be observed . the first is the unperturbed flow region ( @xmath168 ) where @xmath169 . the second is the rarefied region ( @xmath170 ) where @xmath13 increases . the perturbed region does not fill the whole space ; there is a maximum angle @xmath171 the so - called prandtl - meyer angle , or the tail of the rarefaction leaving the rest of the area ( @xmath172 ) void . the pressure equilibrium at the contact discontinuity between the flow and the void space ( @xmath122 ) implies that the thermal and magnetic pressures vanish . consequently the flow is ballistic along the streamline that pass through the corner , and the whole energy flux has been already transferred to kinetic energy flux ( @xmath167 ) . all other streamlines are starting to bent when they cross the head of the rarefaction and asymptotically they become parallel to the tail . during this phase the flow is accelerating , reaching @xmath167 asymptotically . the spatial scale in which this acceleration takes place strongly depends on the magnetization of the flow , something that has important consequences for the applications of the model . the bottom panels of fig . [ raref_gammacont ] show that the cold / magnetized case ( first column ) is accelerated much faster compared to the hydrodynamic case ( third column ) , with the intermediate case ( second column ) lying between these two limiting cases as expected . for example , when the streamline starting from @xmath173 reaches @xmath174 it has already @xmath175 in the former case , while in the later @xmath176 . axis ( @xmath177 ) , for various @xmath83 , and for two models : a cold / magnetized case ( left ) and a hydrodynamic case ( right ) . the scale of the distances is arbitrary.,title="fig:",width=154 ] axis ( @xmath177 ) , for various @xmath83 , and for two models : a cold / magnetized case ( left ) and a hydrodynamic case ( right ) . the scale of the distances is arbitrary.,title="fig:",width=154 ] , but for two solutions with @xmath178.,title="fig:",width=154 ] , but for two solutions with @xmath178.,title="fig:",width=154 ] + , but for two solutions with @xmath178.,title="fig:",width=154 ] , but for two solutions with @xmath178.,title="fig:",width=154 ] the numerical results are in a perfect agreement with the analytical relations given in appendix [ appendixb ] and summarized below . for the cold / magnetized case which is the most important and most efficient , the head of the rarefaction wave is located at @xmath179 ( corresponding to the half - opening angle of the mach cone for the fast - magnetosonic waves , see appendix [ appendixa ] ) . the tail is located at @xmath180 . note that this angle is always smaller than @xmath181 . if the flow inside the progenitor star is magnetically accelerated then its half - opening angle is expected to be @xmath181 @xcite . since @xmath182 , the rarefaction increases the lorentz factor without affecting much the opening angle , meaning that the product of the lorentz factor with the half - opening angle increases up to the value @xmath183 when the lorentz factor attains its maximum value @xmath62 . + as shown in the appendix [ appendixb ] during the acceleration the magnetization decreases as @xmath184 , where @xmath185 is proportional to the distance from the corner . a streamline starting at @xmath177 on the @xmath156 axis crosses the head of the rarefaction at @xmath186 . thus , @xmath187 and we get an analytical approximate expression for the lorentz factor @xmath188 the distance @xmath189 spans a range from zero corresponding to the corner up to a maximum value corresponding to the distance between the corner and the rotation axis of the jet , i.e. , the jet radius , which can be approximated as @xmath190 . at distance @xmath191 from the corner along each streamline ( i.e. , for each @xmath177 ) , the lorentz factor reaches half of its maximum value . as expected , fluid parcels on streamlines that are closer to the corner accelerate faster . + in terms of the streamline shape , equation ( [ b10 ] ) gives the analytical approximate relation between the lorentz factor and the angle @xmath82 between the flow speed and its initial orientation , @xmath192 in the hydrodynamic case the angle @xmath193 is smaller because the sound speed is smaller compared to the fast - magnetosonic speed . as a result the acceleration phase starts later and needs larger distances to reach completion . during the acceleration a combination of equations ( [ b11 ] ) and ( [ b13 ] ) gives the lorentz factor as a function of @xmath0 . for @xmath194 and @xmath195 we get the approximate result @xmath196 and thus @xmath197 , with constant @xmath198 . from this expression it is evident that the acceleration is much slower compared to the magnetized case . [ lraref_energy ] shows the result of the acceleration across the jet , for two models . clearly the cold / magnetized case ( left panel ) is much faster accelerated compared to the hydrodynamic case ( right panel ) . choosing the radius of the jet @xmath199 as the unit of distances we can find the lorentz factors in dimensional @xmath83 , and also estimate the efficiency of the acceleration in the whole jet ( which equals to the mean value of @xmath200 over @xmath138 ) . for example , at @xmath201 @xmath202 cm the mean @xmath13 is @xmath203 in the cold / magnetized case , and the total efficiency @xmath204 . the last column of fig . [ raref_gammacont ] corresponds to a cold / magnetized case with smaller @xmath132 and higher @xmath133 ( such that @xmath205 remains the same as in the other cases ) . it is interesting to note that , since @xmath206 is approximately the same as before , the dependence of @xmath13 on @xmath0 remains the same , see equation ( [ gamma - r ] ) . [ lraref_gammacont ] shows two solutions with @xmath207 , one cold / magnetized ( first column ) and one hydrodynamic ( second column ) . the initial flow in not uniform now , with the density increasing as we move away from the corner along constant @xmath86 . this allows for redistribution of the streamlines and acceleration even before the head of the rarefaction is crossed . this is indeed seen in the figures , in both the initial increase of the lorentz factor as well as in the bending of the flow . however , besides the initial phase the flows are very similar to the corresponding rarefaction cases @xmath106 . in the present paper we develop a model for the steady - state , relativistic , magnetohydrodynamic rarefaction wave . we use the method of self - similarity to reduce the system of partial differential equations to ordinary ones , which we then solve numerically . the model is a generalization of existing works for unmagnetized and nonrelativistic gas and can be applied in cases where plasma flows around a corner ( equivalently it loses its external support at some position ) . we apply the model to long / soft grb jets , which are formed inside the progenitor star and lose their external support when they cross the stellar surface . in particular , we used the model and successfully interpret the results of recent numerical simulations that show a spurt of acceleration in these jets , and more generally , whenever a contact discontinuity with a relativistic flow along its plane is present . between models with the same energy - to - mass flux ratio we find that the _ rarefaction acceleration _ is much faster in magnetized than in hydrodynamic flows . analytical scalings derived in appendix [ appendixb ] helped to quantify this behavior . for the cold / magnetized case we find that the flow reaches @xmath208 ( half of its maximum value , i.e. , 50% efficiency of acceleration ) at distance @xmath209 the above rough estimation corresponds to @xmath210 being the distance of the corner from the rotation axis , and @xmath211 for the stellar radius . ( for the hydrodynamic case this distance is a few orders of magnitude larger . ) our model assumes planar geometry and symmetry , which only locally hold near the points where the surface of the jet intersects the stellar surface . improvements include axisymmetric studies , and also to take into account the reflection of the wave on the rotation axis , which will possibly cause the lorentz factor to saturate at a value smaller than the maximum ( a crude approximation of the time needed for the information to start from a fluid parcel passing the corner , hit the axis and come back at the same parcel is @xmath212 ) . axisymmetric studies are inherently nonuniform ( one of the reasons being that the magnetization vanishes on the rotation axis where the azimuthal magnetic field should be zero ) . for this reason comparison of numerical simulations of axisymmetric jets with our model that assumes a uniform jet initially should be done with caution at distances far away from the corner . another limitation is the assumption of a zero external pressure outside the progenitor star . a finite external pressure will create a standing shock and a contact discontinuity between the jet and its environment , and also limit the terminal lorentz factor to some value smaller than its maximum . our model can not capture this inherently non - self - similar geometry . nevertheless it describes the basic physics of the mechanism and gives quantitatively correct results for most of the rarefaction acceleration phase , till the point where the shock is crossed . since the pressure contrast inside and outside the progenitor star is expected to be high , only the small shocked outflow part can not be described by our model . all our findings are very similar to the ones discussed in @xcite . this is surprising at first , since their study is time dependent and one dimensional in space , while ours is steady - state and two dimensional in space . the reason for this similarity is the so - called frozen pulse approximation , first introduced by @xcite for a relativistic hydrodynamic flow and extended by @xcite for the full relativistic mhd case . according to this approximation , when a time dependent flow is ultrarelativistic and superfast - magnetosonic , it can be described using steady - state equations . the full mathematical proof can be found in appendix [ appendixc ] . the physical reason is that each part of the flow moves practically with @xmath213 and can not communicate with neighboring parts through fast - magnetosonic waves ( which also move at most with @xmath213 ) . thus , a possible time dependence of the flow quantities at some point of space is carried with the flow as a frozen pulse , and the motion of each part is effectively time independent . we thank the referee for many helpful comments . this research has been co - financed by the european union ( european social fund esf ) and greek national funds through the operational program `` education and lifelong learning '' of the national strategic reference framework ( nsrf ) - research funding program : heracleitus ii . investing in knowledge society through the european social fund . nv acknowledges partial support by the special account for research grants of the national and kapodistrian university of athens ( `` kapodistrias '' grant no 70/4/8829 ) . m. a. , rezzolla l. , 2006 , apj , 640 , l115 s. b. , frail d. a. , harrison f. a. , kulkarni s. r. , nakar e. , chandra p. c. , butler n. r. , et al . 2010 , apj , 711 , 641 h. h. , 1973 , physics of fluids , 16 , 825 a. , 1982 , physics of fluids , 25 , 1165 j. , komissarov s. s. , spitkovsky a. , 2011 , mnras , 411 , 1323 kolosnitsyn n. , stanyukovich k. , 1984 , journal of applied mathematics and mechanics , 48 , 96 s. s. , vlahakis n. , knigl a. , 2010 , mnras , 407 , 17 s. s. , vlahakis n. , knigl a. , barkov m. v. , 2009 , mnras , 394 , 1182 a. , 1980 , phys . fluids , 23 , 1083 landau l. d. , lifschitz e. m. , 1987 , _ fluid mechanics_. pergamon press , oxford , 109 e .- w . , racusin j. l. , zhang b. , zhang b .- b . , burrows d. n. , 2008 , apj , 675 , 528 y. , sari r. , 2001 , apj , 555 , 540 m. , 2011 , mnras , 411 , 422 j. , masada y. , shibata k. , 2012 , apj , 751 , 140 y. , hardee p. , hartmann d. h. , nishikawa k .- i . , zhang b. , 2008 , apj , 672 , 72 a. , zhang b .- b . , ryde f. , mcglynn s. , zhang b. , preece r. d. , kouveliotou c. , 2012 , mnras , 420 , 468 t. , shemi a. , narayan r. , 1993 , mnras , 263 , 861 j. l. , liang e. w. , burrows d. n. , falcone a. , sakamoto t. , zhang b. b. , zhang b. , evans p. , osborne j. , 2009 , apj , 698 , 43 a. , narayan r. , mckinney j. c. , 2010 , newa , 15 , 749 n. , knigl a. , 2003 , apj , 596 , 1080 s. , hesse m. , klimas a. , 2010 , apj , 712 , 951 b. , 2011 , comptes rendus physique , 12 , 206 b. , peer a. , 2009 , apj , 700 , l65 x .- h . , li z. , bai j .- , 2011 , apj , 726 , 89 suppose that we study a magnetosonic disturbance on the poloidal plane . its phase speed in the comoving is @xmath214 with @xmath215 ( e.g. , using the expressions given in appendix c of @xcite for propagation normal to the magnetic field , @xmath216 ) . here @xmath217 and @xmath218 . the corresponding proper speed ( over @xmath213 ) is @xmath219 since the propagation is isotropic , the group velocity is equal to the phase velocity , @xmath220 . transforming the dispersion relation to the central object s frame , we get @xmath221 or equivalently @xmath222 the group velocity in the central object s frame can be found from the transformation of the lorentz factors @xmath223 , or , @xmath224 where @xmath225 is the angle between @xmath226 and the flow direction . the above equation can be solved for @xmath227 : @xmath228 all directions @xmath225 which give real values for the group velocity form a mach cone around the flow direction , with half - opening @xmath116 ( the maximum allowed @xmath225 ) given by @xmath229 note that the result is a direct generalization of the nonrelativistic @xmath230 , with the proper speeds replacing their newtonian counterparts @xcite . an alternative way to find @xmath116 follows : + assume a system of coordinates on the poloidal plane such that @xmath231 is along the flow velocity and @xmath232 normal to it . ( note that this is not the same with the @xmath27 system of coordinates adopted in the main body of the paper , in which the velocity makes an angle @xmath82 with the @xmath83 axis . ) consider a disturbance starting at @xmath233 from the line @xmath234 ( for all @xmath235 ) . in the comoving frame the disturbance starts at @xmath236 from the line @xmath237 , and after some time @xmath238 affects a cylindrical regime @xmath239 , since its group velocity is @xmath220 ( given by [ omegaco ] ) . in the central object s frame that regime is lorentz transformed to @xmath240 , or equivalently to the elliptic cylinder @xmath241 an equation of the form @xmath242 . the area to which the disturbance is propagating is limited by the envelope of these elliptic cylinders . solving the system @xmath243 we find the two characteristic planes @xmath244 , and thus the angle @xmath116 between the envelope and the flow velocity is given by @xmath245 an expression equivalent to [ mach_opening ] . + ( the substitution of @xmath246 and @xmath247 in equation [ front ] is an alternative way to find equation [ vglab ] for the group velocity in each direction . ) here we give the equations that characterize the rarefaction regime @xmath248 ( for the case @xmath106 ) . the head of the rarefaction corresponds to @xmath126 . since @xmath249 , @xmath250 ( using expression [ mach_opening ] of appendix [ appendixa ] ) . here subscripts `` j '' refer to the uniform initial phase . using the normalized density @xmath251 as the independent variable , equation ( [ x - m - v ] ) gives @xmath252 equation ( [ ode2 ] ) ( which simplifies to @xmath253 ) gives @xmath254 with @xmath255 , + equation ( [ odethe0 ] ) gives @xmath82 through @xmath256 with @xmath257 , + and the differential equation ( [ ode10 ] ) implies @xmath258 at the tail of the rarefaction wave @xmath259 the thermal and magnetic energy fluxes vanish ( @xmath260 and @xmath261 ) while @xmath262 and @xmath263 . the position of the tail is @xmath122 with @xmath264 the @xmath156 component of the velocity is @xmath265 with @xmath266 and @xmath267 . for a highly superfast - magnetosonic and ultrarelativistic flow @xmath268 the relation @xmath269 simplifies to @xmath270 where @xmath271 @xmath272 \left[1+(2-{\gamma})({h}-1)\right ] } \frac{d\varrho}{\varrho } } { { h}_j { \gamma}_j ( 1+\sigma_j ) } \,.\end{aligned}\ ] ] in that limit @xmath273 the previous expressions can be greatly simplified . we find @xmath274 , @xmath275 , @xmath276 , @xmath277 , and if the flow is highly superfast - magnetosonic and ultrarelativistic @xmath278 , @xmath279 for the head @xmath280 we get @xmath281 , and for the tail @xmath282 we find @xmath283 . + the direction of the flow is given by @xmath284 + the streamlines in the rarefaction regime @xmath285 are ( in polar coordinates ) @xmath286 , or , @xmath287 different values of @xmath288 give different streamlines . for a streamline that crosses the angle @xmath289 at @xmath290 we get @xmath186 and @xmath291 . for the unmagnetized case ( @xmath261 ) similar approximations yield @xmath292 and if the flow is highly superfast - magnetosonic and ultrarelativistic @xmath293 @xmath294 } \frac{d\varrho}{\varrho } } { { h}_j { \gamma}_j } \quad \\ = \frac { { \cal i } \left[\left(2-{\gamma}\right)\left({h}_j-1\right)\right ] -{\cal i } \left[\left(2-{\gamma}\right)\left({h}_j-1\right ) \varrho^{{\gamma}-1}\right ] } { { h}_j { \gamma}_j \left({\gamma}-1\right)^{1/2}\left(2-{\gamma}\right)^{1/2 } } -\frac{{h}u_s}{{h}_j { \gamma}_j } \ , , \nonumber\end{aligned}\ ] ] where @xmath295 \equiv \zeta^{1/2 } \left(1+\zeta\right)^{1/2 } + \ln \left[\zeta^{1/2}+\left(1+\zeta\right)^{1/2}\right ] $ ] . + for the head @xmath296 we get @xmath297 and for the tail @xmath298 we find @xmath299 } { { h}_j { \gamma}_j \left({\gamma}-1\right)^{1/2}\left(2-{\gamma}\right)^{1/2 } } \,.$ ] + for the distance form the corner we get @xmath300 + note that for @xmath301 , @xmath302 , and @xmath303 we recover the relations of section [ coldapp ] for the cold magnetized limit . this is because , for a transverse magnetic field , the magnetic pressure @xmath304 is proportional to the square of the rest mass density ( see equation [ [ phipsi ] ] ) , and thus it is analogous to a polytropic relation with index @xmath305 . for @xmath28 , @xmath306 , and @xmath307 , the electric field is @xmath308 and equations ( [ maxwell][entropy ] ) become , after some manipulation , @xmath309 @xmath310 @xmath311 @xmath312 @xmath313 ( the last two equations correspond to the components of the momentum equation along and normal to the flow . ) + for ultrarelativistic flows with @xmath314 we can simplify the above system , by ( i ) using @xmath315 , ( ii ) dropping the right - hand side of equation [ momentumappv ] ( since the left - hand side includes much larger terms note that @xmath316 ) , and ( iii ) noting that @xmath317 reveals that the assumption @xmath318 holds only in the superfast - magnetosonic regime of that equation we kept the term @xmath319 but not the @xmath320 , something that is correct if @xmath321 , or , @xmath322 . ] . the resulting system gives three integrals of motion @xmath323 ( which in principle are different for different parts of the flow ) , and the equations @xmath324 @xmath325 by inspection of the previous equations we see that the derivatives @xmath326 and @xmath327 always come as a combination @xmath328 , and thus the variables @xmath83 and @xmath329 can be interchanged . the steady state problem where @xmath330 and the flow depends on @xmath83 and @xmath156 is mathematically equivalent to the time - dependent one - dimensional problem where @xmath331 and the flow depends on @xmath332 and @xmath156 . the above is a manifestation of the `` frozen pulse '' behavior of an ultrarelativistic flow , first introduced by @xcite for hydrodynamic flows and extended by @xcite in the mhd case . due to the ultrarelativistic and superfast - magnetosonic velocity of the flow , any possible disturbance is traveling with it and can not affect the neighboring parts . as a result the evolution of each fluid parcel is essentially steady - state . in fact , changing variables from @xmath333 to @xmath334 where @xmath335 , we transform equations [ frozen1 ] , [ frozen2 ] to @xmath336 @xmath337 which are the same with the steady - state equations in the same ( ultrarelativistic ) limit . note however that the partial derivatives @xmath327 , @xmath338 are now taken keeping @xmath339 ( and not @xmath340 ) constant . since the motion is relativistic in the @xmath83 direction the variable @xmath339 is practically constant for each fluid parcel and corresponds to the time in which it passed a certain position @xmath341 . without loss of generality we can set @xmath342 ; in that case @xmath343 . the absence of @xmath339 and @xmath344 in equations [ frozen1s],[frozen2s ] means that they do not constrain the @xmath339 dependence on any flow quantity @xmath345 . this dependence is determined by the initial / boundary conditions only , i.e. , by the values of the flow quantities for each fluid parcel at time @xmath346 when it passes @xmath347 . in other words , we can find the evolution of a time - dependent flow by applying steady - state solutions to each part of the flow passing from @xmath347 at time @xmath343 , by changing only the boundary conditions ( see an example in section 4.1.1 of @xcite ) . in the particular case of the relativistic rarefaction wave , the frozen pulse approximation obviously holds indeed holds in the @xmath348 case in which the flow depends only on @xmath349 . ] . as a result , the steady - state solutions considered in this work can be used for the description of a time - dependent flow , and this can be achieved by simply writing the similarity variable as @xmath350 . thus , we only need to make the substitution @xmath351 ( with constant @xmath346 for each part of the flow ) in order to recover the equations of the time - dependent rarefaction wave with ultrarelativistic velocity in the @xmath83 direction , considered in @xcite . the worldlines of all fluid parcels passing at time @xmath352 from the plane @xmath347 ( at various @xmath353 ) are equivalent to the streamlines of the steady - state model . this is indeed the case in the numerical results . choosing initial conditions as the ones in fig . 4 of @xcite we get practically identical results , by just substituting @xmath354 .

the qcd phase diagram is a topic which currently sees a lot of attention in experimental and theoretical work . a particularly interesting ( and rather unexplored ) part is the region at low temperatures and medium to high densities ( medium to high values of the chemical potential @xmath1 ) . while for large chemical potential exotic states of matter are conjectured , for small chemical potential and vanishing temperature one expects physics to be independent of @xmath1 : as long as the chemical potential is smaller than the mass of the state it couples to no excitations can be generated and physics must be independent of @xmath1 . this is an interesting non - perturbative effect which has come to be known as silver blaze problem @xcite . in principle lattice qcd would be an adequate method to study such non - perturbative phenomena . however , when coupling a chemical potential the fermion determinant becomes complex and can not be used as a probability weight in a monte carlo simulation . this so called complex action problem is a common feature in many field theories at non - zero @xmath1 and has considerably held up the study of qcd and other systems at finite density . using new ideas such as worm algorithms @xcite , designed for updating systems with local constraints , several interesting models at finite density were rewritten in terms of a flux representation where the complex action problem is avoided and worm - based monte carlo simulations are possible ( see , e.g. , @xcite for examples ) . this opens the possibility to explore the silver blaze phenomenon in an ab - initio lattice study . in this article we analyze the silver blaze problem for a charged scalar @xmath0 field . for scalar fields a suitable flux representation @xcite can be found which avoids the complex action problem at finite @xmath1 and monte carlo simulations at high densities become possible . the model has been analyzed previously using the complex langevin approach and mean field techniques @xcite . in this paper we systematically study the transition at the end of the silver blaze region at zero temperature and attempt to follow the transition line for non - zero temperatures . this work sheds light on silver blaze type of phenomena and explores new strategies for overcoming the complex action problem in lattice calculations at finite density . the article is organized as follows : in the next section we present the flux representation and our observables . in section 3 the worm algorithm is described ( see also the appendix for a pseudo - code listing ) and tested . the numerical analysis at vanishing temperature is discussed in section 4 , followed by the results for non - zero temperature in section 5 . a summary and discussion section completes the paper . in the conventional representation the lattice action of the charged scalar field with a @xmath0-interaction is given by @xmath10 \right ) .\ ] ] the first sum is over the sites @xmath11 of a @xmath12 lattice with periodic boundary conditions , the second sum is over the four euclidean directions @xmath13 , and @xmath14 denotes the unit vector in @xmath15-direction . the degrees of freedom are the complex valued field variables @xmath16 attached to the sites @xmath11 of the lattice . @xmath17 denotes the combination @xmath18 , where @xmath4 is the bare mass parameter . the @xmath0 coupling is denoted by @xmath19 and the chemical potential by @xmath1 . all parameters are in units of the lattice spacing @xmath20 , in other words the lattice spacing is set to @xmath21 throughout this paper , except for some places where we display @xmath20 explicitly for clarity . the partition sum @xmath22 is obtained by integrating the boltzmann factor @xmath23 over all field configurations , @xmath24 e^{-s}$ ] . the measure is a product over all lattice sites @xmath11 , with @xmath16 being integrated over the complex plane , i.e. , @xmath25 = \prod_x \int_\mathds{c } d \phi_x / 2\pi$ ] . the normalization with @xmath26 will be useful later . note that in its present form the model has a complex action problem : for @xmath27 the sum of the temporal nearest neighbor terms is complex . thus the boltzmann factor can not be used as a probability weight in a monte carlo simulation . the problem can be overcome by using alternative degrees of freedom , i.e. , a flux representation . the first step to obtain the flux representation of the partition sum is an expansion of the boltzmann factors for the nearest neighbor terms @xmath28 } \right ) \ ! \left ( \prod_{x,\nu}\ ! \big(\phi_x^\star \phi_{x+\widehat{\nu}}\big)^{n_{x,\nu } } \ , \big(\phi_x \phi_{x+\widehat{\nu}}^\star\big)^{\overline{n}_{x,\nu } } \ ! \right ) \!= \nonumber \\ \hspace*{-6 mm } & & \sum_{\ { n , \overline{n}\ } } \!\ ! \left ( \prod_{x,\nu}\ ! \frac{1}{n_{x,\nu } ! \ , \overline{n}_{x,\nu } ! } \right ) \left(\ ! \prod_x e^{\,\mu [ n_{x,4 } - \overline{n}_{x,4 } ] } \ , { \phi_x^{\,\star}}^{\sum_\nu [ n_{x,\nu } + \overline{n}_{x-\widehat{\nu},\nu } ] } \ , { \phi_x}^{\sum_\nu [ \overline{n}_{x,\nu } + n_{x-\widehat{\nu},\nu } ] } \right ) , \nonumber \end{aligned}\ ] ] where by @xmath29 we denote the sum over all configurations of the expansion variables @xmath30 , i.e. , the multiple sum @xmath31 the complex field variables are now written in polar form , @xmath32 . rewriting also the integration measure in polar coordinates one finds for the partition sum @xmath33 } \ ! \right ) \nonumber \\ & \!\!\!\times\!\!\!&\hspace{-2 mm } \left(\ ! \prod_x e^{\,\mu [ n_{x,4 } - \overline{n}_{x,4 } ] } \int_{0}^\infty\!\!\ ! dr_x \ ; r_x^{1 + \sum_\nu [ n_{x,\nu } + n_{x-\widehat{\nu},\nu } + \overline{n}_{x,\nu } + \overline{n}_{x-\widehat{\nu},\nu } ] } \ ; e^{-\eta r_x^2 - \lambda r_x^4 } \right)\ ! . \end{aligned}\ ] ] the integrals over the phase give rise to kronecker deltas , which for notational convenience we will write as @xmath34 . the integrals over the modulus we abbreviate as @xmath35 they can easily be computed numerically . the partition sum now reads : @xmath36 \ ! \right ) \right ) \nonumber \\ & \!\ ! \times \!\ ! & \!\ ! \prod_x e^{\,\mu [ n_{x,4 } - \overline{n}_{x,4 } ] } \ ; w\!\left ( \sum_\nu \big [ n_{x,\nu } + \overline{n}_{x,\nu } + n_{x-\widehat{\nu},\nu } + \overline{n}_{x-\widehat{\nu},\nu } \big ] \right ) \right)\ ! . \end{aligned}\ ] ] in this form the complex phase problem is gone . all weight factors for configurations of the @xmath37 and @xmath38 variables are real and non - negative . however , configurations of the @xmath37 and @xmath38 can still have weight 0 . this is always the case if at a site @xmath11 the constraint enforced by the kronecker deltas is violated , i.e. , when at least for one @xmath11 one has @xmath39 \neq 0 $ ] . in the current representation the constraints mix both , the @xmath37 and the @xmath38 variables . the structure of the constraints can be simplified by introducing new variables @xmath40 and @xmath41 . they are related to the old variables by @xmath42 the partition sum turns into @xmath43 \ ! \right ) \right ) \nonumber \\ & \!\ ! \times \!\ ! & \!\ ! \left(\ ! \prod_x e^{\,\mu k_{x,4 } } \ ; w\!\left ( \sum_\nu \big [ |k_{x,\nu}| + |k_{x-\widehat{\nu},\nu}| + 2 ( l_{x,\nu } + l_{x-\widehat{\nu},\nu } ) \big ] \right ) \right)\ ! . \label{zfinal}\end{aligned}\ ] ] in the final form ( [ zfinal ] ) , which we now refer to as flux representation , the constraints no longer mix the two types of flux variables . obviously only the @xmath44-fluxes are subject to conservation of flux at each site @xmath11 , i.e. , only they must obey @xmath45 = 0 $ ] for all @xmath11 . also observables need to be mapped to the flux representation . for thermodynamic observables which are obtained as derivatives of @xmath46 with respect to the various couplings this is particularly simple . we will need the following observables which we refer to as particle number density @xmath37 , the particle number susceptibility @xmath47 , the derivative @xmath48 of @xmath47 with respect to @xmath1 , the field expectation value @xmath49 , the corresponding susceptibility @xmath50 and its derivative @xmath51 with respect to @xmath1 . they are defined as @xmath52 all observables are normalized by the spatial volume @xmath53 in order to make them intensive quantities . we remark , that the observables ( [ obslist ] ) are dimensionful ( except @xmath54 ) . since we here use a fixed scale approach ( the lattice spacing @xmath20 is set by @xmath17 and @xmath19 and kept fixed ) and drive the temperature @xmath55 by changing @xmath56 , we will often show results in lattice units . where a comparison of results at different scales is necessary we make quantities dimensionless using suitable powers of @xmath2 , the critical value of the chemical potential at zero temperature ( see the discussion below ) . it is straightforward to work out the derivatives that define the observables ( [ obslist ] ) also for the flux representation of @xmath22 and we only discuss two examples to illustrate the general structure . particularly simple is the particle number density . one immediately sees that the flux representation of @xmath37 is given by @xmath57 i.e. , the particle number is simply the net flux of @xmath44 around compactified time (= the 4-direction ) . the expectation value @xmath58 is now of course understood as a vacuum expectation value in the flux representation . not much more difficult is the flux representation for @xmath59 . note that @xmath17 enters only through the weight factors @xmath60 defined in ( [ ww ] ) which have the obvious property @xmath61 . application of the product rule to the product of the factors @xmath62 in ( [ zfinal ] ) and subsequent completion of the remaining products to the full boltzmann factor gives rise to @xmath63 where we use the abbreviation @xmath64 $ ] . in a similar way the other observables listed in ( [ obslist ] ) can be obtained as higher moments of the fluxes and the sum of ratios of @xmath62 . having discussed the flux representation of the partition sum and the observables , we now address the task of finding a suitable algorithm . in the final representation ( [ zfinal ] ) of the partition sum we use two different sets of variables . one of them , the variables @xmath65 , are not subject to a constraint . for these we simply apply sweeps of local metropolis updates : each link @xmath66 is visited and we generate a trial configuration by increasing or decreasing @xmath65 by @xmath67 with equal probability . the change is accepted with the usual metropolis probability given by the ratio of the weights of new and old configuration . a negative trial link is always rejected . considerably more interesting is the update of the flux variables @xmath68 where a suitable update needs to take into account the local constraints @xmath45 = 0 $ ] . for that purpose we use a generalization of the prokofev - svistunov worm algorithm @xcite . the worm starts at a randomly chosen site of the lattice and then grows by adding links to form a chain of links . at each position @xmath11 the worm chooses randomly the next link from the 8 links attached to @xmath11 . for that link it proposes to change the corresponding variable @xmath44 by @xmath67 and accepts that change with a local metropolis step . at the starting point and the head of the worm the constraint is violated , and the worm must continue until its head reaches the starting point such that along the closed contour of changed links all constraints are intact again . a possible initial configuration for the @xmath69 is to set @xmath70 at all links , a configuration that clearly obeys the constraints . the situation for a worm update of the variables @xmath68 is a little bit tricky : the local weight for a variable @xmath68 not only has factors that live on the link ( the factorials in @xmath71 and the contribution from the chemical potential ) , but also two factors at the two endpoints of the link , the terms @xmath72 and @xmath73 ( where again @xmath74 $ ] ) . the contributions from the factors @xmath62 at the endpoints need to be suitably distributed among subsequent steps of the worm . in the appendix we provide pseudo - code for the initialization of the worm and the subsequent steps , illustrating how we solved the weight assignment for the @xmath62 contributions . worms for the variables @xmath68 are alternated with the local metropolis sweeps for the unconstrained variables @xmath65 . in our simulations we use @xmath12 lattices with @xmath75 ranging from 4 to 24 , and @xmath56 from 2 to 100 . a full update sweep is the combination of one sweep for the @xmath65 variables and one worm for the @xmath68 fluxes . we typically use between 500.000 and 1.000.000 measurements separated by 5 full update sweeps . the number of equilibration sweeps is typically 25.000 . all errors we quote are statistical errors determined with the jackknife method . the weights @xmath62 defined in ( [ ww ] ) were evaluated numerically using mathematica and pre - stored for the monte carlo simulation . from the flux simulation at @xmath76 ( symbols ) to the exact result obtained from fourier transformation ( curves ) . we plot @xmath37 as a function of @xmath17 for different values of @xmath1 using lattices of size @xmath77 . plot : results for the expectation value @xmath78 as a function of @xmath17 for different values of @xmath19 , again on lattices of size @xmath77 . the chemical potential is set to @xmath79 and we compare the results from the flux simulation ( triangles , circles , squares ) to the outcome of a conventional @xmath79 simulation ( crosses).,scaledwidth=100.0% ] the algorithm was thoroughly tested by comparing it to the free case ( @xmath76 ) where exact results can be obtained using fourier transformation , and to a conventional simulation at @xmath79 . in fig . [ checks ] we show examples for the comparison using @xmath77 lattices . the lhs . plot shows the particle number density @xmath37 as a function of @xmath17 for different values of the chemical potential @xmath1 for the case of @xmath76 . the symbols are the data from our flux simulation and the full curves the results from fourier transformation . the rhs . plot compares the @xmath79 results for the expectation value @xmath78 from the flux simulation ( triangles ) to those of a conventional simulation ( crosses ) . we plot @xmath78 as a function of @xmath17 for different values of @xmath80 . the two plots show that the results from the flux simulation very nicely match the exact curves for the free case and the data from a conventional simulation at @xmath79 . further comparisons were done for other observables and more values of the parameters all with very good agreement . finally we performed a comparison with the complex langevin results at finite @xmath1 by aarts @xcite , and we found excellent agreement of our data with the complex langevin results that are available on lattices of sizes @xmath81 to @xmath82 . it is interesting to note that these results are very well described also by mean - field theory @xcite . in that study the quartic coupling was fixed to @xmath83 , and the boson mass @xmath84 satisfied @xmath85 . as a result of this heavy mass , the temperatures reached on lattices @xmath86 to @xmath82 are in the range @xmath87 . at the same time , the system is large , since @xmath88 . so the observed mean - field behavior should come as no surprise . our comparison with @xcite confirms the findings by aarts . to summarize : our tests establish that the flux representation is correct , that the algorithm works , and that the observables were correctly mapped to the corresponding expressions in terms of the fluxes . we begin our study of the phase diagram with an analysis of the situation at zero temperature . for this analysis we set the temporal extent of the lattice to @xmath89 , such that in lattice units the temperature is @xmath90 . although not exactly 0 , this is much smaller than all other involved scales . to study the approach of our observables to the thermodynamic limit we analyze different spatial volumes @xmath91 using @xmath92 and 24 . we consider two different combinations of the couplings @xmath17 and @xmath19 . the two points in parameter space are @xmath93 and @xmath94 . for @xmath95 and @xmath96 . in the lhs . column of plots we show @xmath37 , @xmath47 and @xmath97 as a function of @xmath1 ( top to bottom ) . in the rhs . column we show @xmath78 , @xmath50 and @xmath51 . we compare the results for @xmath98 and @xmath99 . , scaledwidth=110.0% ] in fig . [ all_k450_t001 ] we show the @xmath95 , @xmath96 results for @xmath37 , @xmath47 and @xmath97 as a function of @xmath1 in the lhs . column of plots , and for @xmath78 , @xmath50 and @xmath51 on the rhs . the particle number density @xmath37 vanishes up to a critical value of @xmath100 which marks the end of the silver blaze region where observables are independent of @xmath1 . for values @xmath101 the particle number density rises essentially linearly . consequently the particle number susceptibility @xmath47 is discontinuous at @xmath2 , jumping from @xmath102 to a finite value at @xmath2 . the comparison of the different spatial volumes illustrates how the discontinuity emerges in the thermodynamic limit . finally , when looking at the third @xmath1-derivative @xmath48 of the free energy we observe that for increasing volume a strongly peaked maximum emerges that very precisely marks the value of @xmath100 . furthermore , the height of the peak grows with increasing volume @xmath53 ( for a quantitative assessment see below ) , indicating the emergence of a true singularity in the thermodynamic limit . the observables @xmath78 , @xmath50 and @xmath51 behave similarly to the derivatives with respect to @xmath1 : the expectation value @xmath78 is constant below @xmath2 and for @xmath101 increases with constant slope . the corresponding susceptibility @xmath50 thus also develops a step in the thermodynamic limit . finally , another derivative with respect to @xmath1 , i.e. , @xmath51 , shows the corresponding peaks , which again grow with the volume . for @xmath103 and @xmath96 . in the lhs . column of plots we show @xmath37 , @xmath47 and @xmath97 as a function of @xmath1 ( top to bottom ) . in the rhs . column we show @xmath78 , @xmath50 and @xmath51 . we compare the results for @xmath104 and @xmath99 . , scaledwidth=111.0% ] the results for the second parameter set , @xmath103 , @xmath96 are shown in fig . [ all_k372_t001 ] . again we display @xmath37 , @xmath47 , @xmath97 , @xmath78 , @xmath50 and @xmath51 as a function of @xmath1 . in the plots we omit the data at @xmath105 since for these the transition point in @xmath1 is shifted out of the @xmath1-range of our plots due to finite volume effects . for the larger volumes the behavior of the observables is the same as for the @xmath95 runs , but now the critical value @xmath2 that marks the end of the silver blaze region is lower at @xmath106 in lattice units . again we observe scaling of the maxima of the third derivatives with the volume and conclude that also at @xmath103 , @xmath96 we find a second order transition at @xmath106 . we remark that the drop of @xmath100 in lattice units at @xmath95 to @xmath106 at @xmath103 is due to a considerable decrease of the lattice constant @xmath20 . this implies that at @xmath103 we work with a smaller lattice spacing @xmath20 and thus with a smaller physical spatial extent @xmath107 . thus we expect to see more severe finite volume effects at @xmath103 than at @xmath95 . on the other hand with the finer lattice spacing we will be able to go to higher temperatures ( see the next section ) . to further confirm the finding of a singularity at @xmath2 we analyzed log - log plots of the height of the maxima of @xmath48 and @xmath51 versus the 3-volume @xmath53 . for @xmath95 the corresponding figure is fig . [ log_log_k450 ] with the @xmath89 data we discuss in this section labeled by @xmath108 , while the @xmath103 results are shown in fig . [ log_log_k372 ] with the @xmath89 data labeled by @xmath109 . the numerical results are represented by the symbols and we fit them with straight lines . the numbers next to the lines give their slope determined from the fits . at @xmath95 we find that the maxima of @xmath48 and @xmath51 nicely follow a straight line with slopes that are close to 0.5 , both for @xmath48 and @xmath51 . for @xmath103 the finite volume effects are more severe , as discussed above , and we only show the data for our larger volumes , i.e , for @xmath110 and @xmath111 . some of the data show an upward curving trend and if one uses @xmath112 and 24 for the fit , one finds slopes of 0.39(4 ) for the maxima of @xmath48 and 0.43(11 ) for @xmath51 . although for @xmath103 the data are not as clean due to finite volume effects ( and also the temperature in units of @xmath2 is not as low as for @xmath95 ) the data indicate that also at @xmath103 for the lowest temperature the maxima of @xmath48 and @xmath51 diverge as @xmath113 . the analysis implies that at zero temperature the third derivatives diverge in the continuum limit and the susceptibilities develop a jump , indicating a second order transition at the end of the silver blaze region . in the first derivatives of the free energy the transition is characterized by the onset of a non - vanishing particle number density @xmath37 at @xmath2 and a linear response of the expectation value @xmath78 to changing @xmath1 for @xmath101 . we conclude the discussion of the @xmath114 phenomenology with an important cross - check . the silver blaze regions ends at the point where the chemical potential becomes equal to the mass @xmath4 of the excitation it couples to , in other words one expects @xmath115 . in order to check this we performed a small spectroscopy calculation on @xmath116 lattices using the conventional representation at @xmath79 . the mass was extracted from the correlator of the zero momentum fourier transform of @xmath117 , i.e. , the 2-point function @xmath118 in fig . [ effmass ] we show our results for the effective masses @xmath119 of that correlator for both , @xmath95 and @xmath103 . the effective masses are represented by the symbols and the horizontal lines are the respective values of @xmath3 . it is obvious that the effective masses agree very well with the values of @xmath3 data are more difficult to analyze , since the large mass @xmath120 leads to a very rapid exponential decay of the correlator ( [ corr ] ) and the relative statistical error quickly overlays the signal . thus for @xmath95 the effective mass plateau is rather short and the errors are large . ] . an exponential fit of the correlator for @xmath103 gives a mass of @xmath121 which is in excellent agreement with @xmath122 . for @xmath95 we find @xmath123 which agrees with @xmath124 within error bars . we conclude that we properly understand the silver blaze behavior at @xmath114 . ( top set of data ) and @xmath103 ( bottom ) . the effective masses are compared to the values of the critical chemical potential @xmath3 ( horizontal lines ) and we annotate the values for @xmath3 as well as for the mass @xmath125 obtained from an exponential fit of the 2-point function ( [ corr ] ) . , scaledwidth=75.0% ] having studied the transition at zero temperature we now explore how the transition line extends in the temperature direction of the phase diagram . as discussed , we work at fixed values of the couplings ( again using our two values @xmath95 and @xmath103 at @xmath96 ) , and drive the temperature by changing @xmath56 . this has the advantage that we always work at a fixed scale and do not mix temperature and discretization effects , but on the other hand restricts the accessible temperatures to a discrete set . for our finite temperature studies we use the values @xmath89 , 20 , 10 , 9 , 8 , 7 , 6 , 5 , 4 , 3 and 2 . in lattice units the temperatures @xmath126 thus range between @xmath127 and @xmath128 . more interesting is the range of temperatures in units of the critical chemical potential at @xmath114 , i.e. , the value @xmath2 which we determined in lattice units in the previous section this refers to the critical chemical potential at zero temperature . ] . for @xmath95 , where @xmath129 , the largest temperature we can reach with @xmath130 is @xmath131 . for the finer @xmath103 lattices ( @xmath106 ) we can go considerably higher , up to @xmath132 , although at the price of more severe finite volume effects . as the temperature is increased above zero , physics can not be completely independent of @xmath1 below some critical value of the chemical potential . still , at least for not too large temperatures , one should observe only very weak @xmath1-dependence of physical observables at sufficiently small chemical potential ( `` pseudo silver blaze behavior '' ) . in order to define the curves in the phase diagram that separate the pseudo silver blaze behavior from a condensed phase characterized by strong @xmath1-dependence , we determined the position of the maxima of the third derivatives of the free energy as a function of @xmath1 for different temperatures , using our @xmath133 lattices for @xmath95 and @xmath134 for @xmath103 . for the @xmath95 data we find that the positions of the maxima of @xmath48 and @xmath51 always agree exactly , while for the @xmath103 data , where the finite size effects are more severe , we find agreement only within error bars . and @xmath51 in the @xmath1-@xmath135 plane . we combine data from @xmath95 runs for @xmath133 ( circles connected with dashed lines ) and @xmath103 runs at @xmath134 ( triangles ) . @xmath135 and @xmath1 are expressed in units of @xmath3.,width=491 ] in fig . [ phasediagram ] we show the positions of the maxima of @xmath48 and @xmath51 in the @xmath1-@xmath135 plane . for the @xmath95 data we use filled circles connected with dashed lines ( the plot also contains points from @xmath103 ( triangles ) which we will discuss below ) . we express both the temperature @xmath135 and the chemical potential @xmath136 in units of the critical value @xmath100 determined in the zero temperature runs using @xmath89 . in these units the largest temperature we can reach with @xmath95 is @xmath137 . when turning on the temperature , the phase boundary curves slightly to the right reaching @xmath138 at @xmath137 , the highest temperature accessible for @xmath95 . this phase boundary separates pseudo silver blaze behavior from strong @xmath1-dependence . , @xmath47 , @xmath78 and @xmath50 for @xmath95 and @xmath133 as a function of @xmath1 for several temperatures corresponding to @xmath139 and 2 . for @xmath135 and @xmath1 we use units of @xmath2 , while @xmath37 , @xmath47 , @xmath78 and @xmath50 are given in lattice units at a fixed scale ( set by @xmath140 ) . , scaledwidth=110.0% ] in order to assess the behavior across the phase boundary , in fig . [ all_k450_allt ] we show the observables @xmath37 , @xmath47 , @xmath78 and @xmath50 from the @xmath95 runs at @xmath141 as a function of @xmath1 for several temperatures ( again using for @xmath135 and @xmath1 units of @xmath2 ) . in other words we show the observables along horizontal slices through the phase boundary in fig . [ phasediagram ] . while the first derivatives @xmath37 and @xmath78 were almost perfectly constant up to @xmath100 for our zero temperature runs ( compare fig . [ all_k450_t001 ] ) , in fig . [ all_k450_allt ] we now observe a deviation from constant behavior which becomes more prominent as the temperature is increased . however , the transition at the critical value of @xmath1 is still rather pronounced , and the corresponding susceptibilities @xmath47 and @xmath50 still seem to develop a discontinuity . the position of the discontinuity shifts towards larger values of @xmath1 in agreement with the curvature of the phase boundary in fig . [ phasediagram ] . in order to analyze the order of the transition also at @xmath142 we repeat the analysis of the volume dependence of the maxima of the third derivatives @xmath48 and @xmath51 as function of the 3-volume @xmath143 . the results of this analysis ( done on @xmath144 lattices ) for different temperatures are shown in fig . [ log_log_k450 ] . the symbols represent the numerical results and we fit them to straight lines , with the numbers next to the lines giving the corresponding slopes . it is obvious that the data are well described by a linear behavior with slopes near 0.45 for @xmath48 and slopes slightly above 0.5 for @xmath51 . the plots show that also for our finite temperature values the third derivatives keep developing singularities in the thermodynamic limit . we conclude that for the temperature range we cover with our @xmath95 runs , i.e. , for temperatures up to @xmath145 a second order transition separates the pseudo silver blaze region from the condensed phase with strong @xmath1 dependence . for @xmath95 and @xmath146 and 20 . we compare different temperatures corresponding to @xmath147 and 2 . the straight lines are linear fits to the data and the numbers near the lines are the respective slopes.,width=415 ] we remark that the scaling behavior of the transition is expected to be different for the @xmath114 and @xmath142 cases . for @xmath114 one expects that the transition is of mean field type , which was indeed confirmed by aarts in @xcite and our dual simulation perfectly reproduces this finding ( as already discussed in section 3 ) . due to dimensional reduction for @xmath142 one expects a behavior in the universality class of the 3-d xy model . however , it seems that it is difficult to explicitly extract this behavior from a lattice simulation with the resources available to us . this is reflected in the finding by aarts @xcite that the results on lattices ranging from @xmath81 to @xmath82 are equally well described by mean field behavior which illustrates that the change in the critical behavior must be very subtle . also an analysis in the continuum @xcite shows that the non - linear contribution to @xmath37 versus @xmath148 that is predicted for finite @xmath135 is very small in the range of parameters we can access . the finite volume analysis in fig . [ log_log_k450 ] is not accurate enough to extract such a small effect . an interesting question is of course where the phase boundary continues for larger temperatures . with the data taken at @xmath95 we reach a maximal temperature of @xmath137 and found the second order line to persist up to that value . for temperatures above that number we need to use the @xmath103 configurations where @xmath149 in lattice units and temperatures of @xmath150 can be reached . however , as we have already discussed , the @xmath103 data suffer from more severe finite volume effects which spoil the accuracy of the analysis ( e.g. , the peaks of the third derivatives of the free energy become wider and it is more difficult to exactly determine the phase boundary and the order of the transition ) . nevertheless we try to explore the trend for the behavior along the phase boundary towards larger @xmath135 . the phase diagram in fig . [ phasediagram ] also contains a few points from @xmath103 runs on @xmath151 lattices ( triangles ) . as for @xmath152 we use the positions of the maxima of the third derivatives @xmath48 and @xmath51 to determine the phase boundary . however , while for @xmath95 the positions of the maxima of @xmath48 and @xmath51 always match perfectly , for @xmath103 this is the case only for temperatures up to @xmath153 . consequently , at @xmath103 we use different symbols for the maxima of @xmath48 ( upward pointing triangles ) and @xmath51 ( triangles pointing downward ) . the two @xmath103 points with @xmath154 show matching maxima for @xmath48 and @xmath51 and fall on the curve set by the @xmath95 data . those points confirm ( although with larger error bars ) the picture of a second order line up to at least @xmath137 from the @xmath95 runs . for the larger temperatures @xmath155 the third derivatives @xmath48 and @xmath51 peak at slightly different values . however , the positions of the maxima agree within error bars and we attribute the small discrepancy to the more severe finite volume effects for the @xmath103 ensembles . the corresponding observables are shown in fig . [ all_k372_allt ] where we plot @xmath37 , @xmath47 , @xmath78 and @xmath50 for @xmath103 as a function of @xmath1 for several temperatures . in fig . [ log_log_k372 ] we repeat the finite volume study of fig . [ log_log_k450 ] now for @xmath103 . we show the log - log plot of the maxima of @xmath48 and @xmath51 versus the spatial volume @xmath53 . it is obvious that here the data do not as nicely fall on a straight line as in the corresponding plots for @xmath95 shown in fig . [ log_log_k450 ] . we fit only the data points for the three largest volumes and as in fig . [ log_log_k450 ] denote the slopes next to the lines . it is quite interesting to observe that the slope for the maxima of @xmath51 remains at values close to 0.5 up to the largest temperatures , indicating that the susceptibility @xmath50 manages to develop a discontinuity in the thermodynamic limit also for our highest temperatures . the stronger finite volume effects at @xmath103 make the analysis more difficult for @xmath48 : the slopes for @xmath48 are smaller for the smaller volumes but show a tendency towards pointing upwards for larger @xmath143 , indicating second order behavior for the @xmath1-derivatives also up to the highest temperatures we can access . one expects that the second order line persists for large @xmath135 due to the breaking of the u(1 ) symmetry at the transition to the condensed phase . in fig . [ step ] we attempt a plausibility test for this scenario : we plot the height of the maxima of the third derivatives @xmath156 and @xmath157 ( both made dimensionless with suitable powers of @xmath3 ) as a function of the inverse temperature in units of @xmath3 . we find that the maxima are non - zero at finite @xmath135 and only for @xmath158 the curves reach 0 . furthermore the maxima for the @xmath95 runs are larger than those at @xmath103 as expected from the larger volume at @xmath159 . although this analysis is only an exploratory test at a fixed volume , it supports the existence of a phase transition throughout the phase diagram . , @xmath47 , @xmath78 and @xmath50 for @xmath103 and @xmath134 as a function of @xmath1 for several temperatures corresponding to @xmath160 and 6 . for @xmath135 and @xmath1 we use units of @xmath2 , while @xmath37 , @xmath47 , @xmath78 and @xmath50 are given in lattice units at a fixed scale ( set by @xmath161).,scaledwidth=109.0% ] for @xmath103 and @xmath110 and 24 . we compare different temperatures corresponding to @xmath162 and 3 . the straight lines are linear fits to the data ( using only @xmath163 ) and the numbers near the lines are the respective slopes.,width=415 ] and @xmath157 ( both made dimensionless with suitable powers of @xmath3 ) as a function of the inverse temperature in units of @xmath3 . we compare the results for @xmath95 on @xmath144 lattices and @xmath103 on @xmath151 . , scaledwidth=75.0% ] , @xmath47 , @xmath78 and @xmath50 for @xmath95 and @xmath133 as a function of @xmath164 for two different values of @xmath165 . , scaledwidth=110.0% ] we conclude our presentation of the finite temperature results with the discussion of observables along two vertical cuts in the phase diagram fig . [ phasediagram ] . using the @xmath95 data at @xmath133 , in fig . [ condensation ] we show the observables @xmath37 , @xmath47 , @xmath78 and @xmath50 as a function of @xmath164 and compare two different values of the chemical potential . the vertical cut at @xmath166 hits the phase boundary at @xmath167 , while the cut at @xmath168 continues all the way to @xmath169 without entering the condensed phase ( compare fig . [ phasediagram ] ) . this behavior is clearly reflected in the observables : for the @xmath168 data the particle number density @xmath37 and the susceptibility @xmath47 both drop to zero as the temperature is decreased to @xmath169 , as expected for a value of the chemical potential in the silver blaze domain , i.e. , @xmath170 . the vacuum expectation value @xmath50 drops towards 0.237 , while the corresponding susceptibility @xmath50 vanishes as @xmath171 . for the cut at @xmath166 the behavior is quite different : when decreasing the temperature down to @xmath167 , where the cut hits the phase boundary of the condensed phase , the susceptibilities show a drastic increase an then remain essentially constant for @xmath172 . the particle number density @xmath37 has a non - vanishing value at @xmath169 and @xmath78 settles at a value of 0.285 . the plots nicely demonstrate the temperature driven condensation of the system for @xmath166 and its absence at @xmath168 . in this work we have studied a charged scalar @xmath0 field at finite density using the lattice formulation . the complex action problem of the standard form was overcome by mapping the partition sum to a flux representation where only real and non - negative terms appear . the system was updated with a generalized worm algorithm and we studied thermodynamical observables as a function of temperature and chemical potential . for zero temperature the system undergoes a second order phase transition at some critical value @xmath2 that marks the end of the silver blaze region . for @xmath173 the observables are independent of @xmath1 , while above a strong @xmath1-dependence sets in . the critical value @xmath2 could very cleanly be identified from the narrow peaks of two 3rd derivatives of the free energy and in a small spectroscopy calculation we showed that @xmath3 coincides with the mass of the scalar field . a finite volume analysis established the second order nature of the transition . increasing the temperature we found that the critical line bends towards larger values of @xmath1 . for temperatures up to @xmath174 we could reliably establish the continuation of second order behavior for finite @xmath135 . at small but non - zero temperatures we found only a very weak dependence of observables on the chemical potential for values of @xmath1 below the critical value ( pseudo silver blaze behavior ) . at the critical line we saw an abrupt change of physical observables marking the transition into the condensed phase characterized by strong @xmath1-dependence . when studying our observables as a function of the temperature we could nicely illustrate the temperature driven condensation for @xmath175 and the absence of condensation at @xmath168 . for the analysis of temperatures @xmath176 we had to rely on finer lattices ( @xmath103 ) which , however , show considerably more severe finite size effects . for small temperatures the results match the results from our coarser ( @xmath95 ) lattices very well , but for @xmath176 some of the observables the ones related to the particle number density were seriously affected by finite volume effects and thus gave inconclusive results about the nature of the transition between pseudo silver blaze behavior and the condensed phase with strong @xmath1-dependence . the observables related to derivatives of the free energy with respect to the mass parameter , i.e. , @xmath177 and @xmath51 , on the other hand indicate that a second order line could persist to temperatures considerably higher than @xmath174 . lattice studies on large and fine lattices will be necessary to settle this issue . * acknowledgments : * we thank gert aarts , shailesh chandrasekharan , ydalia delgado mercado , hans gert evertz , david kaplan , anyi li and dam thanh son for discussions . c. gattringer thanks the int at the university of washington in seattle , where part of this work was done , for their hospitality and the dr . heinrich jrg foundation of the karl - franzens - university graz for financial support . this work was partly supported by the dfg sfb trr 55 , `` hadron physics from lattice qcd '' . 12 t.d . cohen , phys . * 91 * ( 2003 ) 222001 . n. prokofev , b. svistunov , phys . * 87 * ( 2001 ) 160601 . s. chandrasekharan , arxiv:1205.0084 . s. chandrasekharan , a. li , arxiv:1202.6572 ; phys . * 108 * ( 2012 ) 140404 ; arxiv:1111.5276 ; jhep * 1012 * ( 2010 ) 021 ; jhep * 1101 * ( 2011 ) 018 . d. banerjee , s. chandrasekharan , phys . d * 81 * ( 2010 ) 125007 . s. chandrasekharan , phys . d * 82 * ( 2010 ) 025007 . cecile , s. chandrasekharan , phys . d * 77 * ( 2008 ) 014506 . s. chandrasekharan , a.c . mehta , phys . * 99 * ( 2007 ) 142004 . s. chandrasekharan , f .- j . jiang , phys . d * 74 * ( 2006 ) 014506 ; phys . d * 68 * ( 2003 ) 091501 . y. delgado mercado , h.g . evertz , c. gattringer , phys . lett . * 106 * ( 2011 ) 222001 ; acta phys . * 4 * ( 2011 ) 703 ; comput . * 183 * ( 2012)1920 . y. delgado mercado , h.g . evertz , c. gattringer , d. gschl , pos lattice2011 ( arxiv:1111.0916 ) . c. gattringer , nucl . b * 850 * ( 2011 ) 242 . y. delgado mercado , c. gattringer , nucl . b * 862 * ( 2012 ) 737 . w. unger , p. de forcrand , j. phys . g g * 38 * ( 2011 ) 124190 . de forcrand , m. fromm , phys . * 104 * ( 2010 ) 112005 ; pos lattice * 2008 * ( 2008 ) 191 . m. fromm , j. langelage , s. lottini , o. philipsen , jhep * 1201 * ( 2012 ) 042 . t. sterling , j. greensite , nucl . phys . b * 220 * ( 1983 ) 327 . m. panero , jhep * 0505 * ( 2005 ) 066 . v. azcoiti , e. follana , a. vaquero , g. di carlo , jhep * 0908 * ( 2009 ) 008 . t. korzec , u. wolff , pos lattice * 2010 * ( 2010 ) 029 . u. wolff , nucl . b * 832 * ( 2010 ) 520 ; nucl . b * 824 * ( 2010 ) 254 [ erratum - ibid . * 834 * ( 2010 ) 395 ] ; nucl . b * 814 * ( 2009 ) 549 ; nucl . b * 810 * ( 2009 ) 491 . o. b " ar , w. rath , u. wolff , nucl . b * 822 * ( 2009 ) 408 . u. wolff , nucl . b * 789 * ( 2008 ) 258 . endres , phys . rev . d * 75 * ( 2007 ) 065012 ; pos lat06 ( 2006 ) 133 . m. cristoforetti _ et al . _ , phys . d * 86 * ( 2012 ) 074506 . s. chandrasekharan , pos lattice * 2008 * ( 2008 ) 003 . m. hogervorst , u. wolff , nucl . phys . b * 855 * ( 2012 ) 885 . t. korzec , i. vierhaus , u. wolff , comput . commun . * 182 * ( 2011 ) 1477 . p. weisz , u. wolff , nucl . b * 846 * ( 2011 ) 316 . g. aarts , phys . * 102 * ( 2009 ) 131601 . g. aarts , jhep * 0905 * ( 2009 ) 052 . j. bernstein and s. dodelson , phys . rev . lett . * 66 * ( 1991 ) 683 . k. m. benson , j. bernstein and s. dodelson , phys . d * 44 * ( 1991 ) 2480 . in this appendix we provide pseudo - code for the initialization of the worm and its subsequent steps . the worm algorithm that updates the variables @xmath178 is set up in a fixed background of the variables @xmath65 generated with local metropolis sweeps ( see section [ sec_worm ] ) . the starting point of a worm will be denoted by @xmath179 , the position of its head by @xmath11 , and @xmath180 will be used for the set of all lattice sites . @xmath181 is an equally distributed random variable with @xmath182 that determines whether the worm tries to increase or decrease the flux along its path . by @xmath183 we denote the direction of the next attempted step of the worm , and in case @xmath15 is negative ( @xmath184 ) we use the convention @xmath185 for the assignment of the corresponding link variables . as in the main text we use the abbreviation @xmath64 $ ] , and by @xmath186 we denote the same expression but evaluated with the variable @xmath187 from the trial offer . finally , rand ( ) denotes a random generator for uniformly distributed random numbers in the interval @xmath188 . while worm has not started + + @xmath189 randomly + @xmath190 randomly + @xmath191 randomly + @xmath192 + + @xmath193 then + @xmath194 + + @xmath195 + + + @xmath196 then + @xmath197 + @xmath198 + + + + end while while @xmath199 + + @xmath200 randomly + @xmath201 + @xmath202 + + @xmath203 then + @xmath204 + + @xmath205 + + + @xmath206 then @xmath207 + + @xmath196 then + @xmath208 + @xmath209 + + + end while

the physics of hadrons began about one century ago , when it became appealing to build nuclei from protons and neutrons . the experimental search for the neutron was soon resumed along with the search for the particle predicted by yukawa as being responsible for the tight binding of protons and neutrons . see , e.g. , @xcite for the history of this early period . when interacting with nucleons or among themselves , the pions produce many resonances . the exchange of meson resonances , and the formation of nucleon resonances in the intermediate states are crucial for our understanding of the nuclear forces beyond the single - pion exchange proposed by yukawa . for instance , the existence of the vector mesons , nowadays known as @xmath0 and @xmath1 , was somewhat anticipated from the need for isospin - dependent spin - orbit components in the nucleon - nucleon interaction . see , e.g. , @xcite for an introduction into the early literature on this subject . the discovery of numerous meson or baryon resonances stimulated intense efforts to understand the underlying mechanisms that generate the hadrons . for instance , the @xmath0 meson can be described as a @xmath2 resonance in a @xmath3-wave , with the @xmath4-wave counterpart being the scalar meson responsible for the spin - isospin averaged attraction observed in nuclei . in 1955 , chew and low described the @xmath5 as the result of a strong @xmath6 interaction @xcite , and today , the scattering of pions off nuclei is still pictured as series of @xmath7-hole formation and annihilation . an ever more ambitious picture was proposed , called `` nuclear democracy '' or `` bootstrap '' , in which every hadron consists of all possible sets of hadrons compatible with its quantum numbers @xcite . unfortunately , neither the models based on nucleons and pions as elementary bricks , nor the bootstrap , succeeded in providing an overall picture . for instance , the degeneracy of isospin @xmath8 and @xmath9 mesons such as @xmath0 and @xmath1 , does not emerge naturally , if one assumes that their content is dominated by the nucleon - antinucleon component @xcite . from the very beginning , the attention was focused on hadrons with properties that look unusual , in contrast to the regularity patterns of the rest of the spectrum . the exotic character of a hadron is of course a time - dependent concept : once a striking property of a hadron is confirmed by further measurements , and understood within a plausible theoretical framework , the hadron ceases to be exotic . among the most famous examples , there is the @xmath10 meson , a peak of mass 1.02gev with a remarkably small width of about 4mev @xcite , decaying preferentially into a pair of kaons , though the decay into pions is energetically much easier . the decay patterns of the @xmath10 are now understood from its quark content @xmath11 , where @xmath4 denotes the strange quark , with very little mixing . the suppression of the internal annihilation of @xmath11 into light mesons has been codified in terms of the so - called `` zweig rule '' , eventually justified from the asymptotic - freedom limit of quantum chromo - dynamics ( qcd ) . the process took , however , several years , and was not mature enough when the @xmath12 particle was discovered at the end of 1974 ; it is a meson of mass about 3.1gev and width much less than 1mev @xcite . why the @xmath12 was not immediately identified with @xmath13 ? the great surprise was that the zweig rule works too well ! as compared to the @xmath14 , the @xmath12 is three times heavier , so its phase - space for decay into light mesons is considerably larger , and yet its width is smaller ! for the @xmath15 , the effect is even more pronounced , but when the @xmath15 was discovered in 1977 @xcite , the lesson was learnt , and the @xmath15 was never considered as exotic . today , any attempt to provide with a definition of an `` exotic '' hadron starts from the quark model . a meson with quantum numbers that can not be matched by any spin - orbital state of a quark and an antiquark is obviously exotic . states with unusual properties but ordinary quantum numbers are named `` crypto - exotic '' states . in particular , ordinary baryons can reach any set of values of the spin - parity @xmath16 with half - integer spin @xmath17 and parity @xmath18 , so a baryon can not be exotic except for its flavour content . a consensus has thus been reached that a hadron is exotic if it can not be understood as mainly made of quark - antiquark nor quark - quark - quark , but `` mainly '' remains vague . while simple quark models will propose a reference with a frozen number of constituents , more advance pictures include an expansion in the fock space , with additional quark - antiquark pairs in the wave function . the hope is that the expansion converges , and that a clean separation emerges between exotic and non - exotics . this review is organised as follows . in sec . [ se : mesons ] , we shall review states with unusual properties in the light ( sec . [ sse : mesons - light ] ) , open heavy - flavour ( sec . [ sse : mesons - open1 ] ) , double heavy flavour ( sec . [ sse : mesons - open2 ] ) , and hidden heavy - flavour ( sec . [ sse : mesons - hidden ] ) sectors . the baryon spectrum is discussed in sec . [ se : baryons ] , with , again a separation between the non - exotic light sector ( sec . [ sse : baryons - light - ns ] for non strange and sec . [ sse : baryons - light - negs ] for strange baryons ) , the exotic light pentaquark with positive strangeness in sec . [ sse : baryons - light - poss ] , the case of open heavy - flavour , single ( sec . [ sse : baryons - open1 ] ) or double single ( sec . [ sse : baryons - open2 ] ) , the anticharmed pentaquark in sec . [ sse : antich - baryons ] , and the case of hidden heavy - flavour ( sec . [ sse : baryons - hidden ] ) with in , particular , the recent pentaquark states found by the lhcb collaboration . the multibaryon systems will be discussed in sec . [ se : multibaryons ] . the various theoretical models will be reviewed in sec . [ se : theory ] , with some emphasis on string dynamics and potential models . some limits have to be set . the case of light hadrons will be just outlined , to focus more on the most recent states containing heavy quarks or antiquarks . in the former sector , the literature is enormous , and can be traced back from the conference proceedings that will be cited in the next sections , from some well - known reviews such as @xcite , or from study groups for future facilities @xcite . for the recent exotics with hidden charm , there exists already excellent and even comprehensive reviews , such as @xcite . our aim here is not to compete with these reviews , nor to duplicate them , but rather to discuss critically the recent findings and the models that they inspired . so , this paper will be more subjective . hopefully , it will trigger some discussion , and , perhaps , attract newcomers into the field , where innovative ideas are welcome . the bibliography for such a subject can not be comprehensive . some past review articles are cited , which give access to the early papers . we apologise in advance in the event where some important contributions are inadvertently omitted . there are many hot issues on light mesons : why are the pseudoscalars so light ? what is the status of the many scalar mesons ? why ideal mixing and zweig rule work so well in some sectors , and not in other sectors ? we refer , e.g. , to @xcite for a critical survey of the claims for exotics , and to @xcite for a thorough discussion within qcd and the bibliography , and to the last hadron conferences for an update @xcite . one of the puzzles arises from the @xmath19 , measured in several experiments , in particular at l3 @xcite : the comparison of the different charge modes suggest a sizeable isospin @xmath20 contribution . there are interesting theoretical studies on the production mechanism @xcite , but missing is a simple understanding of why there is strong attraction in that channel in terms of quark dynamics . there is an intense activity in the search of glueballs and light hybrids , either as supernumerary states with ordinary quantum numbers @xmath21 , or as genuine exotics with @xmath21 that can not be matched by ordinary quantum numbers . in particular , evidence has been found for @xmath22 states between 1 and 2gev@xmath23 by ves at serpukhov@xcite , by e852 at brookhaven @xcite , and by compass at cern @xcite , and the search is resumed actively , in particular by gluex at jlab @xcite . for a thorough review of the other contributions , see , e.g. , @xcite . the charmed mesons @xmath24 , @xmath25 have been discovered shortly after the charmonium , and the mesons with charm and strangeness @xmath26 , @xmath27 have followed . relevant for this review on exotics are the so - called @xmath28 states . as summarised in @xcite , the @xmath29 was discovered at babar @xcite and the @xmath30 at cleo @xcite , with a mass remarkably lower than predicted in the `` bible '' @xcite and in similar quark - model calculations . note that the tuning of the spin - dependent forces in such models has been often debated , and no consensus has ever been reached , see , e.g. , @xcite . to explain the low mass and the properties of the @xmath31 states , the coupling to the nearby @xmath32 thresholds certainly plays a role . these mesons have even been considered as molecules , for instance , in @xcite . after , the discovery of charm , it has been early recognized that configurations including both light and heavy quarks or antiquarks might be particularly interesting , for instance , in the case of baryon - antibaryon or multibaryon systems @xcite . another pioneering contribution was due to isgur and lipkin @xcite who considered the flavour exotic @xmath33 configuration . shortly after , even more exotic mesons were studied , @xmath34 , with two heavy quarks and two light antiquarks . as explained later in this review , their existence is a prediction shared by many theorists using different approaches , which will be discussed in the next sections . doubly - heavy exotics , which are surprisingly omitted in some `` reviews '' on multiquarks , have not yet been searched for in any experiment , but will be one of the goals of future experiments , and several theorists estimated the rate of production at hadron @xcite or electron @xcite colliders , or in the decay of hadrons with beauty @xcite . the production of two charm - anticharm pairs is somewhat disconcerting , with the good surprise of a rather abundant production of double charmonium , leading for instance to the identification of the @xmath35 recoiling against the @xmath12 @xcite , and , on the other hand , the double - charm baryons have not been confirmed in recent experiments , see sec . [ sse : baryons - open2 ] . the spectrum is quarkonia , @xmath36 and @xmath37 , is now rather well known and well understood . for a while , the most delicate experimental sector was the one of spin - singlet states , such as @xmath38 , @xmath39 . for instance , a first candidate for the @xmath40 was announced about 300mev below the @xmath12 , triggering some premature announcements on the quark dynamics , but eventually replaced by a more reasonable @xmath40 about 112mev below @xmath12 , see , e.g. , the reviews @xcite . a first indication for the @xmath41 was given a by @xmath42 formation experiment at cern @xcite , with another indication in a similar experiment at fermilab @xcite . it was eventually firmly identified at cleo @xcite . the @xmath35 was seen at belle in @xmath43 decay @xcite and in double - charmonium production @xcite . it also took some time until the spin - singlet states of @xmath37 were identified @xcite . above the threshold , @xmath44 for charmonium and @xmath45 for bottomonium , the spectrum was known to be more intricate , with broader peaks , and an interplay of @xmath4- and @xmath46-states mixed with the @xmath47 continuum , and it was a somewhat a daily grind to improve the measurements . then came the shock of the discovery of the @xmath48 by belle @xcite , in the @xmath49 mass spectrum of the reaction @xmath50 . neither the mass nor the decay properties were according to the expectations of the charmonium models , in particular for the @xmath51 level . on the other hand , such a state was announced in the molecular models , as discussed in sec . [ sse : molec - xyz ] . not only the experts on quark- or hadron - dynamics commented this new state , but also some non - experts became attracted to the subject . as well summarised in a recent review @xcite , a second shock was the discovery of the @xmath52 by babar @xcite , in @xmath53 with isr .. ] this @xmath54 sector of charmonium was already thoroughly explored , with radial excitations , with @xmath4- and @xmath46-waves mixed between them and with the continuum . so this new state was immediately associated with an exotic , or , say crypto - exotic structure . the @xmath52 was seen also at cleo and belle @xcite . a third shock was the discovery of charged states , with minimal content @xmath55 or c.c . the first announcement was the @xmath56 by the belle collaboration , in the @xmath57 mass distribution of the decay @xmath58 @xcite . it is hardly possible to list all the states of @xmath59 or @xmath60 type without duplicating the very comprehensive reviews such as @xcite . let us stress a few points : * the @xmath61 and @xmath62 have been seen in double - charmonium production @xcite , which gives a clear signature of their @xmath13 content , * some states have been seen in @xmath63 , indicating a charge conjugation @xmath64 , and very likely @xmath65 or @xmath66 . this is the case for @xmath67 , @xmath68 and @xmath69 @xcite . * some states , in particular among the above ones , are just candidates for radial excitations of charmonium . if confirmed , they indicate the supernumerary character of other states sharing the same quantum numbers . * the @xmath70 has been seen as an enhancement in the @xmath71 spectrum of the isr reaction @xmath72 @xcite . many references about this state can be found in a recent paper @xcite . as stressed in this literature , the decay into a baryon and an antibaryon might indicate a `` baryonium '' type of structure , though other scenarios have been proposed @xcite . * there is sometimes a clustering of states with similar masses . it might the same state with a peak that depends on the production and decay process . * most quark models and , to a lesser extent , molecular models predict an isospin @xmath9 partner of the @xmath8 state , analogous to @xmath0 vs. @xmath1 . this stimulated searches for charged partner , such as @xcite . * while the @xmath48 has been seen in many experiments , @xmath52 in a few experiments , some states have been seen in just a single experiment . in between , @xmath73 has been seen by belle and @xcite , but these experiments share many collaborators , and used the same reaction @xmath74 , with a slight variant for the confirmation by cleo @xcite . so , the negative result of a search by compass in photoproduction is particularly important for the existence or at least the coupling properties of this state @xcite . * the lightest @xmath75 states have minimal quark content @xmath76 , where @xmath77 . in several models , their @xmath78 with hidden strangeness are expected . recently , the lhcb collaboration re - analysed the @xmath79 mass spectrum from the decay @xmath80 @xcite , and found evidence for @xmath81 and @xmath82 both with @xmath83 , in disagreement with some models ( cited in @xcite ) and in partial or full agreement with others , in particular the extension by stancu @xcite of the chromomagnetic model proposed for the @xmath48 @xcite . * the models somewhat differ about the existence and location of the @xmath84 analogues of the exotic mesons with hidden charm : the heavier @xmath85-quark mass , for instance , suppresses the chromomagnetic effect but enhances the chromoelectric ones , while the coupling schemes are sensitive to the gap between bare quarkonia and meson pairs . the list of candidates is rather scarce for the time being , and are due to belle only . the charged @xmath86 and @xmath87 have been seen in @xmath88 , where @xmath37 stands here for some lower @xmath89 or @xmath90 or @xmath91 . later , belle also got some candidate for the neutral @xmath92 . on the other hand , a search for the heavy partner of @xmath48 , decaying into @xmath93 or @xmath94 , has been so far unsuccessful . the history of the quark model is intimately linked to baryon spectroscopy . one of the motivation for su(3 ) flavour symmetry , nowadays denoted su(3)@xmath95 , was the observation that the hyperons have properties similar to that of non - strange baryons . the quark model was developed first in the baryon sector , with the pioneering works by greenberg , dalitz , etc . , and pushed very far by the subsequent works by dalitz et al . , hey et al . , isgur and karl , and many others . for a review , see , e.g. , @xcite . obviously , the field has been much enriched by the study of baryons with charm or beauty , and baryons carrying two or three heavy flavoured quarks are eagerly awaited . there are many issues there , that have been presented in conferences such as `` baryons '' , `` nstar '' , etc . for years , one of the most striking issues was the signature for the `` symmetric quark model '' with two degrees of freedom , vs. the quark diquark model . clearly , the recent progress , as e.g. , summarised at the recent conference in tallahassee @xcite , confirms the existence of states such as @xmath96 which can not be accommodated in a naive quark - diquark picture with a frozen diquark . the lightest state with an excitation in each of the jacobi variables , say , @xmath97 and @xmath98 , has isospin @xmath99 , spin @xmath100 and orbital momentum and parity @xmath101 . in the particular case of the harmonic oscillator , its wave function reads @xmath102 and is antisymmetric under the permutations . thus , if associated with an antisymmetric colour wave function @xmath103 and an antisymmetric spin - isospin wave function , it fits perfectly the requirement of fermi statistics . note that an antisymmetric spin - isospin wave function is possible for @xmath99 ( @xmath104 ) , but not for @xmath105 ( @xmath106 ) nor @xmath8 ( @xmath107 ) . experimentally , the search for doubly - excited baryons is difficult . as often underlined ( see , e.g. , @xcite ) photon or meson scattering off a nuclear target favours the production of states with a single quark being excited . and in amplitude analyses , doubly excited baryons , if any , come in a region that is already well populated ! another hot issue is the roper resonance , an excitation with the same quantum numbers as the ground state . in potential models , this is mainly a hyper - radial excitation , but in explicit model calculations , it is found higher than the negative - parity excitation , while the experimental roper state lies degenerate or even slightly below the lowest negative - parity excitation . chiral dynamics , as implemented in specific models or in lattice simulations , accounts for this property , which sets limits to naive potential models . here , again , is the problem of missing states , with slow but continuous progress to fill the holes . a very striking property is the low mass of the @xmath108 , discussed first by dalitz @xcite and now in an abundant literature ( see , e.g. , @xcite for a recent summary ) . in the case of the nucleon spectrum , the spin - orbit splittings , such as between @xmath109 with @xmath110 and @xmath111 with @xmath112 , are very small , and isgur and karl , for instance , have set the spin - orbit potential to zero in their fit @xcite . hence the large @xmath113 splitting of the strange analogues appears as somewhat a surprise in a pure quark model . a minimal solution explains the discrepancy by a shift due the coupling to the nearby @xmath114 threshold . similarly , the @xmath115 is shifted down by the nearby @xmath44 threshold , and the hyperfine splitting @xmath116 is slightly smaller than expected . similarly , the @xmath117 state of charmonium gets some admixture of @xmath118 in conservative models of the @xmath48 , etc . now , a more appealing solution is also suggested , in which there are two poles with @xmath119 near @xmath108 @xcite , a mixing of a @xmath120 molecule and a quark - model state . as often underlined in the case of mesons @xcite , the mixing scheme might be more involved than in ordinary quantum mechanics , with the absorptive part , i.e. , the coupling the decay channels , playing the major role . for years , there has been debates about baryons carrying positive strangeness . in the 60s and 70s , much effort was devoted to exploit the analyticity properties of the @xmath121-matrix , and to reconstruct the amplitudes from the scattering data . this program was rather successful in the case of the @xmath6 and @xmath120 reactions , and led to the identification of a number of baryon resonances with strangeness @xmath122 and @xmath123 , respectively . the case of @xmath124 scattering with @xmath125 was much more delicate and controversial . some early analyses , based only on differential cross - sections ( and just a few data on the recoil polarisation ) , suggested the possibility of resonances , see , for instance , @xcite . the resonances were named @xmath126 or @xmath127 , where @xmath8 or @xmath128 is the isospin . then came data on the analysing power , obtained using a polarised target , as in @xcite . not surprisingly , the uncertainties were narrowed @xcite , and the @xmath60 resonances tended to fade away . for years , the particle data group ( pdg ) , summarised the state of art , as in the 1980 edition @xcite , where one finds candidates for several exotic baryons with @xmath125 . there are @xmath129 and @xmath130 in the isospin @xmath8 sector , and @xmath131 , @xmath132 and @xmath133 for @xmath9 . then the @xmath60 baryon disappeared from the review , and the letter @xmath60 was given to another particle . note that other risky amplitudes analyses were done later , for instance for @xmath134 or @xmath135 before the advent of data taken on the polarised target , or for the elastic @xmath42 scattering with just data on the polarisation , while the simpler @xmath136 case notoriously requires several spin observables . act two of the play is better known . meanwhile the name `` pentaquark '' was introduced in the anticharmed sector , as will be discussed shortly , in sec . [ sse : antich - baryons ] . it reflects that the minimal content of the @xmath126 baryon is @xmath137 . in 2003 , nakano et al . found a narrow peak in the @xmath138 distribution of the reaction @xmath139 this state , named @xmath140 , was predicted in a model by diakonov et al . @xcite based on some developments of the chiral dynamics . it was of course rather remarkable to have a relatively small experiment revealing a major discovery . but the sequels quickly went out of control . on the theory side , a flurry of papers , with hasty calculations and ad - hoc assumptions . on the experimental side , some collaborations discovered that they had data on tape , that were never analysed , and contained in fact the pentaquark ! other positive results came , both for the @xmath141 , and for other members of the anti - decuplet of su(3)@xmath95 that was predicted . however , the trend became inverted , and experiments with better statistics and particle identification did not confirm the existence of the these light pentaquarks . this is reviewed for instance in @xcite . the theory part also came under better scrutiny . in ref . @xcite , some lattice qcd calculations following the discovery of the @xmath141 are criticised . in , e.g. , @xcite , some inconsistencies are pointed out in the estimate of the width in the model of diakonov et al . , a crucial property of the putative light pentaquark . in the last 20 years , much experimental progress has been done in the sector of baryons with single charm or beauty . when available , the comparison of @xmath85- and @xmath142-baryons is compatible with the idea of a flavour - independent confinement . for instance , the first evidence for @xmath143 came from d0 at fermilab , with a mass about @xmath144mev , corresponding to a double - strangeness excitation @xmath145 substantially larger than its charm analogue . further measurements by cdf at fermilab and lhcb at cern gave a mass around @xmath146 in closer agreement with the expectations @xcite . a remarkable feature of charmed baryons is their narrowness , with , e.g. , @xmath147mev for @xmath148 , @xmath149mev for @xmath150 , @xmath151mev for @xmath152 @xcite . of course , this gives a better view on the spectrum than in the case of overlapping resonances . an optimistic columnist of pdg @xcite even stated : `` models of baryon - resonance spectroscopy should now _ start _ with the narrow charmed baryons , and work back to those broad old resonances . '' among the current discussions , there is the suggestion that @xmath153 is mainly a @xmath154 molecule @xcite . the same narrowness is observed for @xmath85-baryons , as listed in the latest pdg review @xcite or revealed in the latest lhcb papers @xcite . thirty years ago , double charm baryons were considered as just a straightforward extension of the quark model , to be checked just for the sake of completeness . at most , some experts stressed the co - existence within the same hadron of the slow , adiabatic , charmonium - like motion of the two heavy quarks , and the fast , @xmath24-meson - like motion of the light quark around them . in the literature , @xmath155 is often described as a quark - diquark @xcite ; however , the first excitations occur between the two heavy quarks and thus one has to introduce `` excited diquarks '' . an alternative is the born - oppenheimer method @xcite , which , as discussed later , might also be adequate in the @xmath75 sector and for hybrids . some evidence for double - charm baryons was found at fermilab , in the selex experiment @xcite , but unfortunately not confirmed in other experiments at fermilab or elsewhere @xcite . this becomes particularly annoying . double - charm baryons will certainly receive more attention in the near future @xcite . it is an important question to understand why double - charm baryons have not yet been seen . if the golden modes such as @xmath156 or @xmath157 are suppressed , the detection scheme should be revised . and , perhaps , new production mechanisms should be exploited . our ability to detect double - charm baryons and measure their properties is eagerly watched by people waiting for exotic tetraquarks and dibaryons . as pointed out in @xcite , the binding and the correlation of the two charmed quarks in @xmath155 gives some idea about @xmath158 and @xmath159 . and , in most production experiments , they share the same triggers , so that a simultaneous search for double - charm baryons , mesons and dibaryons can be envisaged . two simultaneous papers by gignoux et al . and lipkin @xcite , in which the word `` pentaquark '' was first used , suggested the existence of stable or metastable states @xmath160 with @xmath161 denoting @xmath162 , @xmath163 or @xmath164 . this pentaquark was searched for in an experiment at fermilab , which was not conclusive @xcite . at the time of the light pentaquark , the h1 collaboration at hera looked at the @xmath165 mass spectrum , i.e. , states of minimal content @xmath166 , without strangeness contrary to the prediction in @xcite , and found a peak near 3.1gev with a width of about 12mev @xcite , which was confirmed neither by zeus @xcite , also operating at hera , nor by babar @xcite at slac , nor in the analysis by h1 of their heraslowromancap2@ data . the anticharmed pentaquark was initially one the goals of the compass experiment @xcite , but it has disappeared from the agenda . this is one of the latest findings , inducing an intense phenomenological activity supplementing the discussions about @xmath75 mesons . the rumour is that , due to the bad reputation of the light pentaquark , the lhcb collaboration tried during months to kill the evidence , with more checks than usual for this type of discovery . the result published in @xcite deals with the weak decay @xmath167 of the lowest baryon with beauty . the dalitz plot , as well as the @xmath168 mass spectrum , indicates two peaks , one with a mass of @xmath169mev and a width of @xmath170mev , an another with a mass of @xmath171mev and a width of @xmath172mev . as the @xmath173 mass is always reconstructed , some ambiguities are removed . the background is fitted as @xmath174 , with many @xmath175 excitations , up to high mass and high spin . when two resonances are added , the preferred quantum numbers are spin @xmath176 and @xmath177 , and opposite parities , namely @xmath178 and @xmath179 . see , also , @xcite . the process is schematically described in fig . [ fig : lhcb - penta1a ] , as well as the companion reaction ( cabibbo suppressed ) @xmath180 . decay ( left ) and @xmath180 decay ( right),title="fig : " ] decay ( left ) and @xmath180 decay ( right),title="fig : " ] this process @xmath180 has been measured by lhcb and published before the discovery of hidden - charm pentaquarks @xcite , but the statistics was not sufficient to identify peaks . this reaction has been discussed in a few papers , as well as other weak decays involving the hidden - charm pentaquark @xcite , shown in fig . [ fig : lhcb - penta2 ] . a more recent analysis of the reaction @xmath180 provides some support for the contribution of hidden - charm pentaquarks @xcite . -hadrons producing a hidden - charm pentaquark.,title="fig : " ] -hadrons producing a hidden - charm pentaquark.,title="fig : " ] it has been proposed @xcite to check the existence of the lhcb pentaquarks and provide more information on their properties by studying the photoproduction reaction @xmath181 , and the experiment will be done at jlab . the situation in the non - strange sector has been recently reviewed at the conference baryon 2016 @xcite . there are regularly some claims for resonant behaviour of some partial waves @xcite . negative results include : @xcite in the isospin @xmath20 sector , @xcite for multineutrons bound by a few @xmath182 . the most serious candidates are the state of mass 2380mev and width @xmath183mev by the wasa collaboration , in the @xmath184 coupled partial waves , corresponding to @xmath185 @xcite , which was expected by various models @xcite , and the more recent state at 2.15gev which is under investigation at jlab @xcite . in the sector of single strangeness @xmath123 , there are recurrent discussions about the possibility of dibaryons . see , e.g. , @xcite and refs . there . note that a study of the atomic @xmath186 system could provide some indirect information . in the two - body case , say @xmath187 , the atomic spectrum is much modified if the strong - interaction @xmath188 potential generates a bound state near threshold . this was realised by zeldovich in the 50 s and has been much studied @xcite . the case of three - body exotic atoms with a combination of coulomb interaction and short - range hadronic interaction remains to be clarified , but probably deserves a detailed study . the most advertised case deals with strangeness @xmath189 , following the work by jaffe@xcite on the coherent chromomagnetic attraction in @xmath190 with isospin and spin @xmath191 . more than 20 experiments have searched without success for the @xmath192 , ranging from the absence of @xmath192 in the decay of double hypernuclei @xcite to the mass spectra emerging from heavy - ion collisions @xcite . new experiments are currently under construction . the status of theoretical studies will be reviewed in sec . [ ssse : chromo ] . the speculations have been extended to the @xmath193 sector , see , e.g. , @xcite . if the @xmath192 , considered as @xmath194 system , is close to binding , as well as some other baryon - baryon systems , it is natural to guess that some multi - baryon configurations will be bound , as an illustration of the quantum phenomenon of borromean binding @xcite . there are for instance debates about the existence of @xmath195 or @xmath196 @xcite , as well as @xmath197 or @xmath198 @xcite . the field of hypernuclear physics is rather active , with new facilities contributing , and is reviewed regularly at dedicated conferences . almost immediately after the discovery of charm , the possibility of new species of hypernuclei was envisaged , with hyperons replaced by charmed baryons @xcite . in atomic physics , the study of multi - electron systems looks at first rather straightforward : the same non relativistic dynamics is at work , and one has `` just '' to solve the @xmath199-body equations with increasing @xmath199 . even there , the progress is slow , and the calculations , which are very delicate , have to be cross - checked . for instance , the last bound state of h@xmath200 , the excitations of the positronium molecule or the existence of borromean molecules were discovered only recently @xcite . in nuclear physics , the situation is even more delicate , as the basic two - body interaction is presumably modified by the presence of other nucleons , and supplemented by 3-body forces . in qcd , the situation is even worse . if one uses simple constituent models to describe mesons and baryons , their extension to multiquark states is hazardous . if one uses more sophisticated approaches to ordinary hadrons , their application to higher configurations becomes technically very difficult . before discussing in some details a few specific schemes , we review the early history of exotic hadrons . in the 60s , much attention was paid to _ crossing symmetry _ , which relates , e.g. , @xmath201 to @xmath202 . the reaction @xmath203 describes what is exchanged to monitor the direct reaction @xmath204 . it is reasonable to assume an equivalence between the content of the two reactions , as well as that of the third reaction , @xmath205 . this principle of _ duality _ was often used as a constraint in building models . for instance , if one describes @xmath6 elastic scattering with several @xmath206 and @xmath207 resonances , duality dictates to refrain from exchanging @xmath208 , @xmath0 , , in the @xmath209-channel . a consequence of duality is that is the reaction @xmath204 is weak , i.e. , without resonances , it does not benefit from coherent @xmath209-channel exchanges . though duality was first formulated at the hadron level , it is nowadays explained in terms of quark diagrams . we refer to @xcite for a review and references . for instance , in the case of @xmath2 scattering with isospin @xmath8 , the diagram of fig . [ fig : dual1 ] ( left ) shows that the possibility of annihilating a quar - antiquark pair opens the way for @xmath210 exchange in the @xmath209-channel and @xmath210 resonances in the @xmath4-channel . this is why there is a strong @xmath2 interaction with @xmath8 . on the other hand , for isospin @xmath20 , there is @xmath4-channel resonance , and the @xmath209-channel dynamics suffers from a more intricate topology . similarly , @xmath120 scattering , shown in fig . [ fig : dual1 ] ( right ) , benefits from both @xmath211 resonances in the @xmath4-channel and @xmath210 exchanges in the @xmath209-channel . this is not the case for @xmath114 scattering which is notoriously less strong . this led rosner to address the problem of baryon - antibaryon scattering @xcite , which is depicted in fig . [ fig : dual2 ] . the interaction is very strong , and benefits , indeed , from the exchanges of mesons in the @xmath209-channel . hence there should be resonances in the @xmath4-channel . this is how baryonium was predicted . -channel ( intersecting the dotted line ) contains two quarks and two antiquarks . the intermediate state in the @xmath209-channel ( crossing the dashed line ) is made of a quark and antiquark . ] as noted by roy @xcite , once one accepts baryonium , in a _ gedankenexperiment _ colliding a baryonium against a baryon , meson - exchange is dual of a pentaquark . in turn , the analysis of pentaquark - baryon scattering implies the existence of dibaryons ! this is illustrated in fig . [ fig : dual3 ] . -channel ) dual of mesons ( quark - antiquark in the @xmath209-channel ) in baryonium - baryon scattering . right : dibaryon ( six quarks in the @xmath4-channel ) dual of mesons ( quark - antiquark in the @xmath209-channel ) in pentaquark - baryon scattering.,title="fig:"]-channel ) dual of mesons ( quark - antiquark in the @xmath209-channel ) in baryonium - baryon scattering . right : dibaryon ( six quarks in the @xmath4-channel ) dual of mesons ( quark - antiquark in the @xmath209-channel ) in pentaquark - baryon scattering.,title="fig : " ] in line with duality , a string model was developed for hadrons , explaining why the squared masses increase linearly with the spin @xmath17 . this model was echoed by the elongated version of the bag model , the flux - tube model , or some potential models with multi - body forces which will be mentioned later in this section . the string model was first rather empirical , but later received a support from some topological and non - perturbative properties of qcd , as reviewed recently @xcite . the mesons are pictured as a string linking a quark to antiquark , see fig . [ fig : string1 ] ( left ) . the picture of baryons was suggested by artru , and rediscovered in several approaches @xcite . it consists of three strings linking each quark to a junction @xmath17 , which plays somewhat the role of a fourth constituent of each baryon , and , in particular , is submitted to an extension of the zweig rule , which prohibits any internal junction - junction annihilation . in this scheme , the baryonium of rosner contains two junctions , see fig . [ fig : string2 ] and thus widely differs from a set of two disconnected mesons . the decay of baryonium occurs by string breaking via pair creation : in the case of an external leg , it corresponds to the pionic cascade @xmath212 , while the breaking of the central string corresponds to the decay into baryon and antibaryon . several experimental states were good candidates for baryonium @xcite , but they were not confirmed in experiments using better antiproton beams . perhaps the nucleon - antinucleon interaction suffers from too strong an annihilation , while the heavy quark sector gives better opportunities . the model can be extended to more complicated structures , as to include pentaquarks and dibaryons . some examples are given in fig . [ fig : string3 ] . the above string pictures of baryonium was challenged by other schemes : the nuclear model which is reviewed in the next section , and some quark models . one of them describes baryonium states as a diquark and an antidiquark separated by an orbital barrier which suppresses the decay into mesons @xcite . in colour chemistry , there is also the possibility of colour - sextet diquark , and thus of @xmath213 states , named ` mock baryonia' which are predicted to be rather narrow @xcite . this was , indeed , an interesting challenge to identify spectroscopic signatures of the colour degree of freedom . the name has evolved . it used to be _ quasi - nuclear _ , and became _ molecular _ model . it starts with the observation that the meson - exchange interaction of yukawa , which successfully binds the deuteron and other nuclei , holds for other hadrons , in particular these containing light quarks . in the 60s , attempts have been made to describe mesons as baryon - antibaryon states , a simplified version of the bootstrap . fermi and yang @xcite had noticed that the @xmath214 potential is often more attractive than the @xmath215 one , as the cancellations in the latter become coherences in the former . but this approach does not account for some important properties of the meson spectrum @xcite , such as exchange degeneracy , e.g. , the @xmath0 and @xmath1 having the same mass . in the late 70s and early 80s , another point of view was adopted , namely that baryon - antibaryon bound states and resonances correspond to new type of mesons , lying higher in mass than ordinary mesons , and preferentially coupled to the baryon - antibaryon channel . in practice , the @xmath214 potential is deduced from the @xmath215 one by the @xmath216-parity rule : if the exchange of a meson @xmath217 ( or set of mesons @xmath217 ) contributes as @xmath218 to the @xmath215 potential in isospin @xmath219 , then its contributes as @xmath220 to the @xmath214 in isospin @xmath219 @xcite . when annihilation is neglected , a rich spectrum of bound states and resonances is found @xcite , with the hope than some of them will not become too broad when annihilation is switched on . before the discovery of the antiproton , annihilation was believed to be be a short - range correction . however , the annihilation part of the @xmath214 integrated cross - section was measured to be about twice the elastic one . to fit the cross- sections , one needs an annihilation acting up to about @xmath221fm , and this makes most quasi - nuclear baryonium states very broad . the molecular model might have better chances to predict exotics in channels with heavy hadrons whose interaction is less absorptive . the yukawa model made a come back in the heavy quark sector . the @xmath222 or @xmath223 potential , where @xmath224 , is perhaps weaker than the @xmath215 potential , but being experienced by heavier particles , it can produce binding as well . the first papers were by okun and voloshin and glashow et al . @xcite . in the latter case , the idea was to explain some intriguing properties of some high - lying charmonium states , but it was eventually realised that these properties were due to the nodes of the wave function when the states are described as radial excitations @xcite . the best known application of the molecular model deals with @xmath75 states . almost simultaneously , a few papers came , stressing the attraction due to one - pion - exchange in some @xmath222 or @xmath223 channels @xcite . of course , the @xmath48 discovery at belle was greeted as a success of this approach . today , the @xmath48 is generally considered as having both a quarkonium and a multiquark component @xcite . for a clarification of the molecular approach , see @xcite . the anticharmed pentaquark has been studied with molecular dynamics , as a light baryon linked to an anticharmed meson . see , for instance , @xcite . note that the @xmath225 or @xmath163 and @xmath164 content of the light sector in the chromomagnetic model is not necessarily the same when mesons are exchanged between a @xmath226 or a @xmath227 and a baryon . after the publication of the lhcb results on the hidden - charm pentaquark @xcite , it was reminded or discovered that there were some predictions of bound states of a charmed baryon and an anticharmed meson , in particular by hofman and lutz @xcite , oset and collaborators @xcite , and z.c . yang et al . the game was even initiated earlier for the case of hidden - strangeness @xcite . of course , the model has been refined recently and compared to the lhcb results . the literature is comprehensively reviewed in @xcite . one might wonder why the hidden - charm pentaquark was not searched for before , given the above predictions . 1 . there are many predictions vs. just a few experiments , which are not primarily devoted to spectroscopy . 2 . in the molecular model , there are many states of this type , and it is not easy to focus on the configuration giving the best signature for the molecular dynamics . for instance , the binding of a charmed baryon and an anticharmed antibaryon was already envisaged in 1977 @xcite , shortly after the meson - meson case @xcite , but he meson - baryon case was unfortunately overlooked in the 70s . 3 . there is rather generally more effort and more interest in the meson than in the baryon sector . the negative parity states , either @xmath112 or @xmath228 , can be accommodated in the molecular model , with a main @xmath4-wave between a meson and a baryon , and a small admixture of @xmath46-wave , as in the more familiar case of the deuteron . the orbital excitation with positive parity is probably much higher in mass . in @xmath223 , the annihilation and creation of light quarks opens a coupling to @xmath13 , leading to description of some isoscalar @xmath75 states as a mixture of molecule and quarkonium states . isovector states do not mix with quarkonium , except for some minor effects of isospin violation . the sector of double - charm molecules has also received much attention , from the very beginning . see for instance , @xcite . it is immediately realised that if the exchange of mesons bind @xmath222 , @xmath223 , and @xmath229 their beauty or mixed charm - beauty analogues and other configurations could as well be formed by this mechanism , such as some meson - baryon and baryon - baryon states @xcite . we already mentioned the prediction of hidden - charm pentaquarks as molecules such as @xmath230 , before their discovery by lhcb . one could also include @xmath231 or @xmath232-nuclei states , see , e.g. , @xcite , where the possibility of @xmath233 states is also envisaged . the non - strange heavy pentaquark @xmath234 is related to the non - confirmed state claimed by h1 @xcite . in particular , there are many studies of bound states of charmed or doubly - charmed baryons with nucleons or with other singly or doubly charmed baryons @xcite . one can even speculate about a new periodic table , where the proton is replaced by @xmath235 and the neutron by @xmath236 @xcite . of course some rearrangement into triple charm and single charm should be taken into account . in several model calculations , many interesting hadron - hadron systems are estimated at the edge between binding and non - binding . this means that 3-body or larger systems are probably bound as long as they are not suppressed by the fermi statistics . so , accounting for the @xmath75 states implies many other states . in qcd , the gluon field carries a colour charge and coupled to itself . this lead to speculations about the existence of new neutral states , made essentially of gluons , except for @xmath237 loop corrections . this was sometimes hotly debated . i remember for instance that in his talk at the thessaloniki conference on antiprotons , van hove expressed doubts about the necessity of glueballs within qcd , though the written version was somewhat softened @xcite . the estimates for the mass and even the quantum numbers of the lightest glueball or gluonium has somewhat changed during the years . the situation is now stabilised , and is reviewed , e.g. , in @xcite . note that the lowest glueballs with unusual quantum numbers , such as @xmath238 , are sometimes found with rather high masses @xcite . another sector that used to be rather fashionable is the one of hybrids . very schematically , hybrids combine quark(s ) , antiquark(s ) and gluons . there are even explicit constituent models in which the constituent gluon is given a mass , and interacts via an effective potential , see , e.g. , @xcite . the first approaches describes hybrids as an excitation of the string or , say , of the gluon field linking the quarks . some pioneering attempts can be found in @xcite . in @xcite , a born - oppenheimer treatment was made , with a variant of the bag model used to estimate the energy of the gluon field . for a given interquark separation , the shape of the bag is adjusted as to give the lowest gluon energy . the ground state of the gluon field provide with the usual quarkonium potential . the first excited state of the gluon field is linked to the effective @xmath239 potential within a hybrid . the first @xmath240 hybrid was predicted near 4.0gev , and this estimate has been much refined with the bag model replaced by more elaborate lattice calculations @xcite , or effective field theory @xcite . note that in the late 70s and early 80s , the motivation was that the hybrids would show up clearly atop a clean and well - understood charmonium spectrum ! with the proliferation of @xmath75 states , we realise today that identifying hybrids is a little more challenging . the method of qcd sum rules ( qcdsr ) is a very ambitious and efficient approach , initiated by shifman , vainshtein and zakharov @xcite , allowing an interpolation between the rigorous perturbative regime and the long - range regime . for a review , see , e.g. , @xcite . it was first applied to ground - state mesons . there is no limitation , in principle , but the numerical stability become more and more delicate for excitations with the same quantum numbers as the ground state ( say , radial excitations ) . for baryons , the choice of the most appropriate operator is not unique , and triggered some debates . this is of course even more delicate for multiquarks . hence , the choice of the strategy is often dictated by models , such a diquark - antidiquark . a few groups have applied qcdsr to exotic states , in particular at montpellier , so paulo and peking , see , e.g. , @xcite for examples of early attempts , and @xcite for some reviews of the state of art over the years . needless to say that qcdsr calculations of multiquark exotics are extremely delicate and require weeks of cross - checks . so any overnight analysis published just after the release of a new experimental state should have been anticipated long in advance . lattice qcd ( lqcd ) is well known and well advertised , and hardly needs to be presented . for an introduction to the recent literature , see , e.g. , @xcite . dramatic progress have been accomplished on various aspects of strong interaction . there are interesting recent developments , for instance to study how the results behave numerically and have to depend theoretically on the pion mass . another extension deals with _ ab - initio _ nuclear physics , at least for light nuclei , and multiquark physics @xcite , with in particular , a renewed interest in the study of the @xmath192 dibaryon . in the sector of heavy quark spectroscopy , lqcd has confirmed many important results obtained by simpler , but empirical , tools . for instance , the @xmath13 interaction of charmonium , including its spin dependence , has been estimated , in agreement with the current potential models . latter , the charmonium spectrum was calculated directly . as already mentioned , some decisive contributions were made for the study of hybrid mesons with hidden heavy flavour @xcite . a born - oppenheimer potential can be computed for other configurations , e.g. , double - charm baryons , and lqcd is very suited for the treatment of the light degrees of freedom . for multiquarks , pure _ ab - initio _ studies are preceded by preliminary investigations inspired by models . for instance , in the sector of doubly - heavy mesons , an effective @xmath241 potential or an effective @xmath242 or @xmath243 interaction has been studied see , e.g. , @xcite . nowadays , multiquark states are often studied directly on the lattice , without the intermediate step of an effective potential . needless to say , lqcd faces difficulties as qcdsr , for the study of multiquark states . in the early days , the quantum numbers were not always clearly identified , and the genuine states were cleanly separated from the artefacts due to lattice boundaries . nowadays , the field is mature , and an impressive program of multiquark and multihadron physics has been developed by several groups @xcite . in this section , we briefly survey some phenomenological models which have been used to predict some exotic hadrons or describe them after their discovery . it should be stressed , indeed , that if these models are more empirical than qcdsr and lqcd , they can be adapted more flexibly to describe the world of ordinary hadrons and used to extrapolate towards novel structures . in the 70s , the hyperfine splitting among hadron multiplet was interpreted as the qcd analogue of the breit - fermi interaction in qed , namely the spin dependent part of the one - gluon exchange @xcite . the corresponding hyperfine interaction reads @xmath244 to be treated at first order , otherwise it requires some regularisation in the channels where it is attractive . a tensor term also contributes beyond @xmath4-waves . in 1977 , jaffe applied this hamiltonian to the configuration @xmath190 with @xmath191 assuming _ i ) _ su(3)@xmath95 flavour symmetry , and _ ii ) _ that the short - range correlation @xmath245 has the same value as for ordinary baryons @xcite . he found a remarkable coherence in the spin - colour operator , with @xmath246 for @xmath192 , to be compared to @xmath247 for @xmath199 or @xmath248 , i.e. , @xmath249 for the @xmath194 threshold . this means that the @xmath192 is bound below the @xmath194 threshold by half the @xmath7-@xmath199 splitting , about @xmath250mev . the @xmath192 was searched for in many experiments , without success so far . for instance , the double hypernucleus @xmath251[\lambda\lambda]{he}$ ] @xcite was not seen decaying into @xmath252 . the model has been reexamined thoroughly @xcite , with the result that any correction tends to reduce the binding , and eventually , the @xmath192 is pushed in the continuum . these corrections are : su(3)@xmath95 breaking , more realistic estimate of the short - short correlation ( a dibaryon is more dilute than a baryon ) , and proper inclusion of the full hamiltonian ( kinetic energy and spin - independent confinement besides @xmath253 ) . of course , some additional attraction such a scalar - isoscalar exchange between the two @xmath248 can restore the binding . meanwhile , some other configurations were found , for which the chromomagnetic interaction , if alone , induces binding . in 1987 , the grenoble group and h.j . lipkin @xcite proposed independently and simultaneously a heavy pentaquark @xmath254 , where @xmath255 or @xmath164 . with the same hypotheses as jaffe s for the @xmath192 , and assuming an infinite mass for the heavy antiquark @xmath256 , they found an optimal @xmath257 , to be compared to @xmath247 for the threshold @xmath258 , i.e. , the same binding energy @xmath250mev . in a series of papers @xcite , leandri and silvestre - brac applied the chromomagnetic operator @xmath259 with suitable changes for antiquarks , to a variety of configurations , and found several cases , for which the chromomagnetic interaction , if alone , induces binding . as noted by lichtenberg @xcite , the @xmath260 dependence is somewhat extreme , as heavy quarks give more compact wave - functions with larger short - range correlation factor . this led some authors to adopt a more empirical prescription , @xmath261 where the coefficients are taken from a fit to ordinary mesons and baryons @xcite . for instance , in the case of @xmath262 with @xmath83 , the diagonalisation of , supplemented by effective constituent masses , gives a state which corresponds nicely to the mass and properties of the @xmath48 . its narrowness comes from being almost a pure octet - octet state of colour in the channels @xmath263 . as most approaches based on quark dynamics , the chromomagnetic model predicts an isospin @xmath9 partner of @xmath48 , with @xmath264 and @xmath64 for the neutral , which is not yet seen experimentally . [ [ the - role - of - flavour - independence ] ] the role of flavour independence + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in constituent models , one can switch off the spin - dependent part of the interaction , and study the limit of a pure central potential . its main property is _ flavour independence _ , with some interesting consequences , which are independent of the detailed shape of the interaction . in particular , if a constituent mass increases , the kinetic energy decreases , and one gets a lower energy , or , say , more binding . this explains why the @xmath15 family has more bound states than the @xmath265 family below the open - flavour threshold . another well - studied consequence deals with doubly - heavy tetraquarks @xmath266 . the 4-body system benefits from the heavy - heavy interaction , which is absent in the @xmath267 threshold . hence , for a large enough @xmath268-to-@xmath269 mass ratio , the system becomes stable . explicit calculations with a colour - additive model @xmath270 normalised to the quarkonium potential @xmath271 , indicates that in such spin - independent interaction , @xmath158 is probably unbound , but @xmath272 stable @xcite . one hardly finds similar configurations . if , for instance , one considers @xmath273 within , there is no obvious gain with respect to the best threshold made of two baryons , but flavour independence dictates that before any spin corrections , the threshold @xmath274 is lower than @xmath275 . the stability of @xmath266 for large mass ratios is intimately related to the observation that the hydrogen molecule is more deeply bound ( in units of the threshold binding energy ) than the equal - mass molecules such as ps@xmath276 . the hamiltonian for @xmath277 reads @xmath278\\ { } + \left(\frac{1}{4\,m}-\frac{1}{4\,m}\right ) \left({\boldsymbol{p}}_1 ^ 2+{\boldsymbol{p}}_2 ^ 2-{\boldsymbol{p}}_3 ^ 2-{\boldsymbol{p}}_4 ^ 2\right)~. \end{gathered}\ ] ] its @xmath279-breaking term @xmath280 lowers the ground - state energy with respect to the @xmath279-even part @xmath281 . since @xmath281 , a rescaled version of the ps@xmath276 molecule , and @xmath192 , have the same threshold , and since ps@xmath276 is stable , h@xmath276 is even more stable . clearly , the coulomb character of @xmath282 hardly matters in this reasoning . the key property is that the potential does not change when the masses are modified . in the case of a simple spin - independent potential with the colour - additive rule , one does not get binding for @xmath283 , but binding shows up when @xmath284 increases . note that , if one breaks the symmetry of ps@xmath276 differently and considers instead @xmath285 configurations , stability is lost for @xmath286 , or , of course , @xmath287 @xcite . a protonium atom can not polarise strongly enough a positronium atom and attach it . similarly a chromoelectric potential can not produce a bound @xmath288 tetraquark : one needs chromomagnetism or long - range yukawa interaction . [ [ more - detailed - comparison - of - ps_2-and - tetraquark ] ] more detailed comparison of ps@xmath276 and tetraquark + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + another problem is to understand why the positronium molecule ps@xmath276 ( neglecting annihilation ) lies slightly below its dissociation threshold , while a chromo - electric model , associated with the colour - additive rule , does not bind ( at least according to most computations , an exception being @xcite ) . this is due to a larger disorder in the colour coefficients than in the electrostatic strength factors @xmath289 entering the coulomb potential in ps@xmath276 . consider , indeed , a 4-body hamiltonian scaled to @xmath290={\boldsymbol{p}}_1 ^ 2+{\boldsymbol{p}}_2 ^ 2+{\boldsymbol{p}}_3 ^ 2+{\boldsymbol{p}}_4 ^ 2 + \sum_{i < j } g_{ij}\,v(r_{ij})~ , \qquad g=\{g_{ij}\}~,\ ] ] with @xmath291 , having at least one bound state in the symmetric case @xmath292 for which @xmath293 @xmath294 . the variational principle , with the solution of this symmetric case as a trial function , implies that @xmath295)\le \min(h[g_0])~,\ ] ] so that , schematically , the more asymmetric the distribution @xmath216 , the lower the ground - state energy . for the cases of interest , one could be more precise , as the distribution is of the type @xmath296 . the ground - state energy @xmath297 is convex as a function of @xmath298 , and , since maximal at @xmath299 , it is convexly decreasing as a function of @xmath300 for both @xmath301 and @xmath302 . if one compares two values of @xmath298 with opposite sign , then it is a good guess , but not a rigorous statement , to say that the larger @xmath300 , the lower the energy , as @xmath297 is nearly parabolic near @xmath299 . now consider the actual values in table [ tab : coeff - tetra ] . .strength coefficients for the positronium molecule and the tetraquark states with colour wave - function @xmath303 or @xmath304 , as well as for their threshold . the list of coefficients is of the type @xmath305 after suitable renumbering , and summarised as @xmath306 , where @xmath307 is the symmetric case @xmath308 . [ tab : coeff - tetra ] [ cols="^,^,^,^,^",options="header " , ] the asymmetry parameter @xmath300 is much larger for ps@xmath276 than for @xmath309 . this explains why the positronium molecule is bound . means that ps@xmath276 benefits from correlation and anticorrelation effects in the various pairs . ] a tetraquark with a pure @xmath310 colour structure has its asymmetry parameter smaller than that of the threshold and has the same sign , so this toy model is rigorously unbound . for @xmath304 , the signs are different , and the magnitude are comparable , so the analysis is more delicate . we note that @xmath311 for ps@xmath276 results in a fragile binding , so it is not surprising that @xmath312 leaves @xmath304 unbound in numerical studies . to summarise , the tetraquark is penalised by the non - abelian character of the colour algebra in the simple model , and its stability can not rely on the asymmetries of the potential energy . it should use other asymmetries , in particular through the masses entering the kinetic energy , or spin effects , or mixing of @xmath310 and @xmath313 , or the coupling to decay channels , etc . [ [ string - dynamics-1 ] ] string dynamics + + + + + + + + + + + + + + + the prescription is of course rather crude , as it treats colour as a global variable similar to isospin , and assumes arbitrarily the choice of a pairwise interaction . it can be justified in the limiting cases were @xmath314 quarks are clustered and the last quark isolated . so it might provide a reasonable interpolation for other spatial configurations . as mentioned earlier , an alternative is the @xmath315-shape interaction of fig . [ fig : string1 ] , now used to estimate a potential which reads @xmath316 if @xmath317 is the linear confinement of mesons . note that features the minimal path linking three points , an old problem first raised by fermat and torricelli . and the value of the minimal sum of distances @xcite . ] for baryons alone , solving the 3-quark problem with instead of the perimeter - based confinement @xmath318 is technically more involved , but does not change the patterns of the excitation spectrum . for tetraquarks , a precursor of non - pairwise potential was the flip - flop model @xcite , with quadratic confinement . in the more recent linear version , one takes the minimum of the two possible meson - meson pairing configurations and the connected diagram shown in fig . [ fig : tetra - string ] . the good surprise is that this potential is more attractive than the one deduced from the colour - additive rule and thus is more favourable for binding @xcite . the same property is observed for the pentaquark @xcite , fig . [ fig : penta - string ] , the dibaryon , fig . [ fig : hexa - string ] and the system made of 3 quarks and 3 antiquarks @xcite , fig . [ fig:3q-3qbar - string ] . however , this string potential corresponds to a kind of born - oppenheimer approximation for the colour fluxes at given quark positions . it mixes colour @xmath319 and @xmath320 for any pair of quark , to optimise the overall energy . thus it holds for quarks bearing different flavours . in @xcite , an attempt is proposed to combine quark statistics and string dynamics in the case of tetraquarks with identical quarks or antiquarks : not surprisingly , the possibility of binding is reduced . + note that in the above potential models , there is no explicit junction _ la _ rossi and veneziano : the minimal length is the only guidance , and nothing refrain a transition from a disconnected flip - flop to a connected string . it is observed , however , that the ground - state uses mainly the flip - flop term , so there are probably two classes of states : the lowest have the same number of junctions as the thrshold , and thus are rather broad if the fall - apart decay is energetically allowed . the second class might be higher in mass , but benefits from a kind of metastability due to its junction content . these string - inspired potentials , however simplistic , have motivated some studies . for instance , the literature on steiner trees helps in optimising the location of the junctions and the estimate of the minimal strength @xcite . as discussed at length in sec . [ sse : chromoelectric ] , the tetraquark @xmath266 might become stable by the sole effect of chromomagnetic forces , but the stability requires a very large @xmath284 mass ratio . it was also noticed @xcite , that some configurations such as @xmath83 , there is a favourable spin - spin interaction in the light sector that benefits the tetraquark and not the threshold . a calculation by rosina et al . @xcite , using a potential tuned to fit the ordinary hadrons , and the colour - additive rule , demonstrated stability for @xmath321 , and this was confirmed by vijande et al . there are not too many similar configurations , in which both chromoelectric and chromomagnetic effects are favourable . for instance , for the doubly - flavoured dibaryons @xmath273 , the two thresholds @xmath274 and @xmath322 are nearly degenerate , at least in case the charm , @xmath323 . the former threshold benefits from the chromoelectric attraction between the two heavy quarks , while the second can receive a favourable chromomagnetic interaction . it is , however , difficult for the 6-quark configuration to combine astutely the two effects , and double - charm baryons seem hardly stable in constituent models @xcite . once a model is adopted , typically @xmath324~,\ ] ] or a more complicated model involving multi - body forces , one is left with solving the wave equation to obtain the energies and the wavefunctions . in the case of quarkonium , there are dozens of reliable numerical methods , starting with the one proposed by hartree @xcite , consisting of matching a regular outgoing solution and a regular incoming solution of the radial equation . most methods can be extended to the case of coupled equations in the unnatural - parity channels , similar to the @xmath325 coupling of the deuteron in nuclear physics . many rigorous results have been revisited and developed for studies related to the quarkonium : scaling law , virial - like theorems , level order , value of wave - function at the origin for the successive radial excitations , etc . this is reviewed in @xcite . the 3-body problem one has to solve for baryons is often a psychological barrier that refrain to envisage a detailed and rigorous treatment beyond the harmonic oscillator , and there is sometimes a confusion between the `` constituent model of baryons '' and the harmonic oscillator model of baryons . for instance , the negative - parity excitation being half - way of the positive - parity excitation is a property of the latter , but not necessarily of the former . for baryons , there are just a few rigorous results , for instance dealing the level order of states when the degeneracy obtained in the harmonic oscillator or hyperscalar limit are lifted at first order @xcite . as for the numerical solution , various methods have been used , such as hyperspherical expansion , faddeev equations , or variational methods , with a comparison in @xcite . note that among the latter , the expansion on a harmonic - oscillator basis is less and less adopted , as it does not give good convergence for the short - range correlation parameters which enter some decays and some production mechanisms . for multiquark systems , the hyperspherical expansion works rather well @xcite , especially for deeply bound states , if any , for which the various r.m.s . radii are of the same order of magnitude . the tendency is towards the method of gaussian expansion , already used in quantum chemistry . if @xmath326 are jacobi variable describing the relative motion of the @xmath199-body system , then the wave - function is sought as @xmath327\,|a\rangle~,\ ] ] where @xmath328 denotes the basis of internal colour - spin - flavour states needed to construct the most general colour singlet of given overall spin and flavour , and @xmath329 is a @xmath330 definite positive symmetric matrix . for a given set of matrices @xmath329 , the weight factors @xmath331 are given by a generalised eigenvalue equation . then the range parameters entering the @xmath329 are minimised numerically . this latter task is not obvious , and sophisticated techniques have been developed , such as the stochastic search @xcite , and the method of hiyama , kino and kamimura @xcite , in which a specific choice of jacobi variables is chosen in each channel . the variational energies below the threshold are interpreted as bound states . the scaling properties of the states above the threshold are scrutinised to decide whether they are genuine resonances or artefacts of the finite variational expansion . this is somewhat similar to the analyses of states above the threshold in qcdsr or lqcd . [ [ diquarks ] ] diquarks + + + + + + + + hadron spectroscopy with diquarks has ups and downs , or , say , pros and cons . the art of physics is to identity effective degrees of freedom at a certain scale , and any simplification is welcome . the concept of diquark is very useful to describe many phenomenons such a multiple production @xcite , and might be legitimated by boson - fermion symmetry @xcite . in spectroscopy , the diquark model provides a simple physical picture that might be obscured in complicated three - body calculations , and it provides an efficient tool for investigating exotics @xcite . some reservations are however expressed about the abundant literature on diquarks , which is sometimes confusing , as some recent papers do refer to the old ones : * explicit quark model calculations find diquark correlations , but the effect depends on a somewhat arbitrary regularisation of the chromomagnetic interaction @xcite . in other studies , no diquark clustering is found for the ground state @xcite . this means that the diquark picture of low - lying baryons implies a change of dynamics . on the other hand , a dynamical diquark formation is obtained for highly excited baryons : if the total angular momentum @xmath332 increases , it is first shared equally among the 3 quarks , but at larger @xmath332 the third quark takes @xmath333 while the two first are mainly in a relative @xmath4-wave ( of course modulo permutations ) @xcite . * the diquark model becomes a good approximation for the ground state of double - charm baryons @xmath334 provided it is handled and interpreted properly , since a part of the @xmath241 interaction comes from the third quark . in the case of the harmonic oscillator , with the usual definition of the jacobi variables @xmath97 and @xmath335 , one notices that @xmath336 , so the direct @xmath241 interaction @xmath337 gets a 50% increase from the light quark . moreover , the first excitations of @xmath334 ( besides spin ) occur with @xmath338 , so a new diquark has to be estimated for each level . in fact a born - oppenheimer treatment la h@xmath200 is more appropriate . [ [ bornoppenheimer ] ] born oppenheimer + + + + + + + + + + + + + + + + along this review , we stressed the importance of the born - oppenheimer point of view : the quarkonium potential is just an effective interaction integrating out the gluon and light - quark degrees of freedom surrounding a heavy quark and a heavy antiquark ; doubly - heavy baryons @xmath334 have typically two scales of energy for its spectrum , the excitation energy of the @xmath241 pair , and the excitation energy of the light quark , and this comes very naturally in the born - oppenheimer treatment , as it does in atomic physics for h@xmath200 . we also mentioned that hybrid mesons , schematically noted @xmath339 have been studied by estimating an `` excited quarkonium potential '' : for a given @xmath239 separation , the energy of the excited gluon field is computed either from a simple bag model , or from lqcd . a born - oppenheimer treatment of @xmath266 would also be feasible , to cross - check the 4-body calculations discussed in sec . [ sse : chromoelectric ] . recently , a very comprehensive and unifying born - oppenheimer treatment of the @xmath75 states was presented @xcite , with a series of effective @xmath239 potentials adapted to the quantum states of the light constituents ( gluon and light quarks ) . this work could of course be generalised as to include @xmath340 instead of @xmath341 in the light sector , and applied to the lhcb pentaquarks . there are interesting alternatives for the coupling of @xmath239 to light quarks and antiquarks . for instance , a relatively weak interaction between @xmath12 or @xmath342 with nuclei has been proposed , but , once more , this potential is probed by a particle which is heavy enough to get bound . in _ , the compact @xmath239 is trapped inside an extended light hadron . see @xcite . in the future , the field of exotic hadrons might be extended very far , if some new constituents are discovered , for instance at lhc . as stressed by many authors ( see , for instance , @xcite ) , some anomalies in the higgs branching ratios , and the other recent observations at cern might indicate the existence of heavy scalars . among the many possible scenarios , there is the possibility of scalar quarks with colour charge @xmath343 , having the same colour - mediated interaction as the ordinary quarks , except of course for spin and statistics . perhaps the lowest of theses scalars will live long enough to hadronise , and give rise to new hadrons . the lifetime of the hypothetical scalar quarks would of course be crucial . remember that for many years the potential model of charmonium and bottomonium was extrapolated towards topomonium . then it was realised that , when the mass of @xmath209 quark is too large , its weak decay becomes too fast for hadronisation to take place @xcite . if this is the case for any new super - heavy quark or squark , then the realm of exotic hadrons will remain the privilege of low - energy hadron physics . let us consider a scenario with heavy scalar quarks of colour @xmath343 , as predicted by supersymmetry or any other bsm ( beyond standard model ) theory . ( there are speculations about colour @xmath320 or other gauge groups . ) we note them `` squarks '' without necessarily endorsing supersymmetry . then , one gets squarkonia ( @xmath344 ) , and perhaps the scalar ground state could be responsible for the excess tentatively seen at 750gev . see , e.g. , @xcite . i apologise for the lack of space needed to cite the other @xmath345 relevant references , and the lack of expertise to separate the really innovative papers from the `` ambulance chasing '' opportunities @xcite . the lightest charged hadron is likely @xmath346 for @xmath347 with charge @xmath348 or @xmath349 of @xmath350 type , but spin 0 , if @xmath351 with charge @xmath352 . the baryon sector becomes entertaining if several squarks are put together . for instance @xmath353 has quantum numbers @xmath110 as the @xmath354 disquark contains one unit of orbital excitation , to fulfill the bose statistics . the triply - superexotic @xmath355 requires a fully antisymmetric wave function @xmath356 , as per eq . that was proposed 50 years ago for the last member of the @xmath357 multiplet for light baryons , the so - called @xmath358 $ ] multiplet in the notation of , e.g. , @xcite . fifty years after the first baryon conference , the physics has evolved , but one uses the same wave function ( besides some scaling ) . exotic hadrons are coming up and down , or back and forth if one prefers . for outsiders , there is somewhat a discredit on a field in which many states regularly show up rapidly and then disappear within a few years , with many papers hastily written to explain the new hadrons , and never any deep understanding on why the models fail . this is of course , a superficial view . the truth is that the experiments are very difficult , due to a low statistics for a signal above an intricate background . the situation is even worse for broad resonances . as often stressed @xcite , the location of the peak might depend on the production process and on the final state , so that there is often the question on whether two or more candidates correspond to the same `` state '' . so far , several mass ranges and flavour sectors have been investigated with already a few good candidates for exotics such as @xmath48 . the main lesson of the last decades is clearly the richness of the configurations mixing heavy and light constituents . the hidden - charm sector is already much explored , and even if the recent discoveries have to be checked , refined and completed , there is clearly an urgent need for investigating the double charm and other flavor exotic sectors . hopefully , there will be an emulation between electron - positron machines , proton colliders , heavy - ion collisions and intense hadron facilities to develop this program . on the theory side , the fundamental approaches such as lqcd or qcdsr involve lengthy calculations and usually come second , to clarify some predictions . first appear the speculations based on more empirical models , inspired by qcd . in phenomenological qcd , _ arx tarpeia capitoli proxima _ : physicists inventing molecules or quark clusters that account for the latest exotics are praised for their inspiration , invited to many workshops and conferences , hired in committees , etc . , but , when the states disappear , the same model - builders are pointed out as lacking rigour and corrupting the field . to be honest , it often happens that the pioneers have written their predictions rather carefully , with many warnings about the width of states in the continuum , but that the followers have developed the models beyond any reasonable control , with tables and tables of states which simply fall part . for one of the models of multiquarks , a warning was set @xcite that only a few states will show up as peaks , the others , being very broad , just paving the continuum , but the warning was often forgotten . pandora box syndrome _ affects practically all approaches . for instance , the molecular picture was first focused on one - pion exchange , with strict selection rules , that makes the interaction repulsive in any second channel . with the exchange of vector mesons , and the old - fashioned potentials replaced by effective theories , the molecular picture has gained some respectability , but has dramatically widened the spectrum of predictions . in the diquark model , frederikson and jandel @xcite discussed the existence of possible dibaryons , and the problem was rediscovered recently @xcite , that the picture of @xmath75 as diquark - antidiquark states probably implies higher configurations . this is why genuinely exotic states , lying below their dissociation threshold , will always play a privileged role , and the searches in the sector of double heavy flavour should be given the highest priority . in any case , the studies on the mechanisms of confinement and the resulting nuclear forces should be pursued . the strong - coupling limit of qcd and lattice simulations explain how a linear confinement is developed between a quark and an antiquark . it is already more delicate to understand its extension to baryons , with possible interplay between @xmath7-shape and @xmath315-shape configurations for the gluon density , and even more speculative to write down a confining hamiltonian for multiquarks . what remain to be further investigated , based on the already rich literature is how clustering sometimes occur in baryons and higher multiquarks , and how the linear confinement breaks to lead to the actual decay of resonances and the virtual hadron - hadron components , i.e. , to operate the matching of the quark dynamics and the long - range hadron - hadron interaction . this review benefited from stimulating discussions with my past and present collaborators and with many colleagues . it was completed during a visit at riken , in the framework of a jsps - cnrs agreement , and finalised during a visit at ect*. the hospitality provided by emiko hiyama , by jochen wambach , and by their colleagues is gratefully acknowledged . thanks are also due to m. asghar for very interesting comments on the manuscript . 300ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty http://cds.cern.ch/record/100978[__ ] ( , , ) http://cds.cern.ch/record/107749[__ ] ( , , ) http://cds.cern.ch/record/317696[__ ] ( , , ) link:\doibase 10.1103/revmodphys.34.766 [ * * , ( ) ] link:\doibase 10.1103/physrev.101.1570 [ * * , ( ) ] http://cds.cern.ch/record/107781[__ ] ( , , ) link:\doibase 10.1103/physrev.142.1000 [ * * , ( ) ] link:\doibase 10.1088/1674 - 1137/38/9/090001 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1016/j.physrep.2007.07.006 [ * * , ( ) ] , in http://alice.cern.ch/format/showfull?sysnb=0105412[__ ] ( ) p. link:\doibase 10.1016/s0375 - 9474(97)00348 - 5 [ * * , ( ) ] link:\doibase 10.1016/s0375 - 9474(97)00694 - 5 [ * * , ( ) ] link:\doibase 10.1016/j.nuclphysbps.2013.10.040 [ * * , ( ) ] @noop * * , ( ) , link:\doibase 10.1016/j.physrep.2010.07.005 [ * * , ( ) ] , @noop * * , ( ) , link:\doibase 10.1016/j.physrep.2016.05.004 [ ( ) , 10.1016/j.physrep.2016.05.004 ] , https://books.google.fr/books?id=r_27t1yxxqac[__ ] , cambridge monographs on particle physics , nuclear physics and cosmology ( , ) , ed . , http://scitation.aip.org/content/aip/proceeding/aipcp/1735[__ ] , vol . ( ) , , , , and , eds . , http://www.actaphys.uj.edu.pl/findarticle?series=sup&vol=8&page=503[__ ] , vol . ( ) @noop link:\doibase 10.1016/j.physletb.2005.04.011 [ * * , ( ) ] , link:\doibase 10.1016/j.physletb.2003.05.003 [ * * , ( ) ] , link:\doibase 10.1016/j.physletb.2005.08.113 [ * * , ( ) ] , link:\doibase 10.1063/1.1482444 [ * * , ( ) ] , , link:\doibase 10.1103/physrevlett.79.1630 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.104.241803 [ * * , ( ) ] , in https://inspirehep.net/record/1409299/files/arxiv:1512.03699.pdf[__ ] ( ) link:\doibase 10.1016/j.ppnp.2015.03.001 [ * * , ( ) ] , @noop link:\doibase 10.1103/physrevlett.90.242001 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.68.032002 , 10.1103/physrevd.75.119908 [ * * , ( ) ] , , link:\doibase 10.1103/physrevd.32.189 [ * * , ( ) ] link:\doibase 10.1016/0370 - 2693(79)90347 - 2 [ * * , ( ) ] link:\doibase 10.1016/0550 - 3213(82)90140 - 7 [ * * , ( ) ] @noop * * , ( ) , link:\doibase 10.1143/ptp.117.1077 [ * * , ( ) ] , link:\doibase 10.1016/j.nuclphysa.2007.03.173 [ * * , ( ) ] link:\doibase 10.1140/epja / i2006 - 10199 - 0 [ * * , ( ) ] link:\doibase 10.1103/physrevd.91.094020 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.93.034035 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.39.1506 [ * * , ( ) ] link:\doibase 10.1103/physrevd.16.799 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevlett.106.212001 [ * * , ( ) ] , link:\doibase 10.1016/j.physletb.2011.09.096 [ * * , ( ) ] , @noop ( ) , link:\doibase 10.1103/physrevd.88.054029 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.70.071102 [ * * , ( ) ] , @noop * * , ( ) link:\doibase 10.1142/s0217751x09046576 [ * * , ( ) ] link:\doibase 10.1016/0370 - 2693(86)91013 - 0 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.69.2337 [ * * , ( ) ] link:\doibase 10.1103/physrevd.72.092004 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.89.102001 [ * * , ( ) ] , , @noop * * , ( ) , ( ) , @noop * * , ( ) , link:\doibase 10.1103/physrevlett.100.142001 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.98.082001 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.100.202001 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.96.082003 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.104.092001 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.104.112004 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.101.172001 [ * * , ( ) ] , @noop ( ) , link:\doibase 10.1103/physrevd.82.094008 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.71.031501 [ * * , ( ) ] , @noop * * , ( ) , link:\doibase 10.1103/physrevlett.110.252002 [ * * , ( ) ] , link:\doibase 10.1016/j.physletb.2013.10.041 [ * * , ( ) ] , in http://inspirehep.net/record/1415952/files/arxiv:1601.05025.pdf[__ ] ( ) @noop ( ) , link:\doibase 10.1088/0954 - 3899/37/7/075017 [ * * , ( ) ] , @noop * * , ( ) , link:\doibase 10.1088/0034 - 4885/76/7/076301 [ * * , ( ) ] , link:\doibase 10.1142/s0218301392000242 [ * * , ( ) ] link:\doibase 10.1016/0003 - 4916(60)90001 - 4 [ * * , ( ) ] @noop link:\doibase 10.1103/physrevd.18.4187 [ * * , ( ) ] link:\doibase 10.1016/j.nuclphysa.2016.04.013 [ * * , ( ) ] , link:\doibase 10.1142/s0217751x02012715 [ * * , ( ) ] , , link:\doibase 10.1103/physrevlett.18.801 [ * * , ( ) ] link:\doibase 10.1103/physrev.165.1770 [ * * , ( ) ] link:\doibase 10.1016/0550 - 3213(72)90491 - 9 [ * * , ( ) ] link:\doibase 10.1016/0550 - 3213(71)90291 - 4 [ * * , ( ) ] link:\doibase 10.1016/0550 - 3213(75)90104 - 2 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.52.s1 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.91.012002 [ * * , ( ) ] , link:\doibase 10.1007/s002180050406 [ * * , ( ) ] , @noop ( ) link:\doibase 10.1140/epjh / e2012 - 20032 - 0 [ * * , ( ) ] @noop * * , ( ) , link:\doibase 10.1140/epjc / s2004 - 01815 - 4 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.75.114018 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.101.232002 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.80.072003 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.110.182001 [ * * , ( ) ] , link:\doibase 10.1016/j.physletb.2008.07.018 [ * * , ( ) ] link:\doibase 10.1103/physrevd.83.094005 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.114.062004 [ * * , ( ) ] , @noop ( ) , link:\doibase 10.1103/physrevd.66.014008 [ * * , ( ) ] , @noop * * , ( ) @noop * * , ( ) , @noop * * , ( ) , link:\doibase 10.1016/s0920 - 5632(02)01948 - 5 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.74.011103 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.89.052003 [ * * , ( ) ] , link:\doibase 10.1007/jhep12(2013)090 [ * * , ( ) ] , link:\doibase 10.1146/annurev - nucl-102711 - 094949 [ * * , ( ) ] , , link:\doibase 10.1103/physrevd.90.094007 [ * * , ( ) ] , link:\doibase 10.1016/s0370 - 2693(02)03069 - 1 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.74.094003 [ * * , ( ) ] , link:\doibase 10.1016/0370 - 2693(87)91244 - 5 [ * * , ( ) ] link:\doibase 10.1016/0370 - 2693(87)90055 - 4 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.81.44 [ * * , ( ) ] , link:\doibase 10.1016/s0370 - 2693(99)00058 - 1 [ * * , ( ) ] link:\doibase 10.1016/j.physletb.2004.03.012 [ * * , ( ) ] , link:\doibase 10.1140/epjc / s2004 - 02042 - 9 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.73.091101 [ * * , ( ) ] , https://cds.cern.ch/record/298433[__ ] , ( , , ) link:\doibase 10.1103/physrevlett.115.072001 [ * * , ( ) ] , @noop ( ) , link:\doibase 10.1007/jhep07(2014)103 [ * * , ( ) ] , link:\doibase 10.1140/epja / i2015 - 15152 - 6 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.93.094001 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.117.082003 , 10.1103/physrevlett.117.109902 [ * * , ( ) ] , , link:\doibase 10.1103/physrevd.92.031502 [ * * , ( ) ] , @noop ( ) , link:\doibase 10.1103/physrevd.94.034002 [ * * , ( ) ] , link:\doibase 10.1016/j.physletb.2015.11.068 [ * * , ( ) ] , @noop ( ) , @noop link:\doibase 10.1103/physrevd.35.128 [ * * , ( ) ] link:\doibase 10.1103/physrevc.48.1926 [ * * , ( ) ] link:\doibase 10.1103/physrevc.50.1796 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.59.1878 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.63.1570 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.53.423 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.112.202301 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.90.035204 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.39.1889 [ * * , ( ) ] link:\doibase 10.1007/s00601 - 012 - 0390 - 6 [ * * , ( ) ] , link:\doibase 10.1093/ptep / ptv076 [ * * , ( ) ] , @noop * * , ( ) link:\doibase 10.1016/0375 - 9474(96)00211 - 4 [ * * , ( ) ] link:\doibase 10.1142/s0217979207037855 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.38.195 [ * * , ( ) ] , link:\doibase 10.1098/rspa.1989.0115 [ * * , ( ) ] link:\doibase 10.1088/0954 - 3899/28/7/323 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.59.627 [ * * , ( ) ] link:\doibase 10.1103/physrevc.92.065202 [ * * , ( ) ] , link:\doibase 10.1103/physrev.47.903 [ * * , ( ) ] link:\doibase 10.1016/0370 - 1573(93)90141-y [ * * , ( ) ] @noop ( ) , , link:\doibase 10.1007/s00601 - 011 - 0296 - 8 [ * * , ( ) ] link:\doibase 10.1103/physrevc.65.041001 [ * * , ( ) ] , link:\doibase 10.1088/0954 - 3899/42/2/025103 [ * * , ( ) ] link:\doibase 10.1103/physrevc.91.014003 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.94.024002 [ * * , ( ) ] , link:\doibase 10.1209/epl / i2003 - 00176 - 1 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.80.1876 [ * * , ( ) ] , link:\doibase 10.1016/j.nuclphysa.2009.05.118 [ * * , ( ) ] link:\doibase 10.1103/physreva.67.034702 [ * * , ( ) ] , @noop * * , ( ) , link:\doibase 10.1103/physrevlett.21.950 [ * * , ( ) ] link:\doibase 10.1007/jhep06(2016)041 [ * * , ( ) ] , @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1016/0370 - 2693(80)90906 - 5 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1016/0370 - 2693(77)90081 - 8 [ * * , ( ) ] link:\doibase 10.1103/physrevd.17.1444 [ * * , ( ) ] link:\doibase 10.1016/0370 - 2693(78)90872 - 9 [ * * , ( ) ] link:\doibase 10.1103/physrev.76.1739 [ * * , ( ) ] link:\doibase 10.1016/0370 - 1573(78)90190 - 4 [ * * , ( ) ] link:\doibase 10.1016/0003 - 4916(79)90091 - 5 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevd.21.203 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) , @noop * * , ( ) link:\doibase 10.1007/bf01413192 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.88.015207 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.78.034007 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.86.014004 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.92.014005 [ * * , ( ) ] , in @noop _ _ ( ) link:\doibase 10.1016/j.nuclphysa.2005.08.022 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.88.056012 [ * * , ( ) ] , link:\doibase 10.1088/1674 - 1137/36/1/002 , 10.1088/1674 - 1137/36/3/006 [ * * , ( ) ] , @noop * * , ( ) , link:\doibase 10.1103/physrevc.69.055202 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.86.034019 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.115.122001 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.80.034008 [ * * , ( ) ] , link:\doibase 10.1016/j.nuclphysa.2005.01.022 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.85.014015 [ * * , ( ) ] , link:\doibase 10.1016/j.physletb.2011.09.088 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.89.035201 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.84.014031 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.86.014020 [ * * , ( ) ] , link:\doibase 10.1016/j.nuclphysa.2005.03.050 [ * * , ( ) ] , in http://alice.cern.ch/format/showfull?sysnb=0082623[__ ] ( ) link:\doibase 10.1088/0954 - 3899/40/4/043001 [ * * , ( ) ] , link:\doibase 10.1016/j.nuclphysb.2016.01.017 [ * * , ( ) ] , link:\doibase 10.1007/s00601 - 008 - 0263 - 1 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1016/0370 - 2693(80)90491 - 8 [ * * , ( ) ] link:\doibase 10.1016/0370 - 2693(84)90473 - 8 [ * * , ( ) ] @noop * * , ( ) , link:\doibase 10.1103/physrevd.92.114019 [ * * , ( ) ] , link:\doibase 10.1016/0550 - 3213(79)90022 - 1 [ * * , ( ) ] link:\doibase 10.1016/0550 - 3213(79)90023 - 3 [ * * , ( ) ] link:\doibase 10.1016/0550 - 3213(79)90024 - 5 [ * * , ( ) ] link:\doibase 10.1103/physrevd.75.014005 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.78.034012 [ * * , ( ) ] , link:\doibase 10.1016/j.nuclphysbps.2010.10.064 [ * * , ( ) ] , @noop ( ) , , @noop * * , ( ) , link:\doibase 10.1103/physrevlett.106.162002 [ * * , ( ) ] , link:\doibase 10.1103/physrevlett.106.162001 [ * * , ( ) ] , link:\doibase 10.1063/1.3647373 [ * * , ( ) ] , link:\doibase 10.1093/ptep / pts010 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.92.054002 [ * * , ( ) ] , link:\doibase 10.1063/1.4949382 [ * * , ( ) ] @noop ( ) , link:\doibase 10.1103/physrevc.57.3384 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.60.054012 [ * * , ( ) ] , @noop * * , ( ) , link:\doibase 10.1103/physrevd.93.034501 [ * * , ( ) ] , @noop * * , ( ) link:\doibase 10.1103/physrevlett.87.212502 [ * * , ( ) ] link:\doibase 10.1016/0370 - 2693(83)91523-x [ * * , ( ) ] link:\doibase 10.1103/physrevd.33.2043 [ * * , ( ) ] link:\doibase 10.1103/physrevd.36.2079 [ * * , ( ) ] link:\doibase 10.1016/0146 - 6410(90)90023-w [ * * , ( ) ] link:\doibase 10.1103/physrevd.40.2340 [ * * , ( ) ] link:\doibase 10.1103/physrevd.45.4221 [ * * , ( ) ] link:\doibase 10.1103/physrevd.51.3628 [ * * , ( ) ] in http://alice.cern.ch/format/showfull?sysnb=0160281[__ ] ( ) link:\doibase 10.1140/epjc / s10052 - 006 - 0142 - 1 [ * * , ( ) ] , @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physreva.57.4956 [ * * , ( ) ] in @noop _ _ , ( ) p. link:\doibase 10.1103/physrevd.70.014009 [ * * , ( ) ] , link:\doibase 10.1016/j.physletb.2009.03.018 [ * * , ( ) ] , link:\doibase 10.1016/0003 - 4916(86)90088 - 6 [ * * , ( ) ] link:\doibase 10.1103/physrevd.76.114013 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.81.015205 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.85.014019 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.87.034040 [ * * , ( ) ] , link:\doibase 10.1140/epjc / s2003 - 01416 - 9 [ * * , ( ) ] , link:\doibase 10.1016/j.physletb.2009.03.008 [ * * , ( ) ] , link:\doibase 10.1140/epjc / s10052 - 009 - 1050-y [ * * , ( ) ] , link:\doibase 10.1007/bf01413104 [ * * , ( ) ] @noop * * , ( ) , link:\doibase 10.1103/physrevd.76.094027 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.94.034038 [ * * , ( ) ] , http://cds.cern.ch/record/112712[__ ] ( , , ) link:\doibase 10.1016/0370 - 1573(79)90095 - 4 [ * * , ( ) ] link:\doibase 10.1016/0370 - 1573(80)90031 - 9 [ * * , ( ) ] http://cds.cern.ch/record/847188[__ ] , cambridge monographs on particle physics , nuclear physics , and cosmology ( , , ) link:\doibase 10.1016/0550 - 3213(90)90144 - 3 [ * * , ( ) ] link:\doibase 10.1016/0370 - 2693(91)90163-k [ * * , ( ) ] link:\doibase 10.1016/0370 - 1573(92)90078-e [ * * , ( ) ] link:\doibase 10.1103/revmodphys.85.693 [ * * , ( ) ] link:\doibase 10.1016/s0146 - 6410(03)90015 - 9 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.65.1199 [ * * , ( ) ] link:\doibase 10.1142/s0217751x16300295 [ * * , ( ) ] , link:\doibase 10.1088/0954 - 3899/23/8/001 [ * * , ( ) ] link:\doibase 10.1007/bf01344395 [ * * , ( ) ] link:\doibase 10.1103/physrevd.38.1519 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevd.90.014044 [ * * , ( ) ] , link:\doibase 10.1016/j.physletb.2008.07.086 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.87.091501 [ * * , ( ) ] , link:\doibase 10.1142/s0217751x15502012 [ * * , ( ) ] link:\doibase 10.1016/0370 - 2693(86)91275-x [ * * , ( ) ] link:\doibase 10.1103/physrevd.93.055042 [ * * , ( ) ] , @noop ( ) , link:\doibase 10.1016/0370 - 1573(83)90114-x [ * * , ( ) ] link:\doibase 10.1140/epjad / s2004 - 03 - 010 - 4 [ * * , ( ) ] , link:\doibase 10.1088/1674 - 1137/40/4/042001 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.19.2105 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.48.14 [ * * , ( ) ] link:\doibase 10.1016/j.physletb.2015.08.049 [ * * , ( ) ] , link:\doibase 10.1016/j.physletb.2008.07.018 [ * * , ( ) ]

in field theory , currents which are conserved classically may not be quantum mechanically @xcite . for example , in massless @xmath2 the conservation of the axial current is violated by the axial anomaly . the adler - bardeen theorem states that with the proper regularization scheme , the coefficient of the anomaly , as computed at one loop order , is exact to all orders in perturbation theory . moreover , the axial anomaly does not change if the fermions propagate in either a thermal bath or a fermi sea @xcite . in vacuum , one of the most striking manifestations of the axial anomaly is the decay of a neutral pion into two photons , as the amplitude is directly proportional to the coefficient of the axial anomaly in qed @xcite . a natural supposition is then that because the axial anomaly does not change with temperature , neither does the amplitude for @xmath3 @xcite . in this paper we show that the story is more involved . we compute with a gauged nonlinear sigma model @xcite which properly incorporates all anomalies by inclusion of the wess - zumino - witten ( wzw ) term @xcite . the effective lagrangian for @xmath3 is @xmath4 where @xmath5 is the pion decay constant , @xmath6 is the number of colors , _ etc . _ in sec . [ sec : vac ] we start by computing the effects of pion loops on the amplitude of ( [ eq:1.1 ] ) , using the wzw action to one loop order in vacuum @xcite . the form of the wzw action is constrained by topology @xcite , so after the dust of calculation settles , in vacuum the result is trivial : the only effect of the pion loops is to change a bare pion decay constant into a renormalized @xmath7 . we then extend the calculations to soft , cool pions at low temperature @xcite . in this paper we work exclusively with two flavors in the chiral limit . the restriction to two flavors is done for ease of calculation , and is otherwise inessential . the chiral limit , @xmath8 , is assumed because then the pion decay amplitude is directly related to the axial anomaly ; for calculations at @xmath9 at nonzero temperature , see @xcite . we believe that our results are relevant for @xmath9 ( as in vacuum ) , but more detailed analysis is required to establish this . at nonzero temperature , calculations in a background field formalism @xcite show that to @xmath10 , the zero temperature pion decay constant is replaced by a temperature dependent form @xcite , @xmath11 thus a second guess for the change of @xmath12 with temperature would be that the zero temperature @xmath7 is replaced by @xmath13 . since @xmath13 decreases to @xmath10 , if true the amplitude , @xmath14 , would increase to this order . in sec [ sec : lowt ] we evaluate precisely the same diagrams as at zero temperature to @xmath10 . the result , ( [ eq:3.9 ] ) , is the sum of two terms : one has exactly the form of ( [ eq:1.1 ] ) , with @xmath7 replaced by @xmath13 , but there is also a second term , special to nonzero temperature . this type of term was derived recently in nonlinear sigma models in the absence of gauge fields @xcite : it is nonlocal , analogous to the hard thermal loops of hot gauge theories @xcite . for @xmath3 , to order @xmath10 we find that the sum of these two terms is such that instead of increasing , like @xmath14 , the amplitude decreases , like @xmath15 , ( [ eq:3.8a ] ) . in sec . [ sec : ward ] we give a general analysis of the relationship between the chiral ward identities and the amplitude for @xmath3 . as is standard @xcite , we use the anomalous ward identity to relate a three point function of currents to the amplitude for @xmath3 . at zero temperature , this relationship is precise because of the sutherland - veltman theorem @xcite ( in a slight abuse of terminology ) . the proof of the sutherland - veltman theorem depends crucially upon lorentz invariance . a heat bath , however , provides a preferred rest frame ; extending an analysis of itoyama and mueller @xcite , we show that consequently , the sutherland - veltman theorem does not apply at any nonzero temperature . this is why @xmath3 changes with temperature , even though the anomaly does nt : besides the contributions from @xmath3 , because there is no sutherland - veltman theorem at nonzero temperature , there are other terms which enter to ensure that the adler - bardeen theorem is satisfied . in sec . [ sec : adlerbardeen ] we demonstrate these general arguments by computing the correlator between one axial and two vector currents in the nonlinear sigma model to one loop order , @xmath10 . at nonzero temperature , new tensor structures arise in this correlator ; these structures are why the sutherland - veltman theorem is inapplicable at @xmath16 . nevertheless , when all terms are added together , we find that the adler - bardeen theorem remains valid to one loop order . this is a useful and nontrivial check of our result for @xmath3 . technical details are relegated to several appendices . the wzw action is discussed in appendix [ awzw ] . various formulas for hard thermal loops are collected in appendix [ ahtl ] . we use the imaginary time formalism at nonzero temperature in this paper , but show in appendix [ sec : realtime ] how the same results follow in the real time formalism . lastly , in appendix [ gamma3pi ] we compute another anomalous amplitude , that for @xmath17 @xcite , at low temperature . while the principal concern of our work are thermal field theories , we hope that some of our dicussion , especially that in sec . [ sec : ward ] , might be of more general interest . perhaps understanding why @xmath3 is not tied to the axial anomaly at nonzero temperature helps us better understand this relation at zero temperature . we start by computing the effects of pion loops on the amplitude for @xmath18 in vacuum . at one loop order , the relevant diagrams are figs . ( 1.a)-(1.d ) : fig . ( 1.a ) gives the pion field renormalization constant , @xmath19 ; fig . ( 1.b ) , the renormalized pion decay constant @xmath7 , while corrections to the amplitude itself are given by figs . ( 1.c ) and ( 1.d ) . for the `` tadpole '' type diagrams of figs . ( 1.a ) , ( 1.b ) , and ( 1.c ) we use a trick . to compute fig . ( 1.a ) we expand the full lagrangian , ( [ eq : a.10 ] ) , to quartic order in the pion field , @xmath20 + \ldots \end{aligned}\ ] ] in all expressions from the appendix , we need to use the bare pion decay constant , @xmath21 , instead of the renormalized quantity @xmath7 , as one loop effects change @xmath21 into @xmath7 . for the quartic terms , contracting two out of the four pion fields in all possible ways gives @xmath22 where @xmath23 while this integral is quadratically divergent , we ignore regularization , since its actual value is irrelevant for our purposes . in ( [ eq:2.4 ] ) @xmath24 is the renormalized pion field , and so @xmath25 for the pion decay constant , instead of fig . ( 1.b ) we expand the axial current @xmath26 of ( [ eq : a.11 ] ) to cubic order in the pion field , @xmath27 contracting all pairs of pion fields , @xmath28 so that @xmath29 as is typical of nonlinear sigma models , unphysical , off - shell quantities such as @xmath19 , ( [ eq:2.5a ] ) , depend upon the parametrization of the coset space , while physical expressions , such as that for @xmath7 in ( [ eq:2.8 ] ) , do not . turning to the amplitude for @xmath3 , from ( [ eq:1.1 ] ) it is @xmath30 @xmath31 and @xmath32 , and @xmath33 and @xmath34 are the momenta and the polarization vectors of the two photons , both of which lie on the mass shell , @xmath35 . at tree level , the bare coupling @xmath36 satisfies @xmath37 the right hand side of ( [ eq:2.10 ] ) is precisely the coefficient of the axial anomaly in qed @xcite . to evaluate fig . ( 1.c ) we expand the anomalous current for the coupling to two photons , ( [ eq : a.7 ] ) , ( [ eq : a.11 ] ) , and ( [ eq : a.14 ] ) , to cubic order in the pion field , @xmath38 + \ldots \end{aligned}\ ] ] contracting two pion fields , @xmath39 the two distinct diagrams of fig . ( 1.d ) require more effort : @xmath40 where @xmath41 this peculiar separation of terms in @xmath42 is done in anticipation of the results at nonzero temperature . in vacuum , for @xmath43 , @xmath44 , and so because of the levi - civita symbol , @xmath45 does not contribute to @xmath46 . hence ( [ eq:2.13 ] ) reduces to a form proportional to the original term in ( [ eq:2.9 ] ) . putting everything together , the renormalized coupling @xmath47 equals @xmath48 the @xmath49 comes from a factor of @xmath50 for the renormalized pion field , fig . ( 1.a ) and ( [ eq:2.5a ] ) , the @xmath51 from fig . ( 1.c ) , ( [ eq:2.12 ] ) , and the @xmath52 from fig . ( 1.d ) , ( [ eq:2.13 ] ) and ( [ eq:2.14 ] ) . hence at one loop order , @xmath53 comparing ( [ eq:2.10 ] ) and ( [ eq:2.17 ] ) , we see that the anomaly is not renormalized to one loop order @xcite : seperate divergences in @xmath7 and @xmath47 cancel in the product @xcite . we now compute the decay for a cool pion , at a temperature @xmath54 @xcite . the diagrams are identical , the only difference is that we need to compute at @xmath55 . for the tadpole diagrams of figs . ( 1.a ) , ( 1.b ) , and ( 1.c ) , the integral is the analogy of ( [ eq:2.5 ] ) : @xmath56 we use the imaginary time formalism , @xmath57 , @xmath58 , and @xmath59 for integral @xmath60 ; after doing the sum over @xmath60 , the bose - einstein statistical distribution function , @xmath61 , appears : @xmath62 @xmath63 is the value of the integral at zero temperature . henceforth we drop the @xmath64 part of any integrals , assuming that they turn bare into renormalized quantities , such as @xmath21 into @xmath7 , throughout . the calculation of temperature dependent corrections to the pion decay constant proceeds as in the previous section . ignoring ultraviolet renormalization , in ( [ eq:2.8 ] ) we replace @xmath63 by @xmath65 , and @xmath21 by @xmath7 , to obtain @xmath66 which was quoted in the introduction , ( [ eq:1.2 ] ) . thus if anything unusual happens at nonzero temperature , it can only be from the diagram of fig . ( 1.d ) . unlike the tadpole diagrams , this diagram has nontrivial momentum dependence , and so we must be more precise in specifying the external momenta . to compute scattering in a thermal bath , we continue the euclidean momenta @xmath67 to a minkowski energy @xmath68 by @xmath69 . following @xcite we further assume that each momentum is not only cool but soft , taking both @xmath70 . for scattering between soft , cool pions , in @xcite we showed that the leading temperature corrections are directly analogous to the hard thermal loops of hot gauge theories @xcite . we used the background field method , but only in the absence of external gauge fields . while the perturbative calculations which follow are thus less elegant , they illustrate the physics more directly . from the perspective of @xcite , there is nothing special about @xmath3 ; the connection to the axial anomaly will be clarified later . we introduce @xmath71 , @xmath72 with @xmath73 the @xmath74 sign denotes that only the hard thermal loops in the integral are retained , which we denote by @xmath75 and @xmath76 . the hard thermal loops are the terms @xmath77 , and are given explicitly in appendix [ ahtl ] . up to an overall constant , @xmath78 is the same hard thermal loop as appears in the polarization tensor for a photon in thermal equilibrium . for a thermal photon , the screening of time dependent electric and magnetic fields implies that the mass shell is at @xmath79 . for the sake of simplicity we assume that the photons do not thermalize ; then the only photons which propagate are transverse modes on the light cone , @xmath80 . the polarization vectors for these modes are purely spatial vectors which satisfy @xmath81 . from ( [ eq:2.13 ] ) , the contribution to the amplitude from fig . ( 1.d ) , @xmath46 , involves @xmath82 for one of the two photons on their mass shell . using ( [ eq : b.5 ] ) of appendix [ ahtl ] , @xmath83 where @xmath84 is the polarization vector for the photon with momentum @xmath85 . only the first term on the right hand side of ( [ eq : b.5 ] ) contributes , as terms in @xmath86 which are @xmath87 or @xmath88 drop out after contraction with @xmath89 . from the definition of @xmath90 , ( [ eq:3.3 ] ) , @xmath91 consequently , while at zero temperature fig . ( 1.d ) contributes to the amplitude for @xmath3 , to leading order at nonzero temperature its contribution vanishes identically , ( [ eq:3.7 ] ) . knowing that fig . ( 1.d ) does nt contribute , it is then easy to read off the one loop corrections to the coupling @xmath47 to @xmath10 , @xmath92 . as in ( [ eq:3.2a ] ) , we start with ( [ eq:2.16 ] ) , and then replace @xmath63 by @xmath65 , and @xmath21 by @xmath7 . we keep the @xmath49 from fig . ( 1.a ) , the the @xmath51 from fig . ( 1.c ) , but replace the @xmath93 from fig . ( 1.d ) by @xmath94 , to obtain @xmath95 notice that while ( [ eq:2.8 ] ) is precisely analogous to ( [ eq:3.2a ] ) , because of the difference in fig . ( 1.d ) , ( [ eq:2.16 ] ) is not analogous to ( [ eq:3.8 ] ) there is a difference in sign . consequently , @xmath96 as discussed in the introduction , naively one might guess that ( [ eq:2.17 ] ) generalizes to nonzero temperature just by replacing @xmath7 and @xmath47 with @xmath13 and @xmath92 , respectively . this is wrong : to @xmath10 , instead of @xmath97 , as would be guessed from ( [ eq:2.17 ] ) , instead @xmath98 , ( [ eq:3.8a ] ) . we do not know why , to leading order about low temperature , @xmath92 decreases in exactly the same manner as @xmath13 . our result in ( [ eq:3.8 ] ) differs from that found by dobado , alvarez - estrada , and gomez @xcite . these authors consider the same model , but find @xmath99 to @xmath10 . our results agree except for fig . ( 1.d ) , which we believe was treated incorrectly @xcite . before continuing , following @xcite we construct the effective lagrangian for @xmath3 to @xmath10 . in @xmath100 of ( [ eq:3.3 ] ) , the term @xmath101 is easy to include . at zero temperature , ( [ eq:2.14 ] ) , this term is the only part of fig . ( 1.d ) which contributes , @xmath93 in ( [ eq:2.16 ] ) , and turns @xmath102 in @xmath47 into @xmath103 in @xmath104 . thus to @xmath10 , the effect of figs . ( 1.a ) , ( 1.c ) , and the term @xmath101 in @xmath100 of fig . ( 1.d ) is just to change @xmath103 into @xmath105 in the original lagrangian , @xmath12 of ( [ eq:1.1 ] ) . including the term @xmath76 is less trivial . because of ( [ eq : b.1 ] ) , it must be constructed out of transverse quantities . using ( [ eq : b.7 ] ) , we find that to @xmath10 , the effective lagrangian for @xmath3 is @xmath106 @xmath107 in this expression @xmath108 , @xmath109 and @xmath110 the vector @xmath111 and the integration over the angle @xmath112 arediscussed following ( [ eq : b.2 ] ) . the non - local term in ( [ eq:3.9 ] ) is specific to finite temperature . at zero temperature , there is no other term besides ( [ eq:1.1 ] ) that contributes to @xmath113 for photons on the mass - shell . in the terminology of the nonlinear sigma model @xcite , the operator of ( [ eq:1.1 ] ) is @xmath114 , while operators at next to leading order are @xmath115 . these operators , however , are either proportional to @xmath116 or @xmath117 , and so vanish on the photon(s ) mass shell , or @xmath118 , and so vanish in the chiral limit . thus in the vacuum , the only possible change in ( [ eq:1.1 ] ) is the transmutation of @xmath21 into @xmath7 , sec . [ sec : vac ] . at nonzero temperature , however , there are new nonlocal terms which arise , ( [ eq:3.9 ] ) . because they are nonlocal , these new terms are also @xmath114 , and so as important as ( [ eq:1.1])@xcite . this is the technical reason why the amplitude for @xmath119 depends nontrivially upon temperature . in the previous section we found that the amplitude for @xmath1 diminishes to leading order in an expansion about low temperature , ( [ eq:3.8a ] ) . the question we address in this section is why is this amplitude tied to the coefficient of the axial anomaly at zero temperature , ( [ eq:2.17 ] ) , but not at nonzero temperature ? we work in the chiral limit to leading order about zero temperature , @xmath10 , because then we can make certain technical assumptions which simplify the discussion . the general case is considered at the end of this section . define the vector current , @xmath120 , and the axial current in the isospin-@xmath121 direction , @xmath122 . the vector curent is conserved , @xmath123 while the axial current is anomalous , @xmath124 by the adler - bardeen theorem , the coefficient of the right hand side is exact to one loop order @xcite , and is independent of temperature and density @xcite . one quantity we can compute is the ( thermal ) three point green s function between two vector , and one axial vector , current : @xmath125 where @xmath126 is the hamiltonian . then @xmath127 satisfies current conservation , @xmath128 and the anomalous ward identity , @xmath129 @xmath130 . to relate the anomalous ward identity to the amplitude for pion decay we follow shore and veneziano @xcite . at low temperature the pion couples to the axial current as @xmath131 ( this is not valid to @xmath132 ; then the relation is more complicated , @xcite . ) to obtain the amplitude for @xmath133 , we introduce @xmath134 times the matrix element between two @xmath2 currents and a pion , @xmath135 this is related to the pion decay amplitude , ( [ eq:2.9 ] ) , as @xmath136 subtracting the one pion pole term from ( [ eq:4.1 ] ) gives @xmath137 , which by construction is one pion irreducible , @xmath138 again , by current conservation @xmath139 while the anomalous ward identity , ( [ eq:4.2 ] ) , becomes @xmath140 the trick is now to try to deduce general relations using just ( [ eq:4.9 ] ) , ( [ eq:4.8 ] ) , and bose symmetry between the two photons , @xmath141 . we first discuss zero temperature , where we can invoke euclidean invariance . the most general pseudo - tensor @xmath137 which satisfies all of our conditions involves three terms : @xmath142 current conservation , ( [ eq:4.9 ] ) , gives @xmath143 while from the anomalous ward identity , ( [ eq:4.8 ] ) , @xmath144 combining these two relations we obtain @xmath145 putting the photons on their mass shell @xmath146 , the left hand side in ( [ eq:4.13 ] ) reduces to @xmath147 . this is zero on the pion mass shell , @xmath148 , since by definition @xmath149 is one pion irreducible , and so can not have a pole @xmath150 . hence the left hand side of ( [ eq:4.13 ] ) vanishes , and we obtain the desired relation between @xmath47 and the coefficient of the axial anomaly , @xmath151 which is ( [ eq:2.14 ] ) . this analysis , and especially the tensor decomposition of ( [ eq:4.10 ] ) , is identical to the derivation of the sutherland - veltman theorem @xcite . historically , this theorem predated the anomaly , and was used to conclude that @xmath152 . by adding the axial anomaly through the anomalous ward identity of ( [ eq:4.8 ] ) , however , we obtain @xmath153 , ( [ eq:4.14 ] ) . from a modern perspective , then , it is precisely the sutherland - veltman theorem which relates the axial anomaly to the amplitude for @xmath3 , and tells us at zero temperature the left hand side of ( [ eq:4.14 ] ) vanishes . this is no longer true at nonzero temperature . following itoyama and mueller @xcite , we write the most general tensor decomposition of @xmath154 . in a thermal bath , however , euclidean symmetry is lost , and the rest frame of the thermal bath , which we take as @xmath155 , enters . some of the possible tensors include @xmath156 we have only included the terms in @xmath149 which contribute to the ward identities of ( [ eq:4.9 ] ) and ( [ eq:4.8 ] ) . current conservation gives @xmath157 while the anomalous ward identity fixes @xmath158 from which follows @xmath159 @xmath160 putting all fields on their mass shell , @xmath80 , @xmath148 , and assuming as before that @xmath161 has no pole @xmath150 , we obtain @xmath162 since the terms on the left hand side involve only @xmath163 and @xmath164 , there is no reason for them to vanish even if @xmath165 ; thus the direct connection between @xmath92 , @xmath13 , and the anomaly is lost . we stress that , as always , the adler - bardeen theorem remains valid , and gives ( [ eq:4.18a ] ) . it is only the sutherland - veltman theorem which no longer applies at nonzero temperature . the above analysis only applies to leading order in low temperature , @xmath10 . this is because beyond leading order , the pion mass shell is no longer at @xmath166 @xcite ; also , as photons thermalize , their mass shell moves off the light cone this is incorporated by using ( [ eq:4.18a ] ) instead of ( [ eq:4.18 ] ) . for instance , we can understand how the anomalous ward identity is satisfied in a chirally symmetric phase . from explicit calculation in a constituent quark model @xcite , @xmath3 vanishes once chiral symmetry is restored . this does not conflict with the anomalous ward identity since even if @xmath167 , there are other terms which can ensure that ( [ eq:4.18a ] ) is satisfied . for photons which do not thermalize , so @xmath80 , even assuming that at the chiral critical point that the pion mass shell is @xmath166 , the tensors @xmath168 and/or @xmath169 will in general be nonzero . our analysis agrees with the results of contreras and loewe @xcite , who computed the triangle diagram at @xmath55 with massive fermions . their result , ( 1.2 ) , is directly related to @xmath137 . there are two terms : the first is regular , temperature dependent , and gives the amplitude for @xmath119 , while the second is the anomaly , and is independent of temperature . at first sight it might appear peculiar that the sutherland - veltman theorem applies at zero temperature , but fails at any nonzero temperature . even at zero temperature , however , the sutherland - veltman theorem only applies in the chiral limit when both photons are on their mass shell . without all of these conditions , the left hand side of ( 4.13 ) does not vanish , and @xmath47 is not given by ( 4.14 ) . an example of this occurs when one ( or both ) of the photons are off the mass shell @xcite , even if @xmath148 . in particular , in the limit of large @xmath170 , it is known that @xmath171 @xcite . thus on the right hand side of ( 4.13 ) , we can neglect @xmath172 relative to the anomaly term , @xmath173 , and ( 4.13 ) tells us that a combination of @xmath174 and @xmath161 are given entirely by the anomaly , @xmath175 . in this section we calculate the correlator @xmath176 to @xmath10 in the nonlinear sigma model . this allows us to check explicitly the temperature dependence of @xmath13 and @xmath92 found previously in sec . [ sec : lowt ] . it is also illuminating to see exactly which amplitudes enter at one loop order in the nonlinear sigma model , and how they conspire to satisfy the adler - bardeen theorem . at tree level in the nonlinear sigma model , there is no direct coupling between two vector and one axial vector currents , so @xmath177 . thus the anomalous ward identity , ( [ eq:4.2 ] ) , is satisfied entirely by the one pion reducible term , ( [ eq:4.8 ] ) , with @xmath178 . this is illustrated in fig . contributions to @xmath179 are generated at one loop order . in order to compute these , it is necessary to compute corrections to the axial current . we have computed these in two ways . the most direct is to follow the original method of wess and zumino @xcite . besides the the photon field @xmath180 , which couples to the electromagnetic current @xmath120 , we also introduce an external field @xmath181 which couples to the axial current @xmath182 . one then differentiates the generating functional of wess and zumino with respect to the external fields @xmath180 and @xmath181 . at one loop order , the terms required are @xmath183 and @xmath184 ; @xmath185 are the charged pion fields . after lengthy calculation , we find @xmath186 @xmath187 to leading order in @xmath188 , this @xmath189 is invariant under an electromagnetic gauge transformation , @xmath190 the terms in @xmath189 can be viewed as corrections to the axial current . in appendix a we compute these corrections using the noether construction of the axial current . this is somewhat delicate , since it is necessary to start with a lagrangian density ( and not merely a lagrangian ) which is manifestly gauge invariant . the result , ( [ eq : a.19 ] ) , agrees with ( [ eq:5.1 ] ) . through the diagrams of fig . ( 3 ) , the couplings in @xmath191 generate the one pion irreducible terms in @xmath154 . we begin with the results at zero temperature . given the couplings in @xmath192 , the only tensor structure which arises from fig . ( 3 ) is @xmath193 . comparing with the tensors in ( [ eq:4.10 ] ) , then , at one loop order automatically @xmath194 . for @xmath195 , we find @xmath196 the contribution @xmath197 is from fig . ( 3.a ) , that @xmath198 from the two diagrams of fig . ( 3.b ) . because all of the @xmath199 s vanish , current conservation ( [ eq:4.12 ] ) and the anomaly equation ( [ eq:4.11 ] ) , are satisfied rather trivially . that the latter vanishes is equivalent to the sutherland - veltman theorem , ( [ eq:4.14 ] ) this explains our results in sec . [ sec : vac ] , where we found that at zero temperature , the renormalization of @xmath7 and @xmath47 exactly compensate each other . at nonzero temperature we can evaluate @xmath200 using the results of sec . [ sec : lowt ] . ( 3.a ) is a tadpole diagram , and so as at zero temperature , just contributes to @xmath195 . thus the only possibility for a new tensor structure is from the hard thermal loop in fig . for one ordering of momenta , fig . ( 3.b ) gives @xmath201 where @xmath202 is given in ( [ eq:3.3 ] ) . putting the photon momentum @xmath31 on its mass shell , we require the result for @xmath203 in ( [ eq : b.5 ] ) . we drop all terms in @xmath204 , ( [ eq : b.5 ] ) , since they vanish upon contraction with the photon polarization tensor . the term @xmath205 in ( [ eq : b.5 ] ) cancels against the same term in ( [ eq:3.3 ] ) ; thus at nonzero temperature , there is no contribution to @xmath195 from fig . ( 3.b ) , and @xmath206 comparing to ( [ eq:5.1a ] ) , again fig . ( 3.a ) gives the term @xmath207 , and fig . ( 3.b ) the term @xmath208 . that fig . ( 3.b ) does nt contribute is just like the same result for @xmath92 , fig . ( 1.d ) and ( [ eq:3.8 ] ) . ( 3.b ) will contribute , however , through the term @xmath209 in @xmath203 in ( [ eq : b.5 ] ) . comparing with the tensor decomposition of ( [ eq:4.15 ] ) , we find : @xmath210 since these are the only diagrams at one loop order , @xmath211 consequently , @xmath212 thus we can see that our results satisfy current conservation , ( [ eq:4.17 ] ) and the anomaly equation of ( [ eq:4.18 ] ) . last but not least , the latter only holds given @xmath13 and @xmath213 in ( [ eq:1.1 ] ) and ( [ eq:3.8 ] ) , and so provides a nontrivial check of these calculations . notice that while there is a factor of @xmath214 in @xmath169 , it is essentially kinematic in origin , as envisioned by itoyama and mueller @xcite . one factor of @xmath215 arises from the definition of @xmath169 , ( [ eq:4.15 ] ) , while the other can be seen to arise from a directional singularity , @xmath216 , in the integral of ( [ eq : b.5 ] ) . further , notice that there is a ( logarithmic ) collinear divergence in the integral of ( [ eq : b.5 ] ) . this singularity drops out of the full amplitude after contracting with the polarization tensor of the photon . in this paper we have concentrated exclusively on the anomalous decay of @xmath3 . we have done so because it is the most familiar anomalous decay , and because the connection to the axial anomaly is especially close . for the collisions of heavy ions at ultrarelativistic energies , though , this is a purely academic point , since any high temperature plasma flies apart long before a @xmath0 has a chance to decay electromagnetically . our basic point is much more general , however . while the axial anomaly for fermions is independent of temperature , anomalous decays of mesonic fields change with temperature . processes of obvious interest for hadronic systems include @xmath217 into two gluons and the decays of the @xmath68 meson @xcite . electroweak processes include couplings of the axion . whether the changes in these couplings are physically relevant can only be determined after detailed calculation ; what is certain is that they do change . in a cosmological context , the mechanisms of electroweak baryogenesis involve the effective coupling of a pseudoscalar field to the pontryagin density for the @xmath218 gauge field . in vacuum , this coupling is directly related to the anomalous divergence of the baryon current . it was argued in @xcite that this coupling must be suppressed above the electroweak phase transition ; presumably this can be understood from our analysis . lastly , we note that the t hooft anomaly matching conditions @xcite strongly constrain the appearance of massless bound states in confining theories . due to the failure of the sutherland - veltman theorem , sec . iv , these conditions are clearly much less restrictive at nonzero temperature . if nothing else , perhaps this gives us a greater appreciation of the wonder of the fermion axial anomaly . while every other anomalous decays changes in complicated and detailed ways , that alone remains inviolate , always . we thank w. marciano for useful conversations . this work is supported by a doe grant at brookhaven national laboratory , de - ac02 - 76ch00016 . in this appendix we review the wess - zumino - witten ( wzw ) model coupled to an external photon field . along with establishing notation , this also enables us to discuss a novel form of the wzw model , mentioned recently @xcite , and to comment about the two currents in the two - flavor version of the wzw model . for @xmath219 flavors , the model is constructed from a @xmath220 unitary matrix @xmath221 , @xmath222 . in the absence of gauge fields , the action is the sum of two terms , @xmath223 , with @xmath224 the usual action for a nonlinear sigma model , @xmath225 the generators of @xmath226 are the matrices @xmath227 , normalized as @xmath228 . this lagrangian is invariant under global @xmath229 unitary transformations , @xmath230 . for example , under axial transformations , for which @xmath231 , the corresponding conserved current is @xmath232 the second piece of the action is the wess - zumino - witten term , @xmath233 where the integral is over a five - dimensional region whose boundary is four - dimensional spacetime . we wish to couple @xmath221 to a photon field @xmath234 in a gauge invariant manner . for three flavors the charge matrix is @xmath235 , and so we introduce the covariant derivative , @xmath236 $ ] . by construction , both @xmath221 and @xmath237 transform covariantly under local @xmath238 gauge transformations , @xmath239 , @xmath240 , @xmath241 , @xmath242 . as usual , to make @xmath243 gauge invariant we simply replace the ordinary by the covariant derivative , @xmath244 . we can also do this for @xmath245 by replacing @xmath246 with the gauge covariant @xmath247 . using @xmath248 in @xmath245 gives something which is manifestly gauge invariant , but incomplete , since it can be shown that the ensuing equations of motion depend on the fifth dimension . in five dimensions , however , there are several other gauge invariant terms which can be added , involving powers of the ( abelian , gauge invariant ) field strength , @xmath249 . the correct action then follows uniquely by requiring that the equations of motion are independent of the fifth dimension@xcite , @xmath250 this expression is manifestly gauge invariant but not obviously independent of the fifth dimension @xcite . this is in contrast to the usual action , in which the terms which couple to the gauge field are manifestly four dimensional , but not evidently gauge invariant : @xmath251 , where @xmath252 where @xmath253 and @xmath254 the lagrangian [ eq : a.5 ] is the form given by witten @xcite : it is gauge invariant up to boundary terms . henceforth we follow brihaye , pak , and rossi @xcite and restrict ourselves to two flavors . introducing the pion field @xmath255 , @xmath256 then for @xmath257 @xmath258 @xmath259 . the original lagrangian @xmath243 becomes @xmath260 the axial current of ( [ eq : a.2 ] ) is @xmath261 in the wzw term , @xmath262 vanishes , while using the identity , @xmath263 we find that @xmath264 the charge @xmath265 ; in going from ( [ eq : a.6 ] ) to ( [ eq : a.13 ] ) , it turns out that the contribution from the piece @xmath266 drops out . this demonstrates that @xmath267 is directly proportional to the baryon current @xcite . further , we find that @xmath268 is proportional to the axial current in the isospin-@xmath121 direction , ( [ eq : a.11 ] ) , @xmath269 that for two flavors @xmath267 and @xmath268 are proportional to the baryon and axial isospin currents does not seem to have been recognized previously . notice that @xmath268 is only proportional to the axial current for the original lagrangian , @xmath243 ; as is discussed in sec [ sec : adlerbardeen ] , the complete axial current includes contributions from the wzw term . we do not know if @xmath267 and @xmath268 are equal to the analogous currents for three or more flavors . to conclude , we discuss how to compute the axial current @xmath270 of sec . [ sec : adlerbardeen ] using the noether construction . the general form for the gauged axial current is rather involved . since we compute perturbatively , however , we just compute the current in the same way . under an infinitesimal axial rotation , @xmath271 , the pion field transforms nonlinearly , @xmath272 to @xmath273 and @xmath274 . after performing such a transformation , the requisite current is then the coefficient of @xmath275 . in the present case , however , while the lagrangian is ( of course ) gauge invariant , the lagrangian density need not be . in particular , the lagragian density of ( [ eq : a.5 ] ) is not gauge invariant ; under a local gauge transformation , it transforms by a total derivative . to the order at which we compute , we add the following term to the lagrangian density : @xmath276 @xmath277 for example , the first piece contributes a chern - simons term to the current . with the addition of ( [ eq : a.20 ] ) , the lagrangian density is manifestly gauge invariant , and the axial current is @xmath278 when @xmath279 , @xmath182 reduces to the axial vector current in the absence of electromagnetism , @xmath280 , as can be verified by expanding ( [ eq : a.11 ] ) . the terms linear and quadratic in @xmath180 agree with the calculation from the wess - zumino consistency condition , ( [ eq:5.1 ] ) . this ambiguity in the construction of the axial current is familiar . instead of the gauge invariant , anomalous current used in sec . [ sec : ward ] , by subtracting off a chern - simons term , we can choose to work instead with a current which is conserved but not gauge invariant . we collect some results of hard thermal loops , with minor differences in notation from previous work @xcite . the simplest hard thermal loop is the integral of ( [ eq:3.1 ] ) , @xmath281 . after that , there is @xmath86 of ( [ eq:3.3 ] ) and ( [ eq:3.4 ] ) . by definition the hard thermal loop includes only only the terms @xmath77 in the integral , in the limit of soft external momentum @xmath282 . in this approximation , @xmath45 is tranverse : @xmath283 one way of writing @xmath284 is @xmath285 where @xmath286 and @xmath111 ; @xmath112 is a three vector of unit norm , @xmath287 , so that @xmath288 is null , @xmath289 , and @xmath290 . it is a dummy variable , in that one integrates over all directions of @xmath112 , as @xmath291 . in component form , @xmath292 @xmath293 @xmath294 @xmath295 where @xmath296 , and the @xmath297 are legendre functions of the second kind , @xmath298 for @xmath3 , we need the value of @xmath86 near the light cone , @xmath299 , @xmath300 @xmath301 where @xmath302 . an equivalent but useful expression for ( [ eq : b.2 ] ) is @xmath303 \ ; ; \end{aligned}\ ] ] from which follows @xmath304 @xmath305 the abelian field strengths of the vector fields @xmath306 and @xmath307 enter because @xmath78 is transverse , ( [ eq : b.1 ] ) . we found in sec . [ sec : lowt ] that the contribution from fig . ( 1.d ) vanishes to order @xmath309 . in this appendix we check this result using the real time approach . in this instance , the real time method is simple and not problematic . it also shows that the diagram has no temperature dependent terms whatsoever : the entire diagram is equal , identically , to its value at zero temperature . after dropping irrelevant terms proportional to @xmath310 , fig . ( 1.d ) is proportional to the integral @xmath311 @xmath312 , @xmath313 . the physical amplitude is obtained by contraction with the polarization vectors for the two photons . if @xmath314 lies in the @xmath315-direction , say , then @xmath316 must lie along the @xmath317- or @xmath318-directions . the angular @xmath319 integration will then vanish unless @xmath320 . thus the integral reduces to @xmath321 @xmath322 this integral vanishes after integration over @xmath323 ; note that this physical amplitude is free of any collinear divergences . this confirms our results obtained with the imaginary time formalism . in this section we compute the one loop corrections to the amplitude @xmath325 for soft , cool pions . this provides another , less trivial , example of hard thermal loops . corrections to the five dimensional wess - zumino - witten term , ( [ eq : a.3 ] ) , have been computed in @xcite using a background field method ; it would be interesting using this method to compute the one loop corrections to the gauged wzw model , ( [ eq : a.5 ] ) . at tree level , the amplitude for @xmath327 scattering is @xmath328 where @xmath329 is the momentum of the pion with isospin @xmath330 , _ etc _ , and @xmath331 from ( [ eq : a.10 ] ) , the amplitude for the scattering between four pions is @xmath332 \ ; , \nonumber\end{aligned}\ ] ] where @xmath333 , _ etc_. in the vacuum , @xmath336 , up to terms @xmath337 which drop out of the amplitude . to @xmath114 , @xmath338 the @xmath339 comes from @xmath340 , fig . ( 1.a ) and ( [ eq:2.5a ] ) , the @xmath341 from fig . ( 4.a ) and ( [ eq : d.1 ] ) , the @xmath342 from fig . ( 4.b ) and ( [ eq : d.5 ] ) . using ( [ eq:2.8 ] ) , then , to one loop order in vacuum ( [ eq : d.3 ] ) renormalizes with no change in form , @xmath343 analogous to ( [ eq:2.10 ] ) and ( [ eq:2.17 ] ) . at low temperature , @xmath344 is replaced by @xmath345 of ( [ eq:3.3 ] ) and ( [ eq:3.4 ] ) . using ( [ eq : b.7 ] ) , we find that the effective lagrangian for @xmath346 is similar to that for @xmath3 , ( [ eq:3.9 ] ) . one term is as at zero temperature , with @xmath7 replaced by @xmath13 , while the second is a hard thermal loop : @xmath347 @xmath348 where @xmath349 s. l. adler , phys . rev . * 177 * , 2426 ( 1969 ) ; j. s. bell and r. jackiw , nuovo cim . * 60 * , 147 ( 1969 ) ; s.l . adler and w.a . bardeen , phys . rev . * 182 * , 1517 ( 1969 ) ; b. zumino , y. s. wu , and a. zee , nucl . phys . * b239 * , 477 ( 1984 ) ; m. a. shifman , phys . rept . * 209 * , 341 ( 1991 ) . l. dolan and r. jackiw , phys . rev . * d9 * , 3320 ( 1974 ) ; m. reuter and w. dittrich , phys . rev . * d32 * , 513 ( 1985 ) ; f. ruiz ruiz and r. alvarez - estrada , phys . lett . * b180 * , 153 ( 1986 ) ; a. das and a. karev , phys . rev . * d36 * , 623 ( 1987 ) ; r. baier and e. pilon , zeit . * c52 * , 339 ( 1991 ) . h. itoyama and a. h. mueller , nucl . phys . * b218 * , 349 ( 1983 ) . l . liu and g ni , phys . rev . * d38 * , 3840 ( 1988 ) ; x .- g . he and g. c. joshi , _ ibid . _ * 41 * , 3796 ( 1990 ) ; r .- t . wang , jour . of phys . * a23 * , 5555 ( 1990 ) ; a. a. rawlinson , d. jackson , and r. j. crewther , zeit . * c56 * , 679 ( 1992 ) ; a. v. smilga , phys . rev . * d45 * , 1378 ( 1992 ) ; a. gomez nicola and r. f. alvarez - estrada , intl . jour . of mod a9 * , 1423 ( 1994 ) . r. jackiw , in _ current algebra and anomalies _ , ( princeton univ . press , princeton , 1985 ) ; c. itzykson and j .- b . zuber , _ quantum field theory _ , ( mc - graw - hill , new york , 1980 ) g. m. shore and g. veneziano , nucl * b381 * , 3 ( 1992 ) . a. dobado , r. f. alvarez - estrada , and a. gomez nicola , phys . lett . * b319 * , 238 ( 1993 ) . a. dobado , r. f. alvarez - estrada , and a. gomez nicola , phys . lett . * b324 * , 345 ( 1994 ) . s. weinberg , physica * 96a * , 327 ( 1979 ) ; j. gasser and h. leutwyler , ann . of phys . * 158 * , 142 ( 1984 ) ; j. f. donoghue , e. golowich , and b. r. holstein , _ dynamics of the standard model _ ( cambridge univ . press , ny , 1992 ) ; g. ecker , prog . part . . phys . * 35 * , 1 ( 1995 ) . j. wess and b. zumino , phys . lett * b37 * , 95 ( 1971 ) ; e. witten , nucl . phys . * b223 * , 422 ( 1983 ) o. kaymakcalan and j. schechter , phys . rev . * d31 * , 1109 ( 1984 ) ; y. brihaye , n. pak , and p. rossi , nucl . phys . * b254 * , 71 ( 1985 ) ; u. g. meissner , phys . * 161 * , 213 ( 1988 ) . c. m. hull and b. spence , mod . . lett . * a6 * , 969 ( 1991 ) . m. creutz and m. tytgat , phys . lett . * 76 * , 4671 ( 1996 ) ; m. creutz and m. tytgat , proceedings of lattice 96 , bnl preprint bnl-63465 , hep - lat/9608007 b. r. holstein and j. f. donoghue and y. c. r. lin , phys . lett . * 55 * , 2766 ( 1985 ) ; j. f. donoghue and d. wyler , nucl . phys . * b316 * , 289 ( 1989 ) ; j. bijnens , a. bramon and f. cornet z. phys . * c46 * , 599 ( 1990 ) ; r. akhoury and a. alfakih , ann . phys . * 210 * , 81 ( 1991 ) ; j. bijnens , nucl . phys . * b367 * , 709 ( 1991 ) ; int . j. mod * a8 * , 3045 ( 1993 ) , and references therein . r. d. pisarski and m. tytgat , bnl preprint bnl - pt-961 , hep - ph/9611206 . c. contreras and m. loewe , * c40 * , 253 ( 1988 ) ; t. hashimoto , k. hirose , t. kanki , and o. miyamura , phys . rev . * d37 * , 3331 ( 1988 ) ; b. pin - zhen and j. rafelski , mod . a7 * , 2493 ( 1992 ) ; ( e ) _ ibid . _ * a7 * , 3439 ( 1992 ) ; a. gomez nicola and r. f. alvarez - estrada , zeit . * c60 * , 711 ( 1993 ) ; s. p. klevansky , nucl . * a575 * , 605 ( 1994 ) ; d. blaschke , m. jaminon , yu . l. kalinovsky , p. petrow , s. schmidt , and b. van den bossche , _ ibid . _ * a2415 * , 1 ( 1995 ) . p. binetruy and m. k. gaillard , phys * d32 * , 931 ( 1985 ) ; j. gasser and h. leutwyler , phys . lett . * b184 * , 83 ( 1987 ) ; m. dey , v. l. eletsky , and b. l. ioffe , _ ibid . _ * b252 * , 620 ( 1990 ) ; a. bochkarev and j. kapusta , phys . rev . * d54 * , 4066 ( 1996 ) . r. d. pisarski and m. tytgat , phys . rev . * d54 * , 2989 ( 1996 ) . r. d. pisarski , phys . * 63 * , 1129 ( 1989 ) ; e. braaten and r. d. pisarski , nucl . phys . * b337 * , 569 ( 1990 ) ; _ ibid . _ * b339 * , 310 ( 1990 ) ; phys . rev . * d45 * , r1827 ( 1992 ) ; j. frenkel and j. c. taylor , nucl . phys . * b334 * , 199 ( 1990 ) ; _ ibid . _ * b374 * , 156 ( 1992 ) ; j. c. taylor and s. m. h. wong , _ ibid . _ * b346 * , 115 ( 1990 ) ; j. p. blaizot and e. iancu , _ ibid . _ * b390 * , 589 ( 1993 ) ; _ ibid . _ * b417 * , 608 ( 1994 ) ; _ ibid . _ * b434 * , 662 ( 1995 ) ; r. efraty and v. p. nair , phys . rev . * d47 * , 5601 ( 1993 ) ; r. jackiw and v. p. nair , _ ibid . _ * d48 * , 4991 ( 1993 ) ; r. jackiw , q. liu , and c. lucchesi , _ ibid . _ * d49 * , 6787 ( 1994 ) ; p. f. kelly , q. liu , c. lecchesi , and c. manuel , _ ibid . _ * d50 * , 4209 ( 1994 ) . d. g. sutherland , nucl . b2 * , 433 ( 1967 ) ; m. veltman , proc . a301 * , 107 ( 1967 ) . pisarski , in _ from thermal field theory to neural networks : a day to remember tanguy altherr _ , edited by p. aurenche , p. sorba , and g. veneziano . ( world scientific publishing , singapore , 1996 ) ; phys . rev . lett . * 76 * , 3084 ( 1996 ) ; r. baier , m. dirks , and o. kober , phys . rev . * d54 * , 2222 ( 1996 ) . l. ametller , j. bijnens , a. bramon and f. cornet , phys . * b303 * , 140 ( 1993 ) ; j. bijnens , a. bramon and f. cornet , z. phys . * c46 * , 599 ( 1990 ) . @xmath350 , ( 2.13 ) , equals @xmath351 , ( 5 ) of ref . the terms @xmath77 in @xmath350 define @xmath75 , ( 3.4 ) . from fig . ( 1.b ) of ref . @xcite , which is our fig . ( 1.d ) , in ( 12 ) they obtain @xmath352 this does not agree with ( [ eq:3.7 ] ) , for which the coefficient of the term @xmath353 vanishes . s. j. brodsky and g. p. lepage , phys . lett . * 87b * , 359 ( 1979 ) ; phys . rev . * d22 * , 2157 ( 1980 ) ; _ ibid . _ , * d24 * , 1808 ( 1981 ) ; l. arnellos , w. j. marciano , and z. parsa , nucl . * b 196 * , 371 ( 1982 ) ; a. anselm , a. johansen , e. leader , and l. lukaszuk , hep - ph/9603444 . m. dine and s. thomas , phys . lett . * b328 * , 73 ( 1994 ) . g. t hooft , in _ recent developments in gauge theories _ , ed . g. t hooft _ et al . _ ( plenum press , n.y . , 1980 ) . this method also allows one to gauge an arbitary subgroup @xmath126 of @xmath354 , whether or not @xmath126 is anomaly free . if the @xmath126 anomalies do not cancel , however , there is a doubling of the effective number of degrees of freedom , similar to the doubling which occurs for fermions on a lattice @xcite .

signals comprised of stream of pulses play a key role in many engineering applications , such as ultrasound imaging and radar ( see , e.g. @xcite ) . in some applications , the signal under examination is known to be real and non - negative . for instance , in single - molecule microscopy we measure the convolution of positive point sources with the microscope s point spread function @xcite . another example arises from the problem of estimating the orientations of the white matter fibers in the brain using diffusion weighted magnetic resonance imaging ( mri ) . in this application , the measured data is modeled as the convolution of a sparse positive signal on the sphere , which represents the unknown orientations , with a known point spread function that acts as a low - pass filter @xcite . this paper focuses its attention on the model of _ positive stream of pulses_. in this model , the measurements are comprised of a sum of unknown shifts of a kernel @xmath0 with positive coefficients , i.e. @xmath1=\sum_{m}c_{m}\mathbf{g}\left[k - k_{m}\right]+\tilde{\mathbf{n}}[k],\quad k\in\mathbb{z},\quad c_{m}>0,\ ] ] where @xmath2 is a bounded error term ( noise , model mismatch ) obeying @xmath3\right|\leq\delta$ ] . we do not assume any prior knowledge on the noise statistics . the pulse @xmath0 is assumed to be a sampled version of a scaled continuous kernel , namely , @xmath4:=\mathbf{g}\left(\frac{k}{\sigma n}\right)$ ] , where @xmath5 is the continuous kernel , @xmath6 is a scaling parameter and @xmath7 is the sampling interval . for instance , if @xmath0 is gaussian kernel , then @xmath8 denotes its standard deviation . the delayed versions of the kernel , @xmath9 $ ] , are often referred to as _ atoms_. we aim to estimate the set of delays @xmath10 and the positive amplitudes @xmath11 from the measured data @xmath12 $ ] . the sought parameters of the stream of pulses model can be defined by a signal of the form @xmath13:=\sum_{m}c_{m}\boldsymbol{\delta}\left[k - k_{m}\right],\quad c_{m}>0,\label{eq : x}\ ] ] where @xmath14 $ ] is the one - dimensional kronecker delta function @xmath15:= \begin{cases } 1 , & k=0,\\ 0 & k\neq 0 . \end{cases}\ ] ] in this manner , the problem can be thought of as a sparse deconvolution problem , namely , @xmath16=\left(\mathbf{g}\ast\left(\mathbf{x}+\mathbf{n}\right)\right)[k],\label{eq : model1}\ ] ] where @xmath17 denotes a discrete convolution and @xmath18 $ ] is the error term . the one - dimensional model can be extended to higher - dimensions . in this paper we also analyze in detail the model of two - dimensional positive stream of pulses given by @xmath19 & = & \left(\mathbf{g_{2}}\ast\left(\mathbf{x_{2}}+\mathbf{n}\right)\right)\left[\mathbf{k}\right],\label{eq : model2}\\ & = & \sum_{m}c_{m}\mathbf{g_{2}}\left[\mathbf{k}-\mathbf{k}_{m}\right]+\tilde{\mathbf{n}}[\mathbf{k}],\nonumber \end{aligned}\ ] ] where @xmath20\in\mathbb{z}^{2}$ ] and @xmath21 is a two - dimensional pulse . as in the one - dimensional case , the pulse is defined as a sampled version of a two - dimensional kernel @xmath22 by @xmath23=\mathbf{g_2}\left(\frac{k_{1}}{\sigma_{1}n_{1}},\frac{k_{2}}{\sigma_{2}n_{2}}\right)$ ] . the signal @xmath24:=\sum_{m}c_{m}\boldsymbol{\delta}\left[\mathbf{k}-\mathbf{k}_{m}\right],\quad c_{m}>0,\label{eq : x2}\ ] ] defines the underlying parameters to be estimated , where here @xmath25 denotes the two - dimensional kronecker delta function . for the sake of simplicity , we assume throughout the paper that @xmath26 . many algorithms have been suggested to recover @xmath27 from the stream of pulses @xmath28 . a naive approach would be to estimate @xmath27 via least - squares estimation . however , even if the convolution as in ( [ eq : model1 ] ) is invertible , the condition number of its associated convolution matrix tends to be extremely high . therefore , the recovery process is not robust ( see for instance section 4.3 in @xcite ) . suprisngly , the least - squares fails even in a noise - free environment due to amplification of numerical errors . we refer the readers to figure 1 in @xcite for a demonstration of this phenomenon . a different line of algorithms includes the well - known prony method , music , matrix pencil and esprit , see for instance @xcite . these algorithms concentrate on estimating the set of delays . once the set of delays is known , the coefficients can be easily estimated by least - squares . these methods rely on the observation that in fourier domain the stream of pulses model ( [ eq:1 ] ) reduces to a weighted sum of complex exponentials , under the assumption that the fourier transform of @xmath0 is non - vanishing . recent papers analyzed the performance and stability of these algorithms @xcite . however , as the fourier transform of the pulse @xmath0 tends to be localized and in general contains small values , the stability results do not hold directly for the stream of pulses model . furthermore , these methods do not exploit the positivity of the coefficients ( if it exists ) , which is the focus of this work . in recent years , many convex optimization techniques have been suggested and analyzed thoroughly for the task super - resolution . super - resolution is the problem of resolving signals from their noisy low - resolution measurements , see for instance @xcite . the main pillar of these works is the duality between robust super - resolution and the existence of an interpolating polynomial in the measurement space , called _ dual certificate_. similar techniques have been applied to super - resolve signals on the sphere @xcite ( see also @xcite ) and to the recovery of non - uniform splines from their projection onto the space of low - degree algebraic polynomials @xcite . the problem of recovering a general signal @xmath27 ( not necessarily non - negative ) robustly from stream of pulses was considered in @xcite . it was shown that the duality between robust recovery and the existence of an interpolating function holds in this case as well . particularly , it turns out that robust recovery is possible if there exists a function , comprised of shifts of the kernel @xmath0 and its derivatives , that satisfies several interpolation requirements ( see lemma [ lemma : q ] ) . in this case , the solution of a standard convex program achieves recovery error ( in @xmath29 norm ) of @xmath30 , for some constant @xmath31 that depends only on the convolution kernel @xmath0 . in @xcite it was proven that the support of the recovered signal is clustered around the support of the target signal @xmath27 . the behavior of the solution for large @xmath32 is analyzed in detail in @xcite . the main insight of @xcite is that the existence of such interpolating function relies on two interrelated pillars . first , the support of the signal , defined as @xmath33\text{\ensuremath{\neq}}0\right\ } $ ] , should satisfy a _ separation condition _ of the form @xmath34 for some kernel - dependent constant @xmath35 which does not depend on @xmath32 or @xmath8 . in the two - dimensional case , the separation condition gets the form . ] @xmath36 where @xmath37\in\mathbb{r}^2 $ ] and @xmath38 . the second pillar is that the kernel @xmath0 would be an _ admissible kernel_. an admissible kernel is a function that satisfies some mild localization properties . these properties are discussed in the next section ( see definition [ def : kernel ] ) . two prime examples for admissible kernels are the gaussian kernel @xmath39 and the cauchy kernel @xmath40 . in @xcite , the minimal separation constant @xmath41 which is required for the existence of the interpolating function ( and hence , for robust recovery ) was evaluated numerically to be 1.1 and 0.5 for the gaussian and cauchy kernels , respectively . inspired by the recent work on super - resolution of positive point sources @xcite , this work focuses on the model of positive stream of pulses . in contrast to @xcite , we prove that in this case the separation condition is no longer necessary to achieve stable recovery . we generalize and improve the results of @xcite as discussed in detail in section [ sec : main - result ] . particularly , we show that positive signals of the form ( [ eq : x ] ) can be recovered robustly from the measurements @xmath28 ( [ eq : model1 ] ) and the recovery error is proportional to the noise level @xmath42 . furthermore , the recovery error grows exponentially with the density of the signal s support . we characterize the density of the support using the notion of rayleigh - regularity , which is defined precisely in section [ sec : main - result ] . the recovery error also depends on the localization properties of the kernel @xmath0 . a similar result holds for the two - dimensional case . we use the following notation throughout the paper . we denote an index @xmath43 by brackets @xmath44 $ ] and a continuous variables @xmath45 by parenthesis @xmath46 . we use boldface small and capital letters for vectors and matrices , respectively . calligraphic letters , e.g. @xmath47 , are used for sets and @xmath48 for the cardinality of the set . the @xmath49 derivative of @xmath5 is denoted as @xmath50 . for vectors , we use the standard definition of @xmath51 norm as @xmath52\right|^{p}\right)^{1/p}$ ] for @xmath53 . we reserve @xmath7 to denote the sampling interval of ( [ eq:1 ] ) and define the support of the signal @xmath27 as @xmath54\text{\ensuremath{\neq}}0\right\ } $ ] . we write @xmath55 to denote some @xmath43 satisfying @xmath56 . the rest of the paper is organized as follows . section [ sec : main - result ] presents some basic definitions and states our main theoretical results . additionally , we give a detailed comparison with the literature . the results are proved in sections [ sec : proof_main_th ] and [ sec : proof_th2 ] . section [ sec : numerical - experiments ] shows numerical experiments , validating the theoretical results . section [ sec : conclusions ] concludes the work and aims to suggest potential extensions . in @xcite , it was shown that the underlying one - dimensional signal @xmath27 can be recovered robustly from a stream of pulses @xmath28 if its support satisfies a separation condition of the form ( [ eq:1dseparation ] ) . following @xcite , this work deals with non - negative signals and shows that in this case the separation condition is not necessary . specifically , we prove that the recovery error depends on the density of the signal s support . this density is defined and quantified by the notion of rayleigh - regularity . more precisely , a one - dimensional signal with rayleigh regularity @xmath57 has at most @xmath57 spikes within a resolution cell : [ def : rayleigh]we say that the set @xmath58 is rayleigh - regular with parameters @xmath59 and write @xmath60 if every interval @xmath61 of length @xmath62 contains no more that @xmath57 elements of @xmath63 : @xmath64 equipped with definition [ def : rayleigh ] , we define the sets of signals @xmath65 we further let @xmath66 be the set of signals in @xmath67 with non - negative values . if @xmath68 , then @xmath69 . if @xmath70 , then @xmath71 . besides the density of the signal s support , robust estimation of the delays and amplitudes also depends on the convolution kernel @xmath0 . particularly , the kernel should satisfy some mild localization properties . in short , the kernel and its first derivatives should decay sufficiently fast . we say that a kernel @xmath0 is _ non - negative admissible _ if it meets the following definition : [ def : kernel]we say that @xmath0 is a non - negative admissible kernel if @xmath72 for all @xmath45 , and : 1 . @xmath73 and is even . 2 there exist constants @xmath74 such that @xmath75 . there exist constants @xmath76 such that 1 . @xmath77 for all @xmath78 . @xmath79 for all @xmath80 . now , we are ready to state our one - dimensional theorem , which is proved in section [ sec : proof_main_th ] . the theorem states that in the noise free - case , @xmath81 , a convex program recovers the delays and amplitudes exactly , for any rayleigh regularity parameter @xmath57 . namely , the convolution system is invertible even without any sparsity prior . additionally , in the presence of noise or model mismatch , the recovery error grows exponentially with @xmath57 and is proportional to the noise level . [ th : main]consider the model ( [ eq : model1 ] ) for a non - negative admissible kernel @xmath0 as defined in definition [ def : kernel ] . then , there exists @xmath35 such that if @xmath82 and @xmath83 the solution @xmath84 of the convex problem @xmath85 satisfies @xmath86 where @xmath87 and @xmath88 for sufficiently large @xmath41 and @xmath32 , the recovery error can be written as @xmath89 where @xmath90 is a constant that depends only on the kernel @xmath0 and @xmath57 . in order to extend theorem [ th : main ] to the two - dimensional case , we present the equivalent of definitions [ def : rayleigh ] and [ def : kernel ] to two - dimensional signals . notice that the two - dimensional definition of rayleigh regularity is not a direct extension of definition [ def : rayleigh ] and is quite less intuitive . in order to prove theorems [ th : main ] and [ thm : main2 ] , we assume that the support of the signal could be presented as a union of @xmath57 non - intersecting subsets , which satisfy the separation conditions of ( [ eq:1dseparation ] ) and ( [ eq:2dseparation ] ) , respectively . in the one - dimensional case , this property is implied directly from definition [ def : rayleigh ] . however , this property is not guaranteed by the two - dimensional extension of definition [ def : rayleigh ] . see figure [ fig : counter_exp ] for a simple counter - example . therefore , in the two - dimensional case the rayleigh - regularity of a signal is defined as follows : [ def:2dray]@xcite we say that the set @xmath91 is rayleigh - regular with parameters @xmath59 and write @xmath92 if it can be presented as a union of @xmath57 subsets @xmath93 that are not intersecting and satisfy the minimum separation constraint ( [ eq:2dseparation ] ) . namely , * for all @xmath94 , @xmath95 , * for all @xmath96 , @xmath97 satisfies : for all square subsets @xmath98 of side length @xmath99 , @xmath100 . this figure presents an example for a set of 8 points that can not be decomposed into 4 non - intersecting subsets that meet the separation condition as in ( [ eq:2dseparation ] ) , although each resolution cell contains at most 4 points . consider a square resolution cell ( in @xmath101 sense ) of size @xmath102 . note that indeed each resolution cell contains at most 4 points . nonetheless , this set of points can not be described as 4 non - intersecting sets that satisfy the separation condition . specifically , the distance ( in @xmath103 norm ) of the black point is smaller than @xmath104 from the other 7 points , meaning it has to be in a separate subset from the others . on the other hand , there is no triplet of points that comprises a legal subset . therefore , the property of definition [ def:2dray ] is not a consequence of the two - dimensional version of definition [ def : rayleigh ] . ] a two - dimensional non - negative admissible kernel is defined as follows : [ def : kernel2d]we say that @xmath105 is a two - dimensional non - negative admissible kernel if @xmath106 for all @xmath107 and it has the following properties : 1 . @xmath108 and @xmath109 2 . : there exist constants @xmath110 such that @xmath111 , for @xmath112 , where @xmath113 . there exist constants @xmath114 such that 1 . @xmath115 for all @xmath116 satisfying @xmath117 , and @xmath118 for all @xmath116 satisfying @xmath119 . @xmath120 for all @xmath116 satisfying @xmath121 . equipped with the appropriate definitions of rayleigh regularity and non - negative admissible kernel , we are ready to state our main theorem for the two - dimensional case . the theorem is proved in section [ sec : proof_th2 ] . [ thm : main2 ] consider the model ( [ eq : model2 ] ) for a non - negative two - dimensional admissible kernel @xmath105 as defined in definition [ def : kernel2d ] . then , there exists @xmath35 such that if @xmath122 , the solution @xmath123 of the convex problem @xmath124 satisfies ( for sufficiently large @xmath32 and @xmath41 ) @xmath125 where @xmath126 is a constant which depends on the kernel @xmath21 and the rayleigh regularity @xmath57 . to conclude this section , we summarize the contribution of this paper and compare it to the relevant previous works . particularly , we stress the chief differences from @xcite which served as inspiration for this work . * this work deviates from @xcite in two important aspects . first , our stability results is much stronger than those in @xcite . particularly , our main results hold for signals with @xmath57 spikes within a resolution cell . in contrast , the main theorems of @xcite require signals with @xmath57 spikes within @xmath57 resolution cells . second , our formulation is not restricted to kernels with finite bandwidth and , in this manner , can be seen as a generalization of @xcite . this generalization is of particular interest as many kernels in practical applications are not band - limited . * in @xcite , it is proven that robust recovery from general stream of pulses ( not necessarily non - negative ) is possible if the delays are not clustered . here , we show that the separation is unnecessary in the positive case and can be replaced by the notion of raleigh regularity . this notion quantifies the density of the signal s support . * we derive strong stability guarantees compared to parametric methods , such as prony method , matrix pencil and music . nevertheless , we heavily rely on the positiveness of signal and the density of the delays , whereas the parametric methods do not have these restrictions . we also mention that several previous works suggested noise - free results for non - negative signals in similar settings , however they do not derive stability guarantees @xcite . in @xcite it was proven that the necessary separation between the delays drops to zero for sufficiently low noise level . the proof follows the outline of @xcite and borrows constructions from @xcite . let @xmath84 be s solution of ( [ eq : cvx ] ) and set @xmath127 . observe that by ( [ eq : cvx ] ) @xmath128 is finite since @xmath129 . the proof relies on some fundamental results from @xcite ( particularly , see proposition 3.3 and lemmas 3.4 and 3.5 ) which are summarized by the following lemma : [ lemma : q ] let @xmath0 be a non - negative admissible kernel as defined in definition [ def : kernel ] and suppose that @xmath130 . then , there exists a kernel - dependent separation constant @xmath35 ( see ( [ eq:1dseparation ] ) ) and a set of coefficients @xmath131 and @xmath132 such there exists an associated function of the form @xmath133 which satisfies : @xmath134 where @xmath135 and @xmath136 are the constants associated with @xmath0 . furthermore , @xmath137 the non - negativity property , @xmath138 for all @xmath45 , does not appear in @xcite , however , it is a direct corollary of the non - negativity assumption that @xmath72 for all @xmath45 . the interpolating function ( [ eq : q_form ] ) also satisfies the following property which will be needed in the proof : [ claim : qh ] let @xmath84 be s solution of ( [ eq : cvx ] ) and set @xmath127 . let @xmath139 be a union of @xmath57 non - intersecting sets obeying @xmath140 for all @xmath141 . for each set @xmath142 , let @xmath143:=\tilde{\mathbf{q}}_{i}\left(k / n\right)$ ] , @xmath43 , be an associated function , where @xmath144 is given in ( [ eq : q_form ] ) . then , for any sequence @xmath145 we have @xmath146\right)^{\alpha_{i}}\mathbf{h}[k]\\ & \leq\left(c_{0}\left(1+\frac{\pi^{2}}{6\nu^{2}}\right)\right)^{r-1}\left(\frac{6\nu^{2}}{3\mathbf{g}\left(0\right)\nu^{2}-2\pi^{2}c_{0}}\right)^{r}\delta \\ & + c^{*}\nu^{-4}\left\vert \mathbf{h}\right\vert _ { 1 } , \end{split}\ ] ] for some constant @xmath147 that depends on the kernel @xmath0 . we begin by two preliminary calculations . first , we observe from ( [ eq : model1 ] ) and ( [ eq : cvx ] ) that @xmath148\mathbf{h}[k]\right| & \leq & \left\vert \mathbf{y}-\mathbf{g}\ast\hat{\mathbf{x}}\right\vert _ { 1 } \nonumber \\ & + & \left\vert \mathbf{g}\ast \mathbf{x}-\mathbf{y}\right\vert _ { 1 } \nonumber\\ & \leq & 2\delta.\label{eq : delta}\end{aligned}\ ] ] additionally , we can estimate for all @xmath43 ( see section 3.4 in @xcite ) @xmath149 and hence with the properties of admissible kernel as defined in definition [ def : kernel ] we have for @xmath150 @xmath151\right| & \leq & c_{\ell}\sum_{k_{m}\in\mathcal{t}_{i}}\frac{1}{1+\left(\frac{k - k_{m}}{n\sigma}\right)^2}\nonumber \\ & \leq & 2c_{\ell}\left(1+\frac{\pi^{2}}{6\nu^{2}}\right).\label{eq : bound_g1h}\end{aligned}\ ] ] according to ( [ eq : q_form ] ) , the left - hand of ( [ eq : lemma3.4 ] ) can be explicitly written as : @xmath152\right)^{\alpha_{i}}\mathbf{h}[k ] \\ & = \sum_{k\in\mathbb{z}}\prod_{i=1}^{r}\left(\sum_{k_{m_{i}}\in\mathcal{t}_{i}}a_{m_{i}}\mathbf{g}\left[k - k_{m_{i}}\right]+b_{m_{i}}\mathbf{g}^{(1)}\left[k - k_{m_{i}}\right]\right)^{\alpha_{i}}\mathbf{h}[k].\label{eq : qh_explicit } \end{split}\ ] ] this expression can be decomposed into ( at most ) @xmath153 terms . we commence by considering the first term of the expression with @xmath154 and @xmath155 for @xmath156 ( namely , the product of the shifts of @xmath0 ) . using ( [ eq : delta ] ) and ( [ eq : bound_g1h ] ) we get @xmath157\sum_{k_{m_{2}}\in\mathcal{t}_{2}}a_{m_{2}}\mathbf{g}\left[k - k_{m_{2}}\right]\mathbf{h}[k ] \\ & \leq & \left\vert \mathbf{a}\right\vert _ { \infty}^{2}\left|\sum_{_{k_{m_{1}}\in\mathcal{t}_{1}}}\sum_{k\in\mathbb{z}}\mathbf{g}\left[k - k_{m_{1}}\right]\mathbf{h}[k]\sum_{k_{m_{2}}\in\mathcal{t}_{2}}\mathbf{g}\left[k - k_{m_{2}}\right]\right|\\ & \leq & 2\left\vert \mathbf{a}\right\vert _ { \infty}^{2}c_{0}\left(1+\frac{\pi^{2}}{6\nu^{2}}\right)\sum_{n\in\mathbb{z}}\left|\sum_{k\in\mathbb{z}}\mathbf{g}\left[k - n\right]\mathbf{h}[k]\right|\\ & \leq & 4\left\vert \mathbf{a}\right\vert _ { \infty}^{2}c_{0}\left(1+\frac{\pi^{2}}{6\nu^{2}}\right)\delta.\end{aligned}\ ] ] from the same methodology and using ( [ eq : a ] ) , we conclude that for any sequence of coefficients @xmath145 we get @xmath158\right)\mathbf{h}[k ] \\ & \leq\left(c_{0}\left(1+\frac{\pi^{2}}{6\nu^{2}}\right)\right)^{r-1}\left(\frac{6\nu^{2}}{3\mathbf{g}\left(0\right)\nu^{2}-2\pi^{2}c_{0}}\right)^{r}\delta . \end{split}\ ] ] next , using ( [ eq : bound_g1h ] ) we observe that all other @xmath159 terms of ( [ eq : qh_explicit ] ) can be bounded by @xmath160 for some constant @xmath161 and @xmath162 , @xmath163 . hence , we conclude by ( [ eq : a ] ) and ( [ eq : b ] ) that the summation of all these terms is bounded by @xmath164 for sufficiently large constants @xmath147 and @xmath41 . the constant @xmath147 depends only on the kernel @xmath0 . this completes the proof . consider @xmath165 and let us define the sets @xmath166<0\right\ } $ ] and respectively @xmath167\geq0\right\ } $ ] . throughout the proof , we use the notation @xmath168 and @xmath169 to denote some @xmath43 so that @xmath170 and @xmath171 , respectively . observe that by definition , @xmath172 and thus @xmath173 . the set @xmath174 can be presented as the union of @xmath57 non - intersecting subsets @xmath175 , where @xmath176 and @xmath177 . therefore , for each subset @xmath178 there exists an associated function @xmath143=\tilde{\mathbf{q}}_{i}\left(k / n\right)$ ] as given in lemma [ lemma : q ] . the proof builds upon the following construction @xmath179:=\prod_{i=1}^{r}\left(1-\tilde{\mathbf{q}}_{i}[k]\right)-\rho,\label{eq : q}\ ] ] for some constant @xmath180 , to be defined later . the function @xmath181 $ ] satisfies the following properties : [ lem : q_prod]let @xmath182 be as in ( [ eq : q ] ) , let @xmath183 and let @xmath184 where @xmath87 . then , we have @xmath185 & = & -\rho,\quad k_{m}\in\mathcal{n},\\ \mathbf{q}\left[k\right ] & \geq & \rho,\quad k\in\mathcal{n}^{c},\\ \mathbf{q}\left[k\right ] & \leq & 1,\quad k\in\mathbb{z}.\end{aligned}\ ] ] since @xmath186 by lemma [ lemma : q ] there exists for each subset @xmath178 an associated interpolating function @xmath187=\tilde{\mathbf{q}_{i}}(k / n)$ ] . consequently , for all @xmath188 we obtain @xmath189&=\prod_{i=1}^{r}\left(1-\tilde{\mathbf{q}}_{i}\left[k_{m}\right]\right)-\rho \\ & = -\rho , \end{split}\ ] ] and for all @xmath169 we have @xmath190 & = & \prod_{i=1}^{r}\left(1-\tilde{\mathbf{q}}_{i}\left[k\right]\right)-\rho\\ & \geq & \left(\frac{\beta}{4\mathbf{g}(0)\gamma^{2}}\right)^{r}-\rho.\end{aligned}\ ] ] by setting @xmath191\geq\frac{1}{2}\left(\frac{\beta}{4\mathbf{g}(0)\gamma^{2}}\right)^{r},\ ] ] we conclude the proof . note that in order to guarantees @xmath192 , we require @xmath183 . equipped with lemma [ lem : q_prod ] , we conclude that @xmath181 $ ] and @xmath193 $ ] have the same sign for all @xmath43 , and thus @xmath194\mathbf{h}[k]&=\sum_{k\in\mathbb{z}}\vert \mathbf{q}[k]\vert \vert \mathbf{h}[k]\vert \\ & \geq\rho\left\vert\mathbf { h}\right\vert _ { 1}.\label{eq : low - bound } \end{split}\ ] ] to complete the proof , we need to bound the inner product @xmath195 from above . to this end , observe that @xmath196\right)=1+\boldsymbol{\kappa}_{r}[k],\label{eq:1-q}\ ] ] where @xmath197 & : = & \sum_{j=1}^{2^{r}-1}\prod_{i=1}^{r}\left(-\tilde{\mathbf{q}}_{i}\left[k\right]\right)^{\alpha_{i}(j)},\label{eq : q}\end{aligned}\ ] ] for some coefficients @xmath198 . for instance , @xmath199=-\tilde{\mathbf{q}}_{1}\left[k\right]-\tilde{\mathbf{q}}_{2}\left[k\right]+\tilde{\mathbf{q}}_{1}\left[k\right]\tilde{\mathbf{q}}_{2}\left[k\right].$ ] therefore , by ( [ eq : q ] ) and ( [ eq:1-q ] ) we get @xmath200\right)-\rho,\mathbf{h}\right\rangle \label{eq : qh1}\\ & = & \left\langle \left(1-\rho\right)+\boldsymbol{\kappa}_{r},\mathbf{h}\right\rangle \nonumber \\ & = & \left(1-\rho\right)\sum_{k\in\mathbb{z}}\mathbf{h}[k]+\left\langle \boldsymbol{\kappa}_{r},\mathbf{h}\right\rangle .\nonumber \end{aligned}\ ] ] recall that by ( [ eq : cvx ] ) we have @xmath201 and therefore @xmath202+\mathbf{h}[k]\right|\\ & + \sum_{k\in\mathbb{z}\backslash\mbox{supp}(\mathbf{x})}\left|\mathbf{h}[k]\right| . \end{split}\ ] ] by definition @xmath193\geq0 $ ] for all @xmath169 and we use the triangle inequality to deduce @xmath203\\ & + & \sum_{k\in\mbox{supp}(\mathbf{x})\backslash\mathcal{n}}\left(\mathbf{x}[k]+\mathbf{h}[k]\right)+\sum_{k\in\mathcal{n}}\left|\mathbf{x}[k]+\mathbf{h}[k]\right|\\ & \geq & \left\vert \mathbf{x}\right\vert _ { 1}+\sum_{k\in\mathcal{n}^{c}}\mathbf{h}[k]-\sum_{k\in\mathcal{n}}\left|\mathbf{h}[k]\right|,\end{aligned}\ ] ] and thus we conclude @xmath204\leq0.\label{eq : nsp}\ ] ] so , from ( [ eq : q ] ) , ( [ eq : qh1 ] ) , ( [ eq : nsp ] ) and lemma [ claim : qh ] we conclude that @xmath205\right)^{\alpha_{i}(j)}\mathbf{h}[k]\right|\nonumber \\ & \leq & \left(2^{r}-1\right)\left(c_{0}\left(1+\frac{\pi^{2}}{6\nu^{2}}\right)\right)^{r-1}\nonumber\\ & \cdot & \left(\frac{6\nu^{2}}{3\mathbf{g}\left(0\right)\nu^{2}-2\pi^{2}c_{0}}\right)^{r}\delta \nonumber \\ & + & c^{*}\left(2^{r}-1\right)\nu^{-4}\left\vert \mathbf{h}\right\vert _ { 1}. \label{eq : upper - bound}\end{aligned}\ ] ] combining ( [ eq : upper - bound ] ) with ( [ eq : low - bound ] ) and ( [ eq : rho ] ) yields @xmath206 this completes the proof of theorem [ th : main ] . the proof of theorem [ thm : main2 ] follows the methodology of the proof in section [ sec : proof_main_th ] . we commence by stating the extension of lemma [ lemma : q ] to the two - dimensional case , based on results from @xcite : [ lemma : q2 ] let @xmath105 be a non - negative two - dimensional admissible kernel as defined in definition [ def : kernel ] and suppose that @xmath207 . then , there exists a kernel - dependent separation constant @xmath35 and a set of coefficients @xmath208 and @xmath209 such that there exist an associated function of the form @xmath210 which satisfies : @xmath211 for sufficiently small @xmath212 associated with the kernel @xmath105 , and some constants @xmath213 . for sufficiently large @xmath35 and constants @xmath214 , we also have @xmath215 we present now the two - dimensional version of lemma [ claim : qh ] without a proof . the proof relies on the same methodology as the one - dimensional case . [ lem:3.4_2]let @xmath216 be a union of @xmath57 non - intersecting sets obeying @xmath217 for all @xmath141 . for each set @xmath218 , let @xmath219:=\tilde{\mathbf{q}}_\mathbf{i,2}\left(\mathbf{k}/n\right)$ ] , @xmath220 , be an associated function , where @xmath221 is given in ( [ eq : q2_form ] ) . then , for any sequence @xmath145 we have for sufficiently large @xmath41 , @xmath222\right)^{\alpha_{i}}\mathbf{h}[{k}]\leq\tilde{c}_{2}(\mathbf{g},r)\delta+c^{*}\nu^{-6}\left\vert \mathbf{h}\right\vert _ { 1},\label{eq : lemma3.4_2}\ ] ] for some constants @xmath147 and @xmath223 which depends on the kernel @xmath21 and the regularity parameter @xmath57 . let @xmath220 . let us define the sets @xmath224<0\right\ } $ ] and @xmath225\geq0\right\ } $ ] . throughout the proof , we use the notation of @xmath226 and @xmath227 to denote all @xmath220 so that @xmath228 and @xmath229 , respectively . by definition , @xmath230 ( see definition [ def:2dray ] ) and it can be presented as the union of non - intersecting subsets @xmath231 where @xmath232 . therefore , for each subset @xmath233 there exists an associated function @xmath219=\tilde{\mathbf{q}}_\mathbf{i,2}\left(\mathbf{k}/n\right)$ ] given in lemma [ lemma : q2 ] . as in the one - dimensional case , the proof relies on the following construction @xmath234:=\prod_{i=1}^{r}\left(1-\tilde{\mathbf{q}}_\mathbf{i,2}[\mathbf{k}]\right)-\rho,\label{eq : q2}\ ] ] for some constant @xmath180 , to be defined later . this function satisfies the following interpolation properties : [ lem : q_prod2 ] suppose that @xmath235 where @xmath236 is given in lemma [ lemma : q2 ] . let @xmath237 be as in ( [ eq : q2 ] ) and let @xmath238 then , @xmath239 & = & -\rho,\quad\mathbf{k}_{m}\in\mathcal{n}_{2},\\ \mathbf{q_{2}}\left[\mathbf{k}\right ] & \geq & \rho,\quad \mathbf{k}\in\mathcal{n}_{2}^{c},\\ \mathbf{q_{2}}\left[\mathbf{k}\right ] & \leq & 1,\quad\mathbf{k}\in\mathbb{z}^{2}.\end{aligned}\ ] ] since @xmath240 by lemma [ lemma : q2 ] there exists for each subset @xmath233 an associated function @xmath241=\tilde{\mathbf{q}}_\mathbf{i,2}(\mathbf{k}/n)$ ] . consequently , for all @xmath242 we obtain @xmath243=\prod_{i=1}^{r}\left(1-\mathbf{q_{i,2}}\left[\mathbf{k}_{m}\right]\right)-\rho=-\rho.\ ] ] for @xmath244 we get for all @xmath227 @xmath245 & = & \prod_{i=1}^{r}\left(1-\mathbf{q_{i,2}}\left[\mathbf{k}\right]\right)-\rho\geq\left(\frac{c_{1}}{\left(n\sigma\right)^{2}}\right)^{r}-\rho.\end{aligned}\ ] ] by setting @xmath246\geq\frac{1}{2}\left(\frac{c_{1}}{\left(n\sigma\right)^{2}}\right)^{r},\ ] ] we conclude the proof . the condition @xmath247 guarantees that @xmath192 . once we constructed the function @xmath248 $ ] , the proof follows the one - dimensional case . by considering lemmas [ eq : lemma3.4_2 ] and [ lem : q_prod2 ] and using similar arguments to ( [ eq : low - bound ] ) and ( [ eq : upper - bound ] ) , we conclude @xmath249 using ( [ eq : rho2 ] ) we get for sufficiently large @xmath41 that @xmath250 for some constant @xmath251 which depends on the kernel @xmath252 and the rayleigh regularity @xmath57 . we conducted numerical experiments to validate the theoretical results of this paper . the simulated signals were generated in two steps . first , random locations were sequentially added to the signals support in the interval @xmath253 $ ] with discretization step of 0.01 , while keeping a fixed regularity condition . once the support was determined , the amplitudes were drawn randomly from an i.i.d normal distribution with standard deviation sd = 10 . for positive signals , the amplitudes are taken to be the absolute values of the normal variables . the experiments were conducted with the cauchy kernel @xmath254 , @xmath255 . we set the separation constant to be @xmath256 , which was evaluated in @xcite to be the minimal separation constant , guaranteeing the existence of interpolating polynomial as in lemma [ lemma : q ] . figure [ fig : example ] presents an example for the estimation of the signal ( [ eq : x ] ) from ( [ eq : model1 ] ) with @xmath257 . as can be seen , the solution of the convex problem ( [ eq : cvx ] ) detects the support of the signal with high precision in a noisy environment of @xmath258 db . figure [ fig:2d ] presents an example for recovery of a two - dimensional signal from a stream of cauchy kernels with @xmath257 and @xmath259 . example for the recovery of a signal of the form of ( [ eq : x ] ) from stream of cauchy kernels with @xmath255 , rayleigh regularity of @xmath257 , separation constant of @xmath256 and noise level of @xmath260 ( snr=27db ) . ] an example for the recovery of a two - dimensional signal of the form ( [ eq : x2 ] ) from the measurements ( [ eq : model2 ] ) , with @xmath257 , @xmath261 and @xmath259 . the figure presents merely the locations ( support ) of the original and the recovered signals . ] figure [ fig : mean_err ] shows the localization error as a function of the noise level @xmath42 . to clarify , by localization error we mean the distance between the support of the original signal and the support of the recovered signal . figure [ fig : positive - negative ] compares the localization error for positive signals and general real signals ( i.e. not necessarily positive ) from stream of cauchy pulses . for general signals , we solved a standard @xmath29 minimization problem as in @xcite , which is the same problem as ( [ eq : cvx ] ) without the positivity constraint @xmath262 . plainly , the localization error of positive signals is significantly smaller than the error of general signals . figure [ fig : err_r ] shows that the error grows approximately linearly with the noise level @xmath42 and increases with @xmath57 . in this paper , we have shown that a standard convex optimization program can robustly recover the sets of delays and positive amplitudes from a stream of pulses . the recovery error is proportional to the noise level and grows exponentially with the density of signal s support , which is defined by the notion of rayleigh regularity . the error also depends on the localization properties of the kernel . in contrast to general stream of pulses model as discussed in @xcite , no separation is needed and the signal s support may be clustered . it is of great interest to examine the theoretical results we have derived on real applications , such as detection and tracking tasks in single - molecule microscopy . we have shown explicitly that our technique holds true for one and two dimensional signals . we strongly believe that similar results hold for higher - dimension problems . our results rely on the existence of interpolating functions which were constructed in a previous work @xcite . extension of the results of @xcite to higher dimensions will imply immediately the extension of our results to higher dimensions as well . in @xcite , it was shown that for general signals that satisfy the separation condition ( [ eq:1dseparation ] ) , the solution of a convex program results in a localization error of order @xmath263 . namely , the support of the estimated signal is clustered around the support of the sought signal . it would be interesting to examine whether such a phenomenon exists in the positive case as well . the author is grateful to prof arie feuer and prof shai dekel for their comments and support and to veniamin morgenshtern for helpful discussions about @xcite . r. tur , y. c. eldar , and z. friedman , `` innovation rate sampling of pulse streams with application to ultrasound imaging , '' _ ieee transactions on signal processing _ , vol . 59 , no . 4 , pp . 18271842 , 2011 . s. deslauriers - gauthier and p. marziliano , `` spherical finite rate of innovation theory for the recovery of fiber orientations , '' in _ 2012 annual international conference of the ieee engineering in medicine and biology society _ , pp . 22942297 , ieee , 2012 . t. a. klar , s. jakobs , m. dyba , a. egner , and s. w. hell , `` fluorescence microscopy with diffraction resolution barrier broken by stimulated emission , '' _ proceedings of the national academy of sciences _ , vol . 97 , no . 15 , pp . 82068210 , 2000 . e. betzig , g. h. patterson , r. sougrat , o. w. lindwasser , s. olenych , j. s. bonifacino , m. w. davidson , j. lippincott - schwartz , and h. f. hess , `` imaging intracellular fluorescent proteins at nanometer resolution , '' _ science _ , vol . 313 , no . 5793 , pp . 16421645 , 2006 . i. bronstein , y. israel , e. kepten , s. mai , y. shav - tal , e. barkai , and y. garini , `` transient anomalous diffusion of telomeres in the nucleus of mammalian cells , '' _ physical review letters _ 103 , no . 1 , p. 018102 , 2009 . tournier , f. calamante , d. g. gadian , and a. connelly , `` direct estimation of the fiber orientation density function from diffusion - weighted mri data using spherical deconvolution , '' _ neuroimage _ , vol . 23 , no . 3 , pp . 11761185 , 2004 . y. hua and t. k. sarkar , `` matrix pencil method for estimating parameters of exponentially damped / undamped sinusoids in noise , '' _ ieee transactions on acoustics , speech , and signal processing _ 38 , no . 5 , pp . 814824 , 1990 . r. roy and t. kailath , `` esprit - estimation of signal parameters via rotational invariance techniques , '' _ ieee transactions on acoustics , speech , and signal processing _ , vol . 37 , no . 7 , pp . 984995 , 1989 . a. moitra , `` super - resolution , extremal functions and the condition number of vandermonde matrices , '' in _ proceedings of the forty - seventh annual acm on symposium on theory of computing _ , pp . 821830 , acm , 2015 . t. bendory , a. bar - zion , d. adam , s. dekel , and a. feuer , `` stable support recovery of stream of pulses with application to ultrasound imaging , '' _ ieee transactions on signal processing _ , vol . 64 , no . 14 , pp . 37503759 , 2016 . j .- j . fuchs , `` sparsity and uniqueness for some specific under - determined linear systems , '' in _ proceedings.(icassp05 ) . ieee international conference on acoustics , speech , and signal processing , 2005 . _ , vol . 5 , pp . v729 , ieee , 2005 . g. schiebinger , e. robeva , and b. recht , `` superresolution without separation , '' in _ computational advances in multi - sensor adaptive processing ( camsap ) , 2015 ieee 6th international workshop on _ , pp . 4548 , ieee , 2015 .

intensive studies of majorana fermion physics in solid state materials were motivated by possible applications of these states in topological quantum computation @xcite . in condensed matter , initially , majorana fermions were predicted to exist in spinless @xmath2-wave superconductors ( sc ) and fractional qhe @xcite . later , the realizations of majorana bound states ( mbs ) @xcite and 1d charge - neutral modes @xcite were proposed in hybrid structures where helical electronic system has a proximity effect with external @xmath3-wave pairing potential and zeeman exchange field . the helical systems with spin and momentum locking are realized in surfaces and edges of topological insulators ( ti ) , quantum spin - hall films @xcite and semiconductors with strong spin - orbital coupling@xcite . the transport signatures of mbs are associated with anomalous @xmath4-periodic josephson effect @xcite . on the other hand , charge - neutral chiral majorana modes ( @xmath5 mm ) which are , for instance , supported by sc - magnet junctions on 3d ti surface @xcite , are expected to show unconventional interferometry if involved in charge transfer between normal metal leads . various @xmath5mm - based devices operating as mach - zehnder @xcite , fabry - p ' erot @xcite , and hanbury - brown twiss @xcite quantum interferometers have been proposed . in context of quantum computation they can serve as a readout tool of qubit states encoded by mbs in vortex cores @xcite . in the paper we calculate dc josephson current - phase relation ( c@xmath6r ) of a long spinless contact between 1d gapless @xmath5mms . the system under consideration represents a chiral fabry - p ' erot interferometer implemented on a surface of 3d ti partially gapped by sc and magnetic ( m ) islands . similar ideas were explored in refs . . in these works normal dissipative transport of dirac fermions influenced by their splitting to majorana modes and the interference of the latter were studied . in contrast , we study the equilibrium phenomenon of josephson current carried by the andreev states formed in the chiral dirac liquid . the latter connects between two superconducting leads , which are effectively gapless due to chiral majorana modes at their edges . spinless structure of the junction follows from the absence of spin degeneracy of 2d helical surface states . presence of zeeman field breaks time - reversal symmetry and produces gapless chiral channels at the magnetic domain boundaries . in our system the normal conducting region consists of two separated chiral dirac channels supported by magnetic domain walls . we explore two models of ( i ) transparent and ( ii ) tunnel junctions . the first one involves four chiral y - junctions , operating as direct charged - to - neutral fermion converters . the second one has two tunnel contacts formed by thin layers of magnetic material where majorana and normal dirac channels are hybridized . the length of the n - region , formed by two counter propagating 1d dirac modes , is assumed to be larger than coherence length of the induced superconductivity . in other words , the thouless energy of the normal conducting part of the system under consideration is significantly smaller than the proximity gap . we take into account a contribution to the transport from sub - gap 1d chiral states only , neglecting 2d bands . since our superconducting leads are gapless due to the presence of @xmath5mms , the spectral current is continuous and consists of smeared andreev levels . assuming that superconducting leads are large and have a fixed chemical potential , we obtain the conventional @xmath7-periodicity of the c@xmath6r . in other words the non - equilibrium @xmath4-periodic josephson effect ( which is predicted @xcite for systems with zero - energy mbss ) turns out to be irrelevant here . the chiral structure of the normal part consisting of two 1d spinless dirac modes allows to make andreev pairs non - local . this means that the pair resides on two separated dirac channels and one can apply magnetic flux @xmath8 inducing aharonov - bohm phase @xmath9 , where @xmath10 . it follows that the critical current has a fractional @xmath11-periodicity rather than the conventional @xmath12 one . the realization of the normal region by two separated chiral channels in the present setup is one of the most relevant distinctions from the junctions with 2d topological insulator where andreev pairs exist on the same edge @xcite . the paper is organized as follows . in section [ 3dti ] we present main ideas of @xmath5mms realization in hybrid structures on a surface of 3d topological insulator . in section [ mm - n - mm ] we consider the regime of transparent josephson contacts . we derive the @xmath13-matrix of the contacts by means of matching the field operators in the chiral channels . next we calculate josephson c@xmath6r at arbitrary temperature and flux enclosed by dirac modes . the section [ mm - i - n - i - mm ] addresses the tunnelling regime , in which majorana and dirac modes meet at the tunnel contacts formed by magnetic constrictions . in section [ majorana - basis ] we discuss the majorana representation of charged - to - neutral fermion scattering in transparent and tunnel contacts . section [ conclusions ] summarizes the results of this paper . the technical details are provided in appendices [ s - t - app ] and [ appendix ] . as was proposed by fu and kane @xcite , the @xmath5 mm is supported by sc / magnet/3d - ti structure like that shown in figure [ chi - mm ] , with an effective low - energy 1d hamiltonian following from a solution of a 2d bogolyubov - de gennes equation . the 2d surface states are described by the rashba hamiltonian . the system is assumed to have a chemical potential @xmath14 , corresponding to the dirac point . the induced @xmath3-wave sc pairing potential is given by @xmath15 in the @xmath16 half plane , while at @xmath17 there is magnetic material inducing zeeman exchange field @xmath18 . the full hamiltonian reads @xmath19 where the field operator of the 2d surface is @xmath20^t$ ] , and @xmath21 and @xmath22 are pauli matrices in spin and nambu spaces , respectively . the field @xmath23 satisfies the charge conjugation constraint @xmath24 , while the eigenstates obey @xmath25 . the low - energy @xmath26 eigenvalues of the bogolyubov - de gennes equation @xmath27 are single degenerate and correspond to a 1d chiral mode with the linear spectrum @xmath28 single degeneracy of eigenvalues implies that bogolyubov quasiparticle operator @xmath29 represents a majorana mode obeying @xmath30 . the corresponding wave function of the @xmath5 mm reads @xmath31 the momentum @xmath2 here is directed along @xmath32-axis and @xmath33 is the transverse shape of 1d guiding channel . the coherence lengths ( transversal decay lengths ) are given by @xmath34 and @xmath35 . the superscript @xmath36 in eqs . ( [ eq : chip ] ) and ( [ xi ] ) emphasizes the majorana nature of the mode . another building block of chiral interferometers is a domain wall on a surface of a 3d ti where the magnetization sign is changed . if we consider such a @xmath37m/@xmath38 m boundary along the @xmath32-axis , which is described by the term @xmath39 in the bogolyubov - de gennes hamiltonian , we end up with the dirac chiral mode with the same spectrum found for @xmath5 mm ( [ dispersion ] ) . in the nambu notation any of @xmath40 eigenvalues are doubly degenerate and related to orthogonal electron and hole eigenstates @xmath41 and @xmath42 , where @xmath43 and @xmath44 . there are two independent excitations with energy @xmath40 in nambu notation in this @xmath37m/@xmath38 m case the electron with momentum @xmath2 and bogolyubov operator @xmath45 and the @xmath46 hole with @xmath47 . this field is complex @xmath48 and corresponds to a charged mode . we introduce here 1d operators @xmath49 @xmath50 with the anticommutation rules given by @xmath51 @xmath52 @xmath53 the same relations hold in the heisenberg picture at equal times @xmath54 . after that the secondary quantized operators in the low energy range @xmath55 can be found as sums over subgap chiral states . for the charge - neutral majorana excitations we obtain @xmath56 while for the dirac mode the field reads @xmath57 substituting these fields into the second quantized bogolyubov - de gennes hamiltonian and performing the transverse integration we obtain the following 1d hamiltonians : @xmath58 and @xmath59 the hamiltonians ( [ h - m ] ) and ( [ h - d ] ) correspond to coherent propagation of the excitations in 1d guiding channels with the fermi velocity @xmath60 and chirality dependent on the sign of the magnetization . the @xmath61 in the majorana hamiltonian @xmath62 reflects the fact that independent excitations in the @xmath5 mm can be considered either at positive or negative energies only . say , the bottom branch of the chiral mode ( [ dispersion ] ) at @xmath63 is redundant . the wave functions @xmath64 or @xmath65 show that the spin direction is locked to the momentum . more specifically , the spin textures of the guiding channels of majorana and dirac modes are orthogonal to the momentum direction in the particular case of rashba type hamiltonian ( [ h ] ) . as a consequence the spin textures are antiparallel in the counterpropagating channels . more generally , any scattering at a junction will be accompanied by the corresponding spin rotation . in this section we consider the josephson junction shown in figure [ hbt ] . this consists of two counterpropagating chiral dirac modes which scatter at the left and right contacts with the superconducting leads . each of these contacts consists of a pair of chiral y - junctions where electrons convert into a pair of majorana fermions with opposite chiralities or vice versa ( see figure [ hbt - contact ] ) . . light and dark grey bars are magnetic materials which induce exchange fields of the opposite polarizations @xmath66 m . the line of the sign change of m supports chiral charged modes marked as blue . magnetic flux @xmath8 in @xmath37 m region induces aharonov - bohm phase @xmath67 . ] the full @xmath13-matrix of this contact is derived from two operator relations for both upper and lower y - junctions , described by @xmath68 and @xmath69 matrices found in refs . . let us assume first that @xmath70 in the electrode . the matrix @xmath69 involves phase difference between electron and hole converting into two majorana fermions . this phase is denoted as @xmath71 and is included in the scattering matrix as follows @xmath72 @xmath73 the fields entering eq . ( [ eq : sin ] ) are heisenberg operators at a given energy . the @xmath68 matrix was found in ref . by means of time reversal transformation of @xmath69 which changes the magnetization sign @xmath74 @xmath75 we assume here that both y - junctions have identical geometries ( and , in particular , @xmath71 is the same ) . in this work we set the value of @xmath71 as a free parameter which is to be found from a particular geometry of the y - junction . the scattering in the upper y - junction in figure [ hbt - contact ] reads as @xmath76 we match majorana operators @xmath77 and @xmath78 for the given energy @xmath40 as @xmath79 , where the dynamic phase @xmath80 is accumulated by the majorana excitation in course of the propagation from the lower to the upper y - junction separated by the distance @xmath81 . the full @xmath82-matrix of the left contact is found straightforwardly after the exclusion of @xmath83 fields from eqs . ( [ s - in ] ) , ( [ s - out ] ) and their hermitian conjugates . to reinstate a non - zero sc phase of the electrode @xmath84 ( colored as green in figure [ hbt - contact ] ) , we employ the transformation @xmath85 , yielding @xmath86 this unitary @xmath13-matrix has particle - hole symmetry @xmath87 and acts on @xmath88 . it describes the partial andreev reflection in spinless dirac channel , which is combined with neutral majorana excitations . the andreev part of this process is accompanied by a cooper pair absorption by the sc electrode . in the following consideration we assume that chiral dirac channels have equal lengths , are parallel to each other and separated by the distance @xmath81 . using the above approach based on @xmath13-matrix ( [ s - matr - hbt ] ) , we find @xmath89-operators of charged fermions as linear combinations of uncorrelated field operators @xmath90 and @xmath91 of incident majorana modes . the latter are characterized by the fermi distribution function : @xmath92 where @xmath93 is the density of states in the @xmath5 mm channels . we assume @xmath94 everywhere and recover it in final expressions . using operator relations , we calculate chiral local densities of states @xmath95 and currents @xmath96 . the linear spectrum of the system ( [ dispersion ] ) means that chiral current @xmath97 is proportional to charge density and , hence , the josephson current @xmath98 is given by @xmath99 the positive direction of the current is defined from left to right sc , bias voltage is zero and the sc phases on the electrodes are equal to @xmath100 , as shown in fig.[hbt ] . while solving for dirac @xmath101-operators , we take into account the dynamic @xmath102 and the aharonov - bohm @xmath103 phases @xmath104 the thouless energy @xmath105 here equals to inverse dwell time of the interferometer ( or the level spacing in the n - region ) @xmath106 dynamic phase @xmath102 is accumulated by @xmath40-electron or @xmath107-hole enclosing the interference loop of total length @xmath108 . calculations of expressions for @xmath101 are analogous to those presented in appendix [ appendix ] for tunnel junction . within this @xmath13-matrix formalism , valid for energies in the range @xmath109 , we find the following result for the c@xmath6r : @xmath110 where the distribution function is determined by that of the incoming @xmath5mms defined in ( [ d - function ] ) . the spectral current @xmath111 can be associated with the local densities of states in the arms @xmath112 via @xmath113 . the densities of states @xmath114 , discussed in section [ dos - app ] , have dimension of inverse velocity , hence , @xmath111 is dimensionless . we obtain @xmath115 this is a non - singular @xmath7-periodic function which reflects the structure of the broadened andreev levels . as we mentioned in the introduction , conventional @xmath7-periodicity with respect to @xmath6 is related to the assumption of equilibrium ( no transport voltage ) . this means that there are no parity effects resulting in anomalous @xmath4 josephson effect . the integration in eq . ( [ eq : jjepsilon ] ) is reduced to a sum over positive matsubara fermionic frequencies @xmath116 . we note that the dynamical phase takes the values @xmath117 at these matsubara frequencies . evaluating the sum , we obtain the c@xmath6r @xmath118 this is one of the central results of this paper . it describes the dc josephson current at arbitrary temperature and takes into account contributions from the subgap 1d states . at zero temperature the summation in @xmath119 is replaced by an integration over @xmath120 and the result reads @xmath121 the c@xmath6r at zero temperature @xmath122 is presented in figure [ cphr - transp - fig ] for the value of the aharonov - bohm phase chosen as @xmath123 . at this value the first @xmath6-derivative of the current is divergent as @xmath124 at @xmath125 . this divergence illustrates the tendency of the c@xmath6r to have spikes at even phases @xmath126 . in section [ high - t - app ] we discuss the limit of full andreev reflection , where the c@xmath6r has a sawtooth form , also with spikes at @xmath126 . two separated dirac modes connecting the two superconductors form a squid loop . in view of the chirality of the junction the andreev pair belongs to both dirac channels , since a reflection into the same channel is forbidden . considering the junction as a squid loop controlled by a magnetic flux applied to the @xmath37 m light gray bar in fig.[hbt ] , we observe the fractional @xmath127-periodic pattern for the critical current . in figure [ hbt - fig ] we plot the critical current as a function of the flux - induced aharonov - bohm phase @xmath128 . the curve is symmetric under the assumption that @xmath129 . ( we recall that @xmath71 is related to the geometry of the y - junctions . ) the critical current @xmath130 in this plot is normalized by its maximal value @xmath131 . at arbitrary @xmath71 the curve in figure [ hbt - fig ] would be horizontally shifted . the positions of the peaks of @xmath130 are given by @xmath132 . c@xmath6r of the transparent junction at @xmath133 . divergent @xmath6-derivatives at @xmath134 are observed.,title="fig : " ] + superconducting phase , @xmath6 -periodic pattern of critical current @xmath135 at @xmath136 . the critical current @xmath135 is normalized to the maximal value @xmath131 and plotted as a function of aharonov - bohm phase @xmath128 , where @xmath8 is the magnetic flux.,title="fig : " ] + aharonov - bohm phase in n - region , @xmath137 the realization of the tunneling regime of the josephson junction studied in this paper is presented in figure [ setup ] . the left and the right tunnel contacts are implemented as constrictions of the magnetic material ( see figure [ contact ] ) . similar to the previous section [ mm - n - mm ] , the dirac channels @xmath138 and @xmath139 are colored as blue lines and have equal lengths @xmath140 and geometries . in contrast to the transparent regime , in the tunneling regime the dirac mode is not terminated but rather forms a closed loop . hence , we introduce the total phase acquired over the loop a sum of the berry phase @xmath141 and the aharonov - bohm phase , @xmath142 . the constriction of the @xmath38 m magnet plays the role of an insulating barrier . we employ the effective 1d hamiltonians ( [ h - m],[h - d ] ) and add a local tunneling term . the resulting hamiltonian reads @xmath143 this hamiltonian describes counter propagating majorana and dirac modes which mix at @xmath144 . the microscopic structure of the contact is accounted for by the small tunneling amplitude @xmath145 , which is an additional parameter of our theory . in appendix [ s - t - app ] , using hamiltonian ( [ h - contact ] ) , we derive the scattering matrix @xmath146 of the tunnel contact acting on fields @xmath88 . the scattering matrix reads @xmath147 the formal solution ( [ s - matr ] ) obtained in appendix [ s - t - app ] is valid for any value of @xmath148 and the dimensionless parameter @xmath149 characterizing the scattering is given by @xmath150 yet , since the hamiltonian ( [ h - contact ] ) is physically justified only in the weak tunneling limit @xmath151 , we obtain @xmath152 . the amplitude of the dominant process of normal reflection is given by @xmath153 . the scattering to the majorana channel scales as @xmath154 , whereas the andreev reflection amplitude has the lowest amplitude @xmath155 . using the scattering matrix @xmath146 calculated above ( [ s - matr ] ) , we obtain the following results for the spectral current and the c@xmath6r ( for details see appendix [ appendix ] ) : @xmath156 @xmath157 here the thouless energy is given by @xmath158 . from the c@xmath6r in ( [ jj - main ] ) we see that at high temperatures , @xmath159 , only @xmath160 term contributes to the sum ( [ jj - main ] ) . in this limit we observe a sinusoidal c@xmath6r and the critical current @xmath130 is exponentially suppressed : @xmath161 the factor @xmath162 is proportional to the full transparency of the junction being the product of the transparencies of the left and the right contacts . this result is similar to the c@xmath6rs for a conventional s / n / s junction with the n - region being a long quantum wire @xcite . in the low temperature regime , @xmath163 , the c@xmath6r is no longer sinusoidal and the critical current @xmath164 decays as @xmath165 . flux dependent oscillations of @xmath130 are more sharper compared to those in the transparent regime . the resonant shape of the non - fraunhofer @xmath0-periodic modulation is shown in figure [ crit_current_fig ] for weak tunneling @xmath166 . the shift of the maximum of @xmath130 by @xmath141 results from the presence of the berry phase . the solution for the scattering matrix ( [ s - matr ] ) follows formally from the tunneling hamiltonian ( [ h - contact ] ) at any value of @xmath148 ( see appendix [ s - t - app ] ) . yet , it is only physically justified for @xmath167 , or @xmath168 . we ask ourselves what happens if we extend ( [ s - matr ] ) , formally , to an arbitrary value of @xmath169 in the tunneling hamiltonian ( [ h - contact ] ) and consider the scattering matrix ( [ s - matr ] ) at any value of @xmath170 . we observe that at @xmath171 the c@xmath6r obtained for the transparent regime ( [ cphr - hbt ] ) is recovered ( up to the geometry dependent phase @xmath71 ) . in other words , transparent contacts formed by y - junctions correspond to the intermediate strength @xmath171 of tunnel contacts , rather than to the regime @xmath172 . in this subsection we investigate the c@xmath6r and the density of states in the n - region at arbitrary @xmath149 . in particular , we analyze the regime of full transparency , @xmath173 , where the scattering matrix @xmath146 reaches the unitary limit and corresponds to the full andreev reflection in the dirac channel . note , that the relation between @xmath149 and @xmath169 , obtained in appendix [ s - t - app ] , assumes a certain microscopic structure of the contacts and , thus , could also be different . we discuss this in more details in appendix [ s - t - app ] . as mentioned above , the formal result for c@xmath6r for the tunneling regime ( [ jj - main ] ) is identical to that for the transparent regime ( [ cphr - hbt ] ) at @xmath171 , up to geometry dependent phase @xmath71 . in figures [ cphr_fig],[crit_current_fig ] we illustrate the evolution of c@xmath6rs and critical currents @xmath130 upon increase of the transparency parameter @xmath149 . we plot the results at small @xmath174 , related to tunneling approximation , and their continuation to higher @xmath175 and @xmath176 . r found in ( [ jj - main ] ) for tunnel junction at dimensionless tunneling strengths @xmath174 . dashed curves : continuation of the results for c@xmath6r ( [ jj - main ] ) to finite transparencies @xmath175 and @xmath177 . the current @xmath178 is measured in units of the thouless energy @xmath105 with aharonov - bohm phase @xmath133 ( @xmath179 ) and at low temperature @xmath180 . the @xmath177 curve indicates the tendency to the formation of spikes at unusual even phases @xmath1.,title="fig : " ] + superconducting phase , @xmath6 of the tunnel ( @xmath166 ) and transparent at ( @xmath171 and @xmath129 ) junctions as a function of aharonov - bohm phase @xmath181 . maximum of the critical current is shifted due to the presence of berry phase . dashed curves correspond to results at finite @xmath182 found from the tunneling approach . all figures are plotted at low temperature @xmath180.,title="fig : " ] + aharonov - bohm phase in n - region , @xmath103 in the limit @xmath173 the scattering matrix @xmath146 ( [ s - matr ] ) reaches the unitary limit @xmath183 this matrix describes the process of full andreev reflection , where an electron converts into a hole with a phase shift of wave function equal to the phase of the sc electrode . at @xmath136 and @xmath173 , where @xmath130 is independent on @xmath103 , the c@xmath6r ( [ jj - main ] ) shows a sawtooth shape with spikes at _ even _ phases @xmath184 in conventional ballistic spinful 1d s / n / s junctions the c@xmath6rs are known to be sawtooth - like @xcite . in contrast to our result ( [ cphr - un - lim ] ) the spikes are usually _ odd _ @xmath185 and the slope is positive . this difference follows from an additional phase of @xmath141 acquired by a fermion in a spinful junction after two andreev reflections from the contacts . for spinful systems , full andreev reflection is associated with the time - reversal transformation @xmath186 of the fermion wave function . being reflected two times from the left and right contacts , the spinful fermion gets a @xmath141 phase . in spinless junction under consideration this phase is zero . we calculate density of states at arbitrary @xmath149 in @xmath138-branch with the use of @xmath187 from ( [ psi_a ] ) derived in appendix [ appendix ] . the density of states in chiral dirac modes are halfs of the spectral current , because @xmath188 if incoming majorana fermion modes @xmath189 are not correlated . we introduce the local retarded green function ( calculated at an arbitrary coordinate @xmath32 in the @xmath138-branch ) as @xmath190 the fourier transform of this function , @xmath191 gives for the density of states @xmath192 assuming the density of states of @xmath5 mm in the sc gap is constant , i.e. @xmath193 , we obtain @xmath194 in figure [ dos_fig ] we plot the density of states @xmath114 at three values of dimensionless tunneling parameter @xmath149 . at weak tunneling , @xmath195 ( see fig . [ dos_fig ] ( a ) ) , the maxima of @xmath114 are slightly dependent on @xmath6 . horizontal lines resemble smeared mesoscopic levels of an isolated 1d dirac wire of length @xmath196 . these maxima can be shifted vertically by a flux induced aharonov - bohm phase @xmath103 , because their position is given by @xmath197 the result for the intermediate transparency @xmath171 , equivalent to the transparent junction , is shown in fig . [ dos_fig ] ( b ) . we see that the density of states is strongly smeared in this case . the continuation of @xmath114 to high transparency ( @xmath198 ) shows that the spectral density is given by @xmath7-periodic narrow lines of andreev levels . their structure can be found from the singularities of ( [ dos ] ) at @xmath173 as @xmath199 in figure [ dos_fig ] ( c ) we plot the density of states at @xmath200 which consists of half of the full set of smeared andreev levels ( [ and - lev ] ) . note , that level positions are independent of the aharonov - bohm phase . this follows from the fact that the electron and the reflected hole get opposite aharonov - bohm phases , which compensate each other in closed paths . ) of the right movers , @xmath201 in the upper @xmath138-arm of the junction . the energy @xmath40 is counted in units of thouless energy @xmath105 . blue and white colors correspond to low and high densities respectively . * ( a ) * weak tunneling regime , @xmath195 ; the horizontal lines are reminiscent of the quantized levels of the isolated n - region ; * ( b ) * intermediate transparency , @xmath171 , the results coincide with those for a transparent junction at @xmath202 ; * ( c ) * the case @xmath200 illustrates the continuation to the high transparency limit ( full andreev reflection ) , where bright narrow lines are half of full set of smeared andreev levels . aharonov - bohm phase @xmath203 for all three plots . ] one can describe the scattering between the dirac and majorana chiral modes by representing the @xmath89 operators in the n - channel with the help of two auxiliary charged - neutral majorana operators @xmath204 and @xmath205 . these modes do not carry charge separately but their superpositions do . the different sc phases of the opposite contacts result in a fusion of the auxiliary @xmath204 and @xmath206 modes , which is responsible for a cooper pair transfer from one lead to the other . in this section we discuss the scattering in the majorana basis for the tunneling and transparent contacts . we analyze in detail the regime @xmath171 ( keeping @xmath71 as a free parameter ) , since it corresponds to the case of the transparent junction ( [ cphr - hbt ] ) up to geometry dependent @xmath71 and the berry phases . the majorana basis is defined through the following transformation for both the left and the right contacts @xmath207 where the index @xmath208 here stands for @xmath209 and @xmath210 channels . the @xmath146-matrix ( [ s - matr ] ) under this transformation at arbitrary @xmath149 reads @xmath211 according to ( [ m - basis - tunn ] ) @xmath204 mode is always fully decoupled , consistent with the scattering theory of li , fleury and bttiker @xcite . in the weak tunneling limit @xmath168 the modes @xmath5 and @xmath206 scatter into each other with the amplitude @xmath154 . in the opposite unitary limit @xmath173 all modes are decoupled from each other but both @xmath5 and @xmath206 invert signs meaning that the corresponding dirac fermions experience the full andreev reflection . in the intermediate case @xmath171 ( figure [ scattering - mf ] , left ) the modes @xmath5 and @xmath206 fully convert into each other : @xmath212 and @xmath213 . the presence of the scattering phase @xmath71 in the scattering matrix ( [ s - matr - hbt ] ) of the transparent dirac - majorana contacts changes the situations considerably . we apply again the transformation ( [ basis-0 ] ) to @xmath89-operators and obtain the following @xmath13-matrix in the majorana basis ( setting @xmath214 ) @xmath215 this matrix coincides with that of ( [ m - basis - tunn ] ) in the limit @xmath216 only if @xmath217 . for the other values of @xmath218 all the modes are mixing with each other . at arbitrary @xmath71 there is mixing between all of the modes except of @xmath219 and @xmath220 . mixing between @xmath219 and @xmath220 is possible if we a add second contact with a different sc phase . we illustrate the scattering in the particular case of @xmath221 in figure [ scattering - mf ] ( right ) , where the incoming @xmath5-mode converts into @xmath204 mode and vise versa . note that , in this case , @xmath206 mode is not converted to the others but gets an inversion of the sign , @xmath222 . and neutral - charge channels @xmath5 written in terms of auxiliary majorana modes @xmath223 . left : tunnel contact described by @xmath146-matrix ( [ s - matr ] ) at @xmath171 . right : transparent contact described by @xmath13-matrix ( [ s - matr - hbt ] ) at @xmath221.,title="fig : " ] and neutral - charge channels @xmath5 written in terms of auxiliary majorana modes @xmath223 . left : tunnel contact described by @xmath146-matrix ( [ s - matr ] ) at @xmath171 . right : transparent contact described by @xmath13-matrix ( [ s - matr - hbt ] ) at @xmath221.,title="fig : " ] to conclude , we analyzed two limits of 1d long ballistic josephson junctions where the leads are formed by gapless 1d chiral majorana channels.these junctions can be realized as hybrid structures involving 3d topological insulator surface in proximity with s - wave superconducting electrodes and magnetic materials . the normal region of these setups is formed by two chiral dirac modes spaced by a magnetic material . such a normal part is effectively spinless because the spin textures are locked to profiles of magnetic domain walls . in the first part of the work we have calculated the c@xmath6r in the regime of high transparency . in this limit the left and the right contacts of the junction consist each of two dirac - majorana converters built by magnet / superconductor interfaces . we find that this system has a continuous spectral current , a @xmath7-periodic non - sinusoidal c@xmath6r at low temperatures and an @xmath0-periodic dependence of the critical current on the magnetic flux . the critical current amplitude at zero temperature is given by thouless energy which is proportional to inverse dwell time of the normal region . the junction can act as a squid because the two dirac channels are spatially separated . an andreev pair in this case appears to be spatially non - local , which offers a possibility of inducing an internal aharonov - bohm phase . this leads to one of our central results : the critical current shows fractional @xmath0-periodic pattern . in the second part of the work we have studied another realization of the junction where dirac and majorana channels are coupled through a tunnel barrier . the c@xmath6r of such a tunnel junction was found in terms of the tunneling hamiltonian approach and corresponding scattering matrix . the resulting critical current patterns show sharper resonant peaks as compared to those for the transparent junction . we have also studied the formal extension of the tunneling solution to the high transparency regime and compared it with the one obtained in the first part of the work . interferometers involving chiral majorana modes @xcite could find their applications as measuring devices of topological qubits @xcite . in this paper we have explored a dual setup , in which the equilibrium josephson current is carried by interfering chiral dirac electrons between chiral majorana leads . embedded into schemes with vortices and/or magnetic or sc islands supporting zero - energy majorana pairs , a josephson based quantum interferometer might be of advantage , e.g. , for performing quantum readout of topological qubits . the authors thank i.v . protopopov , i.s . burmistrov , a.l . rakhmanov , a.v . rozhkov , p.m. ostrovsky , w.v . pogosov , s.n . artemenko and , especially , yu . makhlin for fruitful discussions . we acknowledge financial support by dfg priority program 1666 , by german - israeli foundation , and by the eu network grant internom . the research of a.d.m . was supported by the russian science foundation under grant no . 14 - 22 - 00281 . the local form of tunneling hamiltonians ( [ h - contact ] ) is an approximation , because , microscopically , such contacts are formed by constrictions of finite size as depicted in figure [ contact ] . if the constriction is wide and the hybridization is negligible , the parallel chiral channels along sc/+m and + m/@xmath37 m junctions have opposite ( orthogonal ) spin textures . in the tunneling area , shown as the constriction of the magnetic material in the black bar , the eigenfunctions can be approximated by superpositions of the eigenfunctions of the isolated dirac and majorana channels , @xmath224 and @xmath64 . this means that an electron coming into the tunneling area starts to oscillate between hole- and majorana - like states with a rate , proportional to a hybridization of the channels , estimated as @xmath225 , where @xmath226 is a transversal constriction length . finally , the incoming electron scatters into a superposition of outgoing electron , hole and neutral excitation in the majorana channel . from this qualitative picture one can conclude that the scattering matrix should be periodic with respect to the phase of these oscillations , given as a product of hybridization energy and the constriction dwell time @xmath227 . here , we derive this periodic behavior of the @xmath146-matrix from the local tunneling hamiltonian ( [ h - contact ] ) . we employ the heisenberg equations of motion @xmath228 where the tunnel matrix is given by @xmath229 we write @xmath230 in these equations in order to make the chirality of the majorana mode the same as that of the charged channel . in this representation we can consider incoming and outgoing states as those at @xmath231 and at @xmath232 respectively . it follows from the @xmath32-integration of ( [ eq2dagger],[t - matr ] ) around the point of contact @xmath233 that the relation between the tunneling matrix @xmath234- and the scattering matrix @xmath146 reads @xmath235.\label{s - t - exp}\ ] ] calculating the exponent we obtain @xmath236 where the dimensionless tunneling strength @xmath237 is given by @xmath150 this tunneling @xmath146-matrix is unitary and obeys the particle - hole symmetry @xmath238 like @xmath13 . the eigenvalues of ( [ s - matr - app ] ) are given by @xmath239 . the solution method leading to ( [ s - matr - app ] ) and ( [ s - t - t ] ) is not universally accepted . rather , a different ansatz was used in various problems on transport in 1d systems such as tunneling between edge states of qhe @xcite , impurity scattering in luttinger liquid at @xmath240 @xcite or resonant andreev reflection from zero - mode majorana bound state @xcite . following these works we should have taken the local @xmath89- and @xmath5-operators at point @xmath144 as @xmath241 relations ( [ psi - chi-1/2 ] ) produce then the same solution for the scattering matrix as in ( [ s - matr - app ] ) . however , the parameter @xmath149 is now different and is given by @xmath242 the two solutions coincide in the weak coupling limit @xmath167 . for larger values of @xmath148 the difference is substantial . for instance , the unitary limit @xmath243 is achieved with ( [ psi - chi-1/2 ] ) and ( [ eq : lambdatilde ] ) at @xmath244 . in contrast , with ( [ s - t - exp ] ) it is reached at @xmath245 . we conjecture that the solution method , leading to ( [ s - t - exp ] ) applies if the constriction is smooth enough , so that the validity of the low energy description provided by ( [ h ] ) and ( [ h - contact ] ) is not violated in any point of the constriction . then , eq . ( [ eq2dagger ] ) is solved as a regular differential equation . on the other hand , the ansatz ( [ psi - chi-1/2 ] ) is probably applicable for sharp enough constrictions . this ambiguity should be resolved by solving 2d bdg equations for the constriction geometry . in this appendix we present technical details of the calculation of the josephson current . we find the c@xmath6r as the difference between the chiral currents in @xmath138 and @xmath139 arms by using the tunnel junction scattering matrix @xmath146 , given by ( [ s - matr - app ] ) . we take the 1-st and the 3-d lines of ( [ s - matr - app ] ) , disregarding the outgoing majorana field @xmath220 ( line 2 ) , and obtain @xmath246 the matrix in the r.h.s . of eq . ( [ m - norm - andr ] ) contains amplitudes of normal and andreev reflection , while the last term describes coupling with the equilibrium majorana @xmath209-channel . this term is responsible for the spectral current ( [ sp_current ] ) being continuous , due to the gapless spectrum of the incoming lead majorana mode . using ( [ s-0 ] ) we formulate the boundary conditions for the left and right contacts . we introduce the field @xmath250_{x}^t$ ] , which depends on @xmath40 and on the coordinate @xmath32 along the dirac channels . indices @xmath138 and @xmath139 stand for the upper and lower dirac modes ( see figure [ setup ] ) . we start from the left contact where @xmath251 and @xmath252 . using ( [ s-0 ] ) and its hermitian conjugate at @xmath253 as well as the property of the real majorana field , @xmath254 , we derive @xmath255 here @xmath256 is the @xmath257 identity matrix in the left upper block and @xmath258^t$ ] . note , that for the left contact @xmath139-components of @xmath259 are @xmath209-fields while @xmath138-components are @xmath210-fields : @xmath260 and @xmath261 . the rank of the 4@xmath2624 matrix in ( [ bc_left ] ) equals @xmath263 and the eigenvalues are equal to @xmath264 regardless of the values of @xmath149 and @xmath6 . the condition for the right contact is obtained from ( [ bc_left ] ) by the replacements @xmath265 the pauli matrix @xmath266 , introduced in ( [ bc_right ] ) , mixes the @xmath257 blocks in ( [ bc_left ] ) acting as a particle - hole transformation , while @xmath267 acts inside of the blocks . the product @xmath268 interchanges @xmath138 and @xmath139 indices in @xmath269 . we find that the fields in the middle of n - region , @xmath270 , can be expressed via the operators at the ends of the dirac channels , @xmath259 and @xmath271 , using the dynamical and geometrical phases : @xmath272 here @xmath273 $ ] and @xmath274 $ ] . the external phase @xmath275 equals to the sum of aharonov - bohm and berry phases and the dynamical phase @xmath102 accumulated an electron of energy @xmath40 or a hole of energy @xmath107 , enclosing the interference loop of the length @xmath276 . here @xmath277 the relations ( [ bc_left ] , [ bc_right ] ) together with ( [ x01 ] , [ x02 ] ) make the problem of finding the four components of @xmath278 algebraically close . the result for the first component reads the dirac field in the @xmath139-channel is obtained using the geometrical symmetry of the setup and is given by @xmath280 the operator relations ( [ psi_a ] ) and ( [ psi - b ] ) are used to calculate @xmath281 which results in the c@xmath6r presented in ( [ jj - main ] ) . 40 j. alicea , rep . . phys . * 75 * , 076501 ( 2012 ) . c. nayak , s. h. simon , a. stern , m. freedman , and s. das sarma , rev . * 80 * , 1083 ( 2008 ) . kitaev , phys . * 44 * ( suppl . ) , 131 ( 2001 ) . n. read and d. green , phys . b * 61 * , 10267 ( 2000 ) . g. moore and n. read , nucl . b360 * , 362 ( 1991 ) . l. fu and c. l. kane , phys . b * 79 * , 161408 ( 2009 ) . y. oreg , g. refael , and f. von oppen , phys . . lett . * 105 * , 177002 ( 2010 ) . j. nilsson , a. r. akhmerov , and c.w . j. beenakker , phys . * 101 * , 120403 ( 2008 ) . maslov , m. stone , p.m. goldbart , d. loss , phys . b * 53 * , 1548 ( 1996 ) . i. o. kulik , zh . 57 * , 1745 ( 1969 ) . r. fazio , f. w. j. hekking , and a. a. odintsov , phys . b * 53 * , 6653 ( 1996 ) . c. de c. chamon , d. e. freed , and x. g. wen , phys . b * 53 * , 4033 ( 1996 ) . r. egger and h. grabert , phys . b * 58 * , 10761 ( 1998 ) . l. fidkowski , j. alicea , n. h. lindner , r. m. lutchyn , and m. p. a. fisher , phys . b * 85 * , 245121 ( 2012 ) . i. o. kulik and a. n. omelyanchuk , zh fiz . * 68 * , 2139 - 2148 ( 1975 ) .

sn 1995n in mcg02 - 38 - 017 ( also known as arp 261 ) was discovered by pollas ( 1995 ) on 5 may 1995 ( ut dates are used throughout this paper ) . its approximate distance is 24 mpc , based on @xmath4 mpc@xmath5 and the measured recession velocity of the host galaxy ( 1834 @xmath0 ; theureau et al . optical spectra of the object obtained by benetti , bouchet , & schwarz ( 1995 ) and by garnavich & challis ( 1995 ) showed it to be a peculiar type ii supernova , with h@xmath2 emission ( full - width at half - maximum intensity [ fwhm ] @xmath6 ) and narrow lines of he i , he ii , [ n ii ] , [ o i ] , [ o iii ] , [ ne iii ] , [ ca v ] , fe ii , [ fe ii ] , [ fe iii ] , [ fe vii ] , and [ fe x ] . the intensity ratios of the narrow oxygen lines suggest a high electron density , @xmath7 @xmath8 ( garnavich & challis 1995 ) . benetti et al . ( 1995 ) estimated it to be at least 10 months old , so we will assume an explosion date of 4 july 1994 , befitting for a supernova ( in the opinion of the u.s . coauthors of this paper ) . radio emission from the supernova was discovered at 3.6 cm on 16 june 1995 ( van dyk et al . 1996 ) . x - ray emission was discovered with _ rosat _ on 23 july 1996 , and the source was subsequently observed with _ asca _ on 20 january 1998 ( fox et al . the unabsorbed x - ray luminosity was @xmath9 in the 0.110 kev band . the narrow emission lines observed from sn 1995n place it in the type iin category ( schlegel 1990 ; filippenko 1991a , b , 1997 ) , which often show a slow optical decline compared to other type ii supernovae ( sne ii ) . baird et al . ( 1998 ) have noted the slow evolution of sn 1995n . the high x - ray and radio luminosities of sn 1995n place it in a subset of sne iin whose other well - studied members are sn 1986j ( rupen et al . 1987 ; weiler , panagia , & sramek 1990 ; leibundgut et al . 1991 ; houck et al . 1998 ) , sn 1988z ( stathakis & sadler 1991 ; filippenko 1991a , b ; turatto et al . 1993 ; van dyk et al . 1993 ; fabian & terlevich 1996 ; aretxaga et al . 1999 ) , sn 1978k ( ryder et al . 1993 ; chugai , danziger , & della valle 1995 ; schlegel , petre , & colbert 1996 ) , and sn 1998s ( leonard et al . 2000 , gerardy et al . 2000 ; fassia et al . all of them have x - ray luminosities @xmath10 , which are at least an order of magnitude higher than the x - ray luminosities of other sne ii . these are also among the most luminous radio sne , although sn 1979c does rival them in luminosity , as does sn 1998bw , a very different kind of object and at a much earlier epoch ( e.g. , galama et al . 1998 ; kulkarni et al . 1998 ) . the early optical observations of sn 1988z ( stathakis & sadler 1991 ) showed velocities up to @xmath11 , and very long baseline interferometry ( vlbi ) radio observations of sn 1986j ( bartel , shapiro , & rupen 1989 ; bartel et al . 1991 ) implied expansion velocities of @xmath12 . at the same time , these sne showed h@xmath2 emission lines with fwhm @xmath13 , and at least sn 1988z exhibited very narrow emission lines whose intensity ratios indicate high gas densities . this combination suggests that the rapidly moving supernova material is running into dense circumstellar clumps and is driving slower shock fronts into them ( chugai 1993 ; chugai & danziger 1994 ; chugai et al . in addition , the combination of high velocities with strong circumstellar interaction requires that the supernova have considerable high - velocity gas in order not to be rapidly decelerated . for sn 1988z , chugai & danziger ( 1994 ) estimate that the total mass in supernova ejecta is only @xmath14 if the explosion energy is the standard @xmath15 ergs . in order to more fully understand the nature of this peculiar subset of sne iin , we have undertaken a detailed spectroscopic study of sn 1995n , one of the most important and thoroughly studied examples . we present the ground - based and _ hubble space telescope ( hst ) _ observations in 2 . in 3 , we discuss the three main kinematic components visible in the emission - line profiles . synthetic spectra are presented in 4 , and the temporal evolution of features is examined in 5 . we discuss the results and their implications in 6 , and we summarize our conclusions in 7 . figure [ fig1a ] shows a 3-minute @xmath16-band image of mcg02 - 38 - 107 and sn 1995n taken with the fors1 instrument on the eso very large telescope ( vlt ) on 11 may 1999 ( ut dates are used throughout this paper ) . figure [ fig1b ] is an _ hst _ wfpc2 @xmath17-band ( f814w ) image of sn 1995n obtained on 22 july 2000 ; the supernova is easily visible at @xmath18 mag . these two images demonstrate that the supernova occurred in a region relatively free of background emission and dust . the main observations of sn 1995n presented in this paper consist of ground - based optical spectra obtained on nine epochs and ultraviolet ( uv ) _ hst _ spectra on one epoch . the log of the observations is given in table [ tab1 ] , where the wavelength ranges , ut dates , and approximate epoch since explosion are summarized . the assumed explosion date of 4 july 1994 is uncertain , but this does not appreciably affect our results because our data were taken long after the explosion ; the spectra cover epochs of 3211799 days . standard procedures were followed for the spectral extractions and reductions , which included sky subtraction , wavelength and flux calibration , and removal of telluric absorption ( e.g. , matheson et al . 2000 ) . the ground - based observations fall in two main groups : within @xmath19 years after discovery , and @xmath20 years after discovery . the former were obtained primarily with the kast spectrograph ( miller & stone 1993 ; @xmath21 slit ) on the 3-m shane reflector at lick observatory , but the low resolution imaging spectrometer ( lris ; oke et al . 1995 ; @xmath22 slit ) on the keck - i telescope was used to acquire an excellent spectrum ( day 1007 ) nearly contemporaneously with one of the lick spectra ( day 1012 ) . all of the lick and keck spectra ( except day 1012 ) are shown in figure [ fig2 ] . one can see that the spectrum evolved slowly during the time interval shown . in this paper , we present a detailed analysis of only two of these spectra ( days 716 and 1007 , which are among the two best ones ) . the very late - time ( @xmath20 years ) ground - based spectra were taken through a @xmath23 slit with fors1 on the vlt at eso . the day 1769 spectrum was obtained with gratings 300v and 300i , giving a resolution of @xmath24 . for the day 1799 spectrum the 600b , 600r , and 600i grisms were used , giving a higher resolution of @xmath25 . the day 1769 observing was done in good conditions , while the data from day 1799 were taken through clouds . the latter spectrum has higher signal - to - noise ratio ( @xmath26 ) than the former , however , so emission - line measurements are reported for it ( scaled to the flux level of the day 1769 spectrum ) . the _ hst _ spectra on day 943 were obtained with the faint object spectrograph ( fos ) with the g400h , g270h , and g160l gratings . total exposures times were 2860 , 4770 , and 2110 s , and the resolution was 3.0 , 2.1 , and 6.9 per diode , respectively . the spectra were calibrated by the fos pipeline at the space telescope science institute . all spectra , except the vlt and _ hst _ spectra , were calibrated to the @xmath27-band and @xmath16-band photometry of baird et al . the keck spectrum ( day 1007 ) has the highest @xmath26 , and in figure [ fig3 ] we show it together with the _ hst _ spectra ( day 943 ) . the keck and _ hst _ spectra were taken at different epochs , but from the day 1037 lick spectrum we find a smooth evolution , and we normalize the flux level of the overlapping region between the _ hst _ g400h spectrum and the keck spectrum to that of the keck epoch . the dirbe / iras maps ( schlegel , finkbeiner , & davis 1998 ) give @xmath28 mag for the galactic reddening of sn 1995n . in addition , there could be internal reddening from both the host galaxy and the circumstellar environment . the host galaxy is a blue compact galaxy , a class which is generally dust - poor , although h ii regions can show substantial reddening ( kong & cheng 1999 ) . the fact that the supernova is located in the outskirts of the galaxy ( figures [ fig1a ] , [ fig1b ] ) also makes it natural to expect a relatively low extinction . from the relative strengths of the higher members of the h i and he i recombination series , there is no indication of strong reddening ( [ sec_result ] ) . the normal intensity ratio of uv o iii ] lines relative to the optical also indicates this . in the following , we use @xmath29 mag for the total reddening . figure [ fig3 ] shows that there are lines spanning a large range of velocities , and we can distinguish at least three kinematic components in the supernova . the narrow , unresolved lines are clearly distinct , while the other two components are less obvious and are shown in the mg ii @xmath30 , [ o iii ] @xmath31 , and h@xmath2 lines at 9431007 days on the same velocity scale in figure [ fig4 ] . the h@xmath2 and [ o iii ] profiles are from the keck day 1007 spectrum and the mg ii profile from the _ hst _ spectrum . the h@xmath2 and [ o iii ] lines have quite different profiles , with the former having a fairly peaked profile , but with weak wings reaching very high velocity , @xmath32 , while the latter has a more boxy profile reaching a velocity of @xmath33 . another strong indication of two separate high - velocity components is the different temporal evolution of the [ o i ] , [ o ii ] , and [ o iii ] lines compared to the others ( [ sec_time ] ) . here we discuss these three components individually . the narrowest lines in the spectra are unresolved , implying fwhm @xmath34 . they are listed in table [ tab2 ] , where it can be seen that the emission is dominated by ions with moderately high ionization . some of the lines provide useful diagnostics for the emitting region , and we begin by considering them . * [ o iii]*. the [ o iii ] @xmath354959 , 5007 lines are by far the strongest narrow lines and , together with [ o iii ] @xmath364363 , provide a useful constraint on the electron temperature @xmath37 and the electron density @xmath38 . the relatively low ratio of @xmath39 implies that collisional deexcitation of the @xmath40 level is important , so that the ratio depends on density as well as temperature ( e.g. , filippenko & halpern 1984 ; osterbrock 1989 ) . the densities corresponding to various assumed temperatures are in the range @xmath41 for @xmath42 ( or @xmath43 for @xmath44 ) , where @xmath45 . on day 1799 the vlt spectrum shows a similar value , @xmath46 , which gives @xmath47 for @xmath48 . * o iii bowen lines*. in the near - uv spectral region , several bowen fluorescence lines are seen , most notably o iii @xmath49 . we measure @xmath50 and @xmath51 . based on the bowen process , kallman & mccray ( 1980 ) give ratios of 0.18 and 0.28 for these lines , respectively . considering that the uncertainties in the fluxes of both lines are at least a factor of two , the agreement is satisfying . the @xmath52 line is severely blended with [ ne iii ] @xmath53 and [ ne v ] @xmath54 , and no reliable flux estimate can be obtained . the ratio @xmath55 can be used to estimate the bowen yield for he ii @xmath56 . equation ( 25 ) in kallman & mccray ( 1980 ) , together with @xmath57 , gives a bowen yield of @xmath58 . correcting for reddening , we find @xmath59 , and therefore @xmath60 for sn 1995n at day 1007 . as a comparison , schachter , filippenko , & kahn ( 1990 ) find a mean @xmath61 for a sample of seyfert nuclei , while kallmann & mccray find that planetary nebulae have @xmath62 . the dispersion in both values is considerable . * [ o ii]*. the spectra show that [ o ii ] @xmath633726 , 3729 are not detected , but that [ o ii ] @xmath637320 , 7330 are weakly present with narrow components . the fact that @xmath64 implies that @xmath65 , provided that @xmath66 k. * [ n ii]*. the auroral [ n ii ] @xmath365755 line is present , but the nebular lines are superposed on the very strong h@xmath2 emission and it is not possible to obtain strong limits on them . we estimate @xmath67 , which implies that @xmath68 for @xmath69 k ( osterbrock 1989 ) . * [ fe vii]*. the observed lines of [ fe vii ] are @xmath633586 , 3759 , 5159 , 5276 , 5721 , 6087 . as discussed by nussbaumer & storey ( 1982 ) and keenan & norrington ( 1987 , 1991 ) , several of the [ fe vii ] lines provide diagnostic information on the emitting region . there is temperature sensitivity of the lines , and collisional deexcitation causes transitions from the low - density limits to the high - density limits in the range @xmath70 . the observed , reddening - corrected ratio @xmath71 in the keck spectrum ( day 1007 ) , and @xmath72 in the 12 june 1999 vlt spectrum ( day 1799 ) . from figure 2 of keenan & norrington ( 1987 ) , this sets a limit @xmath73 k for any density on day 1007 , and @xmath74 k for day 1799 . in the day 1799 spectrum , with the highest resolution , there are three distinct peaks at the positions of [ fe vi ] @xmath75 , [ fe vii ] @xmath76 , and [ fe vi ] @xmath77 . because of blending of these lines , it is difficult to obtain accurate fluxes , and depending on where we estimate the base of the line we get a ratio in the range @xmath78 and @xmath79 on days 1007 and 1799 , respectively . from figure 1 of keenan & norrington ( 1987 ) , the lower limits imply that @xmath80 if @xmath81 k. the lines at @xmath364942 and @xmath364989 are contaminated by the strong complex of [ o iii ] @xmath634959 , 5007 , with the result that interesting constraints on @xmath37 and @xmath38 can not be obtained from these lines ( keenan & norrington 1991 ) . overall , the above results point to a relatively high density for the narrow - line gas , which suggests a circumstellar rather than interstellar origin . if the [ o iii ] and [ fe vii ] lines are assumed to come from gas with similar physical properties , the implied characteristics are @xmath82 and @xmath69 k , although there is some evidence that the [ fe vii ] emission comes from a somewhat hotter region . it is likely , because of the low velocity and the relatively low pressure , that this gas is the unshocked circumstellar wind . presumably it was photoionized by the uv and x - ray radiation emitted by the supernova right after shock breakout . one can derive a lower limit to the radial distance of this gas by taking the product of the highest observed supernova ejecta velocity and the estimated age . as discussed below , the highest velocity is @xmath83 . the wind electron density in front of the shock then becomes @xmath84 where @xmath85 is the mass - loss rate from the progenitor star , @xmath86 is the wind velocity , and it has been assumed that h and he are fully ionized . the reference value of @xmath85 chosen here is close to the highest values that have been observed for galactic stars and to the strong radio emitter sn 1979c ( e.g. , lundqvist & fransson 1988 ) . either the progenitor star of sn 1995n was losing mass at an extraordinary rate , or the density of the observed gas is higher than the average density in the circumstellar wind . the gas density is sufficiently high that the recombination and excitation time scales are less than the supernova age , and approximate equilibrium is expected . the ionizing source is the x - ray emission , which had an unabsorbed x - ray luminosity of @xmath87 in the 0.110 kev band during july 1996 january 1998 ( fox et al . observations give a temperature estimate of 9 kev if the emission is thermal ( fox et al . models of x - ray nebulae have been presented by tarter , tucker , & salpeter ( 1969 ) and kallman & mccray ( 1982 ) . the latter models assume a constant density surrounding a central source and generally take @xmath88 kev for the photoionizing source , but the calculations can be used to approximate the current situation , especially because the [ fe vii ] lines are likely to be formed in a region optically thin to the ionizing continuum . the ionization properties depend on the ionization parameter , which can be defined as @xmath89 , where @xmath90 is the luminosity of the ionizing source . taking @xmath91 , @xmath92 , and @xmath93 cm , we find @xmath94 , which is very close to the ionization parameter where fe vii is expected , for a @xmath88 kev source ( kallman & mccray 1982 ) . a lower temperature for the continuum source would increase the estimated @xmath95 for a fixed @xmath90 . the gas temperature of this region in the photoionization models is @xmath96 k , consistent with the observations . this comparison provides support for the overall model and , in particular , for the estimated radial distance of the emitting region . the intensity of the narrow lines gives another constraint on the physical conditions . the hydrogen lines are the best to use because there are no abundance uncertainties . while no narrow h i lines can be seen in the day 1007 keck spectrum because of the strong broad component , the higher resolution day 1799 vlt spectrum clearly shows h@xmath2 , h@xmath97 , and h@xmath98 . for h@xmath97 , we find a luminosity of @xmath99 at 1799 days . if the emitting region is assumed to be at radius @xmath100 with a thickness @xmath101 and a covering factor @xmath102 , we have @xmath103 where the emissivity of h@xmath97 has been taken for @xmath104 k ( osterbrock 1989 ) . although the density is uncertain , it is clear that the observations imply a small filling factor . another indication of the small filling factor of the @xmath105 gas is the detection of a substantial x - ray flux by _ rosat_. if the gas were pervasive at @xmath106 cm , the resulting hydrogen column density , @xmath107 , would give rise to the complete absorption of the soft x - ray flux . the gas is not fully ionized and thus can effectively absorb the x - ray emission . in their estimate of the x - ray luminosity , fox et al . ( 2000 ) take @xmath108 based on the expected galactic absorption . most of the other ions that are observed in sn 1995n require an ionization parameter that is similar to that of fe vii . the ions fe vi through fe xi , which are observed , occur over a relatively small range of ionization parameter ( e.g. , kallman & mccray 1982 ) . however , the presence of narrow [ o i ] emission implies the existence of a separate , lower - ionization component . the region could be of low ionization because of an especially high density , but it is more likely that the region is neutral because the ionizing photons have been used up in the h ii region . the intermediate - width component is best exemplified by the [ o iii ] @xmath634959 , 5007 ( figure [ fig4 ] ) , [ o ii ] @xmath637320 , 7330 , and [ o i ] @xmath109 , 6363 emission lines , which all show box - shaped profiles extending to @xmath110 . the strongest feature in this component at all epochs is the [ o iii ] @xmath634959 , 5007 blend . other likely members of this component are the high - ionization far - uv lines . mg i ] @xmath364571 is a more uncertain member , showing a red deficit if it is correctly identified ; alternatively , it could be a blend of lines , in particular [ fe ii ] @xmath111 , 4583 . the strongest argument for including mg i ] in this component is its temporal evolution , which is similar to that of the [ o i ] and [ o iii ] lines ( [ sec_time ] ) . there is no indication of hydrogen or helium with this type of profile . table [ tab3 ] lists the fluxes of the lines that we designate as having line profiles coming from the intermediate component . as with the narrow lines , the intensity ratios of some of these lines provide constraints on the electron density and temperature . * o iii*. the [ o iii ] @xmath364959 and @xmath365007 lines are strongly blended with each other , and the @xmath364959 line with h@xmath97 . a feature at @xmath112 appears in both the @xmath365007 and @xmath364959 lines , and the ratio of line strengths appears to be consistent with the 3:1 ratio expected from their spontaneous emission probabilities . the red edge of the @xmath365007 line extends to @xmath113 , but not further . the @xmath365007 line also has a strong feature to the red at @xmath114 ; however , it does not appear in [ o iii ] @xmath364363 or in other lines of this kinematic component , so there is a question of whether the feature originates from [ o iii ] or is a blend . its rapid drop in the late - time spectra ( see [ sec_time ] ) argues for the blend interpretation . likely candidates for this feature are he i @xmath115 or [ fe ii ] . broad [ o iii ] @xmath364363 emission is present , although the blue side of the line is disturbed by h@xmath98 emission . neglecting the red feature in the @xmath365007 line , we find @xmath116 from the day 1007 keck spectrum . in the far - uv , the o iii ] @xmath117 intercombination lines are among the strongest uv lines . based on the wavelength , it is unlikely that the he ii @xmath118 line contributes much to this feature . we find a reddening - corrected ratio [ o iii ] @xmath119 . * o ii*. the [ o ii ] @xmath637320 , 7330 doublet is less affected by blending and yields a clearer line profile . the feature at @xmath112 is present , but that at @xmath114 is not . the line profile is asymmetric , although the maximum velocity is @xmath120 to both the blue and the red . as we show in [ sec_result ] , the asymmetry is mainly caused by blending with he i @xmath121 . * o i*. the [ o i ] @xmath109 , 6364 doublet is severely affected by the blue wing of h@xmath2 . narrow components of the [ o i ] doublet , as well as [ fe x ] @xmath122 , also complicate the profile . in [ sec_synt ] we show that the line profile is compatible with that of the other oxygen lines . the auroral [ o i ] @xmath123 line is likely to be present , although it may be distorted by the corresponding night - sky line . we therefore only consider its flux as an upper limit . the presence of the o i recombination line at @xmath367774 is of special interest because it should be accompanied by strong h i recombination emission with the same line profile , if the emission is from material with `` cosmic '' abundances , since hydrogen is more abundant than oxygen by a factor of @xmath124 in such material . the absence of the corresponding hydrogen emission suggests that the emitting gas is oxygen rich . the o i recombination line at @xmath368446 , which is expected to have a strength comparable to the @xmath367774 line , is blended with fe ii lines , and is also affected by fluorescence ( [ sec_fluor ] ) . the emission may be from the oxygen - rich ejecta in the supernova and may extend to relatively high velocity because of hydrodynamic instabilities with the overlying layer ( fryxell et al . 1991 ) and/or a high expansion velocity of the oxygen core ( [ sec_nature ] ) . the lines observed in the short - wavelength g160l _ hst _ spectrum are noisy , so there is uncertainty in the line identifications . the line widths , @xmath125 , point to memberships of most of the uv emissions with the intermediate kinematic component . the line profiles of the strongest lines ( c iv , o iii , and si iii / c iii ) are all consistent with flat - topped profiles of width @xmath126 . except for the @xmath127 feature , which has a central peak , the lines appear symmetric within the noise level . the c iv and si iv lines are also affected by interstellar absorption in the galaxy and the host galaxy . by far the strongest line is ly@xmath2 , whose profile is distorted by geocoronal ly@xmath2 , and possibly by n v @xmath128 . the c iv @xmath129 , 1551 doublet is clearly seen with a boxy profile . the same is true for the feature at @xmath361900 , which can be identified as a blend of si iii ] @xmath361892 and c iii ] @xmath631907 , 1909 . the n iii ] @xmath3517441754 lines are clearly present , although their flux is uncertain . the n iv ] @xmath130 line is near the limit of detection . in figure [ fig4a ] we show the density and temperature constraints from all the oxygen - line ratios of the _ hst _ and keck day 1007 spectra together . we have used a reddening correction @xmath28 mag to obtain @xmath131}~ { \lambda}7774)/i({\rm [ o~ii]}~ { \lambda \lambda}7320 , 7330 ) = 0.43 $ ] , [ o iii ] @xmath132 , and [ o iii ] @xmath133 . the [ o i ] @xmath134 ratio is only an upper limit , @xmath135 , and the curve shown in figure [ fig4a ] therefore only marks an upper boundary to the temperature . implicit in this analysis is that the o i and o iii zones are of similar extent , temperature , and electron density . while the lines coming from the o ii and o iii zones are all compatible with @xmath136 k and @xmath137 , the [ o i ] lines have to arise in a lower temperature and/or lower electron density zone . as shown in chevalier & fransson ( 1994 , hereinafter cf94 ) , both these alternatives are quite natural , because the [ o i ] lines are expected to arise in a partially neutral zone of lower temperature ( see , e.g. , cf94 figures 4 and 5 ) . the velocities observed in the intermediate component suggest that this emission is from freely expanding ejecta . the ejecta are photoionized by x - rays from the circumstellar interaction . this process has been modeled by cf94 , and there is general agreement between the lines that are predicted to be brightest and those that are observed here . because of the strong h@xmath2 line , the broad component dominates the optical emission from the supernova . the total luminosity in h@xmath2 was @xmath138 on day 1007 , using @xmath28 mag . the line profile of h@xmath2 is shown in figure [ fig4 ] . there is no obvious distinction of different components in h@xmath2 , as we demonstrate in [ sec_result ] , and no clear evidence for any velocity structure within the line , although a slight emission deficit can be seen on the red side of the line . the fwhm of the line is @xmath139 on day 1007 . the wings are very extended and reach at least 10,000 @xmath0 . table [ tab4 ] gives a list of the lines that have profiles similar to that of h@xmath2 , although we can trace the high - velocity wings above @xmath140 only for h@xmath2 . besides h@xmath2 , the best - defined line profile from this component is he i @xmath141 , which is very similar to h@xmath2 up to @xmath142 , where the line is lost in the noise . the broad component is dominated by lines of h i , he i , mg ii , and fe ii . the interpretation of the mg ii @xmath632796 , 2803 doublet is complicated by absorption features arising in gas in our galaxy and in the host galaxy of sn 1995n . the line peak is at negative velocities , at @xmath143 , which may be a result of resonance - line scattering of the @xmath144 photons by the @xmath145 component , as well as the influence of the interstellar absorption lines . the line appears considerably broader than h@xmath2 , and shows a clear deficit of emission to the red of line center ( fig . [ fig4 ] ) . as discussed in [ sec_fluor ] , the extended profile is likely to be caused by blending with mg i @xmath146 and with the strong fe ii resonance lines in this region of the spectrum , expected to be present on either side of the wings ( multiplets 399 , 391 , 380 at @xmath147 , and multiplet 373 at @xmath148 ) , due to the ly@xmath2 fluorescence . although almost all of the many fe ii lines present are blended with each other , it appears that they belong to the broad kinematic component . the only reasonably isolated fe ii line is @xmath149 , whose profile is well fitted by a scaled h@xmath2 profile . in the optical spectra , the region 40005500 contains a large number of fe ii lines , often seen in supernovae and active galactic nuclei ( agns ; e.g. , filippenko 1989 ; osterbrock 1989 ) . comparing with spectra of narrow - line seyfert 1 galaxies such as i zw 1 ( phillips 1976 , 1977 ) , and the symbiotic star rr tel ( crawford et al . 1999 ) , we have identified members of fe ii multiplets 14 , 27 , 28 , 27 , 42 , 48 , 49 , and 74 ( table [ tab4 ] ) . the lines listed in table [ tab4 ] are those that can be identified with reasonable certainty . most of the other members of the multiplets are blended with strong lines . one should also note that the fluxes of all fe ii lines , with the possible exception of the @xmath149 line , are highly uncertain . in the uv , the plateau between @xmath150 and @xmath151 , except for mg ii , is dominated by fe ii resonance lines . because of the extreme blending of lines , which creates a pseudo - continuum , we do not attempt any complete identification in this spectral region . we simply note that there are some distinct features such as fe ii @xmath152 and fe ii @xmath153 . the latter lines are of special interest , because they arise from cascades from highly excited levels , and are usually considered an indicator of fluorescence . except for this , the uv part of the spectrum is similar to that of i zw 1 ( e.g. , fig . 2d in laor et al . 1997 ) . in the near - ir , it is especially interesting to note that many of the features can only be identified with fe ii emission based on models of ly@xmath2 fluorescence of the transitions ( johansson & jordan 1984 ; penston 1987 ; sigut & pradhan 1998a ) . sigut & pradhan ( 1998b ) , who discuss this process in detail for agns , provide a detailed list of the predicted lines , yielding many identifications for sn 1995n . the best example of ly@xmath2 fluorescence is probably the feature at @xmath154 , which is difficult to explain with other mechanisms . based on other paschen lines , a significant contribution of h i is unlikely , and he i lines are also excluded by similar arguments , as well as by their wavelengths . another candidate , mg ii @xmath155 , also has the wrong wavelengths . in [ sec_fluor ] we show explicitly how the inclusion of these lines improves the spectral fit . finally , we point out the very strong he i @xmath156 line in the vlt day 1769 spectrum , which is the only spectrum covering this region . although there are many lines observed with the high - velocity component , the only straightforward diagnostic of the physical conditions is given by the lack of forbidden - line emission . in the case of the broad - line regions of agns , this is taken to indicate a density @xmath157 ( e.g. , osterbrock 1989 ) . this should be taken with some caution , however , because the gas is only partially ionized by the hard x - rays ( cf94 ) ; high - ionization lines like [ o iii ] @xmath158 and c iii ] @xmath159 are therefore unlikely to be present . the only forbidden lines expected are [ o i ] @xmath160 , which are suppressed relative to permitted lines for @xmath161 . as we discuss in [ sec_result ] , the large h@xmath2/h@xmath97 ratio , as well as the ca ii line ratios , strongly argue that @xmath162 , which is a likely lower limit for the broad component . the emission is presumably from a radius that is similar to the radius of the narrow - line emission , and the gas is thus exposed to a radiation field that is comparable to that of the narrow - line gas . the high density results in an ionization parameter that is a factor @xmath163 smaller than that for the narrow - line gas , or @xmath164 . the result is low ionization of the gas , as observed . the ionization is much less than that in the broad - line regions of agns , where lines like c iv @xmath361549 are observed . to determine abundances as well as deblend the most important lines of the intermediate and broad components , we have calculated synthetic spectra based on the combined uv - optical _ hst_/keck spectrum from [ sec_obs ] . the model we employ is a simple two - component model , with constant physical parameters and abundances in each of the components . the temperature , densities , and abundances are taken as input parameters . for each of the two components we assume a given emissivity profile with velocity . the emissivities of the h and he lines are given by case - b recombination , except for h@xmath2 , which is multiplied by some factor to account for collisions and optical - depth effects . the he i @xmath353889 , 7065 lines are also adjusted by a common factor to account for optical - depth effects ( see osterbrock 1989 , p. 105 ) . all lines from metals are calculated as multilevel atoms for a given temperature and density . the ionization fractions for a given ionization stage ( such as c iii , n iii , o iii , and si iii ) are assumed to be the same . the continuum is due to h i , he i , and he ii bound - free and two - photon continua , consistent with the recombination lines . we do not attempt to include the many fe ii lines present , nor lines from the narrow component . the temperature and ionization are not calculated self - consistently , as in cf94 , so only the relative line fluxes for a given ion , or group of ions , are testing the model . the model should therefore mainly be viewed as a diagnostic tool . for the line profiles we adjust the variation of the emissivity with velocity so that a best fit is obtained ( eq . [ eq_a ] ) . for spherical symmetry and for a velocity @xmath165 , the flux at wavelength @xmath36 is given by @xmath166 ) r \ , dr , \label{eq_a}\ ] ] where @xmath167 is the source function and @xmath168 is the optical depth . in the optically thin limit @xmath169 \approx j(r)$ ] , where @xmath170 is the emissivity . the limits of integration are given by @xmath171 , and @xmath172 , where @xmath173 is the maximum ejecta velocity and @xmath174 is the speed of light . this characterization of the line profile is appropriate for freely expanding ejecta ; free expansion is assumed in our synthetic spectral models . as discussed in 6.1.3 , this assumption may not apply to the broad component . in this case , the line profile should be viewed as an empirical fit that is applied to the set of lines . we find that the profile of the h i , he i , mg ii , and fe ii lines can be well reproduced with an emissivity [ or , if @xmath175 , a source function @xmath167 ] which is constant inside 1000 @xmath0 and decreasing like @xmath177 outside , out to about @xmath178 . for the metal lines , except for mg ii and fe ii , we have used a constant - emissivity shell with @xmath179 constant for @xmath180 . in figure [ fig5 ] we show the spectrum of a model where all elements , including h , he , and metals , have the same density and temperature , @xmath181 and @xmath182 k. the total metallicity is close to solar . note that the temperature , density , and metallicity are only indicative , and depend on the assumed total emissivity , @xmath183 . the fit gives a good general representation of most lines , in terms of both the relative fluxes and line profiles . in particular , many of the profiles that seem fairly complicated at first sight are well reproduced as a result of blends by several lines . good examples are the [ o i ] + h@xmath2 complex , and the [ o iii ] + h@xmath97 feature at 48005000 . the model , however , severely underproduces the strength of the important o i @xmath1847774 recombination line . this could have been anticipated already from figure [ fig4a ] , which shows that the relative [ o iii ] ratios can be reproduced either with a high - temperature , low - density model , as in figure [ fig5 ] , or with a low temperature and a high density . the [ o ii ] @xmath357320 + 7330/o i @xmath1847774 ratio can , however , only be reproduced with the latter temperature / density combination . the relative fluxes of the oxygen and hydrogen / helium lines then require a very high oxygen abundance , unless the o ii and o iii zones are unreasonably large compared to the zone responsible for the h i and he i recombination lines . for the same extent , the relative flux in o i @xmath1847774 and h i lines is given by [ @xmath1857774)@xmath186/ [ j_{\rm eff}({\rm h~i } ) n({\rm h~ii})]$ ] . because @xmath185 7774 ) @xmath187 ) , this shows that @xmath188 and @xmath189 must be of the same order of magnitude . as an alternative , we show a model with a greatly increased metal abundance ( fig . [ fig5b ] ) . although these lines are likely to arise in different ionization zones , we find that we can get an acceptable fit to all lines with a density @xmath190 and a temperature @xmath191 k. all metal lines , except mg ii and fe ii , are assumed to have a shell - like emissivity profile . compared to figure [ fig5 ] we see that , because of the high oxygen abundance , now both [ o ii ] @xmath357320 , 7330 and o i @xmath1847774 , as well as the strengths relative to the h i lines , are well reproduced . in addition , the o i @xmath1848446 recombination line contributes appreciably to the feature at @xmath192 . we also note the good fit to the o i @xmath1847774 line profile , further strengthening this identification . to reproduce the observed h@xmath2/h@xmath97 ratio , the h@xmath2 strength has to be increased by a factor of @xmath193 above recombination . the reddening - corrected ly@xmath2/h@xmath2 ratio is @xmath194 , while the case - b ratio is @xmath195 . the flux of ly@xmath2 is , however , highly uncertain , because the line profile is disturbed by geocoronal ly@xmath2 , in spite of the recession velocity of sn 1995n . as the self - consistent models in cf94 show , this occurs because of collisional excitation in combination with a high optical depth for h@xmath2 in the partially ionized zone . as both the calculations in cf94 and similar models for agns ( e.g. , kwan & krolik 1981 ) show , this requires the presence of hard x - rays from the shocked region in combination with a density @xmath196 . the higher - order balmer lines are well explained by case - b recombination . the helium lines are well reproduced by pure recombination , except for he i @xmath1843889 and he i @xmath1847065 . the former is down by a factor of @xmath197 , and consequently the he i @xmath1847065 line is increased by a factor of @xmath198 , which strongly indicates that the he i @xmath1843889 line is optically thick ( e.g. , osterbrock 1989 , p. 105 ) . the relative ratio of the he i and h i fluxes results in a he ii / h ii abundance of @xmath199 by number . because the he ii and h ii ionization zones are likely to coincide ( e.g. , cf94 ) , the he ii / h ii ratio is probably close to the elemental he / h abundance ratio . strong he i lines are also present in sn 1986j ( rupen et al . 1987 ; leibundgut et al . 1991 ) and at early times in many sne iin ( e.g. , filippenko 1997 ) . the jump at @xmath200 is mainly caused by high - order balmer lines and the balmer continuum . the two - photon continuum dominates in most of the uv , as well as in the optical . the form of the balmer continuum is sensitive to the temperature in the zone where this is formed , and a somewhat better fit is obtained for @xmath201 k. in a one - zone model , however , this temperature gives a worse fit to the metal lines . more detailed models ( cf94 ) show that both temperature and ionization vary within the emitting regions . the ca ii lines in the spectrum are interesting . the ir triplet lines at @xmath358498 , 8542 , 8662 are clearly present , while for the h&k lines and the forbidden @xmath357291.5 , 7323.9 lines we can only set upper limits . fransson & chevalier ( 1989 ) and ferland & persson ( 1989 ) have discussed the relative strengths of these lines as a function of density and temperature . using their results we find that the fact that @xmath2027291.5 , 7323.9)/@xmath2028498 , 8542 , 8662 ) @xmath203 and @xmath17(h&k)/@xmath2028498 , 8542 , 8662 ) @xmath204 means that the density must be @xmath157 . this result is only weakly sensitive to the temperature . the parts of the spectrum which are not well fitted by the model can in most cases be identified with regions of strong fe ii emission . this includes the uv between @xmath205 , some regions in the range 4200 to 5300 , the @xmath1847155 line , and the fluorescence lines above @xmath206 ( see below ) . in figure [ fig6 ] we show the line profiles and fits of the h@xmath2 , [ o i ] , [ o ii ] , and [ o iii ] lines to illustrate the fits for our assumed radial emissivity model . the figure also demonstrates the complex spectrum which results from the blending of the individually fairly simple line profiles . turning now to the other metal lines coming from the intermediate - velocity component , the uv line - intensity ratios require n / c @xmath207 and o / n @xmath208 , by number . the high n / c ratio , in combination with the relatively large o / n ratio , strongly indicates incomplete cno processing in this component . while the profiles of the high - ionization c , n , o , and si lines in the far - uv are consistent with the same shell - like emissivity as the oxygen lines , this emissivity model shows a bad fit of the mg ii @xmath352796 , 2803 line . the peaked profile shows that this line is instead likely to originate in the h and he - rich broad - line component , rather than in the metal - rich component . based on the fe ii ly@xmath2 fluorescence calculation by sigut & pradhan ( 1998a , b ) , we have modeled the near - ir spectrum of sn 1995n . we have used the same model as in figure [ fig5b ] , but now added the fe ii lines , with wavelengths and fluxes from the list of sigut & pradhan . the line profiles were assumed to be the same as that of h@xmath2 . in the upper panel of figure [ fig4b ] we show the resulting spectrum , while in the lower panel we show for comparison the spectrum without fe ii lines . the observations are from the day 1769 vlt spectrum , but the day 716 and 1037 lick spectra show the same features . the day 1007 keck spectrum only extends to @xmath209 . as can be seen , there is a striking improvement with the observed features between 8000 and 9500 when the fe ii lines are included . in particular , the feature at @xmath154 can be identified with the fe ii @xmath359071 , 9128 , 9177 lines , and provides a good fit both in wavelength and profile of the blend . one of the strongest expected fluorescence - pumped fe ii lines is at @xmath368451 , and there is indeed a strong feature at this wavelength . another possible identification for this feature is the recombination line of o i @xmath368446 . this line is normally accompanied by o i @xmath367774 , which is not present in the broad component ( although it is visible in the intermediate component ; see [ sec_result ] ) . the recombination contribution to the @xmath1848446 line is expected to be close to the @xmath367774 line , and is therefore insufficient to explain this strong feature . the @xmath368446 line can be produced by fluorescence through the ly@xmath97 line ( grandi 1980 ) . leibundgut et al . ( 1991 ) identified the o i @xmath368446 line in the spectrum of sn 1986j . the ly@xmath97 o i @xmath368446 fluorescence also results in a strengthened o i @xmath1841300 multiplet . due to its narrowness , however , this is likely to be absorbed by the o i in the host galaxy . in our keck day 1007 spectrum of sn 1995n , the center of the emission feature gives a better match to fe ii @xmath368451 than to o i @xmath368446 . an additional argument in favor of this mechanism is that fe ii @xmath352506 , 2508 , which is indicative of ly@xmath2 pumping , is the strongest fe ii feature in the uv . this has previously been seen in @xmath210 carina , rr tel , and kq pup ( baratta , cassatella , & viotti 1995 ; johansson & jordan 1984 ; redfors & johansson 2000 ) . other lines in the uv expected to be enhanced by fluorescence are the fe ii multiplets 399 , 391 , and 380 at 2850 , multiplet 373 at 2770 , and multiplet 363 at 2530 . these most likely contribute to the broad wings of the mg ii lines , but are too blended to be uniquely identified . ir lines at @xmath211 m are also expected to be strong ( johansson & hamann 1993 ) , but are outside our observed wavelength range . given that the model was intended for conditions in an agn , the fit is surprisingly good . only at @xmath212 are there a few features ( such as at @xmath213 ) which are not well reproduced , but the noise level at these wavelengths is considerable . one should also note that the relative strengths depend on the uncertain atomic data , as well as on the exact ly@xmath2 profile . in particular , the larger line width in sn 1995n compared with those assumed in the agn calculation by sigut & pradhan ( 1998a , b ) model can change the results appreciably . sigut & pradhan assumed a static ly@xmath2 profile with only damping included . an expanding medium means that the ly@xmath2 line can feed many more fluorescence channels , which is likely to enhance the fluorescence effects . finally , we note the strong he i @xmath18410830 line in the day 1769 spectrum . although the region beyond @xmath214 becomes increasingly noisy , this feature clearly stands out above the noise level . compared to the strength expected by pure recombination , it is enhanced by collisional excitation by a factor of @xmath193 , which is common in many objects ( e.g. , osterbrock 1989 , p. 111 ) . in figure [ fig6a ] we show the evolution of the reddening - corrected h@xmath2 luminosity with time on a linear - log scale . we see that the evolution can be well approximated with an exponential decay , @xmath215 , where @xmath216 days . this is much longer than the @xmath217co decay time scale , and we conclude that the circumstellar interaction is responsible for the longevity , although there is no obvious explanation for this particular time dependence . an efficient way of displaying the change of the spectrum with time is to plot the ratio of two spectra as a function of wavelength . in figure [ fig7a ] we show a running average taken over 2 of the day 1799 vlt and day 1007 keck spectra . the presence of two separate components in the spectrum is apparent . while most of the spectrum has decreased by a factor of @xmath218 , the intermediate - component [ o i ] , [ o ii ] , and [ o iii ] lines only decreased by a factor of 23 . the o i @xmath1847774 line shows a similar decrease , while the @xmath1848450 feature follows the h i lines , consistent with it being mainly due to fe ii fluorescence lines . in addition , mg i ] @xmath1844571 belongs to the class of slowly declining lines . relative to h@xmath97 , the peak fluxes of the [ o i ] and [ o ii ] lines increased by a factor of @xmath219 between 1997 and 1999 , while the [ o iii ] line increased by a more modest factor of @xmath220 between these observations . the [ o iii ] @xmath354959 , 5007/@xmath1844363 ratio remained nearly constant at @xmath221 , while the o i @xmath1847774/[o ii ] @xmath357320 , 7330 ratio decreased from @xmath222 to @xmath223 . repeating the analysis in [ sec_interm ] ( fig . [ fig4a ] ) , we find that the temperature in the oxygen - rich gas had not changed much , while the electron density decreased from @xmath224 to @xmath225 . this is less than expected from a simple @xmath226 law , but our analysis assumes that the relative sizes of the o ii and o iii zones remain constant , and that the temperatures and densities in these two zones are similar . neither of these conditions is obviously satisfied . in addition , the supernova may be older than our estimate , and the uncertainties in the relative fluxes are considerable . nevertheless , the behavior is qualitatively that expected for expanding ejecta . in figure [ fig7 ] we show the relative change of h@xmath2 , [ o i ] @xmath356300 , 6364 , [ o ii ] @xmath357320 , 7330 , and [ o iii ] @xmath354959 , 5007 on a velocity scale . while the [ o i ] and [ o ii ] line profiles changed by roughly the same factor over the whole velocity range , @xmath227 to @xmath228 , the red side of the [ o iii ] line decreased by a large factor relative to the blue . the most conceivable reason for the rapid [ o iii ] change is that blending with another line , decreasing similarly to the h i and he i component , causes the apparent decrease in the red wing . the most likely candidates for this are the he i @xmath355015 , 5049 and fe ii lines . the blue side behaved in a manner similar to the [ o i ] and [ o ii ] lines . we therefore conclude that there is little change in the overall profiles of the oxygen lines . when comparing the width of the best - isolated [ o ii ] line , we find a nearly constant blue extension , while the width on the red side decreased by @xmath229 during this period . the evolution of the h@xmath2 line profile differs dramatically from that of the oxygen lines . as shown in the upper - left panel of figure [ fig7 ] , the red wing up to @xmath142 decreased by almost a factor of 4 relative to the blue , with a maximum decrease at @xmath230 . to further illustrate the different evolution of the wings of h@xmath2 , we have reflected the blue component in the spectra from 1996 1999 ( fig . [ fighaevol ] ) , so that the blue and red wings can be compared directly . the fluxes have been normalized to the same peak value . we see that the day 716 lick spectrum ( dashed lines ) showed nearly identical red and blue wings . the day 1007 keck spectrum ( solid lines ) shows a clear evolution in that for a given velocity the red wing decreased faster than the blue wing . this is even more pronounced in the 1999 vlt spectrum , where at @xmath231 the flux of the red wing is only @xmath232% that of the blue wing at the same velocity . equivalently , for a given intensity the red wing has a much smaller extension . for example , at @xmath233% of the peak flux the blue wing extends to @xmath234 , while the red wing only extends to @xmath235 . relative to the peak flux , the blue wing shows a steady increase in flux with time , while the red wing displays the opposite trend . similar evolution is seen for both h@xmath97 and he i @xmath1845876 , although the continuum contribution and line blending make it difficult to compare directly to h@xmath2 . baird et al . ( 1998 ) have presented evidence that the line slowly narrows with time . we can therefore confirm this trend to the latest epochs . the h@xmath2/h@xmath97 ratio in the 1999 spectra decreased somewhat relative to the 1997 and 1996 spectra . this is consistent with the decreasing density during this time interval . the intensity ratio of the he i lines to the balmer lines does not change much . the two he i lines least affected by blending , he i @xmath355876 , 7065 , have almost the same ratio relative to h@xmath97 at all epochs , and are all reproduced with a relative he ii / h ii abundance of @xmath199 by number . together with the similar line profiles , this argues for the fact that the he i and balmer lines arise in the same region . their relative flux should therefore reflect their abundance ratio . in most cases the fluxes of the narrow lines are uncertain , especially in the day 1799 spectrum . for the strongest lines , like [ o iii ] @xmath1845007 and [ fe vii ] @xmath1846087 , the decrease is by a factor of 510 , with no systematic distinction between lines from different elements . although the flux of the [ o iii ] @xmath236 line is especially uncertain due to the underlying broad component , there is an indication for an increase in the [ o iii ] @xmath1845007/@xmath1844363 ratio from @xmath237 to @xmath238 . however , because this ratio depends on both density and temperature , one can not directly translate this into a unique change in either of these parameters . the general scenario for the emission from sn 1995n and related supernovae ( generally type iin ) is interaction of the supernova ejecta with circumstellar gas . our spectra of sn 1995n show that the interaction is complex , but that there is sufficient diagnostic information on the emission components to deduce some of the interaction details . all the emission components are likely to be heated and ionized by the energetic radiation from the interaction region . based on the unresolved line widths , the density of @xmath239 , and the low filling factor , the narrow - line emission can be identified with clumps in the preshock circumstellar gas . the velocity is likely to be @xmath240 , typical of the circumstellar media of red supergiants . the intermediate - velocity component is identified with unshocked supernova ejecta . the maximum velocity in this component suggests either that the reverse shock front has a minimum velocity of @xmath241 , or alternatively that the emission is coming from a component which is photoionized by the shock but only extends in abundance to this velocity . the fact that this component is only seen as dense , oxygen - rich gas makes it natural to identify it as processed gas from the core . assuming that the oxygen core expands at @xmath113 , the typical oxygen density will be @xmath242 where @xmath102 is the filling factor of this component in the core , and we have assumed an oxygen - dominated composition . with @xmath243 , as was the case for sn 1987a ( e.g. , spyromilio & pinto 1991 ) , the density is of the same order as found from the [ o iii ] and similar lines . we note that the density deduced from the models is based on the assumption that the temperatures and densities are the same in the o i , o ii , and o iii zones . a velocity of @xmath241 is high for the oxygen core of a type iip or type iil supernova , but in the range found in type ib / ic supernovae , as well as in the type iib sn 1993j ( e.g. , houck & fransson 1996 ; matheson et al . this is an important result because it is consistent with a picture where the progenitor has lost most of its hydrogen - rich envelope before the explosion . while most of the hydrogen - rich gas comes from gas moving at velocities of @xmath244 , there is a smooth extension to @xmath83 . any model must explain the wide range of velocities for this component . in particular , the fact that there appears to be hydrogen - rich , cool gas moving at up to @xmath245 implies that there are some regions where the shock region extends considerably beyond the reverse shock front . to explain this , one may identify three main alternatives for the broad , hydrogen - rich component . the first possibility is that it comes primarily from unshocked , freely expanding ejecta , extending from low - velocity hydrogen in the core up to the maximum ejecta velocity . a second possibility is that it comes from shocked ejecta gas , behind the reverse shock . the final possibility is that it comes from shocked circumstellar gas , either in the form of clumps or a more uniform medium . our synthetic spectral models in 4 are appropriate for the first possibility . before discussing these possibilities , we note that the ca ii lines , the absence of forbidden lines , and the h@xmath2/h@xmath97 ratio indicate a density @xmath157 for this component ( [ sec_result ] ) . a further independent lower limit to the density can be obtained by considering the state of ionization in this gas . the luminosity in h@xmath2 was @xmath246 on day 1007 ( [ sec_broad ] ) . typically , the conversion efficiency from x - rays to h@xmath2 is @xmath247% ( cf94 ) , indicating a total x - ray luminosity of @xmath248 . this is much more than was observed by _ rosat _ ( fox et al . 2000 ) , but as calculations for sn 1993j showed ( fransson , lundqvist , & chevalier 1996 ) , almost all the x - ray emission can be absorbed in the ejecta and shocked gas , and only the hard x - rays escape the supernova . with this luminosity the ionization parameter on day 1007 is @xmath249 . for a uniform density , hydrogen - dominated envelope expanding at @xmath250 , the density is @xmath251 therefore , if the emission came from unshocked ejecta and @xmath252 , the ionization parameter would be very high , @xmath253 , and one would expect strong emission from ions like [ o iii ] from this component , contrary to the observations . instead , it is dominated by mg ii and fe ii emission , in addition to the h i and he i lines , characteristic of a partially ionized , high - density region ( e.g. , cf94 ) . it is therefore unlikely to come from a uniform - ejecta component , and only if clumping of this component is very high ( @xmath254 ) is an ejecta origin likely . we therefore think the first possibility given above is unlikely . the density in the broad component is much higher than that in the narrow - line component , but the temperature is expected to be only slightly smaller . the conclusion is that the pressure in the gas is relatively high and one possibility is shocked circumstellar or ejecta gas . the gas pressure is @xmath255 k , where @xmath256 is the h density in units of @xmath257 and @xmath258 is the temperature in units of @xmath259 k. the pressure generated by a shock front moving into gas with a density of @xmath260 is @xmath261 k , where @xmath262 is the shock velocity in units of @xmath263 . if the emission is from shocks moving into a medium with a density of this order , the required high density can easily be produced . we can identify two possibilities for such a shocked gas : it could be either clumps caused by instabilities in the shocked ejecta behind the reverse shock , or clumpiness and instabilities of the shocked circumstellar medium . this is similar to the two scenarios discussed by chugai & danziger ( 1994 ) for sn 1988z . _ shocked ejecta : _ this is a version of the scenario discussed for `` normal '' interacting supernovae like sn 1979c or sn 1993j ( cf94 ; fransson et al . the fact that the line profiles in sn 1995n are peaked as opposed to the boxy profiles seen in sn 1993j , however , makes it necessary to modify this scenario . one possibility is that the circumstellar medium is highly asymmetric , as in the picture discussed by blondin , lundqvist , & chevalier ( 1996 ) . indications for a highly asymmetric circumstellar medium are seen in some red supergiants in a superwind phase ( see [ sec_progenitor ] ) . most of the hydrogen emission could in this case come from the cool , dense shell expected behind the reverse shock ( cf94 ) . the shape of the h@xmath2 line profile then reflects the decreasing emission measure as one moves away from the equator . the maximum ejecta velocity is likely to occur at the poles , while the strongest interaction , with most of the energy release , occurs in the high - density circumstellar gas at the equator . the deceleration of the ejecta will therefore be most rapid here . if the hydrogen envelope mass is low , the reverse shock at the equator may rapidly recede to the massive oxygen core . this could qualitatively explain the presence of the high - ionization , oxygen - rich shell observed in the intermediate - component line profiles . the hydrogen - rich gas in the cool shell behind the reverse shock could then be excited by the x - rays from the shock , giving rise to an extended partially ionized zone . a prediction is that the maximum velocity of the shocks , and therefore the line widths , should decrease with time . this general scenario for sn 1995n receives support from observations of the related supernova sn 1986j in ngc 891 , which had a strong h@xmath2 line with fwhm @xmath264 and oxygen lines with widths of several @xmath263 ( leibundgut et al . it was a strong radio source , and vlbi images showed a shell with protrusions extending at least a factor of two beyond the shell radius ( bartel et al . 1989 , 1991 ) . images on different dates revealed expansion ; the outer protrusions are moving at @xmath12 , while the shell is expanding at @xmath241 . these images suggest a picture that is consistent with that deduced from the spectra of sn 1995n . the origin of the protrusions is uncertain , but there are computational simulations that suggest that they could arise from asymmetries or inhomogeneities in the circumstellar medium . blondin et al . ( 1996 ) found that supernova expansion into a circumstellar medium with a density deficit in the polar direction could produce polar protrusions with a radius 24 times that of the main shell . the problem in comparison to the images of sn 1986j is that there are more than two protrusions . jun , jones , & norman ( 1996 ) carried out a simulation of the interaction of a supernova with a clumpy medium and found that the rayleigh - taylor instability that develops where the ejecta are decelerated by the surrounding medium is accentuated by the presence of clumps . the rayleigh - taylor fingers reach the outer shock front . the protrusions in the simulation did not reach the extent required for sn 1986j , but the simulation was run for a limited amount of time and the density of the surrounding medium was uniform . the circumstellar wind case remains to be investigated . _ shocked circumstellar clouds : _ in this scenario , the emission comes from the cool gas behind the slow radiative shocks propagating into the same clumpy gas clouds as are responsible for the narrow - line emission . the blast wave may then propagate in the intercloud medium , while the high - density circumstellar clumps are left behind . the maximum ejecta velocity will decrease only slowly , while the shocked clouds will be accelerated on a time scale which depends on the column density of the clouds . as simulations show ( e.g. , klein , mckee , & colella 1994 ; jun et al . 1996 ) , however , the cloud may gradually be evaporated , and finally be dissolved and merge with the ejecta gas . in this scenario one expects a smooth distribution of gas , with velocities ranging from newly shocked clouds with low velocity , or clouds with very high column density , to gas which has been accelerated to the ejecta velocity . this agrees qualitatively with the observed h@xmath2 line profile . the observed gradual increase of the flux in the blue wing with time is also consistent with this scenario . the decrease in the red wing , however , is not naturally explained , unless it is caused by dust absorption , either from pre - existing dust in the circumstellar gas , or from newly formed dust in the ejecta , as discussed for sn 1998s by gerardy et al . ( 2000 ) . an important question is the location of the energy source for the excitation of the gas , and there are several possibilities . first , the source could be the shock separating the high - velocity hydrogen ejecta from the circumstellar gas . in the uniform - medium case this will be at the highest velocity and also maximum ejecta radius . this is the model discussed by cf94 , and applied to sn 1993j by fransson et al . however , it is difficult to understand the presence of the high - ionization , oxygen - rich shell at only @xmath265 , since most of the x - rays will be absorbed by the unprocessed ejecta between the shock and the oxygen core . in the clumped circumstellar medium case the bow shocks may be distributed over a large radius , and immersed in the expanding ejecta , depending on the time of shocking and their column densities . the oxygen - shell emission is again difficult to explain . it may , however , explain the distributed nature of the hydrogen emission . some clues to realistic scenarios for the structure of the circumstellar medium can be obtained from observations of galactic supergiants in very advanced evolutionary stages . while most red supergiants , like @xmath2 ori , have fairly modest winds with mass - loss rates @xmath266 , there is a handful of massive stars having more extreme mass - loss rates , @xmath267 , often referred to as `` superwinds . '' van loon et al . ( 1999 ) find that two out of eight red supergiants in an lmc sample have mass - loss rates @xmath268 , while the others have moderate mass - loss rates @xmath266 . the best - studied superwind cases are vy cma and irc + 10420 . the nature of the superwind phase is not yet clear , but can be related either to a specific evolutionary phase , occurring for most red supergiants , or to a phase only occurring for a specific mass range . a hint that the latter is the case comes from the fact that both vy cma and irc + 10420 are very luminous stars , with @xmath269 . from evolutionary models wittkowski , langer , & weigelt ( 1998 ) estimate a zero - age main sequence ( zams ) mass of 3040 @xmath270 for vy cma . the mass - loss rate is determined to be @xmath271 for a wind velocity of @xmath272 ( danchi et al . 1994 ) . several authors have found an asymmetric distribution of the circumstellar gas and dust around vy cma . from _ hst _ imaging kastner & weintraub ( 1998 ) find an asymmetric core region with wavelength dependent fwhm @xmath273 cm in the optical . the star itself has a radius of @xmath274 cm . outside of this is an asymmetric reflection nebula with an extent of @xmath275 cm , with the core close to its edge . based on symmetry arguments , kastner & weintraub speculate that this is only one half of a bipolar nebula , with the other half obscured by an opaque equatorial dust disk . the same conclusion is reached by monnier et al . ( 1999 ) from high - resolution observations of the core and extended structure . the supergiant irc + 10420 is similar to vy cma in terms of mass - loss rate and mass , but probably in an even later evolutionary state , intermediate between a red supergiant and a wolf - rayet star . there is also evidence for a disk - like geometry in this case , based on line profiles ( jones et al . 1993 ) and on _ hst _ and ir imaging ( humphreys et al . 1997 ) . wittkowski et al . ( 1998 ) propose that the asymmetric mass loss arises as a result of rapid rotation in the phase preceding the evolution off the hayashi line when the major part of the envelope is lost . vy cma is now estimated to have a mass of @xmath276 , meaning that it has lost @xmath277 to the circumstellar medium . the helium fraction at this phase is @xmath278 by mass . this is lower than , although qualitatively consistent with , our results for sn 1995n , but is obviously sensitive to mixing prescriptions , the exact evolutionary status , and other factors . as an alternative , soker ( 2000 ) proposes that the asymmetric distribution is caused by the concentration of magnetic spots in the equatorial plane , which through the lower photospheric temperature causes dust formation preferentially at the equator . based on these cases , it seems probable that sn 1995n , and other type iin supernovae , arise from progenitors similar to vy cma or irc + 10420 . this is consistent with the high mass - loss rates needed to explain the strong interaction , as reflected in the total luminosity . a low hydrogen envelope mass is needed to explain the large oxygen core velocity . the observed high he / h ratio is also consistent with high mass loss . an asymmetric circumstellar distribution will result in faster expansion of the supernova ejecta in the polar directions , while it will be slowed down by the equatorial disk or torus as modeled by blondin et al . the radius of the inner boundary of the dusty disk in vy cma is estimated to be @xmath279 cm ( monnier et al . if the density in the disk is large enough , the radius of the inner edge would not be much affected by the supernova ejecta . the relative size of the ejecta and obscuring disk would therefore increase with time . if the disk is opaque due to dust , an increasingly large fraction of the ejecta on the far side of the supernova would be obscured . this could explain the fact that the red side of the h@xmath2 line decreased considerably faster than the blue side . the exact temporal evolution depends on the degree of asymmetry of the ejecta . evidence for dusty environments comes from several type ii supernovae . especially interesting are the observations of sn 1998s , where gerardy et al . ( 2000 ) find strong evidence for dust emission . as gerardy et al . discuss , although this dust could in principle be formed in the ejecta , perhaps a more likely interpretation is pre - existing circumstellar dust , heated by the supernova ( see also fassia et al . 2001 ) . from observations of h@xmath280o masers in vy cma , richards , yates , & cohen ( 1998 ) estimate a density of @xmath281 in the masing regions , while other species give a density of @xmath282 , indicating a considerable clumpiness of the circumstellar medium . one then speculates that these masing regions could be related to the high - density clumps needed to decelerate the supernova blast wave . finally , the small fraction of supernovae in this class , @xmath283% ( cappellaro et al . 1993 ) , is at least qualitatively consistent with them coming from a fairly small fraction of the high - mass stars . late - time spectral observations of sn 1995n are presented , covering the uv to near - ir regions . three kinematic components are found , which most likely are powered by x - rays from the interaction of the ejecta and the circumstellar medium of the progenitor . spectral modeling shows evidence of an increased helium abundance , as well as cno burning . the first evidence for ly@xmath2 pumped fluorescence of fe ii lines in supernovae is found . it is likely that other supernovae interacting with their circumstellar media show similar signatures , especially in the near - ir . while the narrow lines come from unshocked circumstellar gas and the intermediate - velocity component from processed ejecta , the high - velocity unprocessed gas has a more uncertain origin . different possibilities based on either clumpy or asymmetric circumstellar media are discussed , without strongly favoring either of these possibilities . the temporal evolution of the processed and unprocessed gas is distinctly different , again emphasizing the different origins of these components . we propose that the progenitors of the narrow - line supernovae ( type iin ) are similar to red supergiants in their superwind phase , as most of the hydrogen - rich gas is expelled in the last @xmath284 years before explosion . this explains the high oxygen core velocity . we also present evidence that the circumstellar medium is either asymmetric or clumpy . we are grateful to anil pradhan and aron sigut for sending us their fluorescence results in detailed form . financial support for this work was provided to a.v.f.s group by nsf grants ast-9417213 and ast-9987438 , by the guggenheim foundation , and by nasa through grants go-6043 , go-6584 , and go-8602 from the space telescope science institute , which is operated by aura , inc . , under nasa contract nas 5 - 26555 . support was also provided to r.a.c . by nasa grant nag5 - 8232 , and to c.f.s group by the swedish space board and swedish research council . some of the data presented herein were obtained at the w. m. keck observatory , which is operated as a scientific partnership among the california institute of technology , the university of california , and nasa ; the observatory was made possible by the generous financial support of the w. m. keck foundation . we thank the staffs at the _ hst _ , eso , lick , and keck observatories for their overall assistance , as well as andrea gilbert , jeff newman , and adam riess for their help with some of the observations . lrlcc 24 may 1995 & 321 & lick 3 m & 3120 10490 & 611 + 67 july 1995 & 364365 & lick 3 m & 3120 10490 & 611 + 1 august 1995 & 390 & lick 3 m & 4260 7040 & 6 + 22 june 1996 & 716 & lick 3 m & 3120 9880 & 6 + 6 february 1997 & 943 & hst & 1140 4780 & 27 + 11 april 1997 & 1007 & keck i & 3910 8880 & 10 + 16 april 1997 & 1012 & lick 3 m & 3160 9870 & 6 + 11 may 1997 & 1037 & lick 3 m & 3120 9880 & 611 + 13 may 1999 & 1769 & vlt & 3635 11680 & 13 + 12 june 1999 & 1799 & vlt & 3908 9187 & 5 + lccrrrrl [ fe xi ] & 7891.8 & 7891.9 & & & 1.2 & 0.12 & + ? & & 7610.3 & & & 0.81 & 0.17 & + [ ar iii ] & 7135.8 & 7135.7 & & & 0.26 & 0.12 & + he i ( n ) & 7065.3 & 7065.5 & & & & 0.11 & + [ ar v ] & 7005.7 & 7005.0 & & & 0.21 & & broad ? + h@xmath2 & 6562.8 & 6563.8 & & & @xmath285 2.1 & 0.78 & + [ fe x ] & 6374.5 & 6375.5 & 1.8 & & 1.3 & & + [ o i ] & 6363.8 & 6363.0 & & & 0.47 & & + [ o i ] & 6300.3 & 6301.0 & 3.0 & & 3.0 & 0.22 & + [ fe vii ] & 6087.0 & 6086.8 & 3.7 & & 2.8 & 0.60 & + [ n ii ] & 5754.6 & 5755.5 & & & 2.2 & & broad ? + [ fe vii ] & 5720.7 & 5721.9 & 3.0 & & 1.6 & 0.7 & + [ fe vi ] & 5677.0 & 5677.5 & & & 0.50 & & + [ fe vii]+[fe vi ] & 5276.4 + 5277.8 & 5278.1 & 6.5 & & 1.9 & & bl . + [ fe vi ] & 5176.0 & 5176.2 & & & 3.3 & 0.11 & bl . w. fe ii , fe vii + [ fe vii ] & 5158.9 & 5158.7 & & & & 0.20 & bl . w. fe ii , fe vi + [ fe vi ] & 5145.8 & 5148.2 & & & & 0.19 & bl . w. fe ii , fe vii + [ o iii ] & 5006.8 & 5006.7 & 44 . & & 26 . & 3.5 & + [ o iii ] & 4959.0 & 4958.6 & & & 8.5 & 0.62 & + h@xmath97 & 4861.3 & 4861.6 & & & @xmath2853.6 & 0.51 & + [ ne iv ] & 4725.7 & 4725.4 & & 1.8 & 0.88 & & + [ ne iv ] & 4714.2 - 4715.7 & 4714.8 & & 0.43 ? & 0.41 & & + he ii & 4686 & 4686.0 & & 1.5 & 0.97 & 0.11 & + [ o iii ] & 4363.2 & 4364.0 & 6.5 & 8.3 & 6.4 & 0.53 & + h@xmath98 & 4340.5 & 4341.0 & & & & 0.27 & + [ fe iv ] & 4198.2 & 4198.1 & & & 0.07 & & + ? & & 4187.7 & & & 0.05 & & + [ fe v ] & 4180.6 & 4179.8 & & & 2.1 & & bl . w. he i 4169 + [ fe v ] & 4071.2 & 4070.9 & 3.0 & & 1.3 & 0.11 & + [ fe iii ] ? & 4046.2 & 4045.0 & & & 0.04 & & ident . ? + he i & 4009.3 & 4009.0 & & & 0.10 & & + [ ne iii ] & 3967.5 & 3964.8 & & 3.4 & & & bl . w. he i 3964.1 + [ ne iii ] & 3868.8 & 3868.1 & 20 . & 9.5 & 6.8 & & + [ fe v]+[fe vii ] & 3757.6 + 3758.9 & 3757.7 & 6.3 & 3.4 & 9.0 & & bl . + [ fe vii ] & 3586.3 & 3585.3 & & 2.1 & & & + [ ne v ] & 3425.9 & 3419.0 & 15 . & & & & + [ ne iii ] & 3342.4 & 3343.6 & & 3.8 & & & + lccrrrrl o i & 7771.9 - 7775.3 & 7773.5 & 1.4 & & 2.0 & 1.1 & + [ o ii ] & 7319.9 , 7330.2 & 7326.5 & 5.3 & & 4.6 & 5.4 & bl . w. he i , ca ii + [ o i ] & 6300.3 , 6363.8 & 6335.2 & 4.5 & & 3.2 & 4.4 & + [ o iii ] & 4959.0 , 5006.8 & 5006.7 & 54 . & & 39 . & 9.5 & + mg i ] & 4571.1 & 4574.1 & 1.8 & 1.3 & 1.8 & 1.6 & ( + 4583 ) + [ o iii ] & 4363.2 & 4380.7 & & 5.3 & 4.7 & 1.2 & bl . w. h@xmath97 + mg i & 2852.1 & 2862.3 & & 2.5 & & & bl . w. mg ii & fe ii + si iii ] , c iii ] & 1882.7 , 1892.0 , 1906.7 , 1909.6 & 1899.3 & & 21 . & & & bl . + n iii ] & 1746.8 - 1753.4 & 1752.9 & & 2.1 & & & + o iii ] & 1660.8 , 1666.1 & 1658.8 & & 9.9 & & & + c iv & 1548.9 , 1550.8 & 1548.0 & & 7.6 & & & + n iv ] & 1483.3 - 1487.9 & 1492.5 & & 2.2 & & & flux uncertain + si iv , o iv ] & 1393.8 , 1402.8 , 1397.2 - 1407.4 & 1406.3 & & 6.7 & & & bl . + si ii , o i & 1304.4 , 1309.3 , 1302.2 - 1304.9 & 1310.0 & & 2.1 & & & ident . uncertain + & + lccrrrrl he i & 10830 & 10827.6 & & & & 6.0 & + p@xmath286 , he i,[s iii ] ? & 9545.7 , 9529.3 , 9532.0 & 9544.0 & 3.0 & & & 0.70 & + p@xmath287 & 9229 & & & & & & + fe ii & 9196.9 - 9218.2 & & & & & 0.18 & bl . + fe ii+he i & 9175.9 , 9078.1 , 9174.5 & 9177.5 & 7.0 & & & 1.0&bl . + fe ii & 9122.9 , 9132.4 & 9133.8 & 2.0 & & & 0.39 & bl . + fe ii & 9070.5 , 9077.4 & 9071 & 0.54 & & & 0.07 & bl . + p@xmath210 & 9014 & & & & & 0.07 & bl . + ca ii & 8662.1 & 8668.2 & & & 0.27 & & bl . h i , ca ii + h i & 8598.3 & 8601.7 & & & 0.07 & & bl . + h i , ca ii & 8545.4 + 8542.1 & 8545.9 & 0.20 & & 0.22 & & bl . + ca ii & 8498.0 & 8496.9 & & & 0.80 & & bl . + o i + fe ii & 8446.5 + 8451.0 & 8448.1 & 8.4 & & 3.5 & 0.88 & bl . + he i & 8285.4 & 8287.3 & 0.16 & & 0.12 & 0.05 & + fe ii & 8228.9 & 8227.7 & 0.72 & & 0.53 & 0.11 & + he i & 8155.8 & 8156.3 & & & 0.02 & 0.02 & + he i & 7971.6 & 7970.7 & 0.08 & & 0.22 & 0.05 & + ? & & 7892.0 & & & 0.14 & & narrow ? + ? & & 7610.4 & & & 0.07 & & narrow ? + he i & 7281.0 & 7277.7 & & & 0.46 & & bl . w. [ o ii ] + fe ii & 7155.1 , 7172.0 & 7156.7 & & & 0.21 & & mult . 14 + he i & 7065.3 & 7063.5 & 2.0 & & 1.2 & 0.40 & + he i & 6678.2 & 6675.6 & 0.60 & & 0.33 & & + h@xmath2 & 6562.8 & 6562.4 & 330 . & & 160 . & 30 . & + [ o i ] & 6300.3 , 6363.8 & 6335.2 & 4.5 & & 3.2 & 4.4 & + fe ii & 6148 , 6149 & 6146.3 & & & 0.19 & & mult . 74 + he i & 5875.7 & 5876.4 & 3.8 & & 2.5 & 0.88 & + [ n ii ] & 5754.6 & 5753.5 & 0.39 & & 0.27 & 0.11 & + fe ii & 5534.9 & 5533.2 & & & 0.13 & & mult . 55 , bl . w. [ o i ] 5577 + fe ii & 5414.1 , 5425.3 & 5421.1 & & & 0.18 & & mult . 48 + 49 + fe ii & 5362.9 & 5367.1 & & & 0.22 & & mult . 48 + fe ii & 5316.2 & 5317.9 & & & 0.54 & & mult . 49 + fe ii & 5265 , 5276.0 & 5273.9 & & & 0.10 & & mult . 48 + 49 + fe ii & 5169.0 & 5169.5 & 2.6 & & 1.0 & & mult . w. [ fe vi ] , [ fe vii ] + fe ii & 4629.3 & 4627.6 & & & 0.25 & & mult . 37 + h@xmath97 & 4861.3 & 4861.6 & 8.8 & & 4.3 & 1.2 & + he ii & 4685.7 & 4686.0 & 0.77 & 0.24 & 0.26 & & + he i & 4471.5 & 4470.7 & 1.1&0.63 & 0.68 & 0.26 & + [ o iii ] & 4363.2 & 4380.7 & & 5.4 & 4.7 & 1.2&bl . w. h@xmath97 + h@xmath98 & 4340.5 & 4341.0 & & 4.6 & 3.4 & 1.1&bl . w. [ o iii ] + fe ii & 4233.2 & 4238.9 & & & 0.48 & & mult . 27 + fe ii & 4178.9 & 4180.3 & & & 0.22 & & mult . w. [ fe v ] 4180 + he i & 4120.8 & 4122.4 & & & ? & & + h@xmath288 & 4101.7 & 4102.3 & 1.7 & & 1.6 & 0.78 & + [ ne iii ] , h@xmath286 & 3967.5 , 3970.0 & 3967.8 & 2.2 & & 0.93 & 0.30 & blend w. fe v , h@xmath286 + h@xmath287 & 3889.0 & & & & & & + he i & 3888.6 & & 1.5 & & & & + h@xmath210 & 3835.4 & & 1.2&1.1 & & & + [ ne v ] & 3426 & 3447 & 1.8 & & & & + ? & ? , & 3075.4 & & 0.58 & & & uncert . + ? & ? , & 3001.1 & & 0.35 & & & uncert . + o iii & ? , & 2983.5 & & 0.36 & & & uncert . + fe ii , [ mg v ] & 2926.6 , 2928.3 & 2925.8 & & 0.23 & & & mult . 60 , ident . + mg ii & 2795.5 , 2802.7 & 2793.3 & & 130 . & & & + he i + al ii ] & 2663.4 + 2669.2 & 2666.3 & & 0.71 & & & + fe ii & 2614.9 - 2631.0 & 2617.7 & & 3.9 & & & mult . 1 + 171 + fe ii & 2585.9 -2593.7 & 2592.0 & & 0.85 & & & mult . 1 + 64 + fe ii & 2562.5 , 2563.5 & 2561.3 & & 0.35 & & & mult . 64 + fe ii & 2506.4 - 2508.3 & 2505.6 & & 2.0 & & & + ly@xmath2 & 1215 & 1210.9 & & 300 . & & + .

the mechanical response of materials with discrete micro- or meso - structures such as 3d - printed structures , woven textiles , paper , or foams can be modelled with dissipative lattice networks , cf . e.g. @xcite . the main advantage of these models consists in their conceptual simplicity , because lattice springs or beams can be identified as individual fibres or yarns of the underlying structure . therefore , material parameters such as young s or hardening moduli , and constitutive damage or plasticity laws can be determined in a relatively straightforward manner . furthermore , lattice networks incorporate large deformations and yarn reorientations rather easily compared to phenomenological continuum models , see e.g. @xcite . for heterogeneous cohesive - frictional materials such as concrete , lattice networks are capable of capturing distributed microcracking , the heterogeneity at the microscale , and size effects ; applications and further discussions can be found , e.g. , in @xcite . the main drawback of lattice structures is their considerable computational cost which may well be prohibitive for engineering applications , in which the lattice spacing is generally several orders of magnitude smaller than the problem size . ergo , multiscale , or reduced - order modelling methods are required to tackle realistic applications . a reduced - order modelling method with a concurrent multiscale character is the quasicontinuum ( qc ) method . it was originally designed for conservative atomistic systems at the nano - scale level , see @xcite for its initial formulation and @xcite for various extensions . this approach was further generalized to tackle materials at the meso - scale level by introducing dissipation along with internal variables in a virtual - power - based format by @xcite and in a variational format by @xcite . in its essence , the qc methodology resolves the underlying lattice only in small regions of interest , whereas it coarsens it elsewhere . this results in a considerable reduction of the number of degrees of freedom ( dofs ) , internal variables , and effort to construct the governing equations . the problem of interest in this contribution is the combination of the qc method and crack propagation in damageable lattice structures , as depicted in fig . [ fig : motivation ] . because most of the dissipation and fibre reorientation occurs near the crack tip , this region must be fully resolved . to this end , a mesh indicator that follows the evolution of the crack tip needs to be provided . such an approach typically leaves a trail of the fully resolved region in the wake of the crack if no coarsening is included , cf . it may be clear that , apart from allowing the crack to open up , these fully resolved regions do not contribute to the overall physics and accuracy . it is therefore desirable to coarsen them . however , this coarsening should take into account the effect which the crack has on the local , coarse - scale kinematics of the problem i.e . it should allow an arbitrary amount of crack opening without any mechanical resistance . the aim of this contribution is to provide a refinement _ and _ coarsening strategy that allows for the removal of nodes in the wake of the crack tip to increase the efficiency of the variational qc method . because the damage in the crack wake is highly localized , it can be treated as a `` discontinuity '' , although the system is obviously discrete . it is reasonable , therefore , to introduce techniques known from the extended finite element method ( xfem ) , @xcite , or generalized finite element method ( gfem ) , @xcite , that rely on the partition of unity ( pu ) concept , @xcite . a certain similarity to the particle partition of unity method ( ppum ) , @xcite , can be also observed . namely , the space of interpolation functions is enriched to also include `` jumps '' in kinematic as well as in internal variables . as will be explained below , and demonstrated in the examples section , such an enrichment provides an accurate and efficient framework to model cracks in damageable lattices . in what follows , this approach will be referred to as the _ extended qc _ ( in analogy to continuum systems ) , or _ x - qc _ for short . several previous works have focused on adaptivity and coarsening in qc methodologies as well , but these mainly dealt with conservative , atomistic lattice models , cf . an xfem - type qc framework was reported in @xcite , which treated atomistic models within the scope of molecular dynamics and the bridging domain method ( concurrent coupling between the atomistic and continuum regions ) . in contrast to these previous works , this contribution focuses on structural lattice networks at the meso - scale level . moreover , we employ the _ dissipative _ variational qc with damage , including a _ fully non - local _ formulation . the paper is organized as follows . in section [ sect : varform ] , we briefly recall the theory of rate - independent systems on which the variational qc is built . for simplicity , only time - discrete versions of all equations are provided . having specified the theoretical background , we can introduce the model at hand , its geometry , state variables , and energies driving the evolution . the main developments of this contribution , i.e. the interpolation and summation qc steps revisited from the point of view of lattice structures with localized damage , are discussed in section [ sect : qc ] . namely , the enrichment functions together with the required changes in the summation rules are introduced in detail . the numerical solution strategy is addressed in section [ sect : sol ] . two examples demonstrate the additional efficiency of the proposed method . the results show that the coarsening strategy , in particular , allows one to reduce the number of dofs to approximately 25 75% of that of an adaptive qc without coarsening , with no significant increase in error . the number of dofs reduces to 1 @xmath0 of that of the direct numerical simulations ( dns ) . the corresponding computing times are reduced by a factor of 4 20 , depending on the particular example . the paper closes with a summary and conclusions in section [ sect : conclusion ] . in this section , the basic principles of the variational formulation of rate - independent systems in the context of lattice networks are recalled . for the sake of brevity , we limit ourselves to the time - discrete setting ; the general theory is discussed in @xcite , whereas applications to continuous systems can be found , e.g. , in @xcite . applications to lattice networks and qc are presented in @xcite for isotropic hardening plasticity and localized damage , respectively . an admissible configuration of a rate - independent system of interest is specified by a state variable @xmath1 , where @xmath2 stores the kinematic variables and @xmath3 the internal variables . all admissible configurations are specified by the state space @xmath4 , @xmath5 , @xmath6 , which incorporates , e.g. , prescribed dirichlet boundary conditions or constraints on damage variables . the energetic solution @xmath7 is determined using the following incremental minimization problem at time @xmath8 @xmath9 with an initial condition @xmath10 , where @xmath11 describes a discrete time horizon @xmath12 $ ] , @xmath13 denotes an incremental energy of the form @xmath14 and the inclusion sign @xmath15 indicates that the potential @xmath13 is in general nonsmooth or may have multiple minima . the incremental energy consists of the total ( helmholtz type ) energy @xmath16 \times \mathscr{q } \rightarrow \mathbb{r}$ ] , and the dissipation distance @xmath17 . the dissipation distance @xmath18 measures the minimum dissipation by a continuous transition between two consecutive states @xmath19 and @xmath20 ; for further details see @xcite , section 3.2 . the potential energy reads @xmath21 where @xmath22 \times \mathscr{q } \rightarrow \mathbb{r}$ ] is the internally stored energy , @xmath23 \rightarrow \mathscr{r}^*$ ] represents the column matrix with external loads , and @xmath24 is the space dual to @xmath25 . each time the minimization problem is solved in this contribution , a local minimum that satisfies the energy balance @xmath26 is searched . the energy balance equates the internally stored energy plus the _ dissipated energy _ @xmath27 with the work performed by the external forces @xmath28^\mathsf{t}[{{\boldsymbol{r}}}(t_l)-{{\boldsymbol{r}}}(t_{l-1 } ) ] . \label{sect : varform : eq:2}\ ] ] in eqs . , , and , the symbol @xmath29 indicates the dependence on @xmath30 for @xmath31 . in this section , we briefly specify the geometry , underlying lattice , and the kinematic as well as the internal variables . our considerations below are limited to 2d lattices for notational simplicity , and lattice nodes are called `` atoms '' . first , the reader is referred to fig . [ subsect : geometry : fig:1 ] , where a sketch of the kinematic as well as internal variables is presented . as indicated , we assume an x - braced lattice with the nearest - neighbour interactions . the original location vector of atom @xmath32 is denoted as @xmath33 . for all @xmath34 atoms ( collected in an index set @xmath35 ) , they are stored in the column @xmath36^\mathsf{t}$ ] , @xmath37 . by analogy , the current location vectors @xmath38 are stored in @xmath39^\mathsf{t}$ ] . the entire system contains @xmath40 interactions stored in an index set @xmath41 . each interaction @xmath42 that connects atoms @xmath32 and @xmath43 is of the length @xmath44 where @xmath45 denotes the euclidean norm . for _ damaging _ interactions , cf . [ subsect : geometry : fig:1b ] , the only internal variable is the damage variable @xmath46 $ ] . all of these damage variables are stored in the column @xmath47 , which is hence of length @xmath40 . the internally stored energy associated with damaging lattices reads @xmath48 , \label{subsect : lndamage : eq:1}\ ] ] where @xmath49 denotes the set of the nearest neighbours associated with atom @xmath32 . the elastic energy of a bond @xmath50 is reduced by the damage variable @xmath51 only if the bond is loaded in tension to better reflect physical behaviour of damageable materials . in eq . , we have introduced @xmath52 , @xmath53 , and assumed @xmath54 . the dissipation distance for bond @xmath55 for two consecutive states @xmath56 and @xmath57 is defined as @xmath58 and the total dissipation distance is specified as @xmath59 in definition , @xmath60 is the energy dissipation of a single bond during a unidirectional damage process up to a damage level @xmath61 . this function therefore increases from @xmath62 to @xmath63 , where @xmath64 is the energy dissipated at complete failure . the incremental interaction energy , @xmath65 , and incremental site energy , @xmath66 , then read : [ subsect : lndamage : eq:6 ] @xmath67 in this section we specify the two steps of the standard qc methodology , interpolation and summation , as well as additional procedures that must be adopted to extend it to an efficient description for localized damage . we start with a geometric crack description for lattice systems in section [ subsect : crack ] , proceeding to the interpolation step in section [ subsect : interpolation ] , where the standard framework as well as the mesh refinement and the mesh coarsening are treated . the first part closes with the description of the enrichment functions . the second part is devoted to summation , starting with the standard framework in section [ subsect : summation ] , which is later extended to incorporate also the enrichment functions . finally , the complete x - qc procedure is summarized in section [ subsect : xqc ] in terms of a general algorithm , and energy implications are discussed in section [ subsect : enimplications ] in order to allow for proper verification of the energy equality . in what follows , a crack is defined as an ordered set of points collected in a set @xmath68 , cf . also @xcite : @xmath69 the set @xmath70 collects all crack wake atoms , whereas @xmath68 stores the midpoints of all fully damaged bonds . upon ordering , the set @xmath68 defines an oriented polygon @xmath71 ( a curve parametrized by a curvilinear coordinate @xmath72 ) geometrically representing the crack . the threshold @xmath73 should be close enough to @xmath74 , e.g. @xmath75 . naturally , any duplicates in are eliminated , and the ordering may be implemented in various ways , e.g. by sorting the points in @xmath68 with respect to their @xmath76-coordinates . for future use , it is convenient to introduce the signed distance function @xmath77 , cf . e.g. @xcite , section 3 . its definition reads @xmath78 where @xmath79 denotes the column storing the unit normal to the crack @xmath80 at the closest point @xmath81 . for instance , @xmath79 can be defined such that @xmath82 , where @xmath83 is the tangent vector to @xmath80 at @xmath81 , and @xmath84 is the unit basis vector pointing in the @xmath85-direction of the adopted coordinate system . if @xmath81 is situated at a kink of @xmath80 , entire cone of normals must be considered ; then , the sign of @xmath86 is positive if the vector @xmath87 belongs to the cone of normals and negative otherwise . the function @xmath77 is therefore positive on one side of the crack and negative on the other . because the crack tip is always located inside the fully refined region , no second level set function is required to describe the crack tip , cf . @xcite , section 3.2 . for a pictorial representation of the introduced quantities see fig . [ subsect : crack : fig:1 ] . , definition ) are depicted by red lines ( other interactions are omitted for clarity ) and set @xmath68 is depicted using green dots . the signed distance function @xmath77 is represented by contour lines , and the crack polygon @xmath80 is presented as a green line . atom sites are shown as black dots . ] in the four following subsections , the interpolation step of the qc method and its effect on the incremental energy of the full system ( @xmath13 in ) are detailed . interpolation introduces to the minimization problem the following equality constraints : @xmath88 where @xmath89 denotes a column of generalized dofs located in a kinematically constrained subspace @xmath90 of the fully dimensional space @xmath91 . after substitution of eq . , the incremental energy becomes a function of @xmath89 , i.e. @xmath92 which can be minimized over the subspace @xmath90 . this reduces the computational effort if @xmath93 . in qc approaches , the generalised dofs @xmath89 are chosen as the dofs of a small set of atoms , the so - called _ repatoms_. usually , repatoms are chosen as nodes of a mesh @xmath94 that triangulates the domain @xmath95 , and the interpolation is implemented via piecewise affine ( @xmath96 ) fe shape functions constructed between the repatoms . consequently , @xmath97 contains standard fem shape function evaluations at all atom positions , cf . e.g. @xcite . note that higher - order polynomial approximations inside elements can be adopted as well , see e.g. @xcite , @xcite , or @xcite . in order to distinguish classical qc approach from its extended version presented below , we use the following notation for the classical qc : all @xmath98 repatoms are collected in an index set @xmath99 , whereas their admissible positions @xmath100 , @xmath101 , are stored in the column @xmath102 by analogy to @xmath2 . associated interpolation matrix is denoted @xmath103 . in this contribution , the mesh refinement indicator of @xcite , is adopted with minor changes . it proceeds as follows : a triangle @xmath104 of the current triangulation @xmath105 is endowed with a set of sampling interactions @xmath106 specified as @xmath107 then , the following energy condition is introduced @xmath108 which specifies interactions that are likely to be damaged . in eq . , @xmath50 denotes the pair potential , @xmath109 a refinement safety parameter , and @xmath110 the stored threshold energy at which the internal variable starts to evolve . the mesh refinement indicator then reads @xmath111 having specified a set of triangles marked for refinement , @xmath112 , the current triangulation @xmath105 must be refined . to this end , the backward - logest - edge - bisection algorithm by @xcite is used . before coarsening the current triangulation @xmath105 , first an indicator is presented that marks which triangles need to be coarsened . in analogy to eq . we introduce the following energy condition @xmath113 where the compressive part is now included as well . the rationale behind condition is that in the crack wake the stress is low , and hence also the energy . the mesh coarsening indicator then reads @xmath114 verifying condition for all elements yields a set of triangles marked for coarsening , @xmath115 . note that because the damage remains localized , it yields steep spatial ( interaction ) energy variations in the vicinity of the crack tip . for instance , in fig . [ subsect : nodeprotection : fig:1 ] , the normalized interaction energy can be seen to drop from @xmath116 to @xmath117 over just two to three lattice spacings . from this it may be clear that if the values of @xmath118 and @xmath109 are too close , or if @xmath118 is too high , mesh oscillations may occur in subsequent time steps . in order to achieve good performance , i.e. to avoid mesh oscillations , @xmath118 should be relatively small compared to @xmath109 . in particular , we choose @xmath119 or @xmath120 and @xmath121 in both examples of section [ sect : examples ] . around a crack tip ( based on linearly interpolated data ) of a qc system . regular atom sites are shown as black dots and the existing crack as a green line . ] in order to coarsen the current triangulation @xmath105 , a similar algorithm to those proposed by @xcite and @xcite is adopted . because we use the two - triangle refinement scheme , the entire binary - tree structure of the triangulation history must be stored . coarsening of a patch of triangles is decided based on their nodes , cf . [ subsect : coarsening : fig:1 ] . all nodes ( repatoms ) of the current triangulation are stored in a set @xmath122 , which is divided into three disjoint sets : @xmath123 , @xmath124 , and @xmath125 . the set @xmath123 stores nodes associated with triangles in @xmath126 ( and also other protected nodes , see alg . [ subsect : xqc : alg:1 ] ) . the set @xmath124 collects nodes associated with triangles in @xmath127 , and @xmath125 complements @xmath122 . as a consequence of the mesh hierarchy , only a subset of @xmath124 can be removed . these are identified through their valences @xmath128 ( the number of triangles connected to a node ) : @xmath129 only nodes with valence @xmath130 ( boundary nodes ) or @xmath131 ( internal nodes ) can be removed ( the blue points in fig . [ subsect : coarsening : fig:1 ] ) . the corresponding two or four child elements are subsequently replaced with one or two parent elements , shifting locally the binary - tree hierarchy of the mesh one level up . as valences change during the coarsening process , this procedure is repeated until convergence . . they are found at , and ahead of , the crack tip , and are protected from coarsening . the yellow along with the blue points define the set @xmath124 , and can be coarsened . the green points remain unchanged and correspond to the set @xmath125 . only the blue points satisfy the valence condition and can be removed . regular atoms are shown as black dots and the current triangulation @xmath105 is depicted in black thin lines . the existing crack is indicated as a thick black line . all interactions are omitted for clarity . ] in what follows , we use a right - angled initial triangulation @xmath132 , as already indicated in fig . [ subsect : coarsening : fig:1 ] . because the above described refinement and coarsening algorithms yield self - similar triangulations , all meshes @xmath105 are right - angled as well . this is desirable because the summation error is minimized , cf . @xcite , and poorly shaped triangles are avoided . an alternative approach to the mesh refinement and coarsening indicators , proposed in eqs . and , would be to use _ error estimators _ such as the _ goal - oriented _ error estimator reported e.g. in @xcite . generalizations of this type will not be developed in this contribution and are left as a possible future challenge . in sections [ subsect : genint ] [ subsect : coarsening ] we discussed the qc interpolation and mesh coarsening independently of an existing crack the crack served merely to identify the fully resolved region at its tip and a coarse region elsewhere . in this subsection , we describe how to extend the qc interpolation when the current triangulation @xmath105 is such that a crack cuts through a coarse triangle @xmath104 , and how it influences @xmath97 , @xmath133 , and @xmath89 . similarly to the previous sections , the discussion considers the system at time step @xmath8 . we first specify the cut elements ( also called reproducing , @xcite ) , cf . fig . [ subsect : enrichment : fig:1a ] . a coarse element @xmath134 such that @xmath135 is considered to be cut by a crack . conditions require that at least one point in @xmath68 belongs to @xmath104 and that the signed distance function @xmath77 changes its sign inside @xmath104 . let us assume in what follows that a crack can not cut directly through any of the nodes , atoms , or along domain boundaries . assume then that our system is approximated with a standard qc framework in which the positions of the repatoms are collected in a column @xmath133 . then , the reconstruction in eq . can be rewritten as @xmath136 where @xmath137 represents the @xmath96 fe shape function associated with the repatom @xmath43 ( this function is evaluated at the undeformed configuration of atom @xmath32 ) , and @xmath138 denotes an admissible position of a repatom @xmath43 . this definition does not include an existing crack yet . in order to incorporate a crack , an enrichment using the local pu concept ( standard local xfem approximation ) is introduced . enrichment terms are included as follows @xmath139}_{\varphi_j^\star({{\boldsymbol{r}}}_0^\alpha ) } \ , \widehat{{{\boldsymbol{g}}}}^\star_j , \quad \alpha \in n_\mathrm{ato } , \quad \beta_j \in n^\star , \label{subsect : enrichment : eq:3}\ ] ] where the same @xmath96 fe shape functions for the extrinsic enrichment terms are used as well . in eq . , @xmath140 is a set of @xmath141 enriched repatoms associated with the cut elements , @xmath142 denotes the enrichment function of the @xmath143-th repatom , and @xmath144 denotes an admissible vector of generalized dofs associated with @xmath145-th enriched repatom . let us emphasize that the only enrichment function is @xmath146 , depicted in fig . [ subsect : enrichment : fig:1b ] , and that the approximation in eq . yields continuous , piecewise linear interpolation of the displacement jump . note also that no crack tip enrichments are required , since the crack tip is always located inside the fully resolved region , and that the shifting adopted in eq . guarantees the kronecker-@xmath147 property of the resulting approximation , cf . @xcite or @xcite . furthermore , the shifting implies that there are actually no blending elements , as shifted enrichment functions vanish outside the cut elements . even when shifting is not adopted , no difficulties with blending elements occur , because the interpolation is based on @xmath96 fe shape functions and the enrichment function is piecewise constant . for further details see @xcite , sections 4.3 and 5.1.2 . in continuum xfem , poorly conditioned stiffness matrices occur if the crack path is close to triangle edges and nodes , cf . e.g. @xcite , or @xcite . for lattice networks , the measure for closeness is the lattice spacing @xmath148 . hence , if a crack path is closer than @xmath148 to an element edge or node , the resulting system matrix is rendered singular , cf . [ subsect : enrichment : fig:2 ] . two approaches are used for continua to remedy this situation . the first approach is the use of clustering ( or node gathering ) implemented through flat - top functions , cf . e.g. @xcite . these techniques usually increase the polynomial order of the resulting approximations if the kronecker-@xmath147 property is required . hence , this is not desirable within the qc framework , as more extensive changes in the summation rule would be needed . the second approach is to neglect those @xmath142 that would render the stiffness matrix singular altogether ( * ? ? ? * section 11 and references therein ) , which is adopted in this contribution . the reason is that such shifted enrichments are identical to zero vector for lattice networks and hence , they provide no additional information to the system . finally , upon introducing @xmath149 along with @xmath150 \quad \mbox{and } \quad \widehat{{{\boldsymbol{g } } } } = [ \widehat{{{\boldsymbol{r}}}}_\mathrm{rep},\widehat{{{\boldsymbol{g}}}}_1^\star,\dots,\widehat{{{\boldsymbol{g}}}}_{n^\star}^\star]^\mathsf{t } , \label{subsect : enrichment : eq:8}\ ] ] the approximation in eq . takes the form of eq . . the second step of qc frameworks , summation , is the qc counterpart to numerical integration in the finite element method . it serves two purposes . the first one is to estimate the exact system s incremental energy @xmath13 in an efficient , yet accurate way . this reduces the computational effort related to determining the energies , gradients , and hessians . the second purpose is to reduce the dimensionality of the internal variable @xmath151 , which in combination with the interpolation step realises the reduction of the entire state variable @xmath152 . the exact incremental energy @xmath13 is approximated using a limited number of site- or interaction - energies and their weight factors . these are typically chosen such that the error of the induced approximation is limited , but a low number of sampling atoms / interactions is required . the interested reader is referred for further details to @xcite for atom - based summation rules , to @xcite for an interaction - based summation rule , or to @xcite for an overview of summation rules in atomistic systems . here , we employ the central site - energy - based summation rule @xcite , which considers only the atoms at the element corners plus one near the center . the central one is taken to be representative also of atoms which interact across the element boundaries . if no internal atoms exist , all boundary atoms are considered . the central rule involves a certain degree of approximation , but is significantly more efficient than the exact summation rule not discussed here ; for details see @xcite . the summation rule is implemented by introducing a set of @xmath153 sampling atoms collected in an index set @xmath154 . the @xmath155 interactions connected to these sampling atoms are stored in an index set @xmath156 . then , the summation step yields the following approximation , @xmath157 where @xmath158 are the weight factors associated with atom sites , and @xmath159 are corresponding weight factors associated with interactions . because the approximate incremental energy in eq . only depends on the internal variables associated with @xmath155 interactions , one can introduce a reduced state variable of the qc system in the form @xmath160 , where @xmath161 is a column that stores the internal variables of all sampling interactions . the reduced state space then reads @xmath162 . if the enrichment of section [ subsect : enrichment ] is adopted and a coarse triangle is cut by a crack , eq . still holds , but the selection of the individual sampling atoms @xmath163 and their weights @xmath158 changes . in the central summation rule , all crack wake atoms are treated as additional triangle edges . therefore , a cut triangle is split into two regions , bounded by triangle edges and one crack face . for each of the two subregions , the standard central summation rule is used . hence , if no internal atoms exist , all edge atoms are added as discrete sampling atoms . otherwise , a central sampling atom is selected with the weight corresponding to the sum of the number of internal atoms plus one half of the number of edge atoms . atoms at edge intersections are added as discrete sampling atoms . for the detailed algorithmic description and pictorial representation see alg . [ subsect : xsumrule : alg:2 ] and fig . [ subsect : xsumrule : fig:1 ] . note that the above - described summation rule , which in its current form does not take into account crack closure ( recall eq . and the discussion therein ) , can be easily generalized to deal also with non - monotonous loading and crack closure merely by adding all crack wake atoms @xmath70 as discrete sampling atoms . special summation rule that samples crack faces can also be devised , but this is omitted from our further considerations for the sake of simplicity . because only monotonous loading and crack opening are observed in examples section [ sect : examples ] , the summation rule presented in alg . [ subsect : xsumrule : alg:2 ] is adequate yet efficient . 1 . identify crack wake atoms @xmath164 . 2 . * for each triangle @xmath165 * 1 . * if * triangle is not cut by a crack , use standard central summation rule @xcite . * else * 1 . identify triangle s internal @xmath166 , edge @xmath167 , vertex @xmath168 , and all @xmath169 , atoms . 2 . select vertex atoms @xmath170 as discrete sampling atoms ( @xmath171 ) . if * no internal atoms exist , select all atoms @xmath169 as discrete sampling atoms ( @xmath171 ) . * else * 1 . split @xmath169 in sets @xmath172 and @xmath173 according to the sign of @xmath77 . 2 . define @xmath174 . split @xmath175 into two disjoint sets @xmath176 and @xmath177 according to the sign of @xmath77 and sort both with respect to the number of neighbours inside triangle ( in decreasing order ) . 3 . * if * @xmath178 is empty , all atoms in @xmath179 are discrete sampling atoms * else * * choose @xmath180 as central sampling atom . * for each @xmath181 , @xmath182 . * add all @xmath183 as discrete sampling atoms ( @xmath171 ) . * identify discrete sampling atoms on edges resulting from neighbouring triangles @xmath184 . * for each @xmath185 , @xmath186 . * end * for the central summation rule . large blue dots represent the sampling atoms , and the crack points @xmath68 from are shown as red dots . ] in all previous sections [ subsect : crack ] [ subsect : summation ] , we have assumed an existing crack and triangulation @xmath105 . we are , however , interested in evolving crack tips and fully resolved regions . therefore , we summarize the mesh refinement , coarsening , and additions of enrichments of the total x - qc framework in alg . [ subsect : xqc : alg:1 ] . 1 . initialize the system : apply initial condition @xmath187 and construct initial ( coarse ) mesh @xmath132 with required information , e.g. @xmath188 , @xmath189 , @xmath190 , @xmath191 . 2 . * for @xmath192 apply the boundary conditions at time @xmath8 , @xmath193 , @xmath194 , @xmath195 , @xmath196 , @xmath197 , etc . 2 . equilibrate the unbalanced system , i.e. solve for @xmath198 using @xmath199 of eq . . update crack description @xmath77 , @xmath68 , etc . 3 . for each element @xmath134 evaluate its refinement indicator in and construct @xmath200 . if @xmath201 refine and update @xmath105 4 . protect newly added repatoms from all previous refinements during the current time step @xmath8 ; if necessary , protect also other repatoms , e.g. those near the crack tip . use to identify @xmath202 . if possible , coarsen and update @xmath105 . if * @xmath105 changed in ( iii ) ( iv ) , update the system information @xmath203 , @xmath204 , @xmath205 , @xmath206 , etc . , and return to ( ii ) since the system is unbalanced . + * else if * the mesh has converged ; proceed to ( vi ) . store relevant outputs : @xmath207 , @xmath105 , @xmath203 , @xmath204 , @xmath205 , @xmath206 , @xmath68 , etc . * end * this section discusses the implications of the adaptive scheme summarized in alg . [ subsect : xqc : alg:1 ] from an energetic point of view , and extends the discussion presented in @xcite , section 3.5 . recall that the motivations for energy considerations are threefold . first , the entire theory presented so far is based on energy minimization and the adaptive procedure should be consistent with these principles . second , validity of the solution @xmath152 , as well as of @xmath207 , is decided based on the energy balance , which must hold during the entire evolution of the system . finally , upon taking into account energy implications , the accuracy of the x - qc framework can be assessed from the energetic point of view . when refinement and coarsening occur . only the @xmath208 and @xmath209 components are shown for clarity . ] the issue focused on here is illustrated in fig . [ subsect : enimplications : fig:1 ] . at the end of the previous time step @xmath210 , the system is relaxed for some mesh @xmath211 . after the next increment is applied for mesh @xmath212 and the system is equilibrated ( stage ( ii ) of alg . [ subsect : xqc : alg:1 ] ) , refinement and coarsening may be required ( ( ii ) ( iv ) ) . upon convergence , this yields a triangulation @xmath213 . in order to accurately construct the energy increments , one should distinguish physical increments , computed on mesh @xmath214 and denoted with subscripts `` p '' , and changes in energy which are due to the mesh adaptation from @xmath215 to @xmath216 , denoted with subscripts `` a '' . the artificial energy increments due to mesh adaptation are computed by projecting converged solutions onto an artificial mesh @xmath217 , which establishes communication between the two systems ( note that these two systems have different numbers of dofs and internal variables ) . the artificial mesh therefore contains the union of repatoms and sampling interactions present in both systems and hence , it is locally the finest triangulation of the two , cf . [ subsect : enimplications : fig:1 ] . using the above - described procedure and evaluating energies at time instants @xmath218 provides energy profiles that we will call _ reconstructed_. for further details see @xcite , section 3.5 . let us note that we omit dissipative coarsening ( i.e. the interpolation of the internal variables when the mesh is coarsened ) , and assume that the coarsening occurs only in elastic regions . this assumption is reasonable because damage is localized . consequently , all damaged bonds are captured accurately by the special function enrichments and associated summation rule . in this section , we discuss the minimization of the reduced incremental energy @xmath219 in , considered in step ( ii ) of alg . [ subsect : xqc : alg:1 ] . a fixed triangulation @xmath105 is again assumed . for the solution of the dns , the interested reader is referred to @xcite , section 4.1 . we first reduce the incremental energy @xmath220 see e.g. @xcite , section 3.1.2 , or @xcite , and rewrite as @xmath221 second , is solved by a standard newton algorithm because @xmath222 is assumed to be smooth enough in @xmath89 . using the chain rule in the taylor expansion of @xmath223 , the following stationarity condition is obtained : @xmath224 where [ subsect : sol : eq:2 ] @xmath225 the partial derivatives in eqs . read [ subsect : sol : eq:3 ] @xmath226 where the reduced site energies , @xmath227 , are introduced in analogy to by condensing out the internal variables . . presents a system of linear equations for increments @xmath228 , and iterating eqs . and until convergence of @xmath229 yields @xmath230 . note that explicit expressions for the gradients and hessians can be found in @xcite , appendix a. here , the quantities are derived for lattice structures with isotropic hardening plasticity and hence they have to be multiplied by @xmath231 if a bond is loaded in tension . moreover , the dependencies on all plastic internal variables have to be dropped . before the gradient and hessian in eqs . are computed , the approximate incremental energy must be minimized with respect to @xmath232 according to eq . . to this end , the interaction - based formulation of the incremental energy specified in eq . the minimization then results in @xmath155 independent one - dimensional box - constrained minimization problems @xmath233 for further details see @xcite , section 4.1 and appendix a. in this section the framework is applied to two examples discussed already in @xcite , where coarsening was not considered . consequently , we can make a direct comparison in terms of accuracy and efficiency between the adaptive qc frameworks with and without coarsening . the same material model is used for each lattice spring ( the superscripts @xmath55 are dropped for brevity ) : @xmath234 where @xmath235 is the young s modulus , @xmath236 the cross - sectional area , and @xmath237 the bond strain . the dissipation function @xmath238 in eq . is adopted from @xcite , appendix a , providing an exponential softening rule @xmath239 in eq . , @xmath240 describes the softening branch of the associated stress - strain diagram , @xmath241 is the bond stress , @xmath242 measures the initial slope of the softening branch , and @xmath243 is the limit elastic strain , cf . [ sect : examples : fig:1 ] . the employed parameters are specified in tab . [ sect : examples : tab:1 ] . and @xmath244 , cf . the elastically stored energy , @xmath245 , corresponds to the area of the dashed triangle , and the dissipated energy , @xmath246 , to the area of the upper triangle with the curved edge . ] .dimensionless material and geometric parameters for both examples . [ cols="<,>,<,>,<",options="header " , ] [ sect : examples : tab:1 ] denotes the applied vertical displacement . ] the first example focuses on an l - shaped plate test , see fig . [ subsect : simpleex : fig:0 ] . the lattice properties are homogeneous except for the vicinity of @xmath247 , where a vertical displacement is applied . here , in order to prevent any damage , the young s modulus is @xmath248 times larger than elsewhere and the limit elastic strain @xmath243 is infinite . the reference domain comprises @xmath249 atoms , @xmath250 interactions , and is fixed at the bottom part of the boundary , @xmath251 . the evolution of the system is controlled by the crack mouth opening displacement ( cmod ) through a pair of crack tip atoms highlighted in fig . [ subsect : simpleex : fig:0 ] . the numerical simulation is performed for two x - qc systems , each with a different safety margin @xmath109 , cf . the first system is referred to as the _ moderate x - qc _ ( @xmath119 ) and the second one to as the _ progressive x - qc _ ( @xmath252 ) ; in both cases , @xmath121 , recall eq . . relatively low safety margins are necessary due to the highly fluctuating interaction energies near the crack tip ( recall fig . [ subsect : nodeprotection : fig:1 ] and the discussion therein ) . the results are compared to those of the adaptive version of the variational qc , reported in @xcite , for the same choices of the refinement safety margins , i.e. @xmath119 and @xmath252 , and for the central summation rule . these systems are referred to as the _ moderate adaptive qc _ and the _ progressive adaptive qc_. for completeness , results of the dns are also shown . the deformed configuration computed by the dns at @xmath253 is depicted in fig . [ subsect : simpleex : fig:1a ] . the crack path initiates at the inner corner ( @xmath254 ) and propagates to the left . this is also predicted by the qc frameworks . in fig . [ subsect : simpleex : fig:1b ] , the reaction force @xmath255 is presented as a function of @xmath256 . the responses predicted by the x - qc are practically identical to those of the adaptive qc without coarsening . ( ldeformed ) at ( 0,0 ) ; ( lreact ) at ( 6.5,0 ) ; ( lzoom ) at ( 11.5,-0.72 ) predicted by the dns , ( b ) the force - displacement diagram for the reaction force @xmath255 as a function of @xmath256 . the instants at which mesh refinement and coarsening occur for the first time are indicated by dashed vertical lines of the respective colours.,title="fig : " ] ; ( 8,-1.1 ) rectangle ( 8.7,-0.6 ) ; ( 10.3,-1.3 ) rectangle ( 12.7,0 ) ; ( 8.35,-0.6 ) ( 10.3,0 ) ; ( 8.35,-1.1 ) ( 10.3,-1.3 ) ; ( text ) at ( 11.5,0.2 ) zoom in ; the energy profiles appear in fig . [ subsect : simpleex : fig:2a ] as functions of @xmath256 . the energy balance is clearly satisfied along the entire loading path , i.e. the thin dotted lines presenting the work performed by external forces @xmath257 lie on top of the thick dashed lines representing @xmath258 . second , similarly to the reaction force profiles , we notice that the energy paths are almost indistinguishable from those of adaptive qc , demonstrating good accuracy of the x - qc approach . in fig . [ subsect : simpleex : fig:2b ] , we present the energy components that are exchanged ( the artificial energies ) in the moderate x - qc simulation . their high magnitudes compared to the curves of fig . [ subsect : simpleex : fig:2a ] indicate the importance of the energy reconstruction procedure , discussed in section [ subsect : enimplications ] . [ subsect : simpleex : fig:3a ] presents the relative number of repatoms as a function of @xmath256 . note that the number of enriched repatoms @xmath141 , defined in eq . , is negligible ( in particular , @xmath259 ) . both x - qc systems initially follow the trends of the corresponding adaptive qc systems ( the refinement stage and crack localization ) . after the crack localizes and starts to propagate however , the crack wake unloads and the mesh coarsens , resulting in a substantial drop in the number of repatoms of the progressive x - qc . at the end of the simulation , the specimen is almost stress - free , allowing for a very coarse representation . note that the number of dofs associated with the x - qc approach is less than one half compared to the adaptive qc in the case of the moderate approach , and approximately a quarter in the case of the progressive approach . in both cases , the final relative number of dofs as well as repatoms remain below @xmath260 , whereas they never exceed @xmath261 during the simulation . a similar behaviour can be observed also for the relative number of sampling atoms , presented in fig . [ subsect : simpleex : fig:3b ] . finally , figs . [ subsect : simpleex : fig:4 ] and [ subsect : simpleex : fig:5 ] present some of the meshes of the x - qc and the adaptive qc frameworks associated with fig . [ subsect : simpleex : fig:3 ] . although at the initial stages the progressive and moderate x - qc meshes differ substantially , they are almost identical when the crack localizes . the effect of coarsening in both cases is evident . note that in the progressive case coarsening occurs not only along the crack path , but also along the right edge of the lower part of the plate . this explains the particularly large gain in number of repatoms in this case , i.e. the factor of four visible in fig . [ subsect : simpleex : fig:3a ] . as a function of @xmath256 , and ( b ) relative number of sampling atoms.,title="fig : " ] + + + the second example is a four - point bending test , presented e.g. in @xcite , cf . also fig . [ subsect : complexex : fig:1 ] . the rectangular reference domain @xmath95 is pre - notched at the top to initiate a crack . the otherwise homogeneous x - braced lattice is again locally stiffened where the dirichlet and neumann boundary conditions are applied ( the young s modulus is @xmath248 times larger than elsewhere and the limit elastic strain @xmath243 is infinite to prevent damage ) . the domain comprises @xmath262 atoms connected by @xmath263 interactions and is loaded by a pair of vertical forces @xmath264 and @xmath265 , given by @xmath266 where @xmath267 represents the indirect displacement control parameter , see e.g. @xcite , section 22.2.3 . unlike the l - shaped plate example , the sum of cmod and cmsd ( crack mouth sliding displacement ) is now used to control the simulation . in fig . [ subsect : complexex : fig:2a ] the loading program is specified . again two x - qc systems are studied : the moderate x - qc with @xmath119 , and the progressive x - qc with @xmath252 . the results are again compared to those of the adaptive qc ( with the same values of @xmath109 ) and to those of the dns . the deformed configuration predicted by the dns at @xmath268 is depicted in fig . [ subsect : complexex : fig:2b ] . the x - qc crack paths , shown in fig . [ subsect : complexex : fig:3a ] , are practically identical to those of the adaptive qc ( the maximum distance amounts to one lattice spacing ) . they also match the crack path predicted by the dns , where the distance does not exceed four lattice spacings in all cases . the same quality of approximation is achieved also for the total applied force , @xmath269 , plotted in fig . [ subsect : complexex : fig:3b ] against @xmath270 . some minor differences can be observed between the responses of the x - qc and the adaptive qc around the peak load ( when the crack localizes ) . at this stage , the x - qc meshes start to coarsen . at later stages , all curves become practically indistinguishable as the specimen becomes stress - free . ( bcrack ) at ( 0,0 ) ; ( breact ) at ( 9,0.2 ) ; ( breactzoom ) at ( 4.2,0.7 ) versus @xmath270 . note that the colour scheme applies to both diagrams . the instants at which mesh refinement and coarsening occur for the first time are indicated by vertical dashed lines of the respective colours.,title="fig : " ] ; ( 7,2.3 ) rectangle ( 7.7,1.5 ) ; ( 2.8,-0.8 ) rectangle ( 5.6,2 ) ; ( 7.35,2.3 ) ( 5.6,2 ) ; ( 7.35,1.5 ) ( 5.6,-0.8 ) ; ( text ) at ( 4.2,2.2 ) zoom in ; the energy profiles for all employed approaches are presented in fig . [ subsect : complexex : fig:4a ] . the energy balance clearly holds . one can also notice that the x - qc energies match the adaptive qc and dns energies well . in fig . [ subsect : complexex : fig:4b ] the energy exchanges of the moderate x - qc approach are shown . again , like in the previous example , their contributions are substantial . in fig . [ subsect : complexex : fig:5a ] , the relative number of repatoms of the qc schemes are shown as a function of @xmath270 . the number of enriched repatoms is again negligible compared to the number of repatoms ( @xmath271 ) . before the crack localizes , the number of repatoms rapidly increases . but it drops significantly after crack localization , and converges to approximately @xmath272 for both x - qc approaches . similar behaviour can be observed for the relative number of sampling atoms shown in fig . [ subsect : complexex : fig:5b ] . this result , in combination with the accuracy captured by figs . [ subsect : complexex : fig:3 ] and [ subsect : complexex : fig:4a ] , clearly demonstrates the strength of the x - qc framework . finally , in figs . [ subsect : complexex : fig:6 ] and [ subsect : complexex : fig:7 ] several snapshots capturing the evolution of the triangulations @xmath105 are presented . as a function of @xmath270 , and ( b ) relative number of sampling atoms.,title="fig : " ] + + + + + + + in this contribution , an adaptive mesh algorithm for the variational qc for lattice networks with localized damage is developed , which refines the mesh where necessary , but also coarsens it where possible . to achieve the latter , xfem - like enrichment functions are required . it is shown that the efficiency of the qc framework applied to lattices with localised damage increases significantly . the main results can be summarized as follows : 1 . the two qc steps , interpolation and summation , were generalized to incorporate special enrichment functions , reflecting cracks within the framework of the fully - nonlocal variational qc . 2 . to determine regions to be coarsened , a heuristic marking strategy was proposed with a variable parameter controlling the mesh coarseness . the mesh refinement and coarsening were discussed from an energetic point of view . 4 . for the investigated examples the extended qc methodology , including coarsening , reduces the number of repatoms and sampling atoms with at least @xmath273 of that of a simpler adaptive qc method with refinement only . the accuracy remains the same , however . one can note that the employed coarsening procedure neglected the diffusive dissipation . this was justified for the application to localized damage , but if more ductile behaviour is modelled ( due to elastic - plastic models with damage ) , this must be incorporated . therefore , a further generalization is required . moreover , instead of the heuristic marking strategies used here , goal - oriented error estimators may be used to further improve the accuracy and efficiency of the qc methodology . the variational foundation of the method provides an ideal platform for this , cf . e.g. @xcite . the same x - qc framework can also be employed for the description of heterogeneities in otherwise homogeneous lattice networks . these aspects are , however , outside the scope of this contribution and will be reported separately . this work was supported by the czech science foundation ( gar ) , through project no . 14 - 00420s . 75 natexlab#1#1[1]`#1 ` [ 2]#2 [ 1]#1 [ 1]http://dx.doi.org/#1 [ ] [ 1]pmid:#1 [ ] [ 2]#2 , , , a. . , . http://www.sciencedirect.com/science/article/pii/s0045782516301050 , . , , , , b. . , . http://dx.doi.org/10.1002/nme.4982 , . . , , , . . , http://www.sciencedirect.com/science/article/pii/s0022509615000630 , . , , , . . , http://dx.doi.org/10.1002/nme.2542 , . , , , . . , http://dx.doi.org/10.1002/nme.2675 , . http://www.sciencedirect.com/science/article/pii/s0045782511003082 , . http://dx.doi.org/10.1002/(sici)1097-0207(19970228)40:4<727::aid-nme86>3.0.co;2-n[http://dx.doi.org/10.1002/(sici)1097-0207(19970228)40:4<727::aid-nme86>3.0.co;2-n ] , . http://journals.cambridge.org/article_s0962492901000010 , . , , , , a. . , . http://www.sciencedirect.com/science/article/pii/s0045782514002047 , . http://dx.doi.org/10.1002/nme.3134 , . , , , b. . , . http://www.sciencedirect.com/science/article/pii/s0022509614001100 , . , , , c. . , . http://www.sciencedirect.com/science/article/pii/s0022509613002445 , . , , , d. . , . http://www.sciencedirect.com/science/article/pii/s004578251300279x , . , , , , a. . , . http://www.sciencedirect.com/science/article/pii/s0167663614001495 , . , , , , b. . , . http://dx.doi.org/10.1002/gamm.201510018 , . , , , . . , http://www.sciencedirect.com/science/article/pii/s1359835x13000134 , . http://dx.doi.org/10.1002/(sici)1097-0207(19990620)45:5<601::aid-nme598>3.0.co;2-s[http://dx.doi.org/10.1002/(sici)1097-0207(19990620)45:5<601::aid-nme598>3.0.co;2-s ] , . , , , , . . , http://dx.doi.org/10.1002/1097-0207(20010210)50:4<993::aid-nme164>3.0.co;2-m[http://dx.doi.org/10.1002/1097-0207(20010210)50:4<993::aid-nme164>3.0.co;2-m ] , . , , , a. . , . http://www.sciencedirect.com/science/article/pii/s016766361500157x , . , , , b. . , . http://www.sciencedirect.com/science/article/pii/s0020768314004600 , . , . . , . . , , , http://www.sciencedirect.com/science/article/pii/s0022509699000289 , . , , , . . , http://epubs.siam.org/doi/abs/10.1137/080741033 , . , , , . . , , , . . , . , , , , , . . , http://www.sciencedirect.com/science/article/pii/s0045782505003956 , . , , , , . . , http://www.sciencedirect.com/science/article/pii/s0013794415000053 , . , . http://dx.doi.org/10.1002/nme.2259 , . http://dx.doi.org/10.1002/nme.3299 , . , , . http://dx.doi.org/10.1002/nme.2914 , . , , , . . , http://dx.doi.org/10.1002/nme.2488 , . http://www.sciencedirect.com/science/article/pii/s0020768309004752 , . http://dx.doi.org/10.1137/s1064827599355840 , , http://arxiv.org/abs/http://dx.doi.org/10.1137/s1064827599355840 [ arxiv : http://dx.doi.org/10.1137/s1064827599355840 ] . , , . . http://dx.doi.org/10.1007/s10704-012-9753-8 , . , , . http://www.sciencedirect.com/science/article/pii/s0022509610002425 , . http://www.sciencedirect.com/science/article/pii/s0020768315002620 , . http://www.sciencedirect.com/science/article/pii/s0167663612000683 , . , , , , . . , http://dx.doi.org/10.1002/nme.1370 , . , , , . . , http://www.sciencedirect.com/science/article/pii/s016784421000039x , . http://journals.cambridge.org/article_s0962492913000068 , . http://www.sciencedirect.com/science/article/pii/s0045782596010870 , . , , , . . , http://dx.doi.org/10.1007/s00466-015-1127-4 , . , , , . . , http://www.sciencedirect.com/science/article/pii/s004578251400423x , . , . http://dx.doi.org/10.1007/s00161-003-0120-x , . http://www.springer.com/us/book/9781493927050 , . , , , . http://dx.doi.org/10.1007/s002050200194 , . http://dx.doi.org/10.1023/a%3a1026098010127 , . , , , . . , http://dx.doi.org/10.1002/(sici)1097-0207(19990910)46:1<131::aid-nme726>3.0.co;2-j[http://dx.doi.org/10.1002/(sici)1097-0207(19990910)46:1<131::aid-nme726>3.0.co;2-j ] , . http://www.sciencedirect.com/science/article/pii/002076839190004y , . http://www.sciencedirect.com/science/article/pii/s0021999102971834 , . http://www.sciencedirect.com/science/article/pii/s1359835x04002593 , . http://www.sciencedirect.com/science/article/pii/s002250961100055x , . , , , . . , http://www.sciencedirect.com/science/article/pii/s0045782598002308 , . , , , . http://www.sciencedirect.com/science/article/pii/s0022509610001341 , . , . . , . . , , , , a. . http://arxiv.org/abs/1601.00625[arxiv:1601.00625 ] . , , , , b. . http://arxiv.org/abs/1601.00625[arxiv:1601.00625 ] . , . . ph.d . thesis . technische universiteit delft . . http://www.sciencedirect.com/science/article/pii/095894659290004f , . , , , , , , . . , http://www.sciencedirect.com/science/article/pii/s0022509698000519 , . http://www.sciencedirect.com/science/article/pii/s0045782599000729 , . , , , b. . , . http://dx.doi.org/10.1002/(sici)1097-0207(20000320)47:8<1401::aid-nme835>3.0.co;2-8[http://dx.doi.org/10.1002/(sici)1097-0207(20000320)47:8<1401::aid-nme835>3.0.co;2-8 ] , . , , , . . , http://dx.doi.org/10.1080/01418619608243000 , , http://arxiv.org/abs/http://dx.doi.org/10.1080/01418619608243000 [ arxiv : http://dx.doi.org/10.1080/01418619608243000 ] . , , , , , . . http://dx.doi.org/10.1002/nme.2387 , . http://dx.doi.org/10.1002/nme.143 , . http://www.sciencedirect.com/science/article/pii/s0045782514003545 , .

elastic scattering from a point defect in a hard - walled , multimode quantum waveguide is a problem which has been the subject of experimental@xcite and theoretical@xcite inquiry in both condensed matter and atomic physics . in this paper , we revisit the problem from an unconventional point of view , reducing it to the scattering of a single effective wavefunction off an array of images . our approach enables us to easily understand complex transport phenomena which have already been observed ( such as conductance dips ) , as well as predict other new ones . the problem of scattering from pointlike impurities in waveguides is a general one , and thus has physical applications in several fields . impurity scattering in quantum wires , and the resulting disorder effects on electron transport , have long been of interest in mesoscopic physics . specifically , the problem arises in the context of two - dimensional electron gases ( 2degs ) confined by quantum wire potentials ; the impurity can represent a defect in the wire . such quantum waveguides can be fabricated at low - temperature @xmath0 interfaces , and are of interest due to their potential role in high - frequency and quantum devices . as carbon nanotubes can behave as few - mode quantum waveguides,@xcite our results are relevant to transport in nanotubes with adsorbed impurities of interest in the study of biosensors . the problem of scattering in confined geometries arises in atom waveguides as well , with potential implications for quantum computing and atom interferometry . in both cases , atoms must maintain coherence as they pass through the waveguide , and the goal is thus to minimize phenomena such as collisional phase shifts.@xcite finally , the wave phenomena arising in quantum waveguides are in direct mathematical correspondence with a number of wave phenomena in other systems : water waves resonantly trapped between an array of cylinders,@xcite large antenna arrays,@xcite and sound waves directed by `` bessel '' line arrays in acoustics . @xcite motivated by the above applications , we examine low - energy scattering of noninteracting particles from impurities in an infinite two - dimensional quantum wire , with hard walls at @xmath1 and @xmath2 the wire ( fig . [ fig : schematic ] ) contains a point impurity , which we model as an @xmath3 wave scatterer of effective radius @xmath4 . our method differs from previous treatments , such as the renormalized @xmath5 matrix method,@xcite in that we use an unconventional approach which combines the following three ideas : 1 . the method of images . 2 . the realization that a confined scatterer , like a free space scatterer , is a rank - one target . we can , therefore , combine the @xmath6 degenerate transverse modes into a basis in which only a single wavefunction scatters . 3 . that the effect of scattering from the waveguide walls is to renormalize this single scattering wavefunction , so that it takes on a new effective value at the scatterer location . using the method of images allows us to explicitly understand how each reflection from the waveguide walls affects the conductance . the fact that the confined scatterer is a rank one target enables us to express the transport properties of the wire entirely in terms of the single scattering wavefunction . finally , the idea of renormalization allows us to fully explain the transport properties of the wire in terms of reflections of this single wavefunction from the waveguide walls . in section [ sec : freespace ] , we review free space multiple scattering , anticipating phenomena which will appear in the wire . section [ sec : wire ] lays out a similar formalism for scattering in the waveguide . in section [ sec : results - and - discussion : ] , we present physical phenomena which we observed in our model , which we interpret in section [ sec : physical ] in terms of interference effects . we relegate mathematical details to appendices [ sec : diffraction]-[sec : kummer ] . in section [ sec : wire ] , we show that the lippmann - schwinger equation for scattering from a confined impurity resembles the free space lippmann - schwinger equation , with the exception that in the wire , a renormalized incoming wavefunction replaces the true one . surprisingly , even for a single confined scatterer , this renormalized wavefunction includes free space _ multiple _ scattering effects . we thus begin by briefly reviewing free space scattering , emphasizing the three phenomena which will reappear in the lippmann - schwinger equation for a single scatterer in the waveguide . consider @xmath6 pointlike impurities in two - dimensional free space , located at @xmath7 let @xmath8 be the potential at @xmath9 due to the @xmath10 scatterer . we seek scattering solutions @xmath11 to the time - independent schrodinger equation@xmath12\psi(\vec{r})=0,\label{eq : schrodinger}\ ] ] under the assumption of low - energy , @xmath3-wave scattering . except as noted , we henceforth use atomic units , where @xmath13 applying the @xmath5 matrix formalism,@xcite we characterize a single scatterer at @xmath14 by its @xmath5 matrix,@xmath15 where @xmath16 is a function of the wavenumber , @xmath17 we choose the functional form of @xmath16 to simulate the scatterer of interest , under the constraint that @xmath16 must satisfy the optical theorem @xmath18 in this paper , we choose@xmath19 which represents a hard disk of scattering length @xmath4 . consider a single scatterer at @xmath20 in our @xmath5 matrix representation , an incident wave @xmath21 scatters into a wavefunction @xmath11 according to the lippmann - schwinger equation@xmath22 where @xmath23\label{eq : freegreens}\ ] ] is the 2d free space advanced green s function satisfying @xmath24 we shall henceforth assume we are using the advanced green s function and omit the superscript on the hankel function ; we shall also omit the implicit @xmath25 dependence in the green s function . any incoming wave @xmath21 must be a solution to the free space schrodinger equation . we can therefore express @xmath21 in the basis of `` cylinder harmonics , '' @xmath26 where @xmath27 correspond to @xmath3,@xmath28@xmath29 waves . an @xmath3 wave scatterer in free space is a rank one perturbation : of all the terms in ( [ eq : freespaceswf ] ) , only the @xmath30 term is nonzero at the scatterer . from ( [ eq : lippmannschwingersingle ] ) , then , only the @xmath3 wave scatters , acquiring a phase shift ; remaining higher partial waves pass through the scatterer unperturbed . while trivial in free space , these facts will reappear more subtly in the wire . the lippmann - schwinger equation describing scattering from a collection of @xmath6 identical point scatterers at @xmath31 is@xmath32 in which we express the @xmath33 recursively as@xmath34 where we have followed foldy s method.@xcite comparing ( [ eq : lippmannschwingermultiple1]-[eq : lippmannschwingermultiple2 ] ) with the single - scatterer version ( [ eq : lippmannschwingersingle ] ) , the @xmath33 are the @xmath35 incoming wavefunctions at each scatterer : @xmath33 is the amplitude incident on the @xmath10 scatterer after scattering from each of the other scatterers . a crucial point is that @xmath33 excludes the singular self - interaction of the @xmath36 scatterer . defining @xmath37 @xmath38 inverting ( [ eq : lippmannschwingermultiple1]-[eq : lippmannschwingermultiple2 ] ) yields@xmath39 where@xmath40 excludes the singular term . an alternate expression of ( [ eq : lippmannschwingermatrix ] ) is as the born series@xmath41\vec{\phi}.\label{eq : expansion}\ ] ] the effective wavefunctions thus have a simple interpretation in terms of interfering paths : the terms in square brackets describe amplitude incident at @xmath14 after interactions with zero , one , two , or three other scatterers respectively . the series continues infinitely . anticipating phenomena that will reappear in the wire , we highlight the following points of this section : 1 . the effect of multiple scattering is to create a new effective incoming wavefunction at each scatterer.[enu : newwf ] 2 . the effective wavefunction at a particular scatterer is a sum of waves scattered from all the _ other _ scatterers , and excludes the ( singular ) self - interaction of the scatterer.[enu : renorm ] 3 . an @xmath3 wave scatterer in free space is a rank one target.[enu : specialwf ] in this section , we examine how confinement in a wire affects the transport properties of a scatterer . applying the method of images , we derive a form of the empty wire green s function . using this form of the green s function , we show that the sole effect of confinement is to renormalize the incoming wavefunction . the image representation allows us to very simply calculate the renormalization coefficient . we define an effective cross section for a confined scatterer , and relate it to the conductance of the wire - impurity system . we derive an effective optical theorem in the wire . finally , applying this formalism , we investigate the behavior of the cross section and conductance as functions of various parameters . in order to write down the lippmann - schwinger equation for an impurity in the wire , we require the green s function @xmath42 of the empty wire . the usual spectral form of the empty wire green s function is@xmath43 where the @xmath44 are the transverse modes for the particular waveguide ( see e.g. datta@xcite ) . in our hard wire , the modes have the form @xmath45 we note that the spectral green s function ( [ eq : dattagreenswire ] ) is a sum over evanescent modes as well as propagating ones . although we could , in principle , proceed using the spectral form ( [ eq : dattagreenswire ] ) of the green s function , the spectral green s function provides little physical insight into the precise mechanism of the scattering . we therefore take an alternative approach , and use the method of images to derive an alternate , equivalent form of ( [ eq : dattagreenswire ] ) for the hard wire . although the problem of an impurity in a hard wire has been the subject of numerous theoretical studies,@xcite the authors are aware of only one work@xcite which applies the method of images : ref . 3 treats the related problem of a point scatterer in a 3d waveguide of rectangular cross section , which the authors reduce to the problem of scattering from an impurity in a finite , 2d box . the method of images has the significant advantage of making multiple scattering effects explicit . we shall thus use the approach of ref . 3 , rather than the more common spectral methods . because our problem of an infinite waveguide is somewhat simpler than the finite box treated in ref . 9 , the role of multiple scattering is more transparent . we are therefore able to explore in detail the considerable effect of multiple scattering on the physical properties of the wire , which , although briefly mentioned , is not examined in ref . 3 . the empty wire green s function @xmath42 satisfies green s equation inside the wire , and is zero on the wire walls : @xmath46 the confined green s function differs from the free space version due to the necessity of including reflections off the walls when describing the response to a point excitation . note the factor of two in ( [ eq : greenwire ] ) , obtained by casting ( [ eq : greens ] ) in atomic units . as in electrostatics , we can linearly combine free space green s functions ( [ eq : freegreens ] ) with point sources in different places , so that the linear combination satisfies ( [ eq : wirebc ] ) . as long as only one of the point sources is inside the wire @xmath47 , our sum will satisfy ( [ eq : greenwire ] ) inside the wire . outside the wire , we discard the solution . if the point source is located at @xmath48 , we use the configuration of image point sources in fig . [ fig : offcenter ] . we reflect the scatterer , creating a series of images such that the position of the @xmath49 image is @xmath50\hat{y}.$ ] this image configuration yields an empty wire green s function of the form @xmath51 the green s function ( [ eq : wiregreens ] ) satisfies ( [ eq : greenwire ] ) inside the wire , and ( [ eq : wirebc ] ) on the wire boundaries . outside the wire we implicitly set the green s function to zero . the green s function satisfies the dirichlet boundary conditions because , as we take infinitely many images , the alternating sum of hankel functions converges to zero on the walls . this series converges extremely slowly , precluding its use for numerical purposes ( see the discussion in morse and feshbach@xcite ) . the lippmann - schwinger equation requires knowledge of both the green s function and the incoming wavefunction . in section [ sub : single - scatterer ] , we calculated the green s function for the wire . in this section , we discuss the incident wavefunction . we show that the effects of confinement resemble free space multiple scattering : confinement renormalizes the effective wavefunction at the scatterer from its true value . in the image formalism , the multiple scattering arises due to scattering off images . applying the free space multiple scattering formalism ( [ eq : lippmannschwingermultiple1],[eq : lippmannschwingermultiple2 ] ) to the array of images in fig . [ fig : offcenter ] , we find that the lippmann - schwinger equation for the wire is@xmath52 where@xmath53 recall from our discussion of free space multiple scattering the physical meaning of this recursive lippmann - schwinger equation : the @xmath33 defined in ( [ eq : wavelets1 ] ) are the effective incoming wavefunctions at the @xmath36 scatterer in the array . in a general multiple scattering problem , we would have to stop here and solve numerically . the image array is , however , a special case : not only is it periodic , but the boundary conditions at the wall impose the antisymmetry@xmath54 any wavefunction , incident or scatterered , must be zero on the walls . thus , inside the wire , in the @xmath55 direction , the wavefunction is a superposition of modes ( [ eq : wiremodes ] ) . replacing the wire with an image array requires that we extend these modes outside the wire , into the region @xmath56 keeping the @xmath44 continuous at the walls , we can extend them so that the wavefunction is either symmetric or antisymmetric across the wire walls ; either set forms a basis . however , we know that in the absence of a wall @xmath57 must be continuous also , as we have eliminated the hard walls . the symmetric extension does not satisfy this condition . we therefore discard the symmetric extension , leaving ( [ eq : as ] ) as the proper boundary condition . combining ( [ eq : wavelets1 ] ) and ( [ eq : as ] ) , the equation for the wavelets thus becomes@xmath58\label{eq : wavelets}\end{aligned}\ ] ] where we have defined the renormalization sum@xmath59 noting that@xmath60 taking @xmath61 in ( [ eq : wavelets ] ) , we find @xmath62 substituting ( [ eq : as],[eq : zerowavelet ] ) into ( [ eq : ms ] ) yields the lippmann - schwinger equation for scattering from an impurity in a wire:@xmath63\sum_{i=-\infty}^{\infty}(-1)^{i}g_{0}(\vec{r},\vec{r}_{i})\label{eq : mswf}\\ & = & \phi(\vec{r})+s\left[\frac{\phi(\vec{r}_{0})}{1-sg_{r}}\right]g_{w}(\vec{r},\vec{r}_{0}).\label{eq : mswf2}\end{aligned}\ ] ] comparing ( [ eq : mswf2 ] ) to the free space lippmann - schwinger equation ( [ eq : lippmannschwingersingle ] ) for a single scatterer , we see that the wire simply renormalizes the incoming wavefunction at the scatterer to have a new effective value @xmath64 and so our final lippmann - schwinger equation is@xmath65 for convenience , we define a renormalization factor@xmath66 which is the ratio between the original and renormalized incoming wavefunctions at the scatterer . in order to cast the renormalized incoming wavefunction in terms of interfering paths , we expand perturbatively , as we did in ( [ eq : expansion ] ) for free space . for shorthand , define@xmath67 we find@xmath68\phi(\vec{r}_{0})\nonumber \\ & = & \left[1+\mathop{\sum_{i=-\infty}^{\infty}}_{i\neq0}sg_{i\rightarrow0}\right.\label{eq : bouncesfeynman}\\ & & \left.+\mathop{\sum_{i=-\infty}^{\infty}}_{i\neq j}\mathop{\sum_{j=-\infty}^{\infty}}_{j\neq0}sg_{i\rightarrow j}\times sg_{j\rightarrow0}\right]\phi(\vec{r}_{0})\nonumber \end{aligned}\ ] ] where we have used ( [ eq : signstrength ] ) . the @xmath69 term of the sum ( [ eq : bouncesfeynman ] ) describes propagation to @xmath70 after scattering from any @xmath71 images , beginning with the first term , which describes free propagation to @xmath70 . alternately , the @xmath69 term corresponds to @xmath72 scattering events combined with any number of reflections off the wall , each of which introduces an additional phase of -1 . [ fig : born ] illustrates the correspondence between scattering from an image and reflecting from the wall . renormalization of the effective wavefunction is simply the manifestation of the identical phenomenon in free space ( see item [ enu : newwf ] in the list of section [ sub : freespacemultiple ] ) . from a semiclassical point of view , the incoming wave , given infinite time to spread , reflects from each possible combination of scatterers before returning to the source . note that @xmath73 is simply the wire s green s function evaluated at the scatterer , with the singular self - interaction of the source removed,@xmath74.\label{eq : gr}\ ] ] this idea , as expressed in ( [ eq : gr ] ) , is the essence of `` renormalized @xmath5 matrix theory.''@xcite the theory , as applied to our case , is equivalent to the idea that in free space multiple scattering , the effective wavefunction at a scatterer excludes the singular self - interaction ( see item [ enu : renorm ] of section [ sub : freespacemultiple ] ) . we note that although we have derived ( [ eq : gr ] ) via images only for the special case of a hard walled guide , the result ( usually derived perturbatively ) is in fact true for an arbitrary guide or confining potential.@xcite we have presented both the image series ( [ eq : wiregreens ] ) and the spectral series ( [ eq : dattagreenswire ] ) for the green s function . because the green s function of a system is unique , these two forms must be equivalent . however , the physical relationship between the method of images and the spectral form is far from obvious . the two forms of the green s function highlight different physical phenomena . for example , reflections from the walls , which appear immediately in ( [ eq : wiregreens ] ) , are far less obvious in the spectral form ( [ eq : dattagreenswire ] ) . contrastingly , the role of evanescent channels in the scattering , while straightforward in the spectral green s function ( [ eq : dattagreenswire ] ) , is less transparent in the image expansion ( [ eq : wiregreens ] ) of the same green s function ( although even the image expansion suggests that evanescent channels will be present in some form , because hankel functions are singular ) . in order to understand the connection between the image and spectral forms of the green s function , we show their equivalence mathematically in appendix [ sec : diffraction ] . in appendix a , we use an integral form of the hankel function to show the equivalence of ( [ eq : wiregreens ] ) and ( [ eq : dattagreenswire ] ) . the physical connection between the image and spectral formalisms is diffraction . the images in fig . [ fig : offcenter ] form a periodic array , which is effectively a diffraction grating . we note one small difference from a typical diffraction situation : due to the wire boundary conditions , the incident modes are superpositions of two plane waves each , rather than a single incident plane wave . this difference is trivial , due to the linearity of schrodinger s equation . when such modes are incident on the effective lattice of images , they diffract : plane waves ( or superpositions of plane waves ) striking the grating scatter , at large distances , into sums of plane waves at the bragg angles . the diffracted spectral orders are evident as the quantized wavenumbers appearing in the spectral form of the green s function . as the green s function for each image scatterer is singular , evanescent waves appear in the fresnel regime , near the lattice . we examine these statements rigorously in appendix [ sec : diffraction ] . while the method of images led quickly to the central results ( [ eq : mswf2],[eq : gr ] ) , these results expressed in terms of image sums converge so slowly as to be impractical for numerics . although the spectral form ( [ eq : dattagreenswire ] ) converges more rapidly than the image series ( [ eq : wiregreens ] ) , its convergence is still not uniform , as the evanescent modes include a logarithmic singularity . in appendix [ sec : kummer ] , we use kummer s method to accelerate the convergence of the green s function , casting it in the form ( [ eq : limitform ] ) suitable for numerical work . a side benefit of applying kummer s method is that we obtain a more rapidly converging expression for @xmath73 : in appendix [ sec : kummer ] , we show that the series ( [ eq : lambda ] ) for @xmath73 ( which is a schlmilch series ) resums to@xmath75+\frac{i}{2}-\frac{\gamma}{\pi}\label{eq : schlomilchgreens}\end{aligned}\ ] ] where @xmath76 is the euler - mascheroni constant . we note that this expression is considerably more complex than the expression ( [ eq : lambda ] ) derived via the method of images . in the image formalism , the expression for the renormalization constant @xmath73 has the intuitive form ( [ eq : lambda ] ) , because removing the contribution of the source simply involves excluding the scatterer itself , while retaining the images . in the spectral formalism , because the singularity is not explicit , its removal is considerably less transparent . in this subsection we define an effective cross section and optical theorem for the confined impurity . as we have shifted from the image formalism ( which applies only to the hard wire ) to the spectral one , where ( [ eq : dattagreenswire ] ) applies generally , the results presented in section [ sub : scattering - in - wire ] apply to arbitrary waveguides , not only hard wires . since the wire is infinite , for calculating transmission / reflection at infinity , we need consider only the open channels . without loss of generality ( due to translational invariance of the wire in the @xmath77 direction ) , suppose the scatterer is at @xmath78 with incident mode @xmath72 , normalized to unit flux , the wavefunction at large distances from the scatterer is@xmath79\label{eq : larger}\end{aligned}\ ] ] from which we can read off a scattering matrix@xmath80 where the reflection and transmission coefficients are@xmath81 one can show algebraically ( or verify numerically ) that @xmath82 in this section , we define an effective scattering cross section for the wire . the usual three - dimensional free space differential scattering cross section is@xmath83 where @xmath84 is the number of particles per area scattered into the solid angle @xmath85 , and @xmath86 is the number of incident particles per unit area . expressed in terms of probability currents , the above relation becomes@xmath87 where @xmath88 is the probability current , @xmath89 is the incoming probability current , and @xmath90 is the radial probability current . integrating over solid angles , we find that@xmath91 the optical theorem in free space relates the forward scattering amplitude to the cross section . here , we modify the above free space relations , in order to define a cross section for our confined geometry . with incident mode @xmath92 , the scattering wavefunction is@xmath93 which yields a probability current in the @xmath77 direction@xmath94 integrated over @xmath95 the cross terms cancel , and this current yields a scattered flux@xmath96 going in each direction . as our basis is normalized to unit flux , in our case , the incident flux is simply 1 . we define a differential cross section for the @xmath69 incoming mode , similar to the free space differential cross section ( [ eq : freespxsec]):@xmath97 where we note the exact analogy between plane waves in our waveguide and partial waves in free space . note also that @xmath98 has the correct dimensions of length . in free space , the cross section of a spherically symmetric scatterer is independent of the direction of the incident plane wave . in the waveguide , however , the cross section depends on the incoming mode , because the waveguide breaks spherical symmetry . we can define a total cross section as the average over incoming directions,@xmath99 we note that @xmath100 represents a sort of fraction of the incoming wavefunctions which scatter . flux conservation imposes the bound@xmath101 the maximal value of @xmath100 will , therefore , be @xmath102 , and its minimal value will be 0 . we had previously mentioned that @xmath103 this reduced rank indicates that , as in free space , a basis exists in which a single wavefunction scatters , while the remaining @xmath104 do not . the quantity @xmath105 thus tells us the fraction of the flux of this scattering wavefunction , and should be between 0 and 1 . henceforth we will speak of @xmath106 as the cross section ( as a fraction of the wire width ) . our final cross section is thus@xmath107 and satisfies@xmath108 we note that for @xmath109 impurities in the wire , assuming there are @xmath110 modes available , @xmath111 by arguments analogous to the single scatterer case , with @xmath109 scatterers present , we could choose a basis where only @xmath109 wavefunctions would scatter , while the remaining @xmath112 would be transmitted without any perturbation . the usual free space optical theorem involves the imaginary part of the forward scattering amplitude . one can show that the free space optical theorem constrains @xmath3 via@xmath113 a similar relation holds in the wire , and as it turns out , flux conservation implies that not every value of @xmath114 is physically permissible . unitarity of @xmath115 implies that@xmath116 using ( [ eq : rmatrix]-[eq : trmatrix ] ) in ( [ eq : unitaritycondition ] ) , after some algebra , we find@xmath117 which requires that @xmath114 satisfy the constraint@xmath118 we can express ( [ eq : optical ] ) as an optical theorem , like the one in free space . define the `` forward scattering amplitude '' to be the amplitude which scatters in the direction of the incoming wave:@xmath119 with this definition , our `` optical theorem '' in the wire becomes@xmath120 which is very similar to the 2-d free space optical theorem , @xmath121 where @xmath122 is the forward scattering amplitude . combining ( [ eq : crossmodes ] ) and ( [ eq : optical ] ) allows us to define a cross section,@xmath123 we note that our cross section is trivially related to the conductance of the wire , calculated as in the landauer formalism@xcite as @xmath124 noting that @xmath125 and letting @xmath6 equal the number of open channels,@xmath126 where we have used ( [ eq : rmatrix ] ) , ( [ eq : optical ] ) , and ( [ eq : sigmacalculation ] ) . in this section , we have laid out the formalism which we shall use to examine scattering in the wire . in section [ sec : results - and - discussion : ] , we use our formalism to examine how the cross section and conductance of the confined scatterer vary as the physical parameters of the wire change . confinement - induced effects on transport , in particular `` conductance dips '' as each new mode opens , have been observed experimentally @xcite as well as in other theoretical investigations@xcite of impurity scattering in quantum wires . changes in the transport due to motion of a single defect are studied in relation to universal conductance fluctuations.@xcite theoretically , the observed phenomena are generally explained via mode - mixing effects,@xcite or loosely attributed to multiple scattering from the walls.@xcite in this section , we discuss some confinement induced phenomena as we specifically observe them in our wire . the first quantity we wish to examine is how the cross section of the confined scatterer relates to the free space cross section of the identical scatterer , as a function of energy . the cross section of a free space scatterer is@xmath127 fig . [ fig : freespace ] shows that one feature resulting purely from confinement is the appearance of sharp discontinuities in the cross section as new modes become available ; the effect is present only if the newly opened mode is nonzero at the scatterer . to the left of each mode opening , the scatterer appears entirely transparent , whereas to the right of the mode opening its cross section is the full width of the wire . another point to note is that , while the free space cross section decreases monotonically with @xmath25 , the cross section of the confined impurity does not , and can be nonzero at arbitrarily high @xmath25 . we wish to examine some of the features of the confined cross section in fig . [ fig : freespace ] in greater detail . in particular , we want to look at the resonances , and understand the effect of varying the scatterer position [ fig : xseccond ] shows how the cross section and conductance vary as functions of energy , for three different values of @xmath128 . we observe the general property that resonances in the cross section appear as new transverse modes open @xmath129 unless the newly opened mode is zero at the scatterer . we further observe that the resonances have a universal structure : if a resonance exists at @xmath130 , then@xmath131 as the scatterer position changes , the character , shape , and number of resonances change also . in fig . [ fig : xseccond]a ) , the scatterer is located at @xmath132 , where each of the modes is nonzero . each resonance obeys ( [ eq : sigmaleft]-[eq : sigmaright ] ) . in fig . [ fig : xseccond]c ) , the scatterer position is @xmath133 because @xmath134 , the resonance at @xmath135 is much narrower than the others , and will in fact vanish if we move the scatterer slightly to @xmath136 . in fig [ fig : xseccond]e ) , we observe such missing resonances with the scatterer in the precise center of the wire , all even modes are zero there , and the resonances at @xmath137 have vanished . figs . [ fig : xseccond]b ) , [ fig : xseccond]d ) , and [ fig : xseccond]e ) show the conductance for each scatterer position , which is related to the cross section by ( [ eq : conductancexsec ] ) . discontinuities such as ( [ eq : sigmaleft]-[eq : sigmaright ] ) in the cross section lead to a continuously varying conductance . the impurity reduces the conductance by at most one stairstep . comparing the conductance with ( heavy line ) and without the impurity ( thin line ) , we observe that to the right of each mode opening , the conductance falls , generally to the value of the next - lowest stairstep . from fig . [ fig : xseccond ] , our observations are as follows : ( 1 ) resonances occur for values of @xmath25 at which a new mode opens , unless the new mode is zero at the scatterer . ( 2 ) the general structure of a resonance is that the scatterer becomes transparent immediately before a mode opens , and that its cross section jumps to one immediately after the mode has opened . ( 3 ) the shape and width of the resonances depend on the scatterer position , @xmath138 in section [ sec : physical ] , we shall prove these statements , and explain them in terms of interference effects . although we have examined the case of a repulsive scatterer in fig . [ fig : xseccond ] , our analytic and numerical results indicate that even for an attractive impurity ( @xmath139 ) the conductance is reduced by a single unit immediately after the subband opening @xmath130 , rather than just below it . similar results about the conductance reduction near the opening of each subband appear in other theoretical investigations of quantum wires ; we cite some representative works.@xcite one result@xcite is for a scatterer in a rectangle , and therefore can not be exactly compared to ours . however , the remaining references@xcite treat the infinite wire , and show a slight difference from ours : in refs . 6 and 7 , which examine only attractive impurities , the reduction in conductance appears slightly _ below _ the subband opening . 10 examines both attractive and repulsive scatterers . for repulsive scatterers , ref . 10 presents results similar to our fig . [ fig : xseccond ] , while for attractive scatterers the results of ref . 10 resemble those in refs . 6 and 7 . because delta functions in more than one dimension do not scatter , many different methods exist of representing a point impurity in two dimensions . we have chosen to use a @xmath5 matrix representation as in refs . 12 , 16 , and 17 , and the particular @xmath5 matrix we have chosen in ( [ eq : optical1 ] ) mimics the @xmath3 wave scattering of a hard disk of radius @xmath4 . the references whose results differ from ours have used explicit limiting forms of delta potentials . we attribute the slight difference in results to the different forms and representations of the point impurity . we have observed that the scatterer position in the wire influences not only the width and shape of the resonances , but can even cause a resonance to disappear . between resonances , we note the opposite phenomenon : the scatterer position barely affects the cross section at all . as we vary the scattering length @xmath4 , the cross section behaves very similarly to the free space cross section ( see fig . [ fig : crosssection ] ) . our general observations from the numerical results are the following : ( 1 ) scattering resonances occur where nonzero modes open . ( 2 ) the cross section jumps from zero to @xmath140 across a resonance . ( 3 ) the width and existence of a resonance is influenced by @xmath138 ( 4 ) away from resonances , the cross section behaves like the free space cross section . we have shown that items [ enu : newwf]-[enu : renorm ] in our list of section [ sub : freespacemultiple ] are characteristics of free space scattering which have analogs in the wire . we have yet to examine whether item [ enu : specialwf ] has an analog as well . in this section , we show that in the waveguide , as in free space , a basis exists in which only a single wavefunction scatters . we write this basis explicitly , and show that the scattering of this single wavefunction allows us to explain the confinement - induced phenomena shown in figs . ( [ fig : freespace]-[fig : crosssection ] ) simply , in terms of reflections from the scatterer and its images . in the basis of plane waves , the scatterer couples the channels so that , in general , an incoming mode scatters into all the modes . however , @xmath141=1,\label{eq : ranks}\ ] ] which implies that a particular choice of basis exists in which only @xmath142 of the incoming wavefunctions scatters at all , and is decoupled from the other @xmath143 basis functions . this wavefunction is the analog of the @xmath3 wave in free space . another way of seeing that only one wavefunction scatters is by the following argument : for a given energy , @xmath6 channels are open . any incoming wavefunction is a linear combination of the @xmath6 basis functions@xmath144 ( we require only the left- or right - moving set ) . any linear combination@xmath145 of the @xmath6 basis functions ( [ eq : inclr ] ) which is nonzero at the scatterer will satisfy@xmath146 equivalently , define an @xmath147 matrix @xmath148 such that @xmath149 and a vector of coefficients @xmath150.$ ] then , for any scattering wavefunction , the @xmath151 will satisfy @xmath152 as @xmath148 is a rank 1 matrix , only one solution exists . that solution is simply @xmath153 that is , the single scattering wavefunction is@xmath154 where the choice of sign determines whether the incoming wave is right or left - moving . suppose we take a linear combination of the two so the scattering wavefunction is symmetric in @xmath155 the scattering wavefunction becomes@xmath156\label{eq : swfplane}\ ] ] ( the other linear combination does nt scatter ) . by comparison with ( [ eq : dattagreenswire ] ) , we recognize the above form of the scattering wavefunction as@xmath157 note that all the definitions in this section , and in particular ( [ eq : speqimg ] ) , are independent of the specific form of @xmath158 , and thus apply to arbitrary guides , not only to the hard wire . as we shall discuss later , ( [ eq : speqimg ] ) is particularly important because it shows that for an arbitrary guide , we can obtain the scattering wavefunction entirely from the green s function , which we can approximate semiclassically . we now return to the specific case of the hard wire , and try to understand the scattering wavefunction physically . from ( [ eq : wiregreens ] ) , an alternate form of the green s function is@xmath159 where the @xmath160 are the image positions . combining ( [ eq : speqimg ] ) and ( [ eq : hankelgreens ] ) , we find an alternate expression for the scattering wavefunction:@xmath161 eq . ( [ eq : specialbessel ] ) shows that the analog of the free space @xmath3 wave in the hard wire is , as we might have expected , simply the free space @xmath3 wave , plus an infinite series of images a `` hall of mirrors @xmath3 wave . '' we can complete the analogy between free space and our wire by expressing the @xmath143 unscattered wavefunctions in terms of a `` mirrors '' basis . we do so by including higher order mirrored waves as in table [ table : swfs ] . @xmath162 ( twelve modes open)@xmath3 & @xmath163 & @xmath164 & @xmath165 & @xmath166 & @xmath167 & @xmath168 & @xmath169 & @xmath170 & @xmath170 & @xmath170 the mirroring , and the signs assigned to each image , are somewhat subtle : only those combinations of free space orbitals which satisfy the boundary conditions , but are also zero at the scatterer , will not scatter . we note that the basis in table [ table : swfs ] is @xmath171 an orthogonal one ; inner products of an @xmath172 wave centered on one image and a @xmath140 wave centered on another image , for example , do not vanish , nor do the various such cross - terms cancel each other . however the important point is that the higher partial waves we have defined are all orthogonal to the mirror @xmath3 wave : higher mirrored partial waves do not scatter , whereas the mirror @xmath3 wave does . we note that the image sums in table [ table : swfs ] converge slowly and are unsuitable for numerical purposes . however , we have the plane wave expansion ( [ eq : swfplane ] ) for the @xmath3 wave , as well as the analogous expansion@xmath173\label{eq : psisplus}\end{aligned}\ ] ] for the @xmath3 wave plus @xmath174 images . we can apply the raising operator @xmath175 to these plane wave expansions to obtain plane wave expansions for the higher partial waves . for example , we can obtain plane wave expansions of the waves in table [ table : swfs ] as@xmath176 & = & \frac{1}{k}\partial_{x}\phi_{s}(\vec{r})\\ \phi_{d_{xy}}(\vec{r } ) & = & { \rm im}\left[\hat{l}_{+}^{2}\phi_{s}^{+}(\vec{r})\right ] & = & \frac{2}{k^{2}}\partial_{xy}\phi_{s}^{+}(\vec{r})\\ \phi_{f_{x^{3}-3xy^{2}}}(\vec{r } ) & = & { \rm re}\left[\hat{l}_{+}^{3}\phi_{s}(\vec{r})\right ] & = & \frac{1}{k^{3}}\left(\partial_{x}^{3}-3\partial_{x}\partial_{y}^{2}\right)\phi_{s}(\vec{r}),\end{array}\ ] ] and we could proceed similarly to obtain higher partial waves . we might expect that the cross section is related to the fraction of the single scattering wavefunction which actually scatters , and we might thus expect the cross section to be proportional to the amplitude of the mirror wavefunction at the scatterer . however , some numerics reveal that the mirror wavefunction alone is insufficient to fully describe the scattering . from ( [ eq : crossmodes ] ) and ( [ eq : swfplane ] ) , we find@xmath177 meaning that the relevant quantity is the @xmath178 mirror wavefunction , and that its value at the scatterer alone explains all the scattering phenomena . ( [ eq : sigmaswf ] ) is again independent of the form of the waveguide . we have already discussed how renormalization of the scattering wavefunction arises from interference between all possible bounces off images . the implication of ( [ eq : sigmaswf ] ) is that the value of the resulting interference pattern at the location of the scatterer determines the cross section : at configurations ( wire widths , incident wavenumbers , etc . ) for which this interference is destructive at the scatterer , the scatterer is transparent . if the system is configured so that the bounces interfere constructively at the scatterer , a resonance results . we wish to highlight another important point , which is that combining ( [ eq : sigmaswf ] ) with ( [ eq : gr ] ) and ( [ eq : speqimg ] ) , we find that@xmath179}\right|^{2}\right\ } .\label{eq : semiclassical}\ ] ] the significance of ( [ eq : semiclassical ] ) , which applies to an arbitrary wire , is that the cross section is fully determined by the green s function of the empty guide . this relation is important because the green s function is a quantity which we can approximate semiclassically . substituting the semiclassical green s function into ( [ eq : semiclassical ] ) will give us a semiclassical approximation to the cross section . semiclassical methods may be of use in determining the transport properties of guiding potentials which are more complicated than the hard wire , and thus not amenable to exact quantum treatment . in figs . [ fig : freespace]-[fig : xseccond ] , we observed a reduction in the conductance , which are similar ( although not identical ) to the conductance dips observed in other theoretical investigations.@xcite using our formalism , we can understand the resonances and conductance dips of figs . [ fig : freespace]-[fig : xseccond ] simply , in terms of interference between different paths . again the results of this section do not depend on the specific form of the transverse modes @xmath158 , and consequently apply to a general waveguide . as we showed in ( [ eq : bouncesfeynman ] ) , the renormalized mirror wavefunction contains contributions that have propagated from each of the image scatterers . we are interested in the behavior of a cross section when a mode is about to open . the mirror wavefunction involves only open channels , and is thus discontinuous across a mode opening . using ( [ eq : swfplane ] ) , we can examine the behavior of the mirror wavefunction near @xmath180 , where the @xmath181 mode opens . suppose that the newly opened mode is nonzero at the scatterer , @xmath182 on the left of the mode opening,@xmath183 is finite . immediately after the mode opening , @xmath184 diverges as @xmath185 that is , the mirror wavefunction is always finite immediately before a nonzero mode opens , and diverges afterwards . the value of @xmath186 determines the width of the resonance ( and the lifetime of the corresponding quasibound state ) . however , while the limiting behavior of the mirror wavefunction influences the shape of the resonances , it does not fully describe their shape . as we discussed , the general structure of a resonance is that @xmath187 drops to zero just before the mode opens ; see e.g. fig . [ fig : xseccond ] . clearly the incoming wavefunction , which includes only modes which are already open , can not explain this transparency . the renormalization factor due to the wire , however , includes @xmath188 modes , both evanescent and propagating and unlike the incoming mirror wavefunction , varies continuously across the mode opening . using the expression ( [ eq : kummerform ] ) for @xmath73 , we find that@xmath189 generally . combining ( [ eq : renormalizedswfgr ] ) with the limits ( [ eq : mirrorleft]-[eq : grright ] ) , and using ( [ eq : sigmaswf ] ) , we can show that@xmath190^{2}}{\sqrt{n^{2}-n^{2}}}\right|^{2}.\label{eq : limitsigmaepsilon}\end{aligned}\ ] ] the value of @xmath191 controls the width and structure of the resonance . examining the right side of the resonance,@xmath192 under the assumption that @xmath193 an important point to note is that when @xmath194 as it does at a resonance , the free space properties of the scatterer have little effect on the cross section ( see for example ( [ eq : sigmaright1 ] ) ) , and so the structure of resonances is universal . we would like to note , in passing , that the transparency of the scatterer at certain energies is reminiscent of the ramsauer - townsend effect in free space . in free space , we define the @xmath3 wave phase shift @xmath195 of an @xmath3 wave scatterer by@xmath196 let us define it in the confined scatterer as the phase shift of @xmath197 @xmath3 wave generated from the source and the images , when the incident wavefunction is our mirror @xmath3 wave:@xmath198 comparing to the form ( [ eq : mswf ] ) of the lippmann - schwinger equation , we find that@xmath199 applying ( [ eq : sigmaswf ] ) , we find that@xmath200 so that @xmath201 when @xmath202 and @xmath203 when @xmath204 as in the free space ramsauer townsend effect , the cross section vanishes when our analog of the @xmath3 wave phase shift does . in conclusion , we can make the following general statements about the cross section : 1 . if @xmath205 the cross section will be discontinuous at @xmath206 the limit on the left hand side will be finite , and depend the value of @xmath207 the limit on the right hand side will be unity . 2 . the width and shape of the resonances in item ( 2 ) will depend on @xmath191 . physically this means that one can tune the resonances by sliding the scatterer up and down in the wire , or by changing the nature of the confining potential . having explored the structure of the resonances , we wish to understand them semiclassically , from an interference point of view . we again turn to the hard wire for insight into the scattering processes . the scattering cross section depends only on a single quantity , the renormalized mirror wavefunction . we have previously expressed this renormalized mirror wavefunction as a sum over the different paths ending on the scatterer ( [ eq : bouncesfeynman ] ) . when the interference between these paths is maximally constructive , the cross section is maximal . when it is maximally destructive , the cross section vanishes . in order to make this interference clearer , we shall make a simple approximation . consider the most straightforward case , with the scatterer in the center of the wire ( the case we examined in fig . [ fig : freespace ] ) . the image positions are @xmath208 in the green s function ( [ eq : wiregreens ] ) , we replace each hankel function by its asymptotic form , which is equivalent to making a semiclassical approximation . this yields an approximate green s function@xmath209 the contribution from each image carries a phase related to the path length from the image to the observation point , as well as a maslov index @xmath210 describing the sign change after @xmath72 reflections from the wire walls . in the limit that @xmath211 , we see that the contribution propagating to the scatterer from the @xmath69 image has relative phase @xmath212 . for values of @xmath25 such that@xmath213 all the scattered wavelets thus interfere constructively . comparing with fig . [ fig : freespace ] , we see that this constructive interference coincides precisely with the resonances which occur as new modes open . to fully explain the structure of the resonances , and understand the scatterer s transparency just before modes open , we would have to consider the effects of renormalization in this approximation . we shall not pursue the semiclassical limit further here , as we have already examined the exact case in great detail in section [ sub : universal - conductance - reduction ] . however , purely from this simple semiclassical argument , we can see that the resonances induced by the wire are indeed related to interference effects between different reflections . combining the method of images with the idea of a single scattering wavefunction , we have developed a formalism with which to treat scattering from a single impurity confined in a quantum guide . we find many similarities between scattering in the confined geometry and scattering in free space : we have defined meaningful analogs of the free space cross section , optical theorem , partial waves , and ramsauer - townsend effect . although the hard guide is particularly useful for insight , many of our results apply to other types of waveguides also . we have examined the transport properties of a confined scatterer , making several general observations about the existence , locations , and and characters of the resonances . additionally , we have outlined a method for approximating the cross section and conductance semiclassically , for arbitrary guides . we have derived an effective optical theorem for general quantum wire potentials . using the fact that a confined target is rank one , we have described the transport properties of the wire entirely in terms of a single scattering wavefunction , renormalized by reflections from the confining potential . our central result is that the cross section of a confined impurity is@xmath214 where @xmath215 is the renormalized single scattering wavefunction , which includes interference effects due to reflections from the waveguide walls . despite the common use of the hard - walled guide as a theoretical model for 2deg quantum wires,@xcite actual confining potentials for both atom waveguides as well as 2deg quantum wires tend to be soft or even parabolic whereas in nanotube quantum wires , the relevant boundary conditions are periodic ( the periodic boundary case is also exactly solvable via the method of images , although we have not presented the derivation here ) . while the hard wall itself is not the best model for 2degs , it is the ideal system in which to examine how effects like conductance reduction , which are independent of the confining potential and occur in more realistic geometries , arise from simple reflection and interference phenomena . the effects of multiple scattering from the impurity are in fact very similar for general confining potentials ( see e.g. the renormalized t matrix formalism@xcite ) . a second limitation of this work includes the inadequacy of the @xmath3 wave scatterer model at high energies , and in particular at energies which are sometimes relevant when imaging electron flow . in a future publication,@xcite we plan to extend our formalism to treat higher partial waves . another possible extension of this work could be to examine the transport properties of many scatterers in the wire ; for example , explicitly explaining the transition from ballistic to diffusive transport in terms of interference between multiple mirror wavefunctions . jyv was funded during the earliest stages of this work by a fellowship from the national defense science and engineering graduate ( ndseg ) fellowship program . subsequent research was funded by a grant from the national science foundation , nsf che-0073544 , and by the nanoscale science and engineering center ( nsec ) under nsf grant phy-0117795 . jyv is grateful to m. stopa for a helpful critical reading of the manuscript . 24 natexlab#1#1bibnamefont # 1#1bibfnamefont # 1#1citenamefont # 1#1url # 1`#1`urlprefix[2]#2 [ 2][]#2 , , , * * , ( ) . , , , * * , ( ) . , * * , ( ) . , , , , * * , ( ) . , * * , ( ) . , * * , ( ) . , , , * * , ( ) . , , , * * , ( ) . , _ _ ( , ) . , * * , ( ) . , , , , , , * * , ( ) . , * * , ( ) . , * * , ( ) . , , , , * * , ( ) . , _ _ ( , , ) . , thesis , ( ) . , , , * * , ( ) . , * * , ( ) . , * * , ( ) . , _ _ , vol . ( , , ) . , * * , ( ) . , , ( ) . , _ _ ( , , ) . , * * , ( ) . in sec . [ sec : greensfunction ] , we used the mathematical equivalence of the wire to a periodic array of image scatterers . further extending the analogy , we now reformulate the problem of the confined impurity as one of diffraction from an infinite , periodic grating , and the scattered wavefunction as a diffracted beam . we wish to note the detail that because of our dirichlet boundary conditions , our incident beam is a sum of two plane waves rather than a single plane wave:@xmath216 due to the linearity of schrodinger s equation , we can consider each of the constituent plane waves as scattering independently . the relevant physics thus reduces to diffraction of a single plane wave incident on a periodic array of scatterers . when our incident wavefunction strikes the diffraction grating , we expect that in the fraunhofer limit , far from the array , the diffracted beam will be a sum of plane waves at the bragg angles ( see fig . [ fig : diffractedbeam ] ) . each order of the diffracted beam will have a different weighting ; our object is to cast our green s function ( [ eq : wiregreens ] ) in a manner that makes these weightings explicit . this formulation is of interest because it is equivalent to determining the transmission and reflection coefficients of our system . namely , if we send in a mode , we would like to know what modes come out and with what intensities , which is precisely what we shall find by casting our scattering wavefunction as a diffracted beam . to begin , we note that bessel functions are regular , and consequently have plane wave expansions . in fact , the well - known integral form of the bessel function@xmath217 expresses the bessel function as an isotropic sum of plane waves emanating in all directions from the origin . we can rewrite ( [ eq : intbessel ] ) in terms of a line integral over @xmath218as@xmath219 where @xmath220 lies along the real axis , going back and forth within the interval @xmath221 . we seek an analogous expression for the hankel function . since @xmath222 is always less than @xmath25 in ( [ eq : intbessel2 ] ) , @xmath223 is always real , and the plane waves in ( [ eq : intbessel2 ] ) do not include evanescent waves . suppose we modify the right - hand side of ( [ eq : intbessel2 ] ) . by permitting @xmath218 to range from @xmath224 to @xmath225 , we may include contributions from all possible evanescent waves , yielding@xmath226 for @xmath227 the corresponding value of @xmath228 , where @xmath229 is imaginary . choosing the positive value makes the evanescent waves die off to the right @xmath230 and diverge to the left @xmath231 therefore , with this choice of sign , we may expect ( [ eq : evanescent_sum ] ) to converge only in the right half - plane . by adding on an evanescent contribution to the bessel function s plane wave expansion , we have in fact arrived ( up to a constant ) at an integral form of the hankel function , valid in the right half - plane only , @xmath232 where@xmath233 eq . ( [ eq : inthankel ] ) is a well - known expansion,@xcite readily obtained by fourier transforming green s equation . consider now the sum@xmath234 which is the green s function for a fully periodic array of scatterers at @xmath235 ( or equivalently , a periodic wire with walls at @xmath236 and a scatterer halfway between ) . based on the physical picture of diffraction ( fig . [ fig : diffractedbeam ] ) , the sum in ( [ eq : althanksum ] ) must be a superposition of plane and evanescent waves at real and complex bragg angles respectively . we wish to find the weighting on each wave ( analogous to the structure factor in x - ray diffraction ) . to find the weights , we calculate the expansion explicitly . using ( [ eq : inthankel ] ) in ( [ eq : althanksum ] ) , we find @xmath237 where we have used@xmath238^{n}=2\pi\sum_{n=-\infty}^{\infty}\delta(k_{y}d-2\pi n),\ ] ] ( similar to stationary phase ) to reach the expression ( [ eq : poisson ] ) . one can , more rigorously , reach ( [ eq : poisson ] ) via the poisson sum formula.@xcite performing the integral in ( [ eq : poisson ] ) yields the final expression@xmath239 where we use the absolute value signs to extend the green s function to converge on @xmath240 and we define@xmath241 where the @xmath242 are the bragg angles@xmath243 and the complex - valued @xmath242 are the result of the logarithmic singularity . the image representation of our wire is not exactly the periodic array of fig . [ fig : diffractedbeam ] . the images alternate in sign , and in general the scatterer is off - center . we can , however , represent our image array as a sum of two arrays , as in fig . [ fig : twoarrays ] . one can easily verify that the green s function for the wire with point source at @xmath244 becomes the sum@xmath245\nonumber \\ & = & -i\sum_{n=1}^{\infty}\frac{1}{k_{x}^{(n)}}e^{ik_{x}^{(n)}|x - x_{0}|}\label{eq : imagegreensequal}\\ & & \hspace{.5in}\times\chi_{n}(y)\chi_{n}(y_{0})\nonumber \end{aligned}\ ] ] where the bragg angles are modified slightly from the case of the periodic wire , and defined by@xmath246 the form ( [ eq : imagegreensequal ] ) is identical to ( [ eq : dattagreenswire ] ) . we have shown that diffraction translates the method of images into the usual eigenfunction expansion of the green s function . form ( [ eq : dattagreenswire ] ) of the green s function converges slowly also . the reason for the slow convergence is the singularity at the source , which is present but disguised in the evanescent modes . we use the kummer method for convergence acceleration,@xcite which also makes the singularity explicit . kummer s method involves adding and subtracting a multiple of the static ( @xmath247 ) green s function . we note that [ 12 ] obtains similar results , via a slightly different procedure . we wish to consider a more rapidly converging form of the green s function:@xmath249\label{eq : quickconverge}\\ & & \hspace{.5in}+\alpha g_{w}(\vec{r},\vec{r}_{0};0)\nonumber \end{aligned}\ ] ] where @xmath250 is a constant which we shall determine . the static green s function happens to be the potential which solves poisson s equation for a point charge between two grounded , conducting plates,@xcite and we can put it in closed form as follows:@xmath251e^{-\frac{m\pi}{d}\left|x - x_{0}\right|}\nonumber \\ & = & \frac{1}{\pi}\textrm{re } \sum_{m=1}^{\infty}\frac{1}{m}e^{-\frac{m\pi}{d}\left|x - x_{0}\right|}\label{eq : g24}\\ & & \hspace{.5in}\times\left[e^{\frac{im\pi(y+y_{0})}{d}}-e^{\frac{im\pi(y - y_{0})}{d}}\right]\nonumber \\ & = & \frac{1}{\pi}\textrm{re } \sum_{m=1}^{\infty}\left(\frac{z_{+}^{m}}{m}-\frac{z_{-}^{m}}{m}\right)\label{eq : gw5}\end{aligned}\ ] ] where@xmath252 using the identities@xmath253 and@xmath254 we find@xmath255\nonumber \\ & = & \frac{1}{2\pi}\ln\frac{\cos\left[\frac{\pi}{d}(y - y_{0})\right]-\cosh\left[\frac{\pi}{d}(x - x_{0})\right]}{\cos\left[\frac{\pi}{d}(y+y_{0})\right]-\cosh\left[\frac{\pi}{d}(x - x_{0})\right]}.\nonumber \end{aligned}\ ] ] the singularity of the green s function arises from a logarithmic singularity in the hankel term at the source . near the scatterer , the contribution from this term becomes@xmath256 where @xmath76 is the euler - mascheroni constant , while from eq . ( [ eq : limitform ] ) , we find that the limiting behavior of the static green s function is ( after some algebra)@xmath257 in order to cancel the logarithmic singularity in green s function , we thus require that@xmath258 so that@xmath259 & = \nonumber \\ & & \hspace{-1.5in}-\frac{1}{\pi}\ln\left[\frac{kd}{\pi}\sin\left(\frac{\pi y_{0}}{d}\right)\right]+\frac{i}{2}-\frac{\gamma}{\pi}\label{eq : limit6}\end{aligned}\ ] ] our final form for the green s function , using ( [ eq : quickconverge]-[eq : gamma ] ) is@xmath260-\cosh\left[\frac{\pi}{d}(x - x_{0})\right]}{\cos\left[\frac{\pi}{d}(y+y_{0})\right]-\cosh\left[\frac{\pi}{d}(x - x_{0})\right]}.\nonumber \end{aligned}\ ] ] this expression is of further use in that it allows us to obtain another form of the renormalization constant ( [ eq : lambda ] ) , which we had shown to have the slowly convergent expansion@xmath261 we can now express @xmath73 in an equivalent , but more rapidly convergent expression , suitable for numerical purposes:@xmath262\label{eq : limit9}\\ & = & \sum_{m=1}^{\infty}\left(\frac{1}{ik_{x}^{(m)}}+\frac{d}{m\pi}\right)\chi_{m}^{2}(y_{0})\nonumber \\ & & \hspace{.25in}-\frac{1}{\pi}\ln\left[\frac{kd}{\pi}\sin\left(\frac{\pi y_{0}}{d}\right)\right]+\frac{i}{2}-\frac{\gamma}{\pi}.\label{eq : kummerform}\end{aligned}\ ] ]

synchronization is a phenomenon observed in many different contexts , ranging from epidemics spreading @xcite to neuro - sciences @xcite . the analysis of simple models pursued in the last two decades has clarified several aspects of synchronization ( for a comprehensive review on the subject see ref . @xcite ) . particularly interesting both from a theoretical and an applied point of view is synchronization of chaotic systems . this is a phenomenon known since many years @xcite with important applications , e.g. , in secure communications of digital signals , laser dynamics etc . a simple but not trivial framework to study chaotic synchronization is represented by discrete - time dynamical systems . within this context , the essence of the phenomenon can be captured already by considering two coupled identical maps : @xmath3 where @xmath4 is a chaotic map of the unit interval and @xmath5 the coupling constant . for coupling larger than a critical value @xmath0 a transition from a desynchronized state to a completely synchronized one along a chaotic orbit @xmath6 is observed , i.e. @xmath7 for @xmath8 . the coupling @xmath0 can be predicted within a linear analysis as the value of @xmath5 for which the transverse lyapunov exponent ( tle ) @xmath1 , ruling the instabilities transversal to the line @xmath9 , vanishes . for model ( [ eq:2map_1 ] ) @xmath10 @xcite , and consequently @xmath11 being @xmath12 the lyapunov exponent ( le ) of the single ( uncoupled ) map . while in low dimensional systems the basic mechanisms ruling synchronization , in the absence of strong nonlinear effects , have been well established since long time @xcite , the studies devoted to synchronization in extended systems are still limited to few specific examples @xcite . in the latter case , two main frameworks have been considered . on one hand , the investigation of mutual synchronization between two replicas of a spatially extended system has been pursued for systems coupled either via local interaction @xcite or via spatio - temporal noise @xcite . in both cases , when the amplitude of the coupling ( noise ) overcomes a certain threshold , synchronization is observed : after a transient , the two systems follow the same spatio - temporal chaotic ( stochastic ) orbit . on the other hand , another commonly observed situation corresponds to self synchronization of elements belonging to the same system . this occurs for instance in large collections of coupled elements such as populations of neurons @xcite , josephson junctions @xcite or cardiac pacemaker cells @xcite . in these systems the interaction among the elements can range from nearest neighbors to globally coupled . therefore , a model able to encompass both these limiting cases is certainly of interest . a good candidate is represented by the following coupled map lattice with power - law interactions @xcite @xmath13 which reduces to globally coupled maps ( gcm s ) @xcite for @xmath14 , and to standard coupled map lattices ( cml s ) with nearest neighbor coupling @xcite in the limit @xmath15 . in equation ( [ eq : cml ] ) , @xmath16 and @xmath17 are the discrete temporal and spatial indexes , @xmath18 is the lattice size ( @xmath19 ) , @xmath20 the state variable , @xmath21 $ ] measures the strength of the coupling and @xmath22 its power - law decay . since the sum extends up to @xmath23 the model is well defined only for odd @xmath18-values , and @xmath24 is a normalization factor . periodic boundary conditions are assumed . the synchronization between replicas of the same spatial system has been the subject of a more systematic study than self synchronization and is now rather well understood @xcite . two different mechanisms of mutual synchronization have been identified according to the predominance of linear versus nonlinear effects . nonlinearly dominated transitions are observed when the local dynamics is ruled by a discontinuous ( e.g the bernoulli map ) or `` almost discontinuous '' maps ( e.g. maps possessing very high values of the first derivatives ) . in these cases a linear analysis is no more sufficient to fully characterize the transition , because the instabilities associated with finite amplitude perturbations may desynchronize the system . in both cases the transition to the synchronized state is typically of the second order . however , depending on the linear or non - linear nature of the prevailing mechanism two different universality classes characterize the transition itself . for continuous maps ( e.g. logistic maps ) critical exponents associated with the multiplicative noise ( mn ) universality class are usually found @xcite , while for ( almost ) discontinuous maps the transition belongs to the directed percolation ( dp ) class @xcite . the aim of the present work is to clarify the effect of strong nonlinearities in the synchronization phenomenon , in particular in the case of self synchronization . as a first point , we show that strong nonlinearities may induce nontrivial effects also in low dimensional systems such as two coupled maps . in particular , we find that even though the critical coupling @xmath0 can be always predicted within the linear framework , the transient dynamics preceding the synchronization can be strongly affected by finite scale instabilities . to properly characterize the latter we introduce a new indicator , the finite size transverse lyapunov exponent ( fstle ) , which generalizes the concept of finite size lyapunov exponent ( fsle ) @xcite to the transverse dynamics . the fstle extends the definition of tle to finite perturbation and is therefore able to discern linearly from nonlinearly dominated transitions . depending on the behavior of the fstle at finite amplitudes two different class of maps are singled out maps characterized by a decreasing fstle at any finite scale , and _ class ii _ maps that present a peak in the fstle for some finite amplitude value . moreover , in the proximity of the transition the shape of the probability density function ( pdf ) , @xmath25 , of the synchronization times @xmath26 depends on the dynamics at finite scales and on the multifractal properties of the map . for class i maps two cases can be identified . for maps , such as the symmetric tent map and the logistic map at the crisis , which are characterized by the vanishing of fluctuations of the finite time le at long times , the pdf exhibits a fast falloff at large @xmath26 s . in particular , for the symmetrical tent map we provide an analytical expression for @xmath25 . for maps , such as the skew tent map , exhibiting _ modulational intermittency _ @xcite ( which is related to the persistence of fluctuations of the finite time le also in the long time limit ) , @xmath25 becomes an inverse gaussian distribution @xcite originating by the diffusive motion of the ( logarithm of the ) perturbation in transverse space . for class ii maps the pdf s display an exponential tail at long times which , differently from the previous case , is due to nonlinear effects . as far as spatially extended systems are concerned , anteneodo et al . @xcite have performed a linear stability analysis of model ( [ eq : cml ] ) obtaining analytically the critical coupling @xmath0 for the transition . this prediction works perfectly for class i maps , as found for the coupled logistic maps @xcite . however , as shown in the present paper , linear analysis may fail in class ii maps . in particular , for large system sizes ( @xmath27 ) an exponential divergence of the synchronization times with @xmath18 can be observed even for negative tle , as a result in the limit @xmath28 the transition takes place at a critical coupling @xmath29 . finite scale instabilities are the key elements for observing this _ nonlinear synchronization transition_. similarly to the case of diffusively coupled map lattices it is possible to link the synchronization transition of cml with power law coupling with non - equilibrium phase transitions . indeed as shown in ref . @xcite synchronization of replicas of cml with short range coupling belong either to the mn or to the dp universality classes . here we shall discuss the connection of self synchronization of model ( [ eq : cml ] ) with long range spreading processes @xcite . the material is organized as follows . [ sec:2 ] is devoted to synchronization of two coupled maps . in sect . [ sec:3 ] coupled maps with power law coupling are examined . finally in sec . [ sec:4 ] we briefly summarize the reported results . to analyze the synchronization of two coupled maps ( [ eq:2map_1 ] ) it is useful to introduce the following variables : @xmath30 in terms of which eq . ( [ eq:2map_1 ] ) can be rewritten as @xmath31\nonumber\\ w^{t+1 } & = & \left ( \frac{1}{2}-\varepsilon \right ) \left[f(u^t+w^t ) - f(u^t - w^t ) \right ] \label{eq : diff}\,.\end{aligned}\ ] ] it is easily checked that the synchronized solution ( corresponding to @xmath32 and @xmath33 ) is an admissible solution of ( [ eq : diff ] ) for any value of the coupling @xmath5 . the stability of such a solution can be studied by considering the linear dynamics of an infinitesimal perturbation @xmath34 of the synchronized state . this evolves according to the linearized equation : @xmath35 where @xmath36 . clearly the stability of the state @xmath32 is controlled by the sign of the transverse ( or conditional ) lyapunov exponent @xmath37 where @xmath12 is the lyapunov exponent of the single map @xmath4 . therefore , by requiring @xmath38 one obtains the critical coupling @xmath0 ( [ eq : epsc ] ) above which the synchronized state is linearly stable . notice that for @xmath39 , @xmath1 coincides with the second lyapunov exponent @xmath40 of the coupled system ( [ eq:2map_1 ] ) . the synchronization transition has been mainly studied for continuous maps , here denoted as _ class i _ maps , in particular for the logistic map @xmath41 at the crisis @xcite and the skew tent map @xcite @xmath42 it should be noticed that inside this class of maps one has to distinguish between two situations according to the behavior of the fluctuations of the finite time lyapunov exponents in the long time limit . the vanishing of such fluctuations is observed for the symmetric tent map and the logistic one at the crisis , while their persistence characterizes the skew tent map that represents the generic case @xcite . recent studies however pointed out that discontinuous or `` almost '' discontinuous maps ( i.e. maps with @xmath43 in some point of the definition interval ) , here termed _ class ii _ maps , may give rise to non trivial interesting phenomena . for instance , the synchronization transition for two replicas of class ii cmls is not driven , as usual , by infinitesimal perturbations , but by the instability ( spreading ) of finite amplitude perturbations @xcite . this nonlinear synchronization is linked to the so - called _ stable chaos _ @xcite : a phenomenon characterized by information propagation even in the absence of chaos @xcite . moreover , as we shall show in the following also the self synchronization of a single cml can be strongly affected by nonlinearities . starting from random initial conditions . from top : bernoulli map with @xmath44 , logistic map at the crisis and the skew tent map with @xmath45 . for all maps , the coupling has been chosen in order to ensure @xmath46 . the dashed straight lines display the @xmath47 decay . ] representative examples of class ii maps are the generalized bernoulli map @xmath48 that is discontinuous and its continuous version @xmath49 with @xmath50 , @xmath51 , @xmath52 . as already mentioned , the peculiarity of such maps is that they are not only unstable with respect to infinitesimal perturbations but also to finite ones ( for ( [ eq : bsc ] ) this is true for sufficiently small @xmath53 values ) . the main differences between synchronization transitions ruled by linear mechanisms with respect to those driven by finite amplitude instabilities can be captured already by considering the evolution of @xmath54 . the typical behavior of this quantity for @xmath55 is reported in fig . [ fig : wt ] for the bernoulli , the logistic and the skew tent map . for the logistic map , apart from ( short time ) fluctuations due to the multiplier variations , an average linear decrease of @xmath56 with slope given by @xmath1 is observed . for the bernoulli map , despite the tle is negative and does not fluctuate in time , @xmath54 exhibits instantaneous jumps to @xmath57 values . these jumps , which are due to finite amplitude instabilities , are more probable for large @xmath54 amplitudes . completely analogous behaviors have been found for the map ( [ eq : bsc ] ) , where the multipliers do fluctuate . as shown in fig.[fig : wt ] for the skew tent map with @xmath58 , @xmath59 displays an intermittent evolution during the transient preceding the synchronization , that is reflected in a diffusive motion of @xmath60 . as a consequence long transients are observed close to the transition . this phenomenon , known as _ modulational intermittency _ @xcite is induced by fluctuations of the finite time tle that do not average out in the long time limit and as a consequence can be explained by linear analysis . for the sake of completeness , we mention that long synchronization transients have been reported also for nonlinearly coupled expanding maps @xcite . in this model resurgences of @xmath54 during the transient are due to the presence of an invariant chaotic repelling set . we stress that for class ii maps the transient is absolutely non chaotic i.e. their transverse dynamics in tangent space is contracting at any time . to quantitatively characterize the different _ transverse_-space dynamics ( i.e. the evolution of @xmath61 shown in fig . [ fig : wt ] ) for maps of class i and ii , let us now introduce the finite size transverse lyapunov exponent . the fstle , @xmath62 , generalizes the concept of transverse lyapunov exponent to finite value of the perturbation @xmath63 . following aurell et al . @xcite ( see also ref . @xcite ) we have defined the fstle as follows . we introduce a set of thresholds @xmath64 with @xmath65 , since on average a transverse expanding ( resp . contracting ) dynamics is expected in the desynchronized ( resp . synchronized ) regime , we choose @xmath66 ( e.g. @xmath67 ) for @xmath68 and @xmath69 ( e.g. @xmath70 ) for @xmath8 . first , starting from a random initial conditions we wait for @xmath6 to relax onto its attractor . then we initialize a second variable @xmath71 as @xmath72 by choosing an initial perturbation @xmath73 ( resp . @xmath74 ) if @xmath75 ( resp . if @xmath76 ) , and let @xmath6 and @xmath77 evolve according to eq . ( [ eq:2map_1 ] ) . care should , of course , be taken to maintain @xmath71 within the interval of definition of the map . during the evolution we record the time , @xmath78 , needed for @xmath54 to pass for the first time from one threshold @xmath79 to the following one @xmath80 and also the value of @xmath81 at the moment of the passage . when the last threshold , @xmath82 , is reached the system is reinitialized with the above described procedure and this is repeated @xmath83 times . the fstle is thus defined as @xmath84 where the average @xmath85\rangle$ ] is performed over the set of @xmath83 different initial conditions . in the limit @xmath86 the fstle converges to @xmath1 . notice that due to its definition , the fstle can not measure at the same time expansion and contraction rates , i.e. we have limited the analysis only to consecutive contractions ( resp . expansions ) in the synchronized ( resp . desynchronized ) regime . this implies that the sign of @xmath62 is always negative or positive in accordance with the investigated situation . in fig . [ fig : fsle ] @xmath62 versus @xmath87 is shown for different maps in the desynchronized regime . as expected , in all cases , for very small values of @xmath54 the tle is recovered . however , at larger values of @xmath54 maps of the class ii display an increase of the growth rate , i.e. , @xmath88 for some finite @xmath87 value , while for maps of class i the fstle is monotonically decreasing with @xmath87 . @xmath62 provides a quantitative measure of the strength of the nonlinear effects that , in maps of class ii , may in principle overwhelm the linear mechanisms as pointed out in ref . @xcite . vs @xmath87 in the desynchronized regime for the bernoulli map with @xmath89 ( empty circles ) , its continuous version ( empty boxes ) with @xmath90 , the logistic map ( filled circles ) and the skew tent map ( filled triangles ) . the coupling constants have been chosen in such a way that in all systems the transverse le is equal to @xmath91 ( solid line ) . ] in fig . [ fig : firstarrival ] we report the behavior of @xmath62 in the synchronized regime . again , while for very small @xmath87 the ( negative ) tle is recovered , at larger @xmath87 values the bernoulli - like maps display an increase which is due to the jumps @xmath92 of the transverse perturbation . vs @xmath87 in the synchronized regime for the bernoulli map at @xmath89 ( empty circles ) , the logistic map ( filled boxes ) . the inset show the logistic map ( filled boxes ) and the skew tent map ( empty boxes ) . the straight lines correspond to the transverse le , here @xmath93 for all systems . ] a closer inspection of the small @xmath87 regime reveals that for the logistic map @xmath94 only for very small scales @xmath95 , while for the skew tent map the asymptotic @xmath94 in never exactly reached . the latter convergence difficulties are probably due to the intermittent behavior along the transverse direction , in the proximity of the transition @xcite . in the regime @xmath96 the finite time effects are less important and the asymptotic behavior is recovered for all the investigated maps . the results presented in this subsection clearly indicate that finite scale instabilities can prevail on infinitesimal ones only for strongly nonlinear maps . notwithstanding the observed differences in the fstle , we found that the critical coupling for synchronization to occur is always given by ( [ eq : epsc ] ) for both classes . this means that , at least for the case of two coupled maps , nonlinear effects do never modify the critical coupling . it is natural to wonder whether other observables related to the synchronization transition could be influenced by the presence of finite amplitude instabilities . as we shall show in this section , this is the case for the synchronization times statistics . let us define the synchronization time @xmath26 , as the shortest time needed to @xmath97 for decreasing below a certain threshold @xmath98 , with the further requirement that @xmath99 for a sufficiently long successive time @xmath100 . however , if the threshold value is small enough ( e.g. @xmath101 ) @xmath26 essentially coincides with the first arrival time to the considered @xmath98 . the quantities of interest are the pdf s , @xmath25 , and their moments ( we shall focus mainly on the first moment ) . in fig . [ fig : pdftent ] we show @xmath25 for the the logistic and the symmetric tent map . the main feature is the presence of an exponential tail at short times and the faster than exponential falloff at long times . this means that synchronization times @xmath102 are not observed . as a first result we derive , by following ref . @xcite , an approximate analytical expression for @xmath25 , which is exact for the symmetric tent map . let us define @xmath103 , and consider its linearized evolution , from eqs . ( [ eq : tang_diff ] ) and ( [ eq : lambdat ] ) one obtains @xmath104 formally the above equation can be applied only in the true linear regime , i.e. when @xmath105 , so that it is not appropriate in the early stages of the evolution . however , as shown in fig . [ fig : wt ] , for maps like the logistic one , usually after a reasonably short transient , the linear regime sets in and the use of eq . ( [ eq : zet1 ] ) is justified for the successive evolution . from eq . ( [ eq : zet1 ] ) it is clear that @xmath25 is related to the pdf @xmath106 of the variable @xmath107 and to that of the local multipliers @xmath108 . the formal solution of eq . ( [ eq : zet1 ] ) up to time @xmath109 can be written as follows : @xmath110 where @xmath111 . for sufficiently large @xmath109 , large deviation theory tells us that the pdf of @xmath112 takes the form @xmath113 , being @xmath114 the cramer function @xcite , which is convex and has its minimum value at @xmath115 . it is now clear the distinction between the generic case in which @xmath114 does not collapse onto a @xmath53-function , and the non - generic one in which it does , as for the symmetric tent map and the logistic one at the ulam point . in the former case the dynamics of @xmath116 becomes a biased brownian motion , with an average drift given by @xmath1 . for the sake of simplicity , let us start from the tent map ( eq . ( [ eq : skt ] ) for @xmath117 ) , for which @xmath118 . in this case @xmath26 is simply given by : @xmath119 and @xmath25 can be directly related to @xmath106 . since , in the proximity of the transition , @xmath120 assumes the form ( analytically derived in ref . @xcite ) : @xmath121 it can be easily obtained the related distribution @xmath122 which perfectly agrees with the numerical results ( fig . [ fig : pdftent]a ) . as a function of the synchronization time @xmath26 : ( a ) tent map for @xmath123 , the dashed line is the analytical expression ( [ eq : p_tau ] ) ; ( b ) logistic map for @xmath124 , the solid line refers to the expression ( [ eq : p_tau_log ] ) with @xmath125 and @xmath126 . in all cases the pdf have been obtained by averaging over @xmath127 different initial conditions and by choosing @xmath128 . ] unfortunately the ( short time ) multiplier statistics of the logistic map is nontrivial , impeding a straightforward derivation of @xmath25 . however , we numerically observed that @xmath129 ( @xmath130 ) with @xmath131 almost constant in a neighborhood of the transition ( namely , @xmath132 for @xmath133 ) . this implies that @xmath134 which is in fairly good agreement with the numerically evaluated pdf ( fig . [ fig : pdftent]b ) . notice that @xmath135 is not simply given by @xmath136 , and a fitting procedure is needed . we found @xmath137 , with @xmath138 and 0.65 for @xmath139 and @xmath140 , respectively . the above reported results allow us also to predict the scaling of the average synchronization time @xmath141 with @xmath142 in the proximity of the synchronization transition . in particular , for the symmetric tent map the following expression can be derived @xmath143 where @xcite @xmath144 being @xmath145 the euler - mascheroni constant . since @xmath146 , approximatively we have @xmath147 where @xmath148 ( being @xmath149 ) . note that ( [ eq : tmed ] ) is in perfect agreement with the numerical results for the tent map ( see fig . [ fig : tmedtent ] ) . for the logistic map , a similar dependence on @xmath150 , namely @xmath151 , has been found . the interesting point in ( [ eq : tmed ] ) is the logarithmic correction to the scaling @xmath152 , which would have not been predicted by a naive guess based on ( [ eq : tau ] ) . vs @xmath153 for the tent map ( empty boxes ) and the logistic one ( empty circles ) . the continuous line is the prediction ( [ eq : tmed ] ) with @xmath128 , the dotted one is obtained by a best fit of the form @xmath154 , with @xmath155 and @xmath156 . the data of the logistic map have been shifted by a factor @xmath157 for plotting purposes . ] we now briefly discuss the generic situation ( see chap . 13 of ref . @xcite for more details ) . as mentioned above , @xmath158 performs a biased brownian motion , so that computing @xmath25 amounts to evaluate the distribution of the first passage times to a threshold of a wiener process with drift that , as a standard result of stochastic process theory , is given by an inverse gaussian density @xcite that for long times @xmath26 becomes @xmath159 where the quadratic dependence from the tle comes from the diffusive dynamics of the perturbation . for class ii maps the situation is completely different . for both the continuous bernoulli map ( fig . [ fig : pdfbsc]a ) and the discontinuous bernoulli shift ( fig . [ fig : pdfbsc]b ) we observe that for @xmath8 the pdf s are characterized by an exponential tail at large @xmath26 , similar to poissonian distributions . moreover , we also observe that the pdf s , once rescaled as @xmath160 , and reported as a function of @xmath161 ( where @xmath162 ) collapse onto a common curve in the proximity of the transition . these results are particularly striking in the case of the bernoulli map for which naively one would have expected a complete similarity with the tent map , since its multiplier @xmath163 is also constant . here the presence of strong nonlinear effects makes the transient time statistics to be dramatically different from ( [ eq : p_tau ] ) . indeed , as seen in fig . [ fig : wt ] , @xmath164 is subject to noticeably nonlinear amplifications induced by the ( almost ) discontinuities in the map . therefore , the expression ( [ eq : zet1 ] ) is no more appropriate to describe the dynamics of @xmath165 , and its full nonlinear dynamics has to be taken into account . the latter is characterized by the fact that , with a finite probability , @xmath54 can jump to @xmath92 values at any time during the transient preceding the synchronization , even if the transient is not chaotic . this idea can be better clarified and its consequences on the @xmath25 better appreciated by considering a simple stochastic model for the dynamics of the transverse variable at @xmath8 . this model was originally proposed in ref . @xcite , and reads as @xmath166 where @xmath167 and @xmath168 is a shorthand notation for with probability . the underlying idea is very simple : the transverse perturbation @xmath61 is usually contracted , but with probability proportional to its amplitude can be re - expanded to @xmath169 values , in agreement with the numerical observations . the jumps occur when @xmath6 and @xmath77 are close but located at the opposite sides of the ( almost ) discontinuity . as a function of @xmath170 ( where @xmath162 ) for the continuous bernoulli shift map ( [ eq : bsc ] ) with @xmath171 and @xmath172 at different distances , @xmath173 , from the critical coupling . note the fairly good collapse , indicating a common decay as @xmath174 ( dashed line ) . the collapse is however only approximate near the peak . ( b ) same as ( a ) for the bernoulli map with @xmath175 at different distances from the critical coupling . the dashed line gives @xmath174 . ] for this simple model , it is possible to derive the corresponding probability density @xmath25 , which displays the same peculiar features of the pdfs reported in figs . [ fig : pdfbsc]a and [ fig : pdfbsc]b . in particular , the long time tail . first , notice that the probability that an initial perturbation @xmath176 is never amplified to @xmath92 up to time @xmath16 is given by @xmath177 \to\ ! \bar q(w_0 ) \!=\!\lim_{t \to \infty } \exp\left [ -\sum_{k=1}^\infty \frac{w_0^k { \rm e}^{-|\lambda_t| k}}{k(1- { \rm e}^{-|\lambda_t| k})}\right ] . \label{eq : p0}\end{aligned}\ ] ] interestingly , for @xmath178 the quantity @xmath179 is always positive and strictly smaller than @xmath180 for any perturbation @xmath181 , i.e. the probability that a perturbation of amplitude @xmath182 can be amplified is finite at any time . the pdf of the times @xmath26 needed to observe @xmath183 can be factorized as @xmath184 where @xmath185 is the probability to receive a kick at time @xmath186 , and @xmath187 is the probability of not being amplified for the following @xmath188 steps , where @xmath188 is given by @xmath189 as shown in ref.@xcite @xmath190 , and therefore at large @xmath26 s eq . ( [ eq : ptau_teo ] ) can be rewritten as @xmath191 } \label{eq : ptau_teo_2}\ ] ] that confirms the poissonian character of the pdf @xmath25 for the maps of class ii . by properly normalizing eq . ( [ eq : ptau_teo_2 ] ) , one obtains @xmath192 , that is in good agreement with the numerical results for the bernoulli map ( in particular , we considered @xmath175 and @xmath193 ) . moreover , in the interval @xmath194 $ ] we found that the decay rate of the pdf is directly proportional to the tle , i.e @xmath195 , this should be contrasted with the quadratic dependence found in the case of class i maps , see eq . ( [ eq : inv_gaussian ] ) . we conclude the investigation of two coupled maps by studying the nature of the transition . for this purpose let us introduce the order parameter @xmath196 defined as ( see also ref . @xcite ) : @xmath197 i.e. for each value of the coupling @xmath5 , one considers @xmath109 different random initial conditions and each one is iterated for a transient @xmath198 after which the time average of @xmath54 over a time lapse @xmath199 is considered and further averaged over all the initial conditions . the value of the coupling giving the synchronization transition is then implicitly defined as @xmath200 . in principle , @xmath2 may differ from @xmath0 defined by ( [ eq : epsc ] ) , since this expression is valid only within a linear approximation formalism . vs @xmath5 for the bernoulli map with @xmath201 and its continuous version ( [ eq : bsc ] ) for different values of @xmath53 . for the bernoulli map one has that according to ( [ eq : epsc ] ) @xmath202 . the order parameter is computed according to ( [ eq : op ] ) averaging over @xmath203 different random initial conditions , discarding the first @xmath127 iteration and averaging over the following @xmath204 iterations . @xmath196 has been estimated with an @xmath5-resolution @xmath205 . the inset shows an enlargement of the critical region done with a coupling resolution @xmath206 , the same number of initial conditions and a longer integration transient @xmath207 iterations . ( b ) the same for the logistic map at the crisis . here ( [ eq : epsc ] ) predicts @xmath208 . the blow up of the critical region ( in the inset ) shows that the transition is continuous . note also that for @xmath209 we observe @xmath210.,title="fig : " ] vs @xmath5 for the bernoulli map with @xmath201 and its continuous version ( [ eq : bsc ] ) for different values of @xmath53 . for the bernoulli map one has that according to ( [ eq : epsc ] ) @xmath202 . the order parameter is computed according to ( [ eq : op ] ) averaging over @xmath203 different random initial conditions , discarding the first @xmath127 iteration and averaging over the following @xmath204 iterations . @xmath196 has been estimated with an @xmath5-resolution @xmath205 . the inset shows an enlargement of the critical region done with a coupling resolution @xmath206 , the same number of initial conditions and a longer integration transient @xmath207 iterations . ( b ) the same for the logistic map at the crisis . here ( [ eq : epsc ] ) predicts @xmath208 . the blow up of the critical region ( in the inset ) shows that the transition is continuous . note also that for @xmath209 we observe @xmath210.,title="fig : " ] in fig . [ fig : op ] we show the behavior of the order parameter ( [ eq : op ] ) as a function of @xmath5 for both classes of systems . two observations are in order . first , the transition to the synchronized state always occurs at the critical coupling defined by ( [ eq : epsc ] ) ( i.e. @xmath211 ) . second , the transition is continuous for the logistic map ( simulations show that also the tent map , which belongs to the same class , has a continuous transition ) and discontinuous for the bernoulli shift . this confirms the results reported in ref . however , since the transition for the continuous version of the bernoulli shift map ( [ eq : bsc ] ) is steep but clearly continuous ( see the inset of fig.[fig : op]a ) , this does not seem to be a generic property for all the maps of class ii . the behavior of the bernoulli shift map can be easily understood by considering the map controlling the dynamics of @xmath61 , i.e. @xmath212 being @xmath213 $ ] and @xmath2141/r:1]$ ] . from the above expression , it is evident that the attractor in the @xmath215 plane has always a finite width along the transversal direction , unless @xmath216 . this explains the discontinuity observed at @xmath0 . in the continuous map ( [ eq : bsc ] ) the attractor in the plane @xmath215 has a `` transverse '' width that decreases continuously to zero for @xmath217 . in conclusion , the discontinuity in the transition is a pathology of the bernoulli map , and possibly of other discontinuous maps . the order parameter ( [ eq : op ] ) can be seen as the time average , once discarded an initial transient , of the following quantity @xmath218 where @xmath219 indicates the average over many different initial conditions . once @xmath220 is initialized @xmath92 , depending on whether @xmath5 is smaller or larger than @xmath0 two different asymptotic behavior are observed . in the desynchronized regime , obviously @xmath221 goes to a finite value @xmath222 , while above the synchronization transition @xmath223 with a decay law determined by the nature of the considered maps . for class i maps , the decay is ruled by the tle : @xmath224 with @xmath167 . for class ii maps @xmath221 shows an initial exponential decay @xmath225 ( with @xmath226 ) , due to the effect of the resurgences averaged over many different initial conditions , followed by a final linear decay @xmath227 . the nonlinear rate @xmath228 can be simply related to the exponential tail of the pdf of the first arrival times , by assuming ( as indeed observed ) the following decay @xmath229}$ ] . since @xmath221 is the average amplitude value of the @xmath61 at time @xmath16 , one has @xmath230 the above result tells us that @xmath231 ( if the integral does not diverge , i.e if @xmath232 ) . this is confirmed by numerical checks . in this section we analyze high dimensional systems , namely the power - law coupled maps defined in eq . ( [ eq : cml ] ) . in particular , for class ii maps we shall show that there exist situations where , due to the nonlinearities , the synchronization time diverges exponentially with the number of maps . as a consequence , the transition takes place at a ( nonlinear ) critical coupling @xmath2 larger than the linear value @xmath233 . a similar phenomenon has been observed for gcm s with nonlinear coupling , where the synchronization time divergence is due to a chaotic transient @xcite . chaotic transients , diverging exponentially with the number of coupled elements , have been reported also for spatially extended reaction - diffusion systems @xcite and for diluted networks of spiking neurons @xcite . however , the emphasis of our work is on non - chaotic transients similar to the stable - chaos phenomenon @xcite . let us start by reviewing the linear theory developed in ref . @xcite , which is able to account for the synchronization transition of class i maps . when nonlinear effects are not sufficiently strong , excluding the pathological cases of chaotic transients , the critical coupling for observing the synchronization transition can be predicted by computing the lyapunov spectrum . in particular , since above the synchronization transition the transverse lyapunov exponent coincides with the second lyapunov exponent , @xmath40 , it suffices to evaluate the dependence of the latter on @xmath5 . in order to compute the lyapunov spectrum of the model ( [ eq : cml ] ) , it is necessary to consider the tangent space evolution : @xmath234 in ref . @xcite it has been shown that the full lyapunov spectrum can be easily obtained for the bernoulli map : @xmath235 where @xmath236 in this notation the maximal lyapunov exponent is given by @xmath237 and the second lyapunov exponent is @xmath40 . for generic maps it is still possible to obtain the lyapunov spectrum in an analytical way , but only in the synchronized state , by substituting @xmath238 with the maximal le of the single uncoupled map @xmath12 in ( [ eq : spectrum ] ) . from a linear analysis point of view one expects that the synchronization transition has to occur when @xmath239 . therefore , the following expression for the critical line in the @xmath240-plane can be derived : @xmath241 as shown in ref . @xcite this prediction is well verified for the logistic map , which is here used as a benchmark for our codes . let us conclude this short review by noticing that in the limit @xmath242 synchronization can be achieved only for @xmath243 @xcite . therefore we shall limit our analysis to this interval . let us now study the critical line for class ii maps , exemplified by the bernoulli shift map and its continuous version . first of all we introduce the main observables and the numerical method employed to determine it . a meaningful order parameter for the transition is represented by time average of the following mean field quantity : @xmath244 this can be operatively defined as follows : firstly the system ( [ eq : cml ] ) is randomly initialized and iterated for a transient time @xmath245 proportional to the system size @xmath18 , then the time average of @xmath246 is computed over a time window @xmath199 , i.e. @xmath247 . finally the state of the system is defined as synchronized if @xmath248 , being @xmath98 a sufficiently small value ( @xmath249 is usually enough ) . the coupling value corresponding to the synchronization transition is then obtained by using a bisection method : chosen two coupling values across the transition line one corresponding to a desynchronized state ( @xmath250 ) and the other to a synchronized case ( @xmath251 ) a third value is selected as @xmath252 . if at this new coupling value the system synchronizes ( resp . not synchronizes ) @xmath253 is identified with the new @xmath251 ( resp . @xmath250 ) . the procedure is then repeated until @xmath254 ( in our simulations @xmath255 ) . finally the critical coupling is defined as @xmath256 . the algorithm , tested on the logistic map , was able to recover ( [ eq : epsc_cml ] ) with the required accuracy . in general one has that , once fixed @xmath22 , @xmath257 will be a function of @xmath245 for a given @xmath18 . at @xmath258 . the solid lines are the analytically estimated `` linear '' transition values eq . ( [ eq : epsc_cml ] ) , while the symbols refer to the numerically obtained values . we used @xmath259 , @xmath260 , @xmath261 and @xmath262 . ] in refs . @xcite it has been employed as a synchronization indicator the time average of the following mean - field quantity : @xmath263 we have verified that this indicator gives results completely analogous to @xmath264 in all the considered cases . in ref . @xcite the authors have reported for logistic coupled maps a very good agreement between @xmath265 given by ( [ eq : epsc_cml ] ) and @xmath257 estimated via @xmath266 . however , this is not the case for the bernoulli map , as shown in fig . [ fig : sync.bern ] . in this case ( depending on the slope of the map @xmath267 , on @xmath22 and on the chain length @xmath18 ) strong disagreements between the linear transition line given by @xmath265 and the numerically obtained values are observed . these disagreements are typical of class ii maps . since the nonlinear effects locally desynchronize the system even if @xmath268 , in general @xmath269 . as a function of @xmath18 for different @xmath245 in @xmath270 for the continuous bernoulli shift map with @xmath201 and @xmath271 at @xmath272 ( a ) and @xmath273 ( b ) . the same for the bernoulli shift with @xmath201 for @xmath272 ( c ) and for @xmath273 ( d ) . the continuous solid lines are the theoretical values obtained in the linear analysis framework ( [ eq : epsc_cml ] ) . ] in obtaining the data reported in fig . [ fig : sync.bern ] , the cpu time restrictions forced us to employ a large but somehow limited transient time ( namely , @xmath259 ) . therefore , we checked for the dependence of the results on @xmath245 for fixed @xmath18 . in particular , we measured @xmath257 for two @xmath22-values only ( namely , @xmath272 and 0.8 ) for several chain lengths and transient times . this analysis has been performed for the logistic map at the crisis , for the bernoulli shift map with @xmath201 and for its continuous version . the results are reported in fig . [ fig : transients ] . for the logistic map we found that , for any value of @xmath18 and @xmath22 , a relatively small value of @xmath274 was sufficient to observe a clear convergence of @xmath257 to the linear value @xmath265 . this is also the case of the bernoulli map and its continuous version for @xmath272 ( though as one can see in fig . [ fig : transients]c the discontinuous map is characterized by a slightly slower convergence than its continuous version ) . on the contrary for @xmath273 with @xmath275 we were unable to observe synchronization even for @xmath276 , while at @xmath277 the logistic map had already converged to @xmath265 . summarizing for @xmath201 we found that for @xmath278 the bernoulli shift map and its continuous version always converge toward the linear critical value , while for @xmath279 at increasing @xmath18 the critical coupling @xmath257 becomes more and more independent of the transient time @xmath245 ( see the right of fig . [ fig : transients ] for @xmath280 ) . however , for a smaller value of @xmath267 and for large @xmath18 ( we investigated the case @xmath281 and @xmath282 ) we observed @xmath29 for any value of @xmath283 $ ] . in particular , we stress that also in the globally coupled case ( corresponding to @xmath284 ) we have clear evidences that @xmath285 . this is due to the fact that for @xmath286 nonlinear finite amplitude instabilities becomes more and more predominant with respect to the mechanism of linear stabilization ( see @xcite for more details ) . on the basis of the previous results , it is natural to conjecture that in the limit @xmath28 the non - linear transition is well defined , i.e. that the limit @xmath287 exists and is typically larger than @xmath288 . note that in the above expression the order of the two limits is crucial , we expect that performing at fixed @xmath18 the limit @xmath289 we should always observe a convergence to @xmath0 . however , as shown in the following , the times to reach synchronization may diverge exponentially fast with @xmath18 in the region just above @xmath0 making rapidly infeasible this limit . as for the case of two coupled maps , we study now the synchronization time statistics . in particular , we consider the system above the linear transition line @xmath8 and measure the first passage times @xmath26 needed for @xmath246 ( [ indicator ] ) to decrease below a given threshold @xmath98 . in this way we determine the corresponding pdf and the related momenta . in the high dimensional case , it is fundamental to analyze the dependence of the synchronization time on the the system size @xmath18 . however , as reported in eq . ( [ eq : epsc_cml ] ) , the critical value @xmath0 itself depends on @xmath18 . therefore to perform a meaningful comparison of systems of different sizes we considered situations characterized by the same linear behavior in the transverse space , i.e. having the same value of @xmath40 this means that for each length @xmath18 we chose the coupling strength according to @xmath290 further tests performed at different lengths but at a fixed distance @xmath291 from the critical line give essentially the same results . vs @xmath26 for different system sizes ( see label ) at fixed @xmath292 for different maps : ( a ) logistic map at the crisis with @xmath273 ( for @xmath272 we obtained qualitatively similar results ) ; skew tent map ( [ eq : skt ] ) with @xmath45 ( b ) ; bernoulli shift map with @xmath201 for @xmath272 ( c ) and @xmath273 ( d ) . in the latter case sizes larger than @xmath293 were not drawn for the sake of clarity of the plot . the pdf s have been obtained in all cases by considering @xmath294 different initial conditions and with @xmath128 . ] as shown in fig . [ fig : pdf_log]a , for the logistic map ( but these results can be extended to all the maps belonging to class i , see fig . [ fig : pdf_log]b ) the pdf s of the synchronization times obtained at constant @xmath40 display a very weak ( almost absent ) dependence on the system size , and are qualitatively similar to that found for two coupled maps ( compare fig . [ fig : pdf_log]a with fig . [ fig : pdftent ] ) . moreover , measurements of the average synchronization times @xmath295 , done by fixing the `` distance '' @xmath296 from the critical line , exhibit a clear tendency to saturate for increasing @xmath18 ( fig . [ fig : tau_log ] ) . therefore , in the limit @xmath28 the synchronization time will not diverge . , as a function of the system size @xmath18 for the logistic map at the crisis with @xmath297 estimated at a fixed distance from the linear threshold @xmath298 . data have been averaged over @xmath204 initial conditions with @xmath128 . ] for the maps of class ii the situation is different . here , as shown in fig . [ fig : pdf_log]c , for values of @xmath22 sufficiently small @xmath25 is weakly dependent on the system size as found for the logistic map . on the other hand , by considering more local couplings ( e.g. @xmath273 in fig . [ fig : pdf_log]d ) , the tail of the pdf becomes more and more pronounced as the system size increases . results obtained by fixing the distance from the critical coupling instead of the value of the tle display qualitatively similar features . this picture is further confirmed by examining @xmath295 as a function of @xmath18 for fixed @xmath40 ( i.e. by choosing @xmath5 according to ( [ eps_l2 ] ) ) . as one can see in fig . [ fig : tau_bernou ] , @xmath295 displays a dramatic dependence on @xmath18 . in particular , at @xmath299 ( see the inset of fig . [ fig : tau_bernou ] ) @xmath300 grows exponentially with @xmath18 , while a power - like scaling is observable for @xmath301 , at least for the chain lengths we could reach . deeper inside the `` linear '' synchronization region , i.e. for @xmath302 , we observed a saturation of @xmath303 with the system size . it is worth stressing here that for @xmath272 and @xmath201 , where no appreciable discrepancies between @xmath2 and @xmath0 have been observed , @xmath295 saturates for large @xmath18 . nevertheless the corresponding pdf exhibits the usual tail at long times characteristic of maps of class ii but with weak dependence on @xmath18 ( see fig . [ fig : pdf_log]c ) . notice also that , in principle , the transition from the exponential dependence to the saturation can be used to estimate @xmath2 ( see ref . @xcite ) , however for the present model a systematic study of this aspect is infeasible due to the required computational resources as a function of @xmath18 for the bernoulli map with @xmath201 and @xmath273 for various values of the second lyapunov exponent ( see label ) . the inset shows the case of @xmath292 in linear scale . data have been averaged over @xmath294 initial conditions and @xmath128 . ] we thus found clear indications that a nonlinear synchronization transition can be observed for class ii maps when finite size nonlinear instabilities are sufficiently strong to overcome the effects of stabilization associated with the local linear dynamics and to the spatial coupling . this suggests that nonlinear effects tend to decouple the single units of the chain . this decoupling can be indeed related , by means of the following simple argument , to the observed exponential divergence with @xmath18 of the synchronization times at @xmath304 . synchronization happens when the difference @xmath305 is contracted for ever in each site until @xmath306 decreases below threshold . the probability that for a single site the initial difference @xmath307 will be never amplified to @xmath92 is given by @xmath308 ( see eq . ( [ eq : p0 ] ) ) , therefore by assuming that each sites is completely decoupled from its neighbors the probability of contraction for the whole chain is given by @xmath309}$ ] ( for simplicity we set @xmath310 without loss of generality ) . and the probability that the system settles onto this `` contracting '' state after @xmath311 steps is @xmath312 this quantity represents a good approximation of @xmath25 . therefore one expects a poissonian decay for the pdf of the synchronization times with an associated average time @xmath313}$ ] exponentially diverging with the chain length , as indeed observed . let us now characterize the synchronization transition in the spatially extended model ( [ eq : cml ] ) within the framework of non - equilibrium phase - transitions . a similar parallel was recently established in a series of works concerning synchronization of two replicas of cml s with nearest neighbor coupling @xcite . in these studies it has been shown that the synchronization transition is continuous and belongs to the multiplicative noise ( resp . directed percolation ) universality class depending on the linear ( resp . nonlinear ) nature of the prevailing mechanisms . in both cases one observes a transition from an active phase ( characterized by @xmath314 ) to a unique absorbing state ( identified by @xmath315 ) . we indicated with @xmath316 the order parameter , however for sake of clarity it should be said that for two replicas of a cml this corresponds to @xmath317 and is not given by expression ( [ indicator ] ) used in the present paper for characterizing self synchronization . the rationale for using the same symbol is that the two definitions embody essentially the same information : @xmath318 is a measure of the density of non synchronized ( active ) sites . in the proximity of the transition ( if continuous ) the scaling behavior of the saturated density @xmath319 of active sites is given by @xmath320 while at the critical point @xmath321 the density of active sites scales as @xmath322 and the average synchronization time diverges as @xmath323 we denoted as @xmath324 the critical coupling to avoid at this stage distinctions between linearly and nonlinearly driven synchronization . since continuous non - equilibrium phase transitions are typically characterized by three independent critical exponents , once ( @xmath325 ) are known all the other scaling exponents can be derived @xcite . for @xmath326 ( filled circles ) @xmath327 ( empty boxes ) and @xmath328 ( crosses ) . the solid straight line displays the power law decay @xmath329 . data refer to the logistic map with @xmath330 and @xmath331 and are obtained after averaging over @xmath332 initial conditions . in the inset the average value of the order parameter @xmath333 as a function of @xmath334 is reported . the straight line shows the scaling behavior @xmath335 . data refer to the logistic map with @xmath272 and @xmath331 the average is performed over @xmath336 different initial conditions . ] let us now analyze cml s with power - law coupling ( [ eq : cml ] ) . we start with class i maps , for which the transition is completely characterized by the linear dynamics , i.e. @xmath337 . in particular , we consider the logistic maps at the ulam point for three different @xmath22-values ( namely , @xmath338 and @xmath339 ) . the first observation is that the critical properties of the model seem to be independent of @xmath22 : we measured exactly the same critical exponents for the all @xmath22 values , namely @xmath340 , @xmath341 ( see fig . [ fig : scaling ] ) and @xmath342 ( see fig . [ fig : tau_log ] ) . these exponents coincide with the mean - field exponents reported in ref . @xcite for a model of anomalous directed percolation ( dp ) . the model differently from standard dp considers the spreading of epidemics in the case of long - range infection , namely in one spatial dimension the probability distribution for a site to be infected at distance @xmath267 decays as @xmath343 . the critical behavior of this model can be obtained by considering a langevin equation with power - law decaying spatial coupling for the coarse grained density of infected sites . in ref . @xcite , the authors found that the critical exponents vary continuously with @xmath344 , but below a critical value ( @xmath345 in 1d ) the exponents coincide with the corresponding mean - field results , namely @xmath346 and @xmath347 . in the limit @xmath348 the latter exponents are identical to the ones we have found for the logistic coupled maps . in the following we shall give an argument to explain these similarities and differences . as shown in refs . @xcite there is a deep connection between the synchronization problem for two replicas of a diffusively coupled cml and nonequilibrium phase transitions . indeed there the difference field @xmath349 can be mapped onto the density of infected ( active ) sites and an appropriate langevin equation describing the evolution of @xmath350 can be derived in proximity of the transition @xcite . in our case of cmls with power - law coupling we expect that the corresponding langevin equation for the spatio - temporal coarse grained defect density @xmath351 should contain a long - range interaction with a power law spatial coupling decaying with an exponent @xmath22 . therefore , in the proximity of the transition , it seems natural to map the dynamics of ( [ eq : cml ] ) onto the model studied in @xcite once the exponent @xmath22 is identified with @xmath352 . however , in ( [ eq : cml ] ) the coupling is rescaled by the factor @xmath353 to avoid divergences in the limit @xmath354 . the rescaling @xmath355 amounts to consider effective interaction on large scales of the type @xmath356 independently of @xmath22 . this may explain the fact that the exponents characterizing the transition of model ( [ eq : cml ] ) coincide with the mean - field values reported in @xcite for @xmath357 . a similar rescaling was performed in ref . @xcite to show , for a chain of power - law coupled rotators , that all the equilibrium properties of the system coincide for @xmath358 , once suitably scaled . let us also remark that since we are dealing with a mean - field case the nature of the noise entering in the langevin equation , which distinguishes directed percolation from multiplicative noise , is irrelevant @xcite . this leads us to conclude that , for continuous synchronization transitions , we would not expect any differences in the measured exponents between linearly and nonlinearly driven transitions . for the bernoulli map with @xmath201 , @xmath273 , @xmath331 , and different coupling strength above and below @xmath2 ( see the labels ) . average is done over @xmath359 initial conditions , tests on shorter times and more initial conditions confirm the reliability of the statistics . ] in examining the critical properties associated to coupled class ii maps , we confronted with hard numerical difficulties related to the fact that @xmath360 can not be measured with the required precision . moreover , due to the presence of exponentially long transients in the proximity of the nonlinear transition , simulations of model ( [ eq : cml ] ) with class ii maps become extremely time consuming . for these reasons we were unable to find conclusive results . however , within these limitations , our analysis ( performed up to @xmath331 and up to integration times @xmath361 ) suggests that for class ii maps ( both the bernoulli map and its continuous version ( [ eq : bsc ] ) with @xmath201 were considered ) the transition may be discontinuous as suggested from the data shown in fig . [ fig : disco ] . in the figure the temporal evolution of the order parameter @xmath246 averaged over many initial conditions is reported in proximity of @xmath2 and no indication of critical behaviour is observable . this indicates that in class ii cmls : short range interactions give rise to dp - like continuous transitions , while power - law coupled maps exhibit discontinuous transitions . we should mention that the introduction of long - range interactions in dp model can similarly modify the nature of the transition from continuous to discontinuous , as shown in a recent paper @xcite . the authors have introduced a generalized directed percolation model , in which the activation rate of a site at the border of an inactive island of length @xmath362 is given by @xmath363 . in particular , in ref . @xcite it has been shown that the transition is continuous for @xmath364 with exponent coinciding with dp , and discontinuous for @xmath365 . in the present paper we have examined the influence of strong nonlinearities in the transverse synchronization transition of low and high dimensional chaotic systems . in particular , for two coupled maps we have shown that finite amplitude nonlinear instabilities can give rise to long transients , preceding the synchronization , even when the dynamics is transversally stable at any instant . this should be contrasted with the results found for continuous maps where long transients may happen only as a result of transient chaotic or intermittent transverse dynamics . in high dimensional systems with long range ( power - law ) interactions strong nonlinearities may invalidate the linear criterion to locate the critical coupling . in this case the transition occurs , due to nonlinear mechanisms , at a larger coupling value . the nonlinear transition is characterized by the emergence of transients diverging exponentially with the system size even above the linear critical coupling . the origin of this transition is closely related to the stable chaos phenomenon . the synchronization phenomena both in linearly and nonlinearly driven systems have been compared with models for long ranged contact processes . we found that , for linearly driven systems , the transition is continuous and the critical exponents are given by a mean field prediction . for nonlinearly driven systems , though the results are not conclusive , evidence of a discontinuous transition have been found . we are grateful to w. just , a. politi and a. pikovsky for useful discussions and remarks and to f. ginelli also for a careful reading of this manuscript . partial support from the italian firb contract n. rbne01cw3m_001 is acknowledged . 99 d. he and l. stone , proc r soc lond b biol sci . * 270 * ( 2003 ) 1519 . p. n. steinmetz , a. roy , p. j. fitzgerald , s. s. hsiao , k. o. johnson , and e. niebur , nature * 404 * ( 2000 ) 187 . a. pikovsky , m. rosenblum , and j. kurths , _ synchronization a universal concept in nonlinear sciences _ , ( cambridge university press , 2001 ) l.m . pecora and t.l . carroll , phys . * 64 * ( 1990 ) 821 . kuznetsov and a. s. pikovsky , radiophys . , * 32 * ( 1989 ) 49 . a. s. pikovsky and p. grassberger , j. phys . a : math . gen . * 24 * ( 1991 ) 4587 . w. just , physica d * 81 * ( 1995 ) 317 . l. baroni , r. livi , and a. torcini , phys . e * 63 * ( 2001 ) 036226 . v. ahlers and a. s. pikovsky , phys . * 88 * ( 2002 ) 254101 . c. anteneodo , s.e . de s. pinto , a. m. batista , and r. l. viana , phys . e , * 68 * ( 2003 ) 045202(r ) ; c. anteneodo , a.m. batista , and r. l. viana , phys . lett . a , * 326 * ( 2004 ) 227 . m. dharmala , v.k . jirsa , and m. ding , phys . * 92 * ( 2004 ) 028101 . k. wiesenfeld , p. colet and s.h . strogatz , phys . * 76 * ( 1996 ) 404 . c. peskin , _ mathematical aspects of heart physiology _ ( courant institute of mathematical sciences , new york university , new york , 1975 ) . g. paladin and a. vulpiani , j. phys . * a 25 * ( 1994 ) 4911 . a. torcini and s. lepri , phys . e , * 55 * ( 1997 ) r3805 . k. kaneko , physica d , * 34 * ( 1989 ) 1 . i. waller and r. kapral , phys . a * 30 * , 2047 ( 1984 ) ; k. kaneko , prog . theor . phys . * 72 * ( 1984 ) 980 . v. ahlers , _ scaling and synchronization in deterministic and stochastic nonlinear dynamical systems _ , phd thesis ( potsdam , 2001 ) f. ginelli et al . e * 68 * ( 2003 ) 065102(r ) . m. a. munz and r. pastor - satorras , phys . * 90 * ( 2003 ) 204101 . m. a. munz , in advances in condensed matter and statistical mechanics " , eds . e. korutcheva _ ( nova science publishers , 2004 , new york ) . h. hinrichsen , adv . * 49 * ( 2000 ) 815 . e. aurell , g. boffetta , a. crisanti , g. paladin and a. vulpiani , phys . * 77 * ( 1996 ) 1262 . g. boffetta , m. cencini , m. falcioni and a. vulpiani phys . rep . * 356 * ( 2002 ) 367 . m. cencini and a. torcini phys . e * 63 * ( 2001 ) 056201 . w. feller , _ an introduction to probability theory and its applications _ ( wiley , new york , 1974 ) ; h.c . tuckwell , _ introduction to theoretical neurobiology - vol . 2 - nonlinear and stochastic theories _ ( cambridge university press , cambridge , 1988 ) . f. ginelli , h. hinrichsen , r. livi , d. mukamel , and a. politi , phys . e * 71 * ( 2005 ) 026121 . a. politi , r. livi , g.l . oppo and r. kapral , europhys . * 22 * ( 1993 ) 571 . a. torcini , p. grassberger and a. politi , j. phys . * a27 * ( 1995 ) 4533 . a. politi and a. torcini , europhys . lett . , * 28 * ( 1994 ) 545 . g. paladin and a. vulpiani , phys . rep . * 156 * ( 1987 ) 147 . h. jeffreys and b.s . jeffreys , in _ methods of mathematical physics _ , ( cambridge , england , cambridge university press,1988 ) pp . 470 - 472 . f. ginelli , r. livi , and a. politi , j. phys . a : math . gen . * 35 * ( 2002 ) 499 . a. lipowski and m. droz , `` synchronization and partial synchronization of linear maps '' , cond - mat/0312067 r. wackerbauer and k. showalter , phys . * 91 * ( 2003 ) 174103 . a. zumdieck , m. timme , t. geisel , and f. wolf , phys . * 93 * ( 2004 ) 244103 . f. ginelli , r. livi , a. politi , and a. torcini phys . e * 67 * ( 2003 ) 046217 . h. hinrichsen and m. howard , eur . j. b * 7 * ( 1999 ) 635 . f. tamarit and c. anteneodo , phys . * 84 * ( 2000 ) 208 .

over the last half - a - dozen years , it has been increasingly realized that the detection of non - gaussianities in the primordial perturbations can considerably help in constraining the inflationary models ( see refs . @xcite ; for early efforts in this direction , see refs . @xcite ) . in particular , the detection of a high value for the dimensionless non - gaussianity parameter @xmath0 that is often used to describe the amplitude of the scalar bi - spectrum can rule out a wide class of models . for instance , if the extent of non - gaussianity actually proves to be as large as the mean values of @xmath0 arrived at from the wmap data ( see refs . @xcite ; in this context , also see refs . @xcite ) , then canonical scalar field models that lead to slow roll inflation and nearly scale invariant primordial spectra will cease to be consistent with the data . however , the evaluation of the scalar bi - spectrum in a generic inflationary model proves to be a non - trivial task @xcite . often , it is the slow roll approximation that is resorted to in order to arrive at analytical expressions for the scalar bi - spectrum and the non - gaussianity parameter @xmath0 @xcite . while the slow roll approximation can actually encompass a relatively wide class of models , evidently , the approach will cease to be valid when departures from slow roll occur . in such situations , one is often forced to approach the problem numerically . in this work , we shall present a new and accurate fortran code , the bi - spectra and non - gaussianity operator ( bingo ) , to numerically evaluate the scalar bi - spectrum and the non - gaussianity parameter @xmath0 for single field inflationary models involving the canonical scalar field . although , some partial numerical results have already been published in the literature , we believe that it is for the first time that a general and efficient code has been put together to evaluate the complete scalar bi - spectrum . though , we shall largely restrict our discussion in this paper to the equilateral case , as we shall illustrate , the code can compute the scalar bi - spectrum and the parameter @xmath0 for any triangular configuration of the wavevectors . also , as we shall demonstrate , bingo can also compute _ all _ the different contributions to the bi - spectrum . moreover , it is worth noting that , under certain conditions , the code can arrive at the results within a matter of a few minutes . we should mention here that we have made a limited version of bingo , viz . one that focuses on the equilateral limit , available online at https://www.physics.iitm.ac.in/~sriram/bingo/bingo.html . as an immediate application , we shall utilize the code to examine of the power of the non - gaussianity parameter @xmath0 to discriminate between different inflationary models that admit deviations from slow roll and generate similar features in the scalar power spectrum . recall that , most single field inflationary models naturally lead to an extended period of slow roll and hence to nearly scale invariant primordial power spectra ( see , for example , any of the following texts @xcite or reviews @xcite ) , which seem to be fairly consistent with the recent data on the anisotropies in the cosmic microwave background ( cmb ) ( in this context , see refs . @xcite ; for a comparison of a class of inflationary models with the data , see , for example , ref . @xcite and references therein ) as well as other observational constraints . however , it has been repeatedly noticed that certain features in the inflationary scalar power spectrum can improve the fit to the cmb data at the cost of some additional parameters ( see , for instance , refs . though the statistical significance of these features remain to be understood satisfactorily @xcite , they gain importance from the phenomenological perspective of comparing the models with the data , because only a smaller class of single field inflationary models , which allow for departures from slow roll , can generate them . interestingly , demanding the presence of features in the scalar power spectrum seems to generically lead to larger non - gaussianities ( see , for example , refs . therefore , features may offer the only route ( unless one works with non - vacuum initial states @xcite ) for the canonical scalar fields to remain viable if @xmath0 turns out to be significant . if indeed the presence of features turns out to be the correct reason behind possibly large non - gaussianities , can we observationally identify the correct underlying inflationary scenario , in particular , given the fact that different models can lead to similar features in the scalar power spectrum ? in other words , to what extent can the non - gaussianity parameter @xmath0 help us discriminate between the inflationary models that permit features ? to address this question , we shall consider a few typical inflationary models leading to features , assuming that they can be viewed as representatives of such a class of scenarios . concretely , we shall consider the starobinsky model @xcite and the punctuated inflationary scenario @xcite , both of which result in a sharp drop in power at large scales that is followed by oscillations . we shall also study large and small field models with an additional step introduced in the inflaton potential @xcite . the step leads to a burst of oscillations in the scalar power spectrum which improve the fit to the outliers near the multipole moments of @xmath1 and @xmath2 in the cmb anisotropies . we shall also consider oscillating inflaton potentials such as the one that arises in the axion monodromy model which lead to modulations in the power spectrum over a wide range of scales @xcite . the plan of this paper is as follows . in the following section , we shall quickly describe a few essential details pertaining to the power spectrum and the bi - spectrum as well as the non - gaussianity parameter @xmath0 in inflationary models involving a single , canonical , scalar field . in sec . [ sec : nc - sbs ] , after demonstrating that the super hubble contributions to the bi - spectrum during inflation prove to be negligible , we shall describe the method that bingo adopts to numerically compute the bi - spectrum and the non - gaussianity parameter @xmath0 . we shall also illustrate the extent of accuracy of bingo by comparing the numerical results with the analytical results available in three situations . firstly , restricting ourselves to the equilateral limit , we shall compare the results from the numerical computations with the spectral dependence of the bi - spectrum expected in power law inflation . secondly , focusing on the case of the archetypical quadratic potential , we shall contrast the numerical results with the analytical expressions obtained using the slow roll approximation for an arbitrary triangular configuration of the wavevectors . thirdly , considering a more non - trivial situation involving a brief period of fast roll , we shall compare the numerical results with the analytical results that have recently been obtained in the case of the starobinsky model in the equilateral limit @xcite . in the succeeding two sections , we shall utilize bingo to examine of the power of the non - gaussianity parameter @xmath0 to discriminate between different inflationary models that permit deviations from slow roll and generate similar features in the scalar power spectrum . after a quick outline of the inflationary models of our interest , we shall discuss the scalar power spectra that arise in these models . we shall then present the main results , and compare the @xmath0 that arise in the various models that we consider . we shall finally conclude in sec . [ sec : d ] with a brief summary and outlook . a few words on our conventions and notations are in order before we proceed . we shall work with units such that @xmath3 , and we shall set @xmath4 . as is often done in the context of inflation , we shall assume the background to be described by the spatially flat , friedmann - lemaitre - robertson - walker line - element . we shall denote the conformal time as @xmath5 , while @xmath6 shall represent the number of e - folds . moreover , an overprime shall denote differentiation with respect to the conformal time coordinate @xmath5 . lastly , @xmath7 shall denote the scale factor , with the corresponding hubble parameter being defined as @xmath8 . in this section , we shall first rapidly summarize the essential definitions of the power spectrum and the bi - spectrum as well as the non - gaussianity parameter @xmath0 . we shall then list the various contributions to the bi - spectrum that arise in the maldacena formalism ( for the original discussions , see refs . @xcite ; for more recent discussions , see , for example , refs . @xcite ) . let us begin by recalling some essential points concerning the power spectrum . on quantization , the operator corresponding to the curvature perturbation @xmath9 can be expressed as @xmath10,\label{eq : cr - d}\end{aligned}\ ] ] where @xmath11 and @xmath12 are the usual creation and annihilation operators that satisfy the standard commutation relations . the modes @xmath13 are governed by the differential equation @xcite @xmath14 where @xmath15 , with @xmath16 being the first slow roll parameter . the dimensionless scalar power spectrum @xmath17 is defined in terms of the correlation function of the fourier modes of the curvature perturbation @xmath18 as follows : @xmath19 where @xmath20 denotes the bunch - davies vacuum which is defined as @xmath21 @xmath22 @xmath23 @xcite . in terms of the modes @xmath13 , the scalar power spectrum is given by @xmath24 as is well known , numerically , the initial conditions are imposed on the modes @xmath13 when they are well inside the hubble radius , and the power spectrum is evaluated at suitably late times when the modes are sufficiently outside the hubble radius ( see , for instance , refs . we shall discuss the details concerning the numerical evolution of the background as well as the perturbations and the computation of the corresponding power spectrum in the next section . the scalar bi - spectrum @xmath25 is related to the three point correlation function of the fourier modes of the curvature perturbation @xmath26 , evaluated at the end of inflation , say , at the conformal time @xmath27 , as follows @xcite : @xmath28 in our discussion below , for the sake of convenience , we shall set @xmath29 the observationally relevant non - gaussianity parameter @xmath0 is introduced through the relation @xmath30 , \label{eq : fnl - i}\ ] ] where @xmath31 denotes the gaussian quantity , and the factor of @xmath32 arises due to the relation between the bardeen potential and the curvature perturbation during the matter dominated epoch . upon making use of the corresponding relation between @xmath33 and @xmath31 in fourier space and the wick s theorem , one obtains that @xcite @xmath34.\end{aligned}\ ] ] this expression can then be utilized to arrive at the following relation between the non - gaussianity parameter @xmath0 and the scalar bi - spectrum @xmath25 or , equivalently , the quantity @xmath35 that we have introduced @xcite : @xmath36^{-1}.\label{eq : fnl - d}\end{aligned}\ ] ] in particular , in the equilateral limit , i.e. when @xmath37 , this expression simplifies to @xmath38 two points need to be emphasized here regarding the above expressions for the quantities @xmath0 and @xmath39 . firstly , it ought to be noted that the non - gaussianity @xmath0 parameter introduced in eq . ( [ eq : fnl - i ] ) [ as well as the expression arrived at in eq . ( [ eq : fnl - d ] ) ] corresponds to the so - called local form of the parameter . secondly , the quantity @xmath39 corresponds to the equilateral limit of the _ local _ @xmath0 parameter , and is intrinsically different from a similar parameter introduced when the bi - spectrum has the equilateral form ( in this context , see , for instance , ref . @xcite ) . in the maldacena formalism @xcite , the bi - spectrum is evaluated using the standard rules of perturbative quantum field theory , based on the interaction hamiltonian that depends cubically on the curvature perturbation . it can be shown that the bi - spectrum that results from the interaction hamiltonian can be expressed as @xcite @xmath40\ ; \cg_{_{c}}(\vka,\vkb,\vkc)\nn\\ & & + \l[f_{\vka}^{\ast}(\ee)\ , f_{\vkb}^{\ast}(\ee)\,f_{\vkc}^{\ast}(\ee)\r]\ ; \cg_{_{c}}^{\ast}(\vka,\vkb,\vkc)\biggr\ } \nonumber \\ & & + g_{7}(\vka,\vkb,\vkc).\label{eq : g}\end{aligned}\ ] ] the quantities @xmath41 with @xmath42 correspond to the six terms in the interaction hamiltonian , and are described by the integrals @xmath43,\label{eq : cg3}\\ \cg_{4}(\vka,\vkb,\vkc ) & = & i\,\int_{\ei}^{\ee } \d\eta\ , a^2\,\epsilon_{1}\ , \epsilon_{2}'\ , \l(f_{\vka}^{\ast}\,f_{\vkb}^{\ast}\ , f_{\vkc}'^{\ast}+{\rm two~permutations}\r),\label{eq : cg4}\\ \cg_{5}(\vka,\vkb,\vkc ) & = & \frac{i}{2}\,\int_{\ei}^{\ee } \d\eta\ , a^2\ , \epsilon_{1}^{3}\ , \biggl[\l(\f{\vka\cdot\vkb}{\kb^{2}}\r)\ , f_{\vka}^{\ast}\,f_{\vkb}'^{\ast}\ , f_{\vkc}'^{\ast } \nonumber \\ & & + { \rm five~permutations}\biggr],\label{eq : cg5}\\ \cg_{6}(\vka,\vkb,\vkc ) & = & \frac{i}{2}\,\int_{\ei}^{\ee } \d\eta\ , a^2\ , \epsilon_{1}^{3}\ , \biggl\{\l[\f{\ka^{2}\,\l(\vkb\cdot\vkc\r)}{\kb^{2}\,\kc^{2}}\r]\ , f_{\vka}^{\ast}\ , f_{\vkb}'^{\ast}\ , f_{\vkc}'^{\ast } \nonumber \\ & & + { \rm two~permutations}\biggr\},\label{eq : cg6}\end{aligned}\ ] ] where @xmath44 is the second slow roll parameter that is defined with respect to the first as follows : @xmath45 . the lower limit of the above integrals , viz . @xmath46 , denotes a sufficiently early time when the initial conditions are imposed on the modes . the additional , seventh term @xmath47 arises due to the field redefinition , and its contribution to @xmath35 is given by @xmath48 in this section , after illustrating that the super - hubble contributions to the complete bi - spectrum during inflation proves to be negligible , we shall outline the methods that bingo adopts to numerically evolve the equations governing the background and the perturbations , and eventually evaluate the inflationary scalar power and bi - spectra . also , we shall illustrate the extent of accuracy of bingo by comparing the results from the code with the expected form of the bi - spectrum in the equilateral limit in power law inflation and the analytical results that are available in the case of the starobinsky model @xcite . we shall also contrast the numerical results with the analytical results obtained under the slow roll approximation in the case of the popular quadratic potential for an arbitrary triangular configuration of the wavevectors . it is clear from the expressions in the previous section that the evaluation of the bi - spectrum involves integrals over the mode @xmath13 and its derivative @xmath49 as well as the slow roll parameters @xmath50 , @xmath44 and the derivative @xmath51 . while evaluating the power spectra , it is well known that , in single field inflationary models , it suffices to evolve the curvature perturbation from an initial time when the modes are sufficiently inside the hubble radius to a suitably late time when the amplitude of the curvature perturbation settles down to a constant value on super - hubble scales ( see , for example , refs . we shall illustrate that many of the contributions to the bi - spectrum prove to be negligible when the modes evolve on super - hubble scales . interestingly , we shall also show that , those contributions to the bi - spectrum which turn out to be significant at late times when the modes are well outside the hubble radius are canceled by certain other contributions that arise . as a consequence , we shall argue that , numerically , it suffices to evaluate the integrals over the period of time during which the curvature perturbations have been conventionally evolved to arrive at the power spectra , viz . from the sub - hubble to the super - hubble scales . in fact , we had recently shown that , for cosmological modes that are on super - hubble scales during preheating , the contributions to the bi - spectrum and , equivalently , to the non - gaussianity parameter @xmath0 , prove to be negligible @xcite . the calculations we had presented in the work can be easily extended to the case of the modes on super - hubble scales during inflation . therefore , we shall rapidly sketch the essentials arguments below , referring the readers to our recent work for further details @xcite . during inflation , when the modes are on super - hubble scales , it is well known that the solution to @xmath13 can be written as @xcite @xmath52 where @xmath53 and @xmath54 are @xmath23-dependent constants which are determined by the initial conditions imposed on the modes in the sub - hubble limit . the first term involving @xmath53 is the growing mode , which is actually a constant , while the term containing @xmath54 represents the decaying mode . therefore , on super - hubble scales , the mode @xmath13 simplifies to @xmath55 moreover , the leading non - zero contribution to its derivative is determined by the decaying mode , and is given by @xmath56 where we have set @xmath57 . it is now a matter of making use of the above solutions for @xmath13 and @xmath49 to determine the super - hubble contributions to the bi - spectrum during inflation . to begin with , note that , each of the integrals @xmath58 , where @xmath59 , can be divided into two parts as follows : @xmath60 the integrals in the first term @xmath61 run from the earliest time ( i.e. @xmath46 ) when the smallest of the three wavenumbers @xmath62 , @xmath63 and @xmath64 is sufficiently inside the hubble radius [ typically corresponding to @xmath65 to the time ( say , @xmath66 ) when the largest of the three wavenumbers is well outside the hubble radius [ say , when @xmath67 . then , evidently , the second term @xmath68 will involve integrals which run from the latter time @xmath66 to the end of inflation at @xmath69 ( in this context , see fig . [ fig : eta ] ) . -10pt in what follows , we shall discuss the various contributions to the bi - spectrum due to the terms @xmath68 . we shall show that the corresponding contribution either remains small or , when it proves to be large , it is exactly canceled by another contribution to the bi - spectrum . let us first focus on the fourth term @xmath70 since it has often been found to lead to the largest contribution to the bi - spectrum when deviations from slow roll occur @xcite . as the slow roll parameters turn large towards the end of inflation , we can expect this term to contribute significantly at late times . however , as we shall quickly illustrate , such a late time contribution is exactly canceled by the contribution from @xmath47 which arises due to the field redefinition @xcite . upon using the form ( [ eq : fk - shs ] ) of the mode @xmath71 and its derivative ( [ eq : fkp - shs ] ) on super - hubble scales in the expression ( [ eq : cg4 ] ) , the integral involved can be trivially carried out with the result that the corresponding contribution to the bi - spectrum can be expressed as @xmath72\ , \biggl[\vert a_{\vka}\vert^2\ , \vert a_{\vkb}\vert^2\ , \l(a_{\vkc}\ , { \bar b}_{\vkc}^{\ast}-a_{\vkc}^{\ast}\ , { \bar b}_{\vkc}\r)\nn\\ & & + \,{\rm two~permutations}\biggr].\label{eq : g4-d - prh}\end{aligned}\ ] ] the wronskian corresponding to the equation of motion ( [ eq : de - fk ] ) and the standard bunch - davies initial condition leads to the relation @xmath73 , which can then be utilized to arrive at the following simpler expression @xcite : @xmath74\nn\\ & & \times\ , \biggl[\vert a_{\vka}\vert^2\ , \vert a_{\vkb}\vert^2 + \,~{\rm two~permutations}\biggr].\label{eq : g4-at - es}\end{aligned}\ ] ] the first of these terms involving the value of @xmath44 at the end of inflation _ exactly _ cancels the contribution @xmath47 [ with @xmath13 set to @xmath53 in eq . ( [ eq : g7 ] ) ] that arises due to the field redefinition . but , the remaining contribution can not be ignored and needs to be taken into account . it is useful to note that this term is essentially the same as the one due to the field redefinition , but which is now evaluated on super - hubble scales ( i.e. at @xmath66 ) rather than at the end of inflation . in other words , if we consider the fourth and the seventh terms together , it is equivalent to evaluating the contribution to the bi - spectrum corresponding to @xmath75 , and adding to it the contribution due to the seventh term @xmath47 evaluated at @xmath66 rather than at the end of inflation . let us now turn to the contribution due to the second term , which can occasionally prove to be comparable to the contribution due to the fourth term @xcite . upon making use of the behavior of the mode @xmath13 and its derivative on super - hubble scales in the integral ( [ eq : cg2 ] ) , one obtains the contribution to the bi - spectrum due to @xmath76 to be @xcite @xmath77,\end{aligned}\ ] ] where the quantity @xmath78 is described by the integral @xmath79 note that , due to the quadratic dependence on the scale factor , actually , @xmath78 is a rapidly growing quantity at late times . however , the complete super - hubble contribution to the bi - spectrum vanishes identically since the integral @xmath78 is a purely real quantity @xcite . hence , in the case of the second term , it is sufficient to evaluate the contribution to the bi - spectrum due to @xmath80 . due to their structure , one finds that the first and the third terms and the fifth and the sixth terms can be evaluated together . on super - hubble scales , one can easily show that the contributions due to the first and the third terms can be written as @xmath81\;i_{13}(\ee,\es),\end{aligned}\ ] ] while the corresponding contributions due to the fifth and the sixth terms can be obtained to be @xmath82\;i_{56}(\ee,\es ) , \label{eq : g56-shc}\end{aligned}\ ] ] where the quantities @xmath83 and @xmath84 are described by the integrals @xmath85 and @xmath86 hence , the non - zero , super - hubble contribution to the bi - spectrum is determined by the complete contribution due to the first , the third , the fifth and the sixth terms arrived at above . in order to illustrate that this contribution is insignificant , we shall now turn to estimating the amplitude of the corresponding contribution to the non - gaussianity parameter @xmath0 . let us restrict ourselves to the equilateral limit for simplicity . in such a case , the super - hubble contributions due to the first , the third , the fifth and the sixth terms to the non - gaussianity parameter @xmath0 can be obtained by substituting the quantities ( [ eq : g13-shc ] ) and ( [ eq : g56-shc ] ) above in the expression ( [ eq : fnl - eq ] ) . it is straightforward to show that the corresponding @xmath0 is given by @xmath87,\end{aligned}\ ] ] where we have made use of the fact that @xmath88 at late times in order to arrive at the power spectrum . to estimate the above super - hubble contribution to the non - gaussianity parameter @xmath89 , let us focus on inflation of the power law form . in power law inflation , the scale factor is given by @xmath90 with @xmath91 and @xmath92 being constants , while @xmath93 is a free index . it is useful to note that , in such a case , the first slow roll parameter is a constant , and is given by @xmath94 . under these conditions , the quantity within the brackets involving @xmath53 and @xmath95 in the expression above can be easily evaluated ( in this context , see ref . @xcite ) to arrive at @xmath96\ , \gamma^2\l(\gamma + \f{1}{2}\r)\ , 2^{2\,\gamma+1}\ , \l(2\,\gamma+1\r)\ , ( \gamma+2)\nn\\ & & \times\,\l(\gamma+1\r)^{-2\,(\gamma+1)}\ , \sin\,(2\,\pi\,\gamma)\ , \l[1-\f{h_{\rm s}}{h_{\rm e}}\,{\rm e}^{-3\,(\ne-\ns)}\r]\nn\\ & & \times\,\l(\frac{k}{a_{\rm s}\ , h_{\rm s}}\r)^{-(2\,\gamma+1)}.\end{aligned}\ ] ] it should be mentioned that , in arriving at this expression , for convenience , we have set @xmath92 to be @xmath97 , which corresponds to @xmath91 being @xmath98 , viz . the scale factor at @xmath66 . moreover , while @xmath99 and @xmath100 denote the e - folds corresponding to @xmath66 and @xmath69 , @xmath101 and @xmath102 represent the hubble scales at these times , respectively . recall that , @xmath66 denotes the conformal time when the largest wavenumber of interest , say , @xmath103 , is well outside the hubble radius , i.e. when @xmath104 . since @xmath105 is expected to be at least @xmath2 for the smallest cosmological scale , it is clear that the factor involving @xmath106 $ ] can be completely neglected . observations point to the fact that @xmath107 . therefore , if we further assume that @xmath108 , where @xmath109 , we find that the above estimate for the non - gaussianity parameter reduces to @xmath110 where , in obtaining the final value , we have set @xmath111 . the inequality above arises due to the fact that , for larger scales , i.e. when @xmath112 , @xmath113 at @xmath66 . in models involving the canonical scalar field , the smallest values of @xmath0 are typically generated in slow roll inflationary scenarios , wherein the non - gaussianity parameter has been calculated to be of the order of the first slow roll parameter @xcite . the above estimate clearly points to fact that the super - hubble contributions to the complete bi - spectrum and the non - gaussianity parameter @xmath0 can be entirely ignored . in summary , to determine the scalar bi - spectrum , it suffices to evaluate the contributions to the bi - spectrum due to the quantities @xmath114 , with @xmath59 , which involve integrals running from the initial time @xmath46 to the time @xmath66 when the smallest of the three modes reaches super - hubble scales ( cf . [ fig : eta ] ) . further , the addition of the contribution due to the field redefinition evaluated at @xmath66 ensures that no non - trivial super - hubble contributions are ignored . in the following sub - section , with the help of a specific example , we shall also corroborate these conclusions numerically . the scalar bi - spectrum and the parameter @xmath0 can be easily evaluated analytically in the slow roll inflationary scenario @xcite . however , barring some exceptional cases @xcite , it often proves to be difficult to evaluate the bi - spectrum analytically when departures from slow roll occur . hence , one has to resort to numerical computations in such cases . bingo solves the background as well as the perturbation equations using rksuite @xcite , which is a publicly available routine to solve ordinary differential equations . we shall treat the number of e - folds as the independent variable , which allows for an efficient and accurate computation . to obtain the power spectrum , we impose the standard bunch - davies initial conditions on the perturbations when the modes are well inside the hubble radius , and evolve them until suitably late times . typically , in the case of smooth inflaton potentials , it suffices to evolve the modes @xmath13 from an initial time when @xmath115 . however , in certain models wherein the potentials contain oscillatory terms , the modes may have to be evolved from deeper inside the hubble radius . for example , in the case of the axion monodromy model that we shall discuss in due course ( see sub - sec . [ subsec : oip ] ) , for values of the model parameters of our interest , we find that we need to evolve the modes from an initial time when @xmath116 , so that the resonance that occurs in the model due to the oscillations in the potential is captured @xcite . the scalar power spectra that arise in the various models of our interest , as displayed in fig . [ fig : sps - all ] , have all been evaluated at super - hubble scales , say , when @xmath117 , which is typically when the amplitude of the curvature perturbations freeze in . having obtained the behavior of the background and the modes , bingo carries out the integrals involved in arriving at the bi - spectrum using the method of adaptive quadrature @xcite . it is useful to note that , in the equilateral limit of the bi - spectrum , which we shall largely concentrate on in this work , we can evolve each of the modes of interest independently and calculate the integrals for the modes separately . the integrals @xmath118 actually contain a cut off in the sub - hubble limit , which is essential for singling out the perturbative vacuum @xcite . numerically , the presence of the cut off is fortunate since it controls the contributions due to the continuing oscillations that would otherwise occur . we should highlight here that such a cut off procedure was originally introduced in the earliest numerical efforts to evaluate the scalar bi - spectrum in situations involving deviations from slow roll ( in this context , see refs . @xcite ) . generalizing the cut off that is often introduced analytically in the slow roll case , we shall impose a cut off of the form @xmath119 $ ] , where @xmath120 is a small parameter . in the previous two sub - sections , we had discussed as to how the integrals need to be carried out from the early time @xmath46 when the largest scale is well inside the hubble radius to the late time @xmath66 when the smallest scale is sufficiently outside . if one is focusing on the equilateral configuration , rather than integrate from @xmath46 to @xmath66 , it suffices to compute the integrals for the modes from the time when each of them satisfy the sub - hubble condition , say , @xmath121 , to the time when they are well outside the hubble radius , say , when @xmath122 . in other words , one carries out the integrals exactly over the period the modes are evolved to obtain the power spectrum . the presence of the cut - off ensures that the contributions at early times , i.e. near @xmath46 , are negligible . furthermore , it should be noted that , in such a case , the corresponding super - hubble contribution to @xmath89 will saturate the bound ( [ eq : fnl - shs - ub ] ) in power law inflation for all the modes . with the help of a specific example , let us now illustrate that , for a judicious choice of @xmath120 , the results from bingo are largely independent of the upper and the lower limits of the integrals . in fact , working in the equilateral limit , we shall demonstrate these points in two steps for the case of the standard quadratic potential [ see eq . ( [ eq : qp ] ) ] . firstly , focusing on a specific mode , we shall fix the upper limit of the integral to be the time when @xmath123 . evolving the mode from different initial times , we shall evaluate the integrals involved from these initial times to the fixed final time for different values of @xmath120 . this exercise helps us to identify an optimal value for @xmath120 when we shall eventually carry out the integrals from @xmath124 . secondly , upon choosing the optimal value for @xmath120 and integrating from @xmath124 , we shall calculate the integrals for different upper limits . for reasons outlined in the previous two sub - sections , it proves to be necessary to consider the contributions to the bi - spectrum due to the fourth and the seventh terms together . moreover , since the first and the third , and the fifth and the sixth , have similar structure in the equilateral limit , it turns out to be convenient to club these terms as we have discussed before . in fig . [ fig : g - delta ] , we have plotted the value of @xmath125 times the different contributions to the bi - spectrum , viz . @xmath126 , @xmath127 , @xmath128 and @xmath129 , as a function of @xmath120 when the integrals have been carried out from @xmath130 of @xmath131 , @xmath132 and @xmath133 for a mode which leaves the hubble radius around @xmath134 e - folds before the end of inflation in the case of the quadratic potential . the figure suggests that the term @xmath128 is fairly independent of the cut - off parameter @xmath120 . this can be attributed to the fact that the integral @xmath135 depends on the quantity @xmath51 , which effectively behaves as a cut off . for this reason , actually , we shall _ not _ introduce the cut off while evaluating the term @xmath136 . moreover , the figure clearly indicates @xmath137 to be a highly suitable value for the other terms . a larger @xmath120 leads to a sharper cut off reducing the value of the integrals . one could work with a smaller @xmath120 , in which case , the figure suggests that , one would also need to necessarily integrate from deeper inside the hubble radius . [ 0][1][1.0]@xmath138 [ 0][1][1.0]@xmath120 -10pt in fig . [ fig : g - nshs - all ] , after fixing @xmath120 to be @xmath139 and , with the initial conditions imposed at @xmath140 , we have plotted the four contributions to the bi - spectrum for a mode that leaves the hubble radius at @xmath141 e - folds before the end of inflation as a function of the upper limit of the integrals . [ 0][1][1.0]@xmath138 [ 0][1][0.9]@xmath142 [ 0][1][0.9]@xmath130 -10pt it is evident from the figure that the values of the integrals converge quickly once the mode leave the hubble radius . for efficient numerical integration , as in the case of the power spectrum , we have chosen the super - hubble limit to correspond to @xmath143 . we have repeated similar tests for the other models of our interest too . these tests confirm the conclusions that we have arrived at above , indicating the robustness of the numerical methods and procedures that we have adopted . it is important that we stress three points further at this stage of our discussion . firstly , in situations involving departures from slow roll , the dominant contribution , viz . the @xmath128 term , does not require the introduction of the cut off at all . secondly , in the absence of analytical expressions to compare the numerical results with , it is imperative that a suitable value for the cut off parameter @xmath120 is arrived at by repeating the exercise that we have carried out in fig . [ fig : g - delta ] for the specific model of interest . in this context , it is further important to appreciate the fact that the integrals involved will depend on both the cut off parameter @xmath120 as well as the initial value of @xmath144 . a complete check seems to be mandatory in order to ensure that a unique result is arrived at . ( as we mentioned above , we have indeed carried out this exercise for the models that we consider here . ) lastly , we also need to emphasize that the above comments and conclusions would not apply to certain special cases wherein the amplitude of the curvature perturbation can evolve on super - hubble scales such as , say , in ultra slow roll inflation ( in this context , see , for instance , ref . @xcite ) . before we go on to consider the bi - spectra generated in the inflationary models of our interest , in order to demonstrate the accuracy of bingo , we shall compare the numerical results from the code with the analytical results that can be arrived at in three situations . in the following sub - sections , we shall compare the results from bingo with : ( i ) the spectral dependence that can be arrived at in the context of power law inflation in the equilateral limit , ( ii ) the slow roll results as applied to the case of the conventional quadratic potential for a generic triangular configuration of the wavevectors , and ( iii ) the analytic results that can be obtained for the starobinsky model in the equilateral limit . the first case that we shall discuss is power law inflation wherein , as we shall soon outline , the spectral shape of the non - zero contributions to the bi - spectrum can be easily arrived at in the equilateral limit . consider the case of power law inflation described by the scale factor ( [ eq : pli ] ) with @xmath145 . in such a case , as we have seen , @xmath50 is a constant and , hence , @xmath44 and @xmath51 , which involve derivatives of @xmath50 , reduce to zero . since the contributions due to the fourth and the seventh terms , viz . @xmath146 and @xmath147 , depend on @xmath51 and @xmath44 , respectively [ cf . ( [ eq : cg4 ] ) and ( [ eq : g7 ] ) ] , these terms vanish identically in power law inflation . as is well known , one can express the modes @xmath13 as @xmath148 , where @xmath149 is the mukhanov - sasaki variable . in power law inflation , one finds that , the variable @xmath149 depends only on the combination @xmath150 ( see , for example , ref . moreover , since @xmath50 is a constant in power law expansion , we have @xmath151 . under these conditions , in the equilateral limit , with a simple rescaling of the variable of integration in the expressions ( [ eq : cg1 ] ) , ( [ eq : cg2 ] ) , ( [ eq : cg3 ] ) , ( [ eq : cg5 ] ) and ( [ eq : cg6 ] ) , it is straightforward to show that the quantities @xmath152 , @xmath153 , @xmath154 , @xmath155 and @xmath156 , all depend on the wavenumber as @xmath157 . then , upon making use of the asymptotic form of the modes @xmath13 , it is easy to illustrate that the corresponding contributions to the bi - spectrum , viz . @xmath158 , @xmath159 and @xmath129 , all behave as @xmath160 . since the power spectrum in power law inflation is known to have the form @xmath161 ( see , for example , refs . @xcite ) , the expression ( [ eq : fnl - eq ] ) for @xmath39 then immediately suggests that the quantity will be strictly scale invariant for all @xmath93 . in fact , apart from these results , it is also simple to establish the following relation between the different contributions : @xmath162 , a result , which , in fact , also holds in slow roll inflation @xcite . in other words , in power law inflation , it is possible to arrive at the spectral dependence of the non - zero contributions to the bi - spectrum without having to explicitly calculate the integrals involved . further , one can establish that the non - gaussianity parameter @xmath39 is exactly scale independent for any value of @xmath93 . while these arguments do not help us in determining the amplitude of the various contributions to the bi - spectrum or the non - gaussianity parameter , their spectral shape and the relative magnitude of the above - mentioned terms provide crucial analytical results to crosscheck our numerical code . in fig . [ fig : pl ] , we have plotted the different non - zero contributions to the bi - spectrum computed using our numerical code and the spectral dependence we have arrived at above analytically for two different values of @xmath93 in the case of power law inflation . we have also indicated the relative magnitude of the first and the third and the fifth and the sixth terms arrived at numerically . lastly , we have also illustrated the scale independent behavior of the non - gaussianity parameter @xmath39 for both the values of @xmath93 . it is clear from the figure that the numerical results agree well with the results and conclusions that we arrived at above analytically . [ 0][1][1.5]@xmath138 [ 0][1][1.5]@xmath163 [ 0][1][1.5]@xmath39 -10pt let us now turn to the example of the conventional quadratic potential . the bi - spectrum in such a case can be evaluated analytically in the slow roll approximation for an arbitrary triangular configuration of the wavevectors @xcite . in the slow roll limit , one finds that , the contributions due to the first , the second , the third and the seventh terms are of the same order , while the remaining terms turn out to be comparatively insignificant . for convenience , let us explicitly write down the contributions to the bi - spectrum due to the first , the second , the third and the seventh terms arrived at in the slow roll approximation . they are given by @xmath164,\label{eq : g1-sr}\\ g_2(\vka,\vkb,\vkc ) & = & \f{h_{_{\rm i}}^4}{16\ , \mpl^4\ , \epsilon_1}\ , \l(\f{1}{k_1\ , k_2\,k_3}\r)^3\,\l(\vka\cdot\vkb+\rm two~permutations\r)\nn\\ & & \times\ , \l[-k_{_{\rm t}}+\f{1}{k_{_{\rm t}}}\ , \l(k_1\ , k_2+k_1\ , k_3+k_2\ , k_3\r ) + \f{k_1\,k_2\ , k_3}{k_{_{\rm t}}^{2}}\r],\\ g_3(\vka,\vkb,\vkc ) & = & -\f{h_{_{\rm i}}^4}{16\ , \mpl^4\ , \epsilon_1}\ , \l(\f{1}{k_1\ , k_2\,k_3}\r)^3\ , \biggl[\l(\vka\cdot\vkb\r)\ , \f{k_3 ^ 2}{k_{_{\rm t}}}\ , \l(2+\f{k_1+k_2}{k_{_{\rm t}}}\r)\nn\\ & & + \,{\rm two~permutations}\biggr],\\ g_7(\vka,\vkb,\vkc ) & = & 2\ , \pi^{4}\ , \epsilon_2\,\l(\f{1}{k_1\ , k_2\,k_3}\r)^3\nn\\ & & \times\ , \l[k_1^{3}\ , { \cal p}_{_{\rm s}}(k_2)\,{\cal p}_{_{\rm s}}(k_3 ) + \rm two~permutations\r],\label{eq : g7-sr}\end{aligned}\ ] ] respectively , with @xmath165 . the quantities @xmath50 and @xmath44 are the first two slow roll parameters which are assumed to be largely constant , while @xmath166 denotes the hubble scale during slow roll inflation . the scalar power spectrum during slow roll can be written as @xcite @xmath167 where @xmath168 and @xmath169 denotes a suitable , so - called , pivot scale . -100pt -50pt -50pt from the above expressions and , upon using the fact that @xmath170 has to be zero , one can arrive at the corresponding non - gaussianity parameter @xmath0 for an arbitrary triangular configuation of the wavevectors . we find that it can be written as @xmath171^{-1}\nn\\ & & \times\,\biggl[\epsilon_1\,{\cal f}_{1}(\vka,\vkb,\vkc ) + \f{\epsilon_1}{2}\,{\cal f}_{2}(\vka,\vkb,\vkc)\nn\\ & & -\,\f{\epsilon_1}{2}\,{\cal f}_{3}(\vka,\vkb,\vkc ) + \frac{\epsilon_2}{2}\,{\cal f}_{7}(\vka,\vkb,\vkc)\biggr ] , \end{aligned}\ ] ] with @xmath172\biggr\},\\ { \cal f}_{2}(\vka,\vkb,\vkc ) & = & \l[1+\l(\frac{k_2}{k_1}\r)^2+\l(\f{k_3}{k_1}\r)^2\r]\nn\\ & & \times\,\biggl[1+\f{k_2}{k_1}+\f{k_3}{k_1}\nn\\ & & -\,\l(1+\f{k_2}{k_1}+\f{k_3}{k_1}\r)^{-1}\ , \l(\f{k_2}{k_1}+\f{k_3}{k_1}+\f{k_2}{k_1}\,\f{k_3}{k_1}\r)\nn\\ & & -\l(1+\f{k_2}{k_1}+\frac{k_3}{k_1}\r)^{-2}\ , \f{k_2}{k_1}\,\f{k_3}{k_1}\biggr],\\ { \cal f}_{3}(\vka,\vkb,\vkc ) & = & \l(1+\frac{k_2}{k_1}+\f{k_3}{k_1}\r)^{-1}\nn\\ & & \times\,\biggl\{\l[2+\l(1+\f{k_2}{k_1}+\f{k_3}{k_1}\r)^{-1}\ , \l(\f{k_2}{k_1}+\f{k_3}{k_1}\r)\r]\nn\\ & & \times\,\l[1-\l(\f{k_2}{k_1}\r)^2 -\l(\frac{k_3}{k_1}\r)^2\r]\nn\\ & & -\,\l[2+\l(1+\f{k_2}{k_1}+\f{k_3}{k_1}\r)^{-1}\ , \l(1+\f{k_2}{k_1}\r)\r]\nn\\ & & \times\,\l[1+\l(\f{k_2}{k_1}\r)^2 -\l(\frac{k_3}{k_1}\r)^2\r]\,\l(\f{k_3}{k_1}\r)^2\nn\\ & & -\,\l[2+\l(1+\f{k_2}{k_1}+\f{k_3}{k_1}\r)^{-1}\ , \l(1+\f{k_3}{k_1}\r)\r]\nn\\ & & \times\,\l[1-\l(\f{k_2}{k_1}\r)^2 + \l(\frac{k_3}{k_1}\r)^2\r]\,\l(\f{k_2}{k_1}\r)^2\biggr\},\\ { \cal f}_{7}(\vka,\vkb,\vkc ) & = & \l(\f{k_1}{k_\ast}\r)^{2\ , \l(n_{_{\rm s}}-1\r)}\ , \biggl[\l(\frac{k_2}{k_1}\r)^{n_{_{\rm s}}-1}\ , \l(\frac{k_3}{k_1}\r)^{n_{_{\rm s}}-1}\nn\\ & & + \l(\f{k_2}{k_1}\r)^3\,\l(\f{k_3}{k_1}\r)^{n_{_{\rm s}}-1 } + \l(\f{k_3}{k_1}\right)^3\ , \l(\f{k_2}{k_1}\r)^{n_{_{\rm s}}-1}\biggr].\end{aligned}\ ] ] in fig . [ fig : qp - ac ] , we have plotted the above analytical expression for @xmath0 in the case of the quadratic potential as well as the corresponding numerical result from bingo for an arbitrary triangular configuration of the wavevectors @xcite . the two plots look remarkably alike . also , in fig . [ fig : qp - ac - c ] , we have plotted the percentage difference between the analytical and the numerical results . the figure suggests that the analytical and the numerical results match within @xmath173 . it is worth stressing that part of the difference can be attributed to the slow roll approximation utilized in arriving at the analytical result . the extent of the agreement reflects the fact that bingo performs rather well . however , we should stress here that the level of accuracy of bingo can be improved by essentially carrying out the integrals from a larger initial value of @xmath144 ( i.e. from an earlier point in time ) and simultaneously working with a smaller value of the cut off parameter @xmath120 . the last of the examples that we shall discuss is the starobinsky model which involves a linear potential with a sharp change in its slope [ cf . ( [ eq : p - sm ] ) ] . in the starobinsky model , under certain conditions , the complete scalar bi - spectrum can be evaluated analytically in the equilateral limit @xcite . as we shall discuss in the next section , in the starobinsky model , the change in the slope causes a brief period of fast roll , which leads to sharp features in the scalar power spectrum ( see fig . [ fig : sps - all ] ) . it was known that , for certain range of parameters , one could evaluate the scalar power spectrum analytically in the starobinsky model , which matches the actual , numerically computed spectrum exceptionally well @xcite . interestingly , it has been recently shown that , in the equilateral limit , the model allows the analytic evaluation of the complete scalar bi - spectrum too @xcite . in fig . [ fig : sm - all ] , we have plotted the numerical as well as the analytical results for the functions @xmath158 , @xmath159 , @xmath128 , and @xmath129 for the starobinsky model . we have plotted for parameters of the model for which the analytical results are considered to be a good approximation @xcite . it is evident from the figure that the numerical results match the analytical ones very well . importantly , the agreement proves to be excellent in the case of the dominant contribution @xmath128 . [ 0][1][1.0]@xmath138 [ 0][1][1.0]@xmath174 a couple of points concerning concerning the numerical results in the case of the starobinsky model ( both in fig . [ fig : sps - all ] wherein we have plotted the power spectrum as well as in fig . [ fig : sm - all ] which contains the bi - spectrum ) require some clarification . the second ( and higher ) derivatives of the potential ( [ eq : p - sm ] ) evidently contain a discontinuity . these discontinuities needs to be smoothened in order for the problem to be numerically tractable . the spectra and the bi - spectra in the starobinsky model we have illustrated have been computed with a suitable smoothing of the discontinuity , while at the same time retaining a sufficient level of sharpness so that they closely correspond to the analytical results that have been arrived at @xcite . in the remainder of this paper , as an immediate application , we shall utilize bingo to examine the power of the non - gaussianity parameter @xmath0 to help us discriminate between inflationary models that lead to similar features at the level of the scalar power spectrum . we shall list the various inflationary models of our interest in this section , and describe the power spectra that arise in these models . in the next section , we shall present the essential results and compare the @xmath0 generated in the different models . broadly , the models that we shall consider can be categorized into three classes . the first class shall involve potentials which admit a relatively mild and brief departure from slow roll . the second class shall contain small but repeated deviations from slow roll , while the third and the last class shall involve a short but rather sharp departure from slow roll . let us now briefly outline the different inflationary models that we shall consider under these classes and discuss the scalar power spectra that are generated by them . under the first class , we shall consider models wherein a step has been introduced in potentials that otherwise admit only slow roll . given a potential , say , @xmath175 , we shall introduce the step by _ multiplying _ the potential by the following function : h _ ( ) = 1 + ( ) , [ eq : step ] as is often done in the literature @xcite . clearly , the quantities @xmath176 , @xmath177 and @xmath178 denote the height , the location and the width of the step , respectively . we shall consider the effects of the introduction of the step in the archetypical quadratic large field model , viz . v ( ) = m^2 ^ 2,[eq : qp ] where @xmath179 represents the mass of the inflaton , and a small field model governed by the potential v ( ) = v_0 [1-()^p ] . [ eq : sfm ] we shall specifically focus on the case wherein @xmath180 and @xmath181 , as it leads to a tilt that is consistent with the observations , and a smaller tensor - to - scalar ratio ( of @xmath182 ) than the above quadratic potential ( in planck units ) can be considered to be a drawback of the small field model . ] . the introduction of the step has been shown to lead to a short period of deviation from slow roll inflation , which , in turn , produces a burst of oscillations in the scalar power spectrum , resulting in an improved fit to the data , in both these models @xcite . the best fit parameters in the case of the quadratic potential ( [ eq : qp ] ) prove to be @xmath183 , @xmath184 , @xmath185 and @xmath186 . the field is assumed to start on the inflationary attractor at an initial value of @xmath187 , so that at least @xmath188 e - folds of inflation takes place . the small field model ( [ eq : sfm ] ) too is found to lead to a very similar spectrum and an almost identical extent of improvement in the fit to the data . the best fit values of the parameters in this case are found to be @xmath189 , @xmath190 , @xmath191 and @xmath192 . further , as in the quadratic case , the field is set to start on the attractor . the starting value of the field is chosen to be @xmath193 , and it rolls through the step towards @xmath194 . [ fig : p ] contains an illustration of the potentials corresponding to the quadratic case ( [ eq : qp ] ) and the small field model ( [ eq : sfm ] ) in the presence of the step ( [ eq : step ] ) for the values of the parameters discussed above . -10pt the second class of models that we shall consider are those which lead to small but repeated deviations from slow roll as in the case of models which contain oscillatory terms in the inflaton potential . the first such model that we shall consider is the conventional quadratic potential ( [ eq : qp ] ) that is modulated by sinusoidal oscillations as follows @xcite : v ( ) = m^2 ^ 2 [1 + (+)].[eq : qpsin ] the second model that we shall consider in this context is the axion monodromy model which is motivated by string theory @xcite . the inflaton potential in such a case is given by v ( ) = [+(+ ) ] . [ eq : amm ] evidently , in these potentials , while the parameters @xmath176 and @xmath195 describe the amplitude and the frequency of the superimposed oscillations , the parameter @xmath196 shifts the oscillations within one period . note that , whereas the amplitude of the oscillation is fixed in the axion monodromy model , the amplitude depends quadratically on the field in the chaotic model described by the potential ( [ eq : qpsin ] ) . the inflaton oscillates as it rolls down these potentials , and these oscillations persist all the way until the end of inflation . the oscillating inflaton leads to small oscillations in the slow roll parameters ( in this context , see , for instance , ref . @xcite ) , which , as we shall soon illustrate , results in continuing oscillations in the primordial scalar power spectrum . it is found that , in the case of the potential ( [ eq : qpsin ] ) , the power spectrum corresponding to @xmath197 , @xmath198 , @xmath199 and @xmath200 leads to a modest improvement in the fit to the cmb data @xcite . whereas , the axion monodromy model ( [ eq : amm ] ) leads to a considerable improvement in the fit for the following values of the parameters @xmath201 , @xmath202 , @xmath203 and @xmath204 . while the field is assumed to start on the inflationary attractor at the initial value of about @xmath205 in the monodromy model , the field is evolved from @xmath206 in the case of the quadratic potential with sinusoidal oscillations @xcite . we have plotted the two potentials ( [ eq : qpsin ] ) and ( [ eq : amm ] ) in fig . [ fig : p ] for the above values of the parameters . we shall consider two models under the last class , both of which are known to lead to brief but sharp departures from slow roll . the first of the inflationary models that we shall consider in this class is described by the following potential containing two parameters @xmath179 and @xmath207 : v ( ) = ^2 - ^n + ^2(n-1 ) . [ eq : mssm - p ] the third quantity @xmath208 that appears in the potential is an integer which takes values greater than two . such potentials are known to arise in certain minimal supersymmetric extensions of the standard model @xcite . it is worthwhile noting here that the case of @xmath209 has been considered much earlier for reasons similar to what we shall consider here , viz . towards producing certain features in the scalar power spectrum @xcite . in the above potential , the coefficient of the @xmath210 term is chosen such that the potential contains a point of inflection at , say , @xmath211 ( i.e. the location where both @xmath212 and @xmath213 vanish ) , so that @xmath177 given by _ 0 = ^. if one starts at a suitable value of the field beyond the point the inflection in the above potential , it is found that one can achieve two epochs of slow roll inflation sandwiching a brief period of departure from inflation ( lasting for a little less than a e - fold ) , a scenario which has been dubbed as punctuated inflation @xcite . in fact , it is the point of inflection , around which the potential exhibits a plateau with an extremely small curvature , which permits such an evolution to be possible . it is found that the following values for the potential parameters results in a power spectrum that leads to an improved fit to the cmb data in the @xmath209 case : @xmath214 and @xmath215 . [ fig : p ] contains a plot of the potential ( [ eq : mssm - p ] ) corresponding to these values of the parameters . it should be added that the field is assumed to start from rest at an initial value of @xmath216 on the potential to arrive at the required behavior . the second model that we shall consider in the last class is the starobinsky model @xcite , which we had briefly discussed in the previous section . as we shall soon illustrate , the model leads to a scalar power spectrum that has a certain resemblance to the spectrum generated by punctuated inflation . the model consists of a linear potential with a sharp change in its slope at a given point , and can be described as follows : @xmath217 we have illustrated this potential in fig . [ fig : p ] . evidently , while the value of the scalar field where the slope changes abruptly is @xmath218 , the slope of the potential above and below @xmath177 are given by @xmath219 and @xmath220 , respectively moreover , the quantity @xmath221 denotes the value of the potential at @xmath222 . a crucial assumption of the starobinsky model is that the value of @xmath223 is sufficiently large so that the behavior of the scale factor always remains close to that of de sitter . the change in the slope causes a short period of deviation from slow roll as the field crosses @xmath218 . however , in contrast to the case of the punctuated inflationary scenario , where one encounters a brief departure from inflation , inflation continues uninterrupted in the starobinsky model . we have not compared the starobinsky model with the data , and we shall work with two different sets of values for the parameters of the model . one set corresponds to values of the parameters that allow for the comparison of the analytical results for the bi - spectrum that have been obtained in this case @xcite with the corresponding numerical ones . the other set shall be chosen to lead to a spectrum that closely mimics the power spectrum encountered in punctuated inflation . in the case of the former , as we had already mentioned in the caption of fig . [ fig : sm - all ] , we shall choose the following values of the parameters : @xmath224 , @xmath225 , @xmath226 and @xmath227 , while , in the case of the latter , we shall work with the same values of @xmath228 and @xmath218 , but shall set @xmath229 , and @xmath230 . also , we shall work with an initial value of @xmath231 in the first instance and with @xmath232 in the second . further , we shall start with field velocities that are determined by the slow roll conditions in both the cases . we shall now discuss the scalar power spectra generated by the different inflationary models that we have listed above . we shall highlight here certain aspects of the power spectra that arise in the different class of models . in fig . [ fig : sps - all ] , we have illustrated the scalar power spectrum that arises in the different models that we consider . as we had discussed in sub - sec . [ subsec : dnm ] , the standard bunch - davies initial conditions are imposed on the perturbations when the modes are well inside the hubble radius . the modes are evolved on to super hubble scales , and the spectrum is evaluated when the amplitude of the curvature perturbation has frozen when they are sufficiently outside the hubble radius . [ 0][1][1.0]@xmath233 [ 0][1][1.0]@xmath163 let us first focus on the power spectra that arise in the starobinsky model @xcite and the punctuated inflationary scenario @xcite . note that , both these models lead to a step like feature as well as a spike in the power spectrum . the spikes arise due to the sharp departure from slow roll that occurs in these models . while the first slow roll parameter @xmath50 remains small in the starobinsky model as the field crosses the transition , the second slow parameter @xmath44 turns very large briefly @xcite . in the case of punctuated inflation , @xmath50 itself grows to a large value thereby actually interrupting inflation for about a e - fold . it is this property that results in a sharper spike in the power spectrum in the case of punctuated inflation than the starobinsky model . the overall step in these models is easier to understand , and it simply arises due to the difference in the hubble scales associated with the slow roll epochs before and after the period of fast roll . both these models also lead to oscillations before the spectra turn nearly scale invariant on small scales . the spectra that arises in punctuated inflation , in addition to leading to a better fit to the outliers at very small multipoles ( because of the drop in power on these scales ) , also provides an improvement in the fit to the outlier at @xmath234 @xcite . let us now turn to the cases of the models with a step in the potential and the potentials which contain oscillatory terms . in contrast to the starobinsky model and the punctuated inflationary scenario , these models lead to much milder deviations from slow roll . as a result , they also lead to smaller deviations from a nearly scale invariant power spectrum . in the context of models with a step in the potential , a step at the correct location and of a suitable amplitude leads to a burst of oscillations in the power spectrum which in turn provides an improved fit to the outliers at @xmath234 and @xmath2 @xcite . it is interesting to note in fig . [ fig : sps - all ] that the spectra from punctuated inflation and the model with a step in the potential match briefly as they oscillate near scales corresponding to @xmath234 . clearly , it is this behavior that leads to a better fit to the outlier in the data at these scales . while all the models that we have discussed on until now result in localized features , the models with oscillating inflation potentials are somewhat special as they contain non - local features , i.e. they lead to continuing oscillations that extend over a wide range of scales in the power spectrum . needless to add , they arise due to the persisting oscillations of the scalar field as it rolls up and down the modulations in the potential . interestingly , these features provide an improved fit to the data over a large range of all scales @xcite . we shall now discuss the bi - spectra and the non - gaussianity parameter @xmath0 arrived at numerically in the various models of our interest . in fig . [ fig : g - all.ps ] , we have plotted the various contributions , viz . @xmath158 , @xmath159 , @xmath128 and @xmath129 ( in the equilateral limit ) for the punctuated inflationary scenario driven by the potential ( [ eq : mssm - p ] ) , the quadratic potential ( [ eq : qp ] ) with the step ( [ eq : step ] ) , and the axion monodromy model ( [ eq : amm ] ) which contains oscillations in the inflaton potential . [ 0][1][1.0]@xmath138 [ 0][1][1.0]@xmath163 -15pt -15pt these plots ( and also the ones in fig . [ fig : sm - all ] ) clearly point to the fact that it is the combination @xmath128 that contributes the most to the scalar bi - spectrum in these cases @xcite . in fig . [ fig : fnl - all ] , we have plotted the quantity @xmath39 due to the dominant contribution that arises in the various models that we have considered . [ 0][1][1.0]@xmath39 [ 0][1][1.0]@xmath235 [ 0][1][1.0]@xmath163 -12pt it is clear from this figure that , while in certain cases @xmath89 can prove to be a good discriminator , it can not help in others , and its ability to discriminate depends strongly on the differences in the background dynamics . for instance , the evolution of the first two slow roll parameters are very similar when a step is introduced in either the quadratic potential or a small field model @xcite . hence , it is not surprising that the @xmath39 behaves in a similar fashion in both these models . whereas , @xmath39 proves to be substantially different in punctuated inflation and the starobinsky model . recall that , in the starobinsky model , the first slow roll parameter remains small throughout the evolution . in contrast , it grows above unity for a very short period ( leading to a brief interruption of the accelerated expansion ) in the punctuated inflationary scenario . it is this departure from inflation that leads to a sharp drop in the power spectrum and a correspondingly sharp rise in the parameter @xmath39 in punctuated inflation . in fact , this occurs for modes that leave the hubble radius just before inflation is interrupted @xcite . however , note that , @xmath39 grows with @xmath163 at large wavenumbers in the starobinsky model . this can be attributed to the fact that @xmath51 , which determines the contribution due to the fourth term , diverges due to the discontinuity in the second derivative of the potential @xcite . but , we should hasten to clarify that this can be shown to be an unphysical artifact arising due to the presence of a heaviside function in the description of the potential . similarly , we find that @xmath39 is rather large in the axion monodromy model in contrast to the case wherein the conventional quadratic potential is modulated by an oscillatory term . the large value of @xmath89 that arises in the monodromy model can be attributed to the resonant behavior encountered in the model @xcite . in fact , we have also evaluated the @xmath39 for the case of quadratic potential modulated by sinusoidal oscillations , which too leads to continuing , periodic features in the scalar power spectrum @xcite . however , we find that the @xmath39 in such a case proves to be rather small ( of the order @xmath236 or so ) . in this work , we have presented bingo , a code for the efficient computation of the bi - spectrum and the non - gaussianity parameter @xmath0 in single field inflationary models involving the canonical scalar field . while there has been previous efforts in the literature towards numerically calculating the bi - spectrum and the non - gaussianity parameter @xmath0 , our effort in arriving at bingo can be said to be more complete for the following four reasons . first and foremost , we have explicitly illustrated that the procedures that bingo adopts to evaluate the integrals involved are robust . on the one hand , we have converged on a suitable form for the cut off required in the sub - hubble domain ( in particular , the value of the cut off parameter @xmath120 ) and the corresponding lower limit of the integrals , after a careful investigation of their effects on the amplitude of the bi - spectrum . on the other , we have shown that the super - hubble contributions to the complete bi - spectrum are negligible , which allows us to arrive at a convenient upper limit for the integrals . secondly , bingo can calculate all the contributions to the bi - spectrum . thirdly , it can evaluate the bi - spectrum for an arbitrary triangular configuation of the wavevectors . lastly , we have compared the numerical results from bingo with the analytical expressions available in certain specific cases , an exercise which illustrates that bingo can be accurate to better than @xmath173 . as we had mentioned in the introductory section , we have made a version of bingo , one that focuses on the equilateral limit , available online at https://www.physics.iitm.ac.in/~sriram/bingo/bingo.html . moreover , if needed , as we have pointed out earlier , the level of accuracy of bingo can be improved by carrying out the integrals from a larger initial value of @xmath144 ( i.e. from an earlier point in time ) and simultaneously working with a smaller value of the cut off parameter @xmath120 . in fact , in the code that is publicly available , these parameters have been allowed to be set by the user , as is desired . after presenting bingo and illustrating the extent of its accuracy , we had made use of the code to examine the power of the non - gaussianity parameter @xmath0 to lift degeneracies between various single field inflationary models involving the canonical scalar field at the level of the power spectrum . with this goal in mind , using bingo , we have evaluated the quantity @xmath39 in a slew of models that generate features in the scalar perturbation spectrum . we find that the amplitude of @xmath39 proves to be rather different when the dynamics of the background turns out reasonably different , which , in retrospect , need not be surprising at all . for instance , models such as the punctuated inflationary scenario and the starobinsky model which lead to very sharp features in the power spectrum also lead to substantially large @xmath0 . such possibilities can aid us discriminate between the models to some extent . we had focused on evaluating the quantity @xmath0 in the equilateral limit . it will be worthwhile to compute the corresponding values in the other limits as well as for the other forms ( such as the squeezed and the orthogonal ones ) of the non - gaussianity parameter @xcite . in particular , it will be interesting to examine if the so - called consistency relation between the local non - gaussianity parameter @xmath0 and the scalar spectral index in the squeezed limit is valid even in situations wherein extreme deviations from slow roll occur ( in this context , see refs . @xcite ) . actually , since many of the models leading to features have interesting bi - spectral shapes , it is important that we arrive at the spectral shapes for the various models for an arbitrary triangular configurations of the wavenumbers . it is also important that , apart from computing the local non - gaussianity parameter as we have done here , one evaluates the corresponding equilateral and the orthogonal parameters as well . we are currently working on extending the scope of bingo to these cases . we would like to conclude by highlighting one important point . having computed the primordial bi - spectrum , the next logical step would be to compute the corresponding cmb bi - spectrum , an issue which we have not touched upon as it is beyond the scope of the current work . while tools seem to be available to evaluate the cmb bi - spectrum based on the first order brightness function , the contribution due to the brightness function at the second order remains to be understood satisfactorily ( in this context , see ref . @xcite and the last reference in ref . this seems to be an important aspect that is worth investigating closer . we would like to acknowledge the use of the high performance computing facilities at the harish - chandra research institute , allahabad , india ( http://cluster.hri.res.in ) . dkh also wishes to acknowledge support from the korea ministry of education , science and technology , gyeongsangbuk - do and pohang city for independent junior research groups at the asia pacific center for theoretical physics . 99 j. maldacena , jhep * 0305 * , 013 ( 2003 ) . d. seery and j. e. lidsey , jcap * 0506 * , 003 ( 2005 ) ; x. chen , phys . d * 72 * , 123518 ( 2005 ) ; x. chen , m .- x . huang , s. kachru and g. shiu , jcap * 0701 * , 002 ( 2007 ) ; d. langlois , s. renaux - petel , d. a. steer and t. tanaka , phys . rev . * 101 * , 061301 ( 2008 ) ; phys . d * 78 * , 063523 ( 2008 ) . x. chen , adv . astron . * 2010 * , 638979 ( 2010 ) . a. gangui , f. lucchin , s. matarese and s. mollerach , astrophys . j. * 430 * , 447 ( 1994 ) ; a. gangui , phys . rev . d * 50 * , 3684 ( 1994 ) ; a. gangui and j. martin , mon . not . * 313 * , 323 ( 2000 ) ; l. wang and m. kamionkowski , phys . d , 063504 ( 2010 ) . j. dunkley _ et al . _ , astrophys . j. suppl . * 180 * , 306 ( 2009 ) ; e. komatsu _ et al . _ , astrophys . j. suppl . * 180 * , 330 ( 2009 ) . d. larson _ et al . _ , astrophys . j. suppl . * 192 * , 16 ( 2011 ) ; e. komatsu _ et al . _ , astrophys . j. suppl . * 192 * , 18 ( 2011 ) . c. l. bennett _ et al . _ , arxiv:1212.5225v1 [ astro-ph.co ] ; g. hinshaw _ et al . _ , arxiv:1212.5226v1 [ astro-ph.co ] . e. komatsu and d. n. spergel , phys . d * 63 * , 063002 ( 2001 ) ; e. komatsu , d. n. spergel and b. d. wandelt , astrophys . j. * 634 * , 14 ( 2005 ) ; d. babich and m. zaldarriaga , phys . d * 70 * , 083005 ( 2004 ) ; m. liguori , f. k. hansen , e. komatsu , s. matarrese and a. riotto , phys . d * 73 * , 043505 ( 2006 ) ; c. hikage , e. komatsu and t. matsubara , astrophys . j. * 653 * ( 2006 ) 11 ( 2006 ) ; j. r. fergusson and e. p. s. shellard , phys . d * 76 * , 083523 ( 2007 ) ; a. p. s. yadav , e. komatsu and b. d. wandelt , astrophys . j. * 664 * , 680 ( 2007 ) ; p. creminelli , l. senatore and m. zaldarriaga , jcap * 0703 * , 019 ( 2007 ) ; a. p. s. yadav and b. d. wandelt , phys . rev . lett . * 100 * , 181301 ( 2008 ) ; c. hikage , t. matsubara , p. coles , m. liguori , f. k. hansen and s. matarrese , mon . not . soc . * 389 * , 1439 ( 2008 ) ; o. rudjord , f. k. hansen , x. lan , m. liguori , d. marinucci and s. matarrese , astrophys . j. * 701 * , 369 ( 2009 ) ; k. m. smith , l. senatore and m. zaldarriaga , jcap * 0909 * , 006 ( 2009 ) ; j. smidt , a. amblard , c. t. byrnes , a. cooray , a. heavens and d. munshi , phys . d * 81 * , 123007 ( 2010 ) ; j. r. fergusson , m. liguori and e. p. s. shellard , arxiv:1006.1642v2 [ astro-ph.co ] . m. liguori , e. sefusatti , j. r. fergusson and e. p. s. shellard , adv . astron . * 2010 * , 980523 ( 2010 ) ; a. p. s. yadav and b. d. wandelt , arxiv:1006.0275v3 [ astro-ph.co ] ; e. komatsu , class . quantum grav . , 124010 ( 2010 ) . h. funakoshi and s. renaux - petel , arxiv:1211.3086v1 [ astro-ph.co ] . e. w. kolb and m. s. turner , _ the early universe _ ( addison - wesley , redwood city , california , 1990 ) ; s. dodelson , _ modern cosmology _ ( academic press , san diego , u.s.a . , 2003 ) ; v. f. mukhanov , _ physical foundations of cosmology _ ( cambridge university press , cambridge , england , 2005 ) ; s. weinberg , _ cosmology _ ( oxford university press , oxford , england , 2008 ) ; r. durrer , _ the cosmic microwave background _ ( cambridge university press , cambridge , england , 2008 ) ; d. h. lyth and a. r. liddle , _ the primordial density perturbation _ ( cambridge university press , cambridge , england , 2009 ) ; p. peter and j - p . uzan , _ primordial cosmology _ ( oxford university press , oxford , england , 2009 ) ; h. mo , f. v. d. bosch and s. white , _ galaxy formation and evolution _ ( cambridge university press , cambridge , england , 2010 ) . h. kodama and m. sasaki , prog . * 78 * , 1 ( 1984 ) ; v. f. mukhanov , h. a. feldman and r. h. brandenberger , phys . rep . * 215 * , 203 ( 1992 ) ; j. e. lidsey , a. liddle , e. w. kolb , e. j. copeland , t. barreiro and m. abney , rev . mod . phys . * 69 * , 373 ( 1997 ) ; d. h. lyth and a. riotto , phys . rep . * 314 * , 1 ( 1999 ) ; a. riotto , arxiv : hep - ph/0210162 ; j. martin , braz . j. phys . * 34 * , 1307 ( 2004 ) ; lect . notes phys . * 669 * , 199 ( 2005 ) ; lect . notes phys . * 738 * , 193 ( 2008 ) ; b. bassett , s. tsujikawa and d. wands , rev . * 78 * , 537 ( 2006 ) ; w. h. kinney , arxiv:0902.1529 [ astro-ph.co ] ; l. sriramkumar , curr . * 97 * , 868 ( 2009 ) ; d. baumann , arxiv:0907.5424v1 [ hep - th ] . j. martin , c. ringeval and r. trotta , phys . d * 83 * , 063514 ( 2011 ) . h. v. peiris _ et al . _ , astrophys . j. suppl . * 148 * , 213 ( 2003 ) . s. hannestad , phys . d * 63 * , 043009 ( 2001 ) ; s. l. bridle , a. m. lewis , j. weller and g. efstathiou , mon . not . astron . soc . * 342 * , l72 ( 2003 ) ; p. mukherjee and y. wang , astrophys . j. * 599 * , 1 ( 2003 ) ; s. hannestad , jcap * 0404 * , 002 ( 2004 ) ; a. shafieloo and t. souradeep , phys . d * 70 * , 043523 ( 2004 ) ; d. tocchini - valentini , y. hoffman and j. silk , mon . not . soc . * 367 * , 1095 ( 2006 ) ; a. shafieloo , t. souradeep , p. manimaran , p. k. panigrahi and r. rangarajan , phys . d * 75 * , 123502 ( 2007 ) ; a. shafieloo and t. souradeep , phys . rev . d * 78 * , 023511 ( 2008 ) ; r. nagata and j. yokoyama , phys . d * 79 * , 043010 ( 2009 ) ; g. nicholson and c. r. contaldi , jcap * 0907 * , 011 ( 2009 ) . b. feng and x. zhang , phys . lett . b * 570 * , 145 ( 2003 ) ; m. kawasaki and f. takahashi , phys . b * 570 * , 151 ( 2003 ) ; r. sinha and t. souradeep , phys . d * 74 * , 043518 ( 2006 ) ; m. j. mortonson and w. hu , phys . rev . d * 80 * , 027301 ( 2009 ) . r. k. jain , p. chingangbam , j .- o . gong , l. sriramkumar and t. souradeep , jcap * 0901 * , 009 ( 2009 ) ; r. k. jain , p. chingangbam , l. sriramkumar and t. souradeep , phys . d * 82 * , 023509 ( 2010 ) . j. a. adams , b. cresswell and r. easther , phys . d * 64 * , 123514 ( 2001 ) ; l. covi , j. hamann , a. melchiorri , a. slosar and i. sorbera , phys . d * 74 * , 083509 ( 2006 ) ; j. hamann , l. covi , a. melchiorri and a. slosar , phys . d * 76 * , 023503 ( 2007 ) ; m. j. mortonson , c. dvorkin , h. v. peiris and w. hu , phys . d * 79 * , 103519 ( 2009 ) . d. k. hazra , m. aich , r. k. jain , l. sriramkumar and t. souradeep , jcap * 1010 * , 008 ( 2010 ) . m. benetti , m. lattanzi , e. calabrese and a. melchiorri , phys . d * 84 * , 063509 ( 2011 ) . m. joy , v. sahni , a. a. starobinsky , phys . d * 77 * , 023514 ( 2008 ) ; m. joy , a. shafieloo , v. sahni , a. a. starobinsky , jcap * 0906 * , 028 ( 2009 ) . hunt and s. sarkar , phys . d * 70 * , 103518 ( 2004 ) ; phys . d * 76 * , 123504 ( 2007 ) ; m. kawasaki , f. takahashi and t. takahashi , phys . b * 605 * , 223 ( 2005 ) ; j .- o . gong , jcap * 0507 * , 015 ( 2005 ) ; a. ashoorioon and a. krause , arxiv : hep - th/0607001v1 ; a. ashoorioon , a. krause and k. turzynski , jcap * 0902 * , 014 ( 2009 ) ; m. kawasaki and k. miyamoto , arxiv:1010.3095v2 [ astro-ph.co ] ; a. achucarro , j .- o . gong , s. hardeman , g. a. palma and s. p. patil , jcap * 1101 * , 030 ( 2011 ) ; k. kumazaki , s. yokoyama and n. sugiyama , jcap * 1112 * , 008 ( 2011 ) . j. m. cline , p. crotty and j. lesgourgues , jcap * 0309 * , 010 ( 2003 ) ; c. r. contaldi , m. peloso , l. kofman and a. linde , jcap * 0307 * , 002 ( 2003 ) ; l. sriramkumar and t. padmanabhan , phys . rev . d * 71 * , 103512 ( 2005 ) ; d. boyanovsky , h. j. de vega and n. g. sanchez , phys . d * 74 * , 123006 ( 2006 ) ; phys . d * 74 * , 123007 ( 2006 ) ; b. a. powell and w. h. kinney , phys . d * 76 * , 063512 ( 2007 ) ; c. destri , h. j. de vega and n. g. sanchez , phys . d * 78 * , 023013 ( 2008 ) . j. martin and c. ringeval , phys . d * 69 * , 083515 ( 2004 ) ; phys . d * 69 * , 127303 ( 2004 ) ; jcap * 0501 * , 007 ( 2005 ) ; m. zarei , phys . rev . d * 78 * , 123502 ( 2008 ) . c. pahud , m. kamionkowski and a. r. liddle , phys . d * 79 * , 083503 ( 2009 ) . r. flauger , l. mcallister , e. pajer , a. westphal and g. xu , jcap * 1006 * , 009 ( 2010 ) ; r. flauger and e. pajer , jcap * 1101 * , 017 ( 2011 ) . t. kobayashi and f. takahashi , jcap * 1101 * , 026 ( 2011 ) . m. aich , d. k. hazra , l. sriramkumar and t. souradeep , arxiv:1106.2798v2 [ astro-ph.co ] m. bridges , f. feroz , m. p. hobson and a. n. lasenby , mon . not . soc . * 400 * , 1075 ( 2009 ) ; h. v. peiris and l. verde , phys . d * 81 * , 021302 ( 2010 ) ; z - k . guo , d. j. schwarz and y .- z . zhang , jcap * 1108 * , 031 ( 2011 ) . x. chen , r. easther and e. a. lim , jcap * 0706 * , 023 ( 2007 ) ; jcap * 0804 * , 010 ( 2008 ) . s. hotchkiss and s. sarkar , jcap * 1005 * , 024 ( 2010 ) ; s. hannestad , t. haugbolle , p. r. jarnhus and m. s. sloth , jcap * 1006 * , 001 ( 2010 ) ; r. flauger and e. pajer , jcap * 1101 * , 017 ( 2011 ) ; p. adshead , w. hu , c. dvorkin and h. v. peiris , phys . d * 84 * , 043519 ( 2011 ) ; p. adshead , c. dvorkin , w. hu and e. a. lim , phys . rev . d * 85 * , 023531 ( 2012 ) . a. gangui , j. martin and m. sakellariadou , phys . d * 66 * , 083502 ( 2002 ) ; r. holman and a. j. tolley , jcap * 0805 * , 001 ( 2008 ) ; w. xue and b. chen , phys . d * 79 * , 043518 ( 2009 ) ; p. d. meerburg , j. p. van der schaar and p. s. corasaniti , jcap * 0905 * , 018 ( 2009 ) ; x. chen , jcap * 1012 * , 003 ( 2010 ) ; a. ashoorioon and g. shiu , jcap * 1103 * , 025 ( 2011 ) ; s. kundu , jcap * 1202 * , 005 ( 2012 ) ; p. d. meerburg , r. wijers and j. p. van der schaar , mon . not . soc . * 421 * , 369 ( 2012 ) . a. a. starobinsky , sov . jetp lett . * 55 * , 489 ( 1992 ) . j. martin and l. sriramkumar , jcap * 1201 * , 008 ( 2012 ) . f. arroja , a. e. romano and m. sasaki , phys . d * 84 * , 123503 ( 2011 ) ; f. arroja and m. sasaki , jcap * 1208 * , 012 ( 2012 ) . d. k. hazra , j. martin and l. sriramkumar , phys . d * 86 * , 063523 ( 2012 ) . t. bunch and p. c. w. davies , proc . lond . a * 360 * , 117 ( 1978 ) . d. s. salopek , j. r. bond and j. m. bardeen , phys . d * 40 * , 1753 ( 1989 ) ; c. ringeval , lect . notes phys . * 738 * , 243 ( 2008 ) . c. dvorkin and w. hu , phys . d * 81 * , 023518 ( 2010 ) ; w. hu , phys . rev . d * 84 * , 027303 ( 2011 ) . r. w. brankin , i. gladwell and l. f. shampine , _ rksuite : a suite of runge - kutta codes for the initial value problem for odes _ , softreport 92-s1 , department of mathematics , southern methodist university , dallas , texas , u.s.a , 1992 . see , for instance , http://www.netlib.org/ and http://www.nag.com/. j. martin , h. motohashi and t. suyama , arxiv:1211.0083v1 [ astro-ph.co ] . l. f. abbott and m. b. wise , nucl . b * 244 * , 541 ( 1984 ) ; d. h. lyth and e. d. stewart , phys . b * 274 * , 168 ( 1992 ) ; j. martin and d. j. schwarz , phys . d * 57 * , 3302 ( 1998 ) . see , http://www.wolfram.com/. d. k. hazra , arxiv:1210.7170v1 [ astro-ph.co ] . r. allahverdi , j. garcia - bellido , k. enqvist and a. mazumdar phys . lett . * 97 * , 191304 ( 2006 ) ; r. allahverdi , k. enqvist , j. garcia - bellido , a. jokinen and a. mazumdar , jcap * 0706 * , 019 ( 2007 ) ; j. c. b. sanchez , k. dimopoulos and d. h. lyth , jcap * 0701 * , 015 ( 2007 ) . h. m. hodges , g. r. blumenthal , l. a. kofman and j. r. primack , nucl . b * 335 * , 197 ( 1990 ) . r. k. jain , p. chingangbam and l. sriramkumar , jcap * 0710 * , 003 ( 2007 ) . p. creminelli and m. zaldarriaga , jcap 0410 , 006 ( 2004 ) . s. renaux - petel , jcap 1010 , 020 ( 2010 ) . x. chen , h. firouzjahi , m. h. namjoo and m. sasaki , arxiv:1301.5699v1 [ hep - th ] . c. pitrou , j .- uzan and f. bernardeau , jcap * 1007 * , 003 ( 2010 ) .

spectral graph theory examines relationships between the structure of a graph and the eigenvalues ( or spectrum ) of a matrix associated with that graph . different matrices are able to give different information , but all the common matrices have limitations . this is because there are graphs which have the same spectrum for a certain matrix but different structure such graphs are called cospectral with respect to that matrix . the following are some of the matrices studied in spectral graph theory : * the adjacency matrix @xmath1 . this is defined by @xmath2 when @xmath3 and @xmath4 are adjacent and @xmath5 otherwise . the spectrum of the adjacency matrix can determine the number of edges and if a graph is bipartite , but it can not determine if a graph is connected ( see @xcite for more information ) . see figure [ fig:1a ] , the `` saltire pair '' @xcite , for an example of two graphs which are cospectral with respect to the adjacency matrix but not any of the other matrices we will discuss . * the laplacian @xmath6 . this matrix is also known as the combinatorial laplacian and is found by taking the difference of the diagonal degree matrix @xmath7 and the adjacency matrix . the spectrum of the laplacian can determine the number of edges and the number of connected components , but it can not determine if a graph is bipartite ( see @xcite for more information ) . see figure [ fig:1b ] @xcite , for an example of two graphs which are cospectral with respect to the laplacian but not any of the other matrices we will discuss . * the signless laplacian @xmath8 . this matrix is found by taking the sum of the diagonal degree matrix @xmath7 and the adjacency matrix . the spectrum of the signless laplacian can determine the number of edges and the number of connected components which are bipartite , but it can not determine if a graph is bipartite or connected ( see @xcite for more information ) . see figure [ fig:1c ] @xcite , for an example of two graphs which are cospectral with respect to the signless laplacian but not any of the other matrices we will discuss . * the normalized laplacian @xmath9 . this matrix is defined by @xmath10 , where by convention if we have an isolated vertex then it will contribute @xmath5 to the spectrum . the spectrum of the normalized laplacian is closely related to the spectrum of the probability transition matrix of a random walk . this spectrum can determine if a graph is bipartite and the number of connected components , but it can not determine the number of edges ( see @xcite for more information ) . see figure [ fig:1d ] for an example of two graphs which are cospectral with respect to the normalized laplacian but not any of the other matrices we will discuss . one way to understand what structure the spectrum of a matrix can not identify is to study cospectral graphs . cospectral graphs for the adjacency matrix @xcite and the laplacian matrix @xcite have been studied , particularly for graphs with few vertices . cospectral graphs for the signless laplacian have been little studied beyond their enumeration and rules which apply to all of these matrices @xcite . little is also known about cospectral graphs for the normalized laplacian compared to other matrices . previously , the only cospectral graphs were bipartite ( complete bipartite graphs @xcite and bipartite graphs found by `` unfolding '' a small bipartite graph in two ways @xcite ) or were regular and cospectral for @xmath1 ( since a regular graph is cospectral for all of these four matrices if it is cospectral for any one of them ) . in table [ table ] , we have listed the number of graphs with a cospectral mate for the normalized laplacian matrix for graphs on nine or fewer vertices , counted using sage ( see appendix [ sagecode ] for example code ) . the number of cospectral graphs for the normalized laplacian have also been counted by wilson and zhu @xcite , though they give percentages and not the count of how many have cospectral mates . we have also included similar counts for the other three matrices which come from @xcite . @xmath11 given the relatively small number of graphs with a cospectral mate with respect to the normalized laplacian , it is surprising that so little is known about forming cospectral graphs for that matrix . the problem is that some of the main tools that are used to form cospectral graphs for other matrices do not generalize to the normalized laplacian . one such example is a technique known as switching , which is accomplished by replacing edges by non - edges and non - edges by edges between two subsets ( see @xcite ; a simple example is shown in figure [ fig : switch ] ) . given some basic assumptions , this is an easy way to construct cospectral graphs for the adjacency matrix . however , switching does not in general work for the normalized laplacian , in particular , it will only be guaranteed to work when the degrees are unchanged ( see @xcite ) . so the graphs shown in figure [ fig : switch ] are _ not _ cospectral with respect to the normalized laplacian . in section [ sec : swap ] we will introduce a method to construct cospectral graphs for the normalized laplacian . this construction will work similarly to switching in that we will make a small local change to the graph by swapping in one bipartite graph with a cospectral mate and show the two graphs still share the same eigenvalues . in section [ sec : other ] we will show that if we add additional constraints to the bipartite graphs which are swapped then the resulting graphs are also cospectral with respect to @xmath1 , @xmath12 and @xmath13 . in section [ sec : applications ] we will show how to use this construction to produce large families of graphs which are mutually cospectral . finally , in section [ sec : comments ] we will give some concluding remarks . the method of finding cospectral graphs for the adjacency matrix by switching reduces to making a local change of the graph from one bipartite subgraph to its complement and showing that the spectrum is unchanged . we will consider something similar , namely a local change , but instead of replacing a bipartite subgraph with its complement we will swap out a bipartite subgraph for a cospectral mate . for a subset @xmath14 of the vertices @xmath15 of @xmath16 , we will let @xmath17 $ ] be the induced subgraph of @xmath16 on the vertex set @xmath14 . [ thm : biswitch ] let @xmath18 and @xmath19 be bipartite , cospectral graphs with respect to the normalized laplacian on the vertex set @xmath20 such that all edges go between @xmath21 and @xmath22 and where all vertices in @xmath21 have degree @xmath23 for both graphs . let @xmath24 be a graph on the vertices @xmath25 where @xmath26 $ ] is an arbitrary graph ; @xmath27 $ ] and @xmath28 $ ] have no edges ; there are no edges going between @xmath1 and @xmath21 , between @xmath1 and @xmath22 , or between @xmath29 and @xmath22 ; @xmath30 $ ] is a complete bipartite graph ; and @xmath31=p_1 $ ] . the graph @xmath32 is defined similarly except that @xmath33=p_2 $ ] . then @xmath24 and @xmath32 are cospectral with respect to the normalized laplacian . if the dimension of the eigenspace associated with @xmath34 intersected with the subspace of vectors that are nonzero only on @xmath21 is the same for both @xmath18 and @xmath19 , then the graphs @xmath35 and @xmath36 are also cospectral with respect to the normalized laplacian , where @xmath35 only differs from @xmath24 in that @xmath37 $ ] is the complete graph and similarly @xmath36 only differs from @xmath32 in that @xmath38 $ ] is the complete graph . an example of the construction described in theorem [ thm : biswitch ] is shown in figure [ fig : exswitch ] . in this case @xmath18 and @xmath19 are the graphs @xmath39 and @xmath40 where we have put the degree @xmath41 vertices into @xmath21 . it is well known that the spectrum of a complete bipartite graph @xmath42 is @xmath43},1^{[p+q-2]},2^{[1]}$ ] , where the exponent indicates multiplicity , hence these graphs are easily seen to be cospectral . moreover it is easy to check that the dimension of the eigenspace for the eigenvalue @xmath41 intersected with the subspace of vectors which are nonzero only on @xmath21 is @xmath44 for both graphs , i.e. , we could also have put a complete graph on the vertices of @xmath21 in the graphs shown in figure [ fig : exswitch ] and still had a cospectral pair . additional examples of graphs @xmath18 and @xmath19 satisfying ( both ) the conditions on theorem [ thm : biswitch ] are @xmath45 and @xmath46 where the vertices of degree @xmath47 are all placed in @xmath21 and @xmath48 . two further examples are shown in figures [ fig : other1 ] and [ fig : other2 ] . to prove theorem [ thm : biswitch ] we will find it convenient to work with the harmonic eigenvectors of the normalized laplacian . namely , if @xmath49 is an eigenvector associated with the eigenvalue @xmath50 , i.e. , @xmath51 , then the corresponding harmonic eigenvector is @xmath52 . this translates the relationship @xmath51 into @xmath53 , which at a vertex @xmath4 becomes @xmath54 where @xmath55 is the degree of the vertex @xmath4 . we will say that two harmonic eigenvectors @xmath56 and @xmath57 are orthogonal if @xmath58 , i.e. , if the corresponding eigenvectors are orthogonal . before we begin the proof of theorem [ thm : biswitch ] it will be useful to make some observations about bipartite graphs that will come up in the proof . [ lem : help ] let @xmath59 be a bipartite graph on the vertices @xmath20 where all edges go between @xmath21 and @xmath22 and the vertices in @xmath21 all have degree @xmath23 . further , if @xmath60 is an eigenvector for the normalized laplacian then we can write it as @xmath61 where @xmath62 is the vector @xmath60 restricted to the vertices of @xmath21 , and similarly for @xmath63 . then the following hold : * if @xmath64 , i.e. , @xmath60 is an eigenvector for the eigenvalue @xmath41 , then @xmath65 and @xmath66 . in other words , we can divide the eigenspace for the eigenvalue @xmath41 between those which are nonzero only on the vertices of @xmath21 and nonzero only on the vertices of @xmath22 . * if @xmath61 is an eigenvector associated with @xmath50 , then @xmath67 is an eigenvector associated with @xmath68 . * if @xmath69 are a set of orthogonal harmonic eigenvectors of @xmath59 , then for @xmath70 we have @xmath71 is orthogonal to @xmath72 and @xmath73 is orthogonal to @xmath74 . in particular , for @xmath70 the sum of the entries of @xmath71 is @xmath5 . parts ( a ) and ( b ) easily follow from . for ( c ) we note that @xmath75 will be orthogonal to the vectors @xmath76 and @xmath77 , also since the vertices in @xmath21 are regular the result on the sum of the entries of @xmath71 follows by the definition of orthogonality . let @xmath69 be a full set of orthogonal harmonic eigenvectors of @xmath18 associated with @xmath78 respectively . let @xmath79 denote the degree of a vertex in @xmath21 for the graph @xmath24 and @xmath23 the degree in @xmath18 . then for @xmath70 we have that @xmath80 is a harmonic eigenvector for @xmath24 associated with eigenvalue @xmath81 . to see this we need to examine what happens for each vertex @xmath4 in @xmath24 using . * for @xmath82 : then all of the entries of the vertex and its neighbors are all @xmath5 and so trivially holds . * for @xmath83 : all of the neighbors in @xmath1 and @xmath29 are @xmath5 while the sum of the entries in @xmath21 will be @xmath5 by lemma [ lem : help](c ) , so both sides again are @xmath5 and holds . * for @xmath84 : we note that all of the nonzero elements of @xmath85 adjacent to @xmath4 are in @xmath22 , and further we have that for the harmonic eigenvector @xmath75 for @xmath18 that @xmath86 so we have for @xmath24 that @xmath87 so again holds . * for @xmath88 : we proceed similarly as we did for vertices in @xmath21 and we get @xmath89 next we note by lemma [ lem : help](c ) that @xmath72 and @xmath74 are orthogonal to each of the vectors @xmath85 for @xmath70 . on @xmath18 , the dimension of the space orthogonal to the @xmath85 for @xmath70 restricted to @xmath18 is two . a basis for the orthogonal complement of @xmath90 is @xmath91 , so any harmonic vector orthogonal to all of these _ must _ be a linear combination of @xmath72 and @xmath74 . in particular , if we let @xmath92 be a harmonic eigenvector of @xmath24 that is orthogonal to all of the @xmath85 for @xmath70 , then @xmath92 restricted to @xmath21 and @xmath22 is orthogonal to the @xmath85 , so we have @xmath93 for some constants @xmath94 and @xmath95 . in other words , the harmonic eigenvector @xmath92 is constant on the vertices of @xmath21 and constant on the vertices of @xmath22 . everything that we have done for @xmath24 carries over to @xmath32 . in particular , since @xmath18 and @xmath19 are cospectral , then the @xmath96 found by generalizing the harmonic eigenvectors of @xmath18 and @xmath19 remain the same . furthermore , every other harmonic eigenvector orthogonal to the @xmath85 must be constant on the vertices of @xmath21 and the vertices of @xmath22 . to finish off the first claim of the theorem , we now only need to observe that _ any _ harmonic eigenvector which is orthogonal to all of the @xmath85 in @xmath24 is also a harmonic eigenvector for @xmath32 for the same eigenvalue . again to see this we only need to consider what happens for each vertex . * for @xmath97 : then all of the entries of the vertex and its neighbors are the same for both graphs , so trivially holds . * for @xmath84 : all of the neighbors in @xmath29 are the same , and we have the exact same number of neighbors in @xmath22 as before with the same value and so again holds . * for @xmath88 : for the graph @xmath24 , let @xmath98 be the fixed value of the vertices in @xmath21 . we have @xmath99 of course the @xmath55 terms cancel and we are left with @xmath100 . so even though the degree of the vertex in @xmath22 might change , it will have no effect on this relationship and so again holds . in summary , we were able to find @xmath101 harmonic eigenvectors for @xmath24 and @xmath32 that gave the same set of eigenvalues . for any other harmonic eigenvector orthogonal to these , the same harmonic eigenvector worked for both graphs and so the remaining set of eigenvalues also agreed . so we can conclude that @xmath24 and @xmath32 are cospectral with respect to the normalized laplacian . we now turn to the second statement of the theorem . let @xmath79 be the degree of a vertex in @xmath21 in the graph @xmath35 . so let @xmath69 be a full set of orthogonal harmonic eigenvectors of @xmath18 associated with @xmath78 respectively . first , we consider the harmonic eigenvectors of the graph @xmath18 in the eigenspace corresponding to @xmath34 . by lemma [ lem : help](a ) we can assume that the vectors @xmath75 are either of the form @xmath71 or @xmath73 . let the corresponding harmonic eigenvector of @xmath35 be @xmath102 , i.e. , we simply expand @xmath75 to be zero outside of @xmath18 . then we have two cases . * if @xmath103 : in this case , we know there are no problems for the vertices in @xmath1 ( every term in is @xmath5 ) or @xmath22 ( since reduces to what was done in @xmath18 ) . for the vertices in @xmath29 we can use lemma [ lem : help](c ) to see that both sides of are @xmath5 . finally , suppose that @xmath4 is a vertex in @xmath21 . then we have @xmath104 in particular , we have that @xmath105 is an eigenvalue , i.e. , that @xmath106 is an eigenvalue , for this harmonic eigenvector . * if @xmath107 : in this case we similarly know there are no problems for vertices in @xmath1 , @xmath29 and @xmath21 . for vertices in @xmath22 , reduces to what we had in @xmath18 and so we can conclude that this is a harmonic eigenvector for the eigenvalue @xmath108 . now suppose that @xmath109 is associated with @xmath110 for @xmath18 . then we now will create two harmonic eigenvectors for @xmath35 , namely @xmath111 these will be associated with the eigenvalues of @xmath112 respectively , for @xmath35 . some simple computations show that the following relationships hold : @xmath113 ( before we proceed to the next step of showing that these are indeed harmonic eigenvectors and eigenvalues we first make the observation that we are not creating more harmonic eigenvectors than we had before . this is because in lemma [ lem : help](b ) we can pair up harmonic eigenvectors and eigenvalues . in particular , we would have generated the same new harmonic eigenvectors and eigenvalues if we had used @xmath114 and @xmath115 . so really we are taking pairs of eigenvectors and eigenvalues to new pairs of eigenvectors and eigenvalues . ) to verify that these are harmonic eigenvectors with the specified eigenvalues we need to examine what happens for each vertex @xmath4 in @xmath35 using . we will step through @xmath116 and @xmath117 , the arguments for @xmath118 and @xmath119 are the same . * for @xmath82 : then all of the entries of the vertex and its neighbors are all @xmath5 and so trivially holds . * for @xmath83 : all of the neighbors in @xmath1 and @xmath29 are @xmath5 while the sum of the entries in @xmath21 will be @xmath5 by lemma [ lem : help](c ) , so both sides again are @xmath5 and holds . * for @xmath84 : we note that all of the nonzero elements of @xmath85 adjacent to @xmath4 are in @xmath21 and @xmath22 , and further we have that for the harmonic eigenvector @xmath75 for @xmath18 that @xmath120 so we have for @xmath35 that @xmath121 so again holds . * for @xmath88 : we proceed similarly as we did for vertices in @xmath21 and we get @xmath122 the remainder of this case now proceeds as before . namely , everything that we did for @xmath35 carries over for @xmath36 ( counting multiplicity of eigenvalues ) . further , any other harmonic eigenvector orthogonal to the ones given must be constant on @xmath21 and @xmath22 and so any other harmonic eigenvector which works for @xmath35 also works for @xmath36 . therefore we can conclude that they have the same set of eigenvalues , i.e. , are cospectral with respect to the normalized laplacian . the graphs generated by theorem [ thm : biswitch ] will generally not give cospectral graphs for other matrices . so it is instructive to examine the proof and try to understand the point at which the fact that we were using the normalized laplacian as compared to some other matrix came into play . the key is understanding how remains true even when the degrees of vertices in @xmath22 change if the degree of a vertex in @xmath22 changes , the @xmath55 term on the right side of will change proportionally to the sum on the left so that the equality still holds . in the case when the degrees of vertices do not change , for example , when @xmath18 and @xmath19 are both @xmath123-biregular , then the proof generalizes . recall that a bipartite graph is @xmath123-biregular if the vertices can be partitioned into @xmath20 where all edges go between @xmath21 and @xmath22 , the vertices in @xmath21 all have degree @xmath23 , and the vertices in @xmath22 have degree @xmath47 . we now have the following theorem . [ thm : biregswitch ] let @xmath18 and @xmath19 be bipartite , cospectral with respect to the adjacency matrix , @xmath123-biregular graphs on the vertex set @xmath20 such that all vertices in @xmath21 have degree @xmath23 and all vertices in @xmath22 have degree @xmath47 and edges go between @xmath21 and @xmath22 . further , the dimension of the eigenspace associated with @xmath124 for the adjacency matrix intersected with the subspace of vectors which are nonzero only on @xmath21 is the same for both @xmath18 and @xmath19 . let @xmath24 be a graph on the vertices @xmath25 where @xmath26 $ ] is an arbitrary graph ; @xmath27 $ ] and @xmath28 $ ] have no edges ; there are no edges going between @xmath1 and @xmath21 , between @xmath1 and @xmath22 , or between @xmath29 and @xmath22 ; @xmath30 $ ] is a complete bipartite graph ; and @xmath31=p_1 $ ] . the graph @xmath32 is defined similarly except that @xmath33=p_2 $ ] . then @xmath24 and @xmath32 are cospectral with respect to the matrix @xmath0 for @xmath125 arbitrary . similarly , the graphs @xmath35 and @xmath36 are also cospectral with respect to @xmath0 for @xmath125 arbitrary , where @xmath35 only differs from @xmath24 in that @xmath37 $ ] is the complete graph and similarly @xmath36 only differs from @xmath32 in that @xmath38 $ ] is the complete graph . note when @xmath126 then @xmath127 , when @xmath128 then @xmath129 , and when @xmath130 then @xmath131 , so that any such pairs of graphs are cospectral with respect to all of the matrices @xmath1 , @xmath12 , and @xmath13 . further , these are also cospectral for @xmath132 since if @xmath133 , we have @xmath134 since the two graphs have the same degree sequence , the product @xmath135 is the same , so the characteristic polynomials also are the same . in order to use this theorem , we must find two @xmath123-biregular bipartite graphs which are cospectral and for which the dimensions of the eigenspace of @xmath5 restricted to @xmath21 agree on both graphs . there is one special case for which this is much easier , namely the theorem does not prohibit the possibility that @xmath18 and @xmath19 are the same graph ( so trivially are cospectral so we only are reduced to checking the dimension of the eigenspaces ) . this is not so trivial as it might sound since we are distinguishing the two parts of the bipartite graph in the proof of the theorem , so that while @xmath18 and @xmath19 are the same graph we are attaching them in two different ways and so the resulting graphs might not be isomorphic . as an example of this consider the graphs shown in figure [ fig:2bi ] . these are both the same graph ( just flipped ) and a simple check shows that the dimension of the eigenspace associated with @xmath5 and restricted to @xmath21 is @xmath136 in both graphs . therefore theorem [ thm : biregswitch ] applies , and any two graphs to which these two graphs are attached are cospectral . but note that in @xmath19 there are two vertices in @xmath22 with the same neighbors ( marked ) , but there are no such pairs in @xmath18 . so the resulting graphs are _ not _ isomorphic as long as @xmath137 . so for example if we consider the graphs @xmath35 and @xmath36 where @xmath138 and we induce a complete graph on the vertices of @xmath21 , then theorem [ thm : biregswitch ] shows that the resulting graphs are cospectral , and again since there are two vertices with degree @xmath139 sharing common neighbors in one graph but not the other they are non - isomorphic . these graphs are shown in figure [ fig : bicom ] . these graphs also happen to be complements of one another , and so give an example of a graph which is cospectral with its complement but not self - complementary . ( note the graph shown in figure [ fig:1c ] is another example of a graph which is cospectral with its complement , but in that case it is only cospectral with respect to @xmath13 and not with respect to @xmath1 , @xmath12 or @xmath9 . ) the proof for theorem [ thm : biregswitch ] will proceed similar to the proof for theorem [ thm : biswitch ] , so we will skip some of the routine computations and provide an outline of the proof . note that if @xmath60 is an eigenvalue associated with eigenvalue @xmath50 , i.e. , @xmath140 , then at a vertex @xmath4 we have @xmath141 we first start with some simple properties of bipartite graphs that will be helpful in the proof . the proofs are similar to lemma [ lem : help ] and we will omit them . [ lem : bihelp ] let @xmath59 be a bipartite @xmath123-biregular graph on the vertices @xmath20 where all edges go between @xmath21 and @xmath22 . further , if @xmath60 is an eigenvector for the adjacency matrix , then we can write it as @xmath61 where @xmath62 is the vector @xmath60 restricted to the vertices of @xmath21 , and similarly for @xmath63 . then the following hold : * if @xmath142 , i.e. , @xmath60 is an eigenvector for the eigenvalue @xmath5 , then @xmath143 and @xmath144 . in other words , we can divide the eigenspace for the eigenvalue @xmath5 between those which are nonzero only on the vertices of @xmath21 and nonzero only on the vertices of @xmath22 . * if @xmath61 is an eigenvector associated with @xmath50 , then @xmath67 is an eigenvector associated with @xmath145 . * if @xmath146 are a set of orthogonal eigenvectors of @xmath59 , then for @xmath70 we have @xmath71 is orthogonal to @xmath72 and @xmath73 is orthogonal to @xmath74 . in particular , for @xmath70 the sum of the entries of @xmath71 is @xmath5 . we will first consider the case when the vertices on @xmath21 induce an empty graph . let @xmath146 be a full set of orthogonal eigenvectors of @xmath18 for the adjacency matrix associated with the eigenvalues @xmath147 respectively . let @xmath79 denote the degree of a vertex in @xmath21 in @xmath24 , and note by construction that @xmath47 will be the degree of a vertex in @xmath22 in @xmath24 . we now show how to extend each of these eigenvectors to an eigenvector for @xmath24 . we have the following : * for an eigenvector of the form @xmath148 ( i.e. , only nonzero on the vertices of @xmath21 ) and @xmath124 then @xmath149 is an eigenvector for @xmath0 on @xmath24 associated with the eigenvalue of @xmath150 . * for an eigenvector of the form @xmath151 ( i.e. , only nonzero on the vertices of @xmath22 ) and @xmath124 then @xmath152 is an eigenvector for @xmath0 on @xmath24 associated with the eigenvalue of @xmath153 . * for an eigenvector @xmath61 associated with an eigenvalue @xmath154 , we can construct two new eigenvectors , namely @xmath155 these are associated with the eigenvalues of @xmath0 for @xmath24 @xmath156 respectively . + ( note , as in the proof of theorem [ thm : biswitch ] , that we are not creating too many new eigenvectors and eigenvalues , i.e. , by lemma [ lem : bihelp](b ) we can pair up the eigenvalues and eigenvectors . in particular , @xmath61 and @xmath50 will create the same new eigenvectors and eigenvalues as @xmath157 and @xmath145 . ) the verification that each of these is an eigenvector / eigenvalue pair for @xmath24 reduces to verifying for each vertex , similarly as was done in theorem [ thm : biswitch ] . we will skip these routine computations . next we note by lemma [ lem : bihelp](c ) that @xmath72 and @xmath74 are orthogonal to each of the new eigenvectors for @xmath70 . since the dimension of the space orthogonal to the new eigenvectors restricted to @xmath18 is two , any vector orthogonal to all of these new vectors _ must _ be a linear combination of @xmath72 and @xmath74 when restricted to @xmath18 . in particular , if we let @xmath92 be an eigenvector of @xmath24 that is orthogonal to all of the new eigenvectors , then @xmath92 restricted to @xmath21 and @xmath22 must be of the form @xmath93 for some constants @xmath94 and @xmath95 . in other words , the eigenvector @xmath92 is constant on the vertices of @xmath21 and constant on the vertices of @xmath22 . everything that we have done for @xmath24 carries over to @xmath32 . in particular , since @xmath18 and @xmath19 are cospectral , then the newly found @xmath96 are the same for both graphs , and further , every other eigenvector orthogonal to the ones found must be constant on the vertices of @xmath21 and the vertices of @xmath22 . to finish off the first case of the theorem , we now only need to observe that _ any _ eigenvector which is orthogonal to all of the new eigenvectors in @xmath24 is also an eigenvector for @xmath32 for the same eigenvalue . this again is done in the exact same way as in theorem [ thm : biswitch ] and we will skip the computations here . in summary , we were able to find @xmath101 eigenvectors of @xmath0 for @xmath24 and @xmath32 that gave the same set of eigenvalues . for any other eigenvector orthogonal to these , the same eigenvector worked for both graphs and so the remaining set of eigenvalues also agreed . so we can conclude that @xmath24 and @xmath32 are cospectral with respect to the matrix @xmath0 . now we turn to the case when the vertices on @xmath21 induce a complete graph . using the same notation as in the previous case , we will extend the eigenvectors of @xmath18 for the adjacency matrix to eigenvectors of @xmath35 for @xmath0 . we have the following : * for an eigenvector of the form @xmath148 ( i.e. , only nonzero on the vertices of @xmath21 ) and @xmath124 then @xmath149 is an eigenvector for @xmath0 on @xmath35 associated with the eigenvalue of @xmath158 . * for an eigenvector of the form @xmath151 ( i.e. , only nonzero on the vertices of @xmath22 ) and @xmath124 then @xmath152 is an eigenvector for @xmath0 on @xmath35 associated with the eigenvalue of @xmath153 . * for an eigenvector @xmath61 associated with an eigenvalue @xmath154 , we can construct two new eigenvectors , namely @xmath159 these are associated with the eigenvalues of @xmath0 for @xmath35 @xmath160 respectively . + ( note , as in the proof of theorem [ thm : biswitch ] , that we are not creating too many new eigenvectors and eigenvalues , i.e. , by lemma [ lem : bihelp](b ) we can pair up the eigenvalues and eigenvectors . in particular , @xmath61 and @xmath50 will create the same new eigenvectors and eigenvalues as @xmath157 and @xmath145 . ) the verification that each of these is an eigenvector / eigenvalue pair for @xmath35 reduces to verifying for each vertex , similarly as was done in theorem [ thm : biswitch ] . we will skip these routine computations . the remainder of this case now proceeds as before . namely , everything that we did for @xmath35 carries over for @xmath36 ( counting multiplicity of eigenvalues ) . further , any other eigenvector orthogonal to the ones given must be constant on @xmath21 and @xmath22 , so any other eigenvector for @xmath0 which works for @xmath35 also works for @xmath36 . therefore we can conclude that they have the same set of eigenvalues , i.e. , are cospectral with respect to @xmath0 . we can use theorem [ thm : biswitch ] to construct large families of mutually cospectral graphs for the normalized laplacian that need not be bipartite nor have the same degree sequences . let @xmath161 , @xmath23 be positive integers such that @xmath162 is a partition of @xmath161 into @xmath23 positive integer parts . let @xmath163 be the graph on @xmath164 vertices @xmath165 , where the @xmath166 induce a complete graph , each @xmath167 is only adjacent to exactly @xmath168 vertices from @xmath169 , and each @xmath166 is adjacent to exactly one of the @xmath167 . the collection of all @xmath163 for partitions @xmath170 give a family of cospectral graphs for the normalized laplacian , i.e. , the spectrum is completely determined for the normalized laplacian by @xmath161 and @xmath23 . we have dubbed these graphs `` fuzzy balls '' since they consist of a complete graph ( the @xmath166 ) with some extra vertices attached ( the fuzz ) . in figure [ fig : fuzzy ] , we have shown the cospectral family that corresponds to @xmath171 and @xmath172 . and @xmath172 . ] to see that these are all cospectral , we observe by theorem [ thm : biswitch ] that @xmath163 is cospectral with @xmath173 by using the pair of cospectral graphs @xmath174 and @xmath175 where @xmath21 induces a complete graph . therefore , applying this idea @xmath176 times , we can conclude that for each partition of @xmath161 into @xmath23 parts we have @xmath163 is cospectral with @xmath177 . in particular , they are all mutually cospectral . let @xmath161 , @xmath23 be positive integers such that @xmath162 is a partition of @xmath161 into @xmath23 positive integer parts . let @xmath178 be the graph on @xmath179 vertices @xmath180 , @xmath181 , @xmath182 , @xmath183 , @xmath184 , @xmath182 , @xmath185 , where the @xmath166 and @xmath180 induce a star graph with @xmath180 as the central vertex , each @xmath167 is only adjacent to exactly @xmath168 vertices from @xmath169 , and each @xmath166 is adjacent to exactly one of the @xmath167 . the collection of all @xmath178 for partitions @xmath170 give a family of cospectral graphs , i.e. , the spectrum is completely determined by @xmath161 and @xmath23 . we have dubbed these graphs `` inflated stars '' . in figure [ fig : stars ] , we have shown the cospectral family that corresponds to @xmath171 and @xmath172 . and @xmath172 . ] again , to see that these are all cospectral , we observe by theorem [ thm : biswitch ] that the graph @xmath178 is cospectral with @xmath186 . therefore , applying this idea @xmath176 times , we can conclude that for each partition of @xmath161 into @xmath23 parts , we have @xmath178 is cospectral with @xmath187 . in particular , they are all mutually cospectral . in both of these examples , we are using partitions to form large families . the number of partitions grows subexponentially with @xmath161 , however it is possible to use theorems [ thm : biswitch ] or [ thm : biregswitch ] to construct large mutually cospectral families that grow exponentially with @xmath161 , either with respect to the normalized laplacian , or more generally with respect to @xmath0 . [ thm : nlfam ] for @xmath161 large , there exists a family of @xmath188 non - isomorphic , mutually cospectral graphs with respect to the normalized laplacian . find a graph @xmath16 on @xmath189 vertices that has a trivial automorphism group ( for @xmath190 this is easy to do ) . for each vertex @xmath4 of @xmath16 , attach one of the two inflated stars shown in figure [ fig : nlfam ] by identifying @xmath4 with the vertex marked @xmath4 in the inflated star . this constructs @xmath191 graphs on @xmath192 vertices . in theorem [ thm : nlfam ] . ] we note that for any one of these graphs we can easily recover @xmath16 and identify which attachment was done at @xmath4 and so these graphs are all non - isomorphic . further , by theorem [ thm : biswitch ] changing our choice of widget at any single vertex does not change the spectrum and so all of the graphs are cospectral . we have not tried to optimize the construction . for example , if instead of using the family of inflated stars with @xmath193 and @xmath194 we used the family of inflated stars with @xmath195 and @xmath196 ( of which there are @xmath197 such non - isomorphic members ) , then we get a family with rate of growth @xmath198 which is faster than @xmath199 . we similarly have the following result for graphs cospectral with respect to @xmath0 . [ thm : atdfam ] for @xmath161 large , there exists a family of @xmath200 non - isomorphic , mutually cospectral graphs with respect to the matrix @xmath0 . find a graph @xmath16 on @xmath201 vertices that has a trivial automorphism group ( for @xmath202 this is easy to do ) . for each vertex @xmath4 of @xmath16 attach one of the two graphs shown in figure [ fig : atdfam ] , which are taken from figure [ fig:2bi ] . this constructs @xmath203 graphs on @xmath204 vertices . in theorem [ thm : atdfam ] . ] we note that for any one of these graphs we can easily recover @xmath16 and identify which attachment was done at @xmath4 , so these graphs are all non - isomorphic . further , by theorem [ thm : biregswitch ] changing our choice of widget at any single vertex does not change the spectrum and so all of the graphs are cospectral . the method of swapping bipartite subgraphs was discovered by examining all cospectral pairs of graphs on at most eight vertices ( all cospectral pairs on at most seven vertices has previously appeared in tan @xcite ) , which included several simple examples of this type . the set of all cospectral graphs on at most eight vertices can be easily constructed in sage ( see appendix [ sagecode ] ) . by looking at the small cases , we also discovered some new graphs which are regular , but are cospectral for the normalized laplacian with graphs which are not regular . we have already seen that the four - cycle and the eight - cycle are members of cospectral pairs ( see figures [ fig:1d ] and [ fig : other1 ] ) . also , @xmath205 is cospectral with @xmath42 for @xmath206 . another example of a graph which is regular , but is cospectral with a graph which is not regular , is shown in figure [ fig : regular ] . these examples show that the normalized laplacian can not , in general , detect whether a graph is regular . as noted in the introduction , the number of edges in a graph is not uniquely determined by the spectrum of the normalized laplacian . as a consequence , it is possible for a graph to be cospectral with one of its subgraphs ; examples of this are shown in figure [ fig : sub ] . further examples of graphs which are cospectral with respect to the normalized laplacian but have differing number of edges are shown in figure [ fig : dif ] ( there is currently no known method to generate cospectral graphs with differing number of edges other than using complete bipartite graphs ) . this also indicates the difficulty in counting the number of cospectral graphs . while for the adjacency and the combinatorial laplacian , we could first subdivide the graphs according to the number of edges and work with this coarsening when finding cospectral pairs ( see @xcite ) , this is no longer possible for the normalized laplacian . it would be interesting to find techniques for constructing cospectral graphs for the normalized laplacian which differ in the number of edges . similarly , it would be interesting to investigate the cospectral graphs for the signless laplacian and see if there are any constructions unique to forming cospectral graphs for that matrix ( similar to what theorem [ thm : biswitch ] is for the normalized laplacian ) . 99 andries e. brouwer and edward spence , _ cospectral graphs on 12 vertices _ , electronic journal of combinatorics * 16 * ( 2009 ) , # n20 , 3pp . steve butler , _ cospectral graphs for both the adjacency and normalized laplacian matrices _ , linear and multilinear algebra * 58 * ( 2010 ) , 387390 . michael cavers , the normalized laplacian matrix and general randic index of graphs , ph.d . thesis , university of regina , 2010 . fan chung , _ spectral graph theory _ , cbms lecture notes , ams , providence , ri , 1997 . drago m. cvetkovi , _ graphs and their spectra _ , univ . * 354 - 356 * ( 1971 ) , 150 . drago m. cvetkovi , michael doob and horst sachs , _ spectra of graphs : theory and application _ , academic press , new york , ny , 1979 . drago m. cvetkovi and slobodan k. simi , _ towards a spectral theory of graphs based on the signless laplacian i _ , institut mathmatique . publications . nouvelle srie * 85(99 ) * ( 2009 ) , 19 - 33 . drago m. cvetkovi and slobodan k. simi , _ towards a spectral thory of graphs based on the signless laplacian ii _ , linear algebra and its applications * 432 * ( 2010 ) , 2257 - 2272 . drago m. cvetkovi and slobodan k. simi , _ towards a spectral thory of graphs based on the signless laplacian iii _ , applicable analysis and discrete mathematics * 4 * ( 2010 ) , 156 - 166 . edwin r. van dam , _ graphs with few eigenvalues : an interplay between combinatorics and algebra _ , ph.d . thesis , tilburg university , center for economic research dissertations series , no . 20 , 1996 . edwin r. van dam and willem h. haemers , _ which graphs are determined by their spectrum ? _ , linear algebra and its applications * 373 * ( 2003 ) , 241272 . edwin r. van dam , willem h. haemers and jack h. koolen , _ cospectral graphs and the generalized adjacency matrix _ , linear algebra and its applications * 423 * ( 2007 ) , 3341 . chris d. godsil and brendan mckay , _ products of graphs and their spectra _ , combinatorial mathematics , iv ( proc . fourth australian conf . adelaide , adelaide , 1975 ) , lecture notes in math . 560 , springer , berlin , 1976 , 6172 . chris d. godsil and brendan mckay , _ constructing cospectral graphs _ , aequationes mathematicae * 25 * ( 1982 ) , 257268 . willem h. haemers and edward spence , _ enumeration of cospectral graphs _ , european journal of combinatorics * 25 * ( 2004 ) , 199211 . charles r. johnson and morris newman , _ a note on cospectral graphs _ , journal of combinatorial theory , ser . b * 28 * ( 1980 ) , 96103 . russell merriss , _ large families of laplacian isospectral graphs _ , linear and multilinear algebra * 43 * ( 1997 ) , 201205 . johan jacob seidel , _ graphs and two - graphs _ , proceedings of the fifth southeastern conference on combinatorics , graph theory and computing ( florida atlantic univ . , boca raton , fla . , 1974 ) , congressus numerantium , no . x , utilitas math . , winnipeg , man . , 1974 , 125143 . william a. stein , et al . , _ sage mathematics software ( version 4.6.1 ) _ , the sage development team , 2011 , http://www.sagemath.org . jinsong tan , _ on isospectral graphs _ , interdisciplinary information sciences * 4 * ( 1998 ) , 117124 . richard wilson and ping zhu , _ a study of graph spectra for comparing graphs and trees _ , journal of pattern recognition , * 41 * ( 2008 ) , 28332841 . the code listing below shows how to use sage version 4.5.2 @xcite to generate and check cospectral graphs with 8 vertices . in the listing below , we specifically generate cospectral graphs with respect to the normalized laplacian . the lines starting with the number sign `` # '' are comment lines .

the term _ cardinal interpolation _ refers to interpolation of data given at the integer lattice ( or multi - integer lattice in higher dimensions ) . it was i. j. schoenberg s work on cardinal spline interpolation that brought about an intense study of the subject . many avenues of study have been explored , including forming interpolation operators from translates of certain radial basis functions ( rbfs ) . works by buhmann , baxter , riemenschneider , and sivakumar @xcite ( see also @xcite and references therein ) have explored many cardinal interpolation operators of this type . some of the radial basis functions that have been considered are the gaussian kernel , the thin plate spline , the hardy multiquadric , and the inverse multiquadric . radial basis cardinal interpolation also enjoys a strong connection with classical sampling theory , as evidenced by much of the aforementioned literature . this connection continues to be explored in recent developments by the second author @xcite , and by parts of this article . as this is one of the most interesting aspects of cardinal interpolation , we provide some of the motivation . recall the one - dimensional whittaker kotelnikov shannon sampling theorem , which states that a bandlimited function , @xmath10 , ( say with band size @xmath11 ) can be recovered uniformly by the series @xmath12 , where the @xmath13 function is suitably defined so that it takes value 1 at the origin , and 0 at all the other integers . one observation about this series is that the @xmath13 function decays slowly ( as @xmath14 ) , and so to approximate the series by truncation for example , one would have to use quite a lot of data points of @xmath10 to get a reasonable degree of accuracy . however , there is a way of approximating the @xmath13 series above : we seek to replace the @xmath13 function with a so - called _ fundamental function _ ( or _ cardinal function _ ) , @xmath15 , that preserves the property that @xmath15 takes value 1 at the origin and 0 at all other integers . we then form a function @xmath16 the trade - off here is that while @xmath17 is not pointwise equal to the function @xmath10 , it does interpolate @xmath10 at the integer lattice , and moreover , the fundamental function @xmath15 may be constructed so that it decays much more rapidly than the @xmath13 function . precisely , one may construct a fundamental function from a given radial basis function , @xmath18 , which has the form @xmath19 in the case of the gaussian kernel , @xmath20 , the fundamental function decays exponentially , whereas the fundamental function for the hardy multiquadric , @xmath21 , decays as @xmath22 . so we give up the pointwise equality of the wks sampling theorem in exchange for a series that converges more rapidly , while also ensuring that @xmath17 is close to @xmath10 in the @xmath23 norm . this paper primarily considers the fundamental functions and cardinal interpolation operators associated with general multiquadrics , @xmath0 , which have thus far only been considered for certain exponents @xmath24 . interpolation with fundamental functions has too long a history to recount here ; however , @xcite offers a good introduction using radial basis functions . using below as a starting point is especially popular since it allows one to solve problems in the fourier transform domain . many authors have used similar techniques for various radial basis functions . in @xcite , riemenschneider and sivakumar proved several results pertaining to the gaussian . multiquadric cardinal interpolation has been studied in a similar way by buhmann and micchelli @xcite , baxter @xcite , baxter and sivakumar @xcite , riemenschneider and sivakumar @xcite , among others . compactly supported radial basis functions have been studied by buhmann @xcite and wendland @xcite . the rest of the paper is laid out as follows . section [ secbasic ] provides the necessary preliminaries and a discussion of applications and calculations of the fundamental functions ; section [ secbaxteranalogue ] shows recovery results for cardinal interpolation of bandlimited functions in any dimension via interpolants of the form discussed above . section [ secpropfundamental ] contains decay rates and information about the univariate fundamental functions associated with the general multiquadrics for a broad range of exponents . in section [ seceastjournal ] , we consider the cardinal interpolation operators acting on data in traditional sequence spaces and calculate decay rates , bounds on the operator norms , and also explore some convergence properties in terms of the parameter @xmath25 . section [ secexamples ] provides some interesting concrete examples based on the theoretical results from the previous section . finally , section [ secproofs ] provides the technical proofs of the statements in section [ secpropfundamental ] . if @xmath26 is an interval , then let @xmath27 , @xmath28 , be the usual lebesgue space over @xmath29 with its usual norm . if no set is specified , we mean @xmath30 . similarly , denote by @xmath31 the usual sequence spaces indexed by the set @xmath32 ; if no index set is given , we refer to @xmath33 . we will use @xmath34 to denote the natural numbers including 0 . let @xmath35 be the space of schwartz functions on @xmath36 , that is the collection of infinitely differentiable functions @xmath18 such that for all multi - indices @xmath24 and @xmath37 , @xmath38 the fourier transform of a schwartz function @xmath18 is given by @xmath39 thus the inversion formula is @xmath40 in the event that these formulas do not hold , then the fourier transform should be interpreted in the sense of tempered distributions . recall that if @xmath10 is a tempered distribution , then its fourier transform is the tempered distribution defined by @xmath41 , @xmath42 . let @xmath43 and @xmath44 be fixed ; then define the @xmath45-dimensional _ general multiquadric _ by @xmath46 where @xmath47 denotes the euclidean distance on @xmath36 . if @xmath48 , the generalized fourier transform of @xmath4 is given by the following ( see , for example , ( * ? ? ? * theorem 8.15 ) ) : @xmath49 where @xcite @xmath50 @xmath51 is called the modified bessel function of the second kind . we note that both @xmath4 and its fourier transform are radial . it is also clear from the definition that @xmath52 is symmetric in its order ; that is , @xmath53 for any @xmath54 . if @xmath55 , then the generalized fourier transform of @xmath4 involves a measure and so can not be expressed as a function . now suppose that @xmath48 is fixed . to define the _ fundamental function _ associated with the general multiquadric , we first define the following function @xmath56 we will see that @xmath57 , and so the function @xmath58 is well - defined and continuous . furthermore , we will show that @xmath3 is a _ fundamental function _ , also called a _ cardinal function _ , which means that @xmath59 where @xmath60 is the kronecker delta . additionally , @xmath3 has the form @xmath61 throughout the paper , we will use @xmath62 to denote an absolute constant due to the use of @xmath25 as the shape parameter of the multiquadric . the value of the particular constant may change from line to line , and we use subscripts to denote dependence on certain parameters when needed . interpolation schemes involving fundamental functions as in have been studied for quite some time , and there are many aspects to the theory . for example , such methods enjoy applications to geoscience @xcite and sampling theory @xcite . recently , investigations have considered interpolation via radial kernels on manifolds @xcite . for a galerkin type method for solving pdes using meshless interpolation on the sphere , see @xcite . given the widespread applications of radial basis function approximation , it is of import to the computational community to determine stable ways of evaluating the approximants . consequently , there is a substantial literature dealing with accuracy and stability of different computational methodologies for radial basis function approximation . we do not claim to list all of these methods , but at least a sampling is in order . we note that approximating is typically very difficult , especially if one dilates the lattice . one way around this is the use of indirect computational methods to approximate the rbf interpolant @xcite ( for a discussion specifically related to multiquadrics , see @xcite ) . another technique involves a change of basis method @xcite , while work by fasshauer and mccourt @xcite uses an eigenfunction decomposition to provide stable reconstruction using gaussians . another quite promising method has recently been considered in which so - called _ local lagrange functions _ are used to approximate rather than the global ones @xcite . many of these results revolve around the situation of interpolation at finitely many data sites , which is of a somewhat different nature than we are considering here . for the interested reader , we also mention that these methods of cardinal interpolation have , at their core , a deep connection to the classical spline theory instigated by schoenberg ( see @xcite and references therein ) and continued by many followers . the underlying principle is that many of the results in spline interpolation theory have natural analogues via rbfs , and the problem at hand may determine which method is more useful . here , we provide some brief illustrations of the fundamental functions we have mentioned above . figure [ figsinc1d ] shows the univariate fundamental function associated with the inverse multiquadric and the sinc function for comparison . @xmath63{acom_sinc_imq - eps - converted - to.pdf}\ ] ] @xmath64{multivariatefundamental - eps - converted - to.pdf}\ ] ] the fundamental function was calculated by truncating the series in and using a fast fourier transform ( fft ) method to approximate @xmath3 . the same method may be applied for the multivariate version , as figure [ figsinc ] shows . when an interpolation scheme depends on a parameter , questions of convergence naturally arise . this question has been addressed by several authors . in @xcite , baxter examines the hardy multiquadric , while the gaussian is studied in @xcite by riemenschneider and sivakumar and in @xcite by hangelbroek , narcowich , madych , and ward . in a more general context , ` increasingly flat ' radial basis functions are the focus of driscoll and fornberg in @xcite , while the second author worked with ` regular families ' of cardinal interpolants in @xcite . in this section , we show that the result obtained by baxter @xcite holds not only for the traditional hardy multiquadric ( corresponding to @xmath65 ) but rather for all @xmath48 . we consider interpolation of bandlimited ( or paley - wiener ) functions in any dimension , and show that the cardinal interpolant converges to the function as the shape parameter @xmath25 tends to infinity . general multiquadrics were not considered for quite some time in this setting , but in @xcite convergence results for cardinal interpolation of bandlimited functions were shown for a restricted range of exponents . however , the analysis there was of a more general nature , so here we show that a more specific analysis yields convergence results for the full range @xmath66 we note that the results of this section have very close ties to classical sampling theory , which studies the reconstruction of certain classes of signals from their discrete samples . as mentioned above , these considerations lead to alternative methods for approximating the sampling series given by the wks sampling theorem for bandlimited signals . let @xmath45 be the dimension , and @xmath48 and @xmath44 be fixed . it is evident from that @xmath67 does not change sign . therefore , @xmath68 for all @xmath69 . from , it is also evident that @xmath70 . to show that @xmath71 we begin with the following lemma . [ lembaxterphihatdecay ] let @xmath72 , @xmath44 , and @xmath73 . then @xmath74 defining @xmath75 , equations and yield the following series of estimates : @xmath76 we note that if @xmath77 , then @xmath78 , and so we have a purely exponential upper bound . [ proplchatintegrable ] let @xmath48 and @xmath44 . then @xmath71 . first , choose an @xmath79 large . then since @xmath80 for all @xmath81 , we have that @xmath82^d}| { \widehat{l_{\alpha , c}}}(\xi)|d\xi\leq ( 2m)^d.\ ] ] now we need to estimate @xmath83^d}| { \widehat{l_{\alpha , c}}}(\xi)|d\xi.\ ] ] to do this , we establish a pointwise estimate for @xmath84 . let @xmath85^d$ ] be fixed . since @xmath86 is large , there exists some @xmath87 such that @xmath88 . additionally , there is some constant @xmath89 which depends on @xmath24 and @xmath45 such that if @xmath90 , @xmath91 ( see , for example , ( * ? ? ? * corollary 5.12 ) ) . therefore , choose @xmath86 large enough so that for @xmath85^d$ ] , we have @xmath92 . then if @xmath93 is the constant from lemma [ lembaxterphihatdecay ] , @xmath94 now depending on the sign of @xmath95 , the above expression is minimized by plugging in @xmath96 or @xmath97 for @xmath98 in the appropriate places . consequently , there is a positive constant @xmath99 such that @xmath100 we also find from ( * ? ? ? * lemma 5.13 ) that for every @xmath101 , @xmath102 consequently , by adjusting @xmath86 if need be so that @xmath103 for @xmath85^d$ ] , we find that there is a positive constant @xmath37 such that @xmath104 . we conclude that @xmath105^d}| { \widehat{\phi_{\alpha , c}}}(\xi)|d\xi\\ & \leq\beta d^{-1}|\lambda|e^{4\pi c } { \displaystyle\int } _ { { \mathbb{r}}^d\setminus[-m , m]^d}\|\xi\|^{-\alpha-\frac{d}{2}}e^{-c\|\xi\|}d\xi.\end{aligned}\ ] ] the integral on the right hand side above is convergent , and the constants outside are all finite , so @xmath71 . now we turn our attention to the function @xmath3 . [ proplfundamental ] let @xmath48 and @xmath44 . then the function @xmath106 is continuous , square - integrable , and satisfies the interpolatory condition @xmath107 , for every @xmath108 . proposition [ proplchatintegrable ] implies that @xmath3 is continuous and square - integrable , and indeed that @xmath109 is its fourier transform . to see the interpolatory condition , first define @xmath110^d$ ] . then we have via the substitution @xmath111 that @xmath112 the interchange of sum and integral in the third line is justified by the dominated convergence theorem , for example . it is an important observation that much of the cardinal interpolation theory for bandlimited functions revolves around the fact that the fundamental functions converge to the function @xmath113 , which is equivalent to the statement that the fourier transform of the fundamental function converges almost everywhere to the indicator function of the cube @xmath114^d$ ] . the story is no different here . defining @xmath115 to be the function that takes value 1 whenever @xmath116^d$ ] , and 0 elsewhere , the following holds . [ propconvergencetosinc ] let @xmath48 . then @xmath117 for all @xmath69 such that @xmath118 . first suppose that @xmath119^d$ ] . then there exists some @xmath120 such that @xmath121 . therefore by lemma [ lembaxterphihatdecay ] , @xmath122 consequently , since @xmath67 is of one sign , @xmath123 since the exponent is negative , the limit of the right hand side as @xmath6 is 0 . therefore , for @xmath119^d$ ] , @xmath124 . now suppose that @xmath125 . then for all @xmath126 , @xmath127 . due to , we may write @xmath128 and therefore it suffices to show that @xmath129 by lemma [ lembaxterphihatdecay ] , @xmath130 the series on the right hand side is convergent and dominated by the convergent series where @xmath25 is replaced by 1 , so @xmath131 as desired . convergence of the series stems from the fact that the series is majorized by @xmath132 , which converges . we now consider interpolation of bandlimited functions at the lattice @xmath133 by translates of the function @xmath134 . define the @xmath45-dimensional paley - wiener space by @xmath135^d\}.\ ] ] we begin our analysis with an @xmath23 version of the poisson summation formula : [ lempsf ] if @xmath136 , then @xmath137 where the second series is convergent in @xmath138 . [ lemcauchysequence ] let @xmath136 . for @xmath139 , define @xmath140 where @xmath141 for @xmath108 . then @xmath142 forms a cauchy sequence in @xmath138 . define @xmath143 via @xmath144 thus , @xmath145 . from lemma [ lempsf ] , it is clear that @xmath146 is a cauchy sequence in @xmath147^d$ ] . so @xmath148^d}|q_m(\xi+2\pi k)-q_\ell(\xi+2\pi k)|^2 \\ & & \hfill\times\left ( { \widehat{l_{\alpha , c}}}(\xi+2\pi k)\right)^2 d\xi\\ \\ & = & { \displaystyle\int}_{[-\pi,\pi]^d}|q_m(\xi)-q_\ell(\xi)|^2 { \underset{k\in { \mathbb{z}}^d } { \displaystyle\sum}}\left ( { \widehat{l_{\alpha , c}}}(\xi+2\pi k)\right)^2 d\xi\\ \\ & \leq & { \displaystyle\int}_{[-\pi,\pi]^d}|q_m(\xi)-q_\ell(\xi)|^2d\xi\;. \\ \end{array}\ ] ] the interchange of sum and integral is valid by tonelli s theorem , and the last inequality follows from the fact that @xmath149 we also used the fact that for @xmath108 , @xmath150 . we conclude that @xmath151 is a cauchy sequence in @xmath138 because @xmath152^d}$ ] , and the latter is cauchy . lemmas [ lempsf ] and [ lemcauchysequence ] allow us to define @xmath153 where the series is convergent in @xmath138 . by a periodization argument similar to that in the proof of lemma [ lemcauchysequence ] , one can show that @xmath154 . thus applying the fourier inversion formula we see that @xmath155 [ thmbandlimitedconvergence ] let @xmath48 . if @xmath136 , then @xmath156 and @xmath157 uniformly on @xmath36 . we will first prove uniform convergence . the proof is the same as in @xcite . again let @xmath115 be the characteristic function of the cube . then we see by the inversion formula and the oft - exploited periodization argument , that @xmath158^d}\widehat{f}(\xi ) { \underset{k\in { \mathbb{z}}^d } { \displaystyle\sum}}\left ( { \widehat{l_{\alpha , c}}}(\xi+2\pi k)-i(\xi+2\pi k)\right)\\ & & \hfill\times e^{-i { \left\langlex,\xi+2\pi k\right\rangle}}d\xi.\\ \end{array}\ ] ] therefore , we find that @xmath159^d}|\widehat{f}(\xi)| { \underset{k\in { \mathbb{z}}^d } { \displaystyle\sum}}\left| { \widehat{l_{\alpha , c}}}(\xi+2\pi k)-i(\xi+2\pi k)\right|d\xi\\ \\ & = & \dfrac{1}{(2\pi)^d } { \displaystyle\int}_{[-\pi,\pi]^d}|\widehat{f}(\xi)|\left(1- { \widehat{l_{\alpha , c}}}(\xi)+ { \underset{k\neq0 } { \displaystyle\sum } } { \widehat{l_{\alpha , c}}}(\xi+2\pi k)\right)d\xi.\\ \end{array}\ ] ] but then by definition , @xmath160 therefore , @xmath161^d}|\widehat{f}(\xi)|(1- { \widehat{l_{\alpha , c}}}(\xi))d\xi.\ ] ] as the integrand is non - negative and bounded by @xmath162^d$ ] , and @xmath163 , the dominated convergence theorem implies that @xmath164 the upper bound is independent of @xmath165 , hence the convergence is uniform . we now turn to the proof of @xmath23 convergence . by parseval s identity , it suffices to show that @xmath166 . this breaks up into two estimates . we first show this for the cube @xmath114^d$ ] . recall that since @xmath167 is an orthonormal basis for @xmath147^d$ ] , we may write @xmath168 . moreover , @xmath169^d } = \|f(k)\|_{\ell_2 ( { \mathbb{z}}^d)}.\ ] ] thus using , @xmath170^d}^2 & = { \displaystyle\int}_{[-\pi,\pi]^d}\left| { \underset{k\in { \mathbb{z}}^d } { \displaystyle\sum}}f(k ) ( { \widehat{l_{\alpha , c}}}(\xi)-1)e^{-i { \left\langlek,\xi\right\rangle}}\right|^2d\xi \\ & = { \displaystyle\int}_{[-\pi,\pi]^d}| { \widehat{l_{\alpha , c}}}(\xi)-1|^2\left| { \underset{k\in { \mathbb{z}}^d } { \displaystyle\sum}}f(k)e^{-i { \left\langlek,\xi\right\rangle}}\right|^2d\xi.\end{aligned}\ ] ] the right hand side is bounded by @xmath171 , and so by the dominated convergence theorem and proposition [ propconvergencetosinc ] , @xmath172^d}=0 $ ] . now for the rest of the space , for @xmath173 define @xmath174\times\dots\times[-\pi-2\pi l_d,\pi-2\pi l_d]$ ] . then we see that since @xmath10 is bandlimited , @xmath175^d ) } = \| { \widehat { { \mathscr{i}}_{\alpha , c}f}}\|^2_{l_2 ( { \mathbb{r}}^d\setminus[-\pi,\pi]^d ) } = { \underset{l\neq 0 } { \displaystyle\sum}}\| { \widehat { { \mathscr{i}}_{\alpha , c}f}}\|_{l_2(q_l)}^2.\ ] ] consequently , @xmath176^d}| { \widehat { { \mathscr{i}}_{\alpha , c}f}}(\xi)|^2d\xi & = & { \underset{l\neq 0 } { \displaystyle\sum } } { \displaystyle\int}_{q_l}\left| { \widehat{l_{\alpha , c}}}(\xi ) { \underset{k\in { \mathbb{z}}^d } { \displaystyle\sum}}f(k)e^{-i { \left\langlek,\xi\right\rangle}}\right|^2d\xi\\ \\ & = & { \displaystyle\int}_{[-\pi,\pi]^d } { \underset{l\neq 0 } { \displaystyle\sum}}| { \widehat{l_{\alpha , c}}}(\xi+2\pi l)|^2\left| { \underset{k\in { \mathbb{z}}^d } { \displaystyle\sum}}f(k)e^{-i { \left\langlek,\xi\right\rangle}}\right|^2d\xi,\\ \\ \end{array}\ ] ] by the monotone convergence theorem . recall that @xmath70 , so @xmath177 , and as calculated above , @xmath178 . consequently , the integrand is bounded by @xmath179 therefore , the dominated convergence theorem and proposition [ propconvergencetosinc ] imply that @xmath180^d)}=0 $ ] , and the proof is complete . to illustrate the convergence given by theorem [ thmbandlimitedconvergence ] above , the following figure shows the inverse multiquadric interpolant of the function whose fourier transform is @xmath181 in dimension 1 . [ figinterp ] @xmath182{interp_mq - eps - converted - to.pdf}\ ] ] for the rest of the paper , we turn our attentions to the one - dimensional cardinal interpolation operator associated with the general multiquadric . this section is devoted to the one - dimensional fundamental function @xmath3 , whose fourier transform can be rewritten as @xmath183^{-1}.\ ] ] the proofs of the results in this section are quite technical , so we postpone them until section [ secproofs ] and simply state our conclusions here . to determine decay rates for @xmath3 , we determine how many derivatives @xmath9 has in @xmath184 , which we accomplish by establishing pointwise estimates . we begin by fixing @xmath185 , so that our estimates fall into three ranges : @xmath186 , @xmath187 , and the @xmath96-length blocks @xmath188 $ ] for @xmath189 . due to the differing behavior of @xmath190 for positive and negative values of @xmath24 , we must make corresponding distinctions in our calculations . following the insightful techniques of riemenschneider and sivakumar found in @xcite , we begin by defining some auxiliary functions to aid in the analysis of the fundamental function . we abbreviate as @xmath191 , where @xmath192 and study the properties of @xmath193 . [ prop l1 ] suppose that @xmath194 , @xmath195 , @xmath196 , and @xmath197 . if @xmath198 and @xmath199 , then there exists a constant @xmath200 such that @xmath201 as @xmath202 . this estimate leads to the following bounds on @xmath9 and its derivatives . [ prop l2 ] suppose that @xmath194 , @xmath195 , @xmath196 , and @xmath197 . if @xmath203 , then there exist constants @xmath204 such that 1 . @xmath205 whenever @xmath198 , 2 . @xmath206 whenever @xmath207 $ ] , and 3 . @xmath208 whenever @xmath209 $ ] and @xmath189 , where @xmath210 as @xmath202 . these pointwise estimates yield the following result . [ thm l1 ] suppose that @xmath194 , @xmath196 , and @xmath197 . if @xmath203 , then there exists a constant @xmath211 such that @xmath212 using standard arguments , we have the following estimate for the growth rate of @xmath3 . [ cor l3 ] if @xmath194 and @xmath213 , then @xmath214 as @xmath215 . analogous estimates can be made for the case that @xmath216 . of interest to us are the following results . [ thm l2 ] suppose that @xmath216 , @xmath196 , and @xmath197 . if @xmath217 , then there exists a constant @xmath211 such that @xmath218 [ cor l4 ] if @xmath216 and @xmath213 , then @xmath219 as @xmath215 . it turns out that the poisson kernel , which is the case @xmath220 , is a special case , and exhibits much better decay because @xmath221 is purely an exponential function . [ thm kpoissonftdecay ] suppose that @xmath222 . then for every @xmath223 , there exists a constant @xmath224 such that @xmath225 [ cor kpoissondecay ] suppose that @xmath222 . then @xmath226 as @xmath215 for every @xmath223 . we may refine the above estimates in the case @xmath227 to find a uniform bound on the @xmath228norm of @xmath229 . [ thm l3 ] suppose that @xmath230\cup ( 0,\infty)\right)\setminus\mathbb{n } $ ] . there exists a constant @xmath231 such that for all @xmath213 , @xmath232 . so far , the upper bounds on @xmath3 may depend on the parameters @xmath24 and @xmath25 ; however , the following still holds . [ lem h3 ] if @xmath233\cup(0,\infty))\setminus { \mathbb{n}}$ ] and @xmath222 , then @xmath234 for all @xmath235 . we end the section with a statement on the zeros of @xmath9 . [ thm h3 ] if @xmath233\cup(0,\infty))\setminus { \mathbb{n}}$ ] and @xmath222 , then @xmath236 for every @xmath237 . in this section , we show that the results of riemenschneider and sivakumar @xcite have analogues for general multiquadrics . in section [ secpropfundamental ] , the decay of @xmath109 is discussed , and we use that information to uncover growth conditions on data that are suitable to cardinal interpolation . recall from corollary [ cor l3 ] and theorem [ thm h3 ] that for @xmath238 and @xmath222 , @xmath239 this decay is not the best one can get for individual @xmath24 , in fact for @xmath65 , buhmann @xcite proves a decay rate of @xmath22 . according to further work by buhmann and micchelli @xcite , it appears that the multiquadrics with exponents @xmath240 for @xmath241 are exceptional cases . for these values , @xmath109 can be shown to have more derivatives than what we have shown for general @xmath24 due to some special symmetry involving the bessel functions in the fourier transforms . moreover , in these cases , decay of the fundamental function is given by @xmath242 for negative exponents , the so - called inverse multiquadrics , the fundamental functions have slightly slower decay ( corollary [ cor l4 ] and theorem [ thm h3 ] ) : @xmath243 as a consequence of the decay of the fundamental functions , we have the following . [ proprsprop33 ] for @xmath244 , the function @xmath245 is a well - defined , 1-periodic , bounded , continuous function . periodicity is apparent , so we need only consider @xmath246 $ ] . if @xmath247 , then gives @xmath248 which yields the result since @xmath3 is continuous on @xmath249 and @xmath250 . on the other hand , if @xmath251 , then @xmath252 as the exponent is less than -1 , the series converges , whence the result . we next define the cardinal interpolation operator acting on sequences , and show that it is well - defined for sequences that grow at a sufficiently slower rate than the decay of the fundamental function . [ thmrsthm34 ] if @xmath253 , and @xmath254 for some fixed positive constants @xmath255 and @xmath62 , then the function @xmath256 is well - defined , continuous on @xmath249 , and satisfies @xmath257 @xmath215 . if @xmath258 , and @xmath259 then @xmath7 as defined in is well - defined , continuous on @xmath249 , and satisfies @xmath260 consider the case @xmath253 . we first show that @xmath261 is continuous on every interval of the form @xmath262 $ ] , for @xmath263 . let @xmath264 $ ] . then @xmath265 the first term on the right hand side is a finite sum of continuous functions , and so it suffices to show that the second sum converges uniformly on @xmath262 $ ] . indeed , the decays of @xmath266 and @xmath3 yield the estimate @xmath267 which converges because @xmath268 and @xmath269 . now to consider the decay of @xmath7 , let @xmath270 be fixed . then let @xmath271 be the unique integer such that @xmath272 . since @xmath273 , @xmath274 , and @xmath275 for every @xmath276 , we have @xmath277 the proof for negative @xmath24 values follows similar reasoning . now we explore the properties of the cardinal interpolation operator @xmath7 acting on traditional sequence spaces ; we first show boundedness , and follow with estimates on its norm . [ thmrsthm35 ] for a fixed @xmath244 and @xmath44 , the cardinal interpolation operator @xmath7 is a bounded linear operator from @xmath8 to @xmath278 for @xmath28 . the proof is the same as in @xcite , which follows the techniques of @xcite . linearity is evident , and the cases @xmath279 and @xmath280 follow from proposition [ proprsprop33 ] and lemma [ lem h3 ] , respectively . therefore , let @xmath281 , and @xmath235 be fixed . as before , let @xmath282 be the unique integer such that @xmath272 . if @xmath283 , then theorem [ thmrsthm34 ] implies that @xmath261 is continuous , and we write @xmath284 to estimate the @xmath278 norm of the first term , notice that since @xmath234 , @xmath285 for the second term , first assume that @xmath253 . then @xmath286 the quantity in the final line above corresponds to the @xmath8 norm of the discrete convolution @xmath287 , where @xmath288 , and the entries of @xmath289 are given by @xmath290 . from ( * , theorem 7.6 ) , we find that @xmath291,\ ] ] where @xmath292 is a non - increasing rearrangement of @xmath289 . it is easily checked that the supremum on the right hand side of is finite , and therefore the second term estimated above is @xmath293 , and the theorem follows . the proof for the case @xmath251 is much the same , except that the power on the elements of @xmath289 will be @xmath294 , which nevertheless results in the supremum of @xmath295 being finite . [ thmrsthm36 ] let @xmath244 and @xmath281 be fixed . then @xmath296 consider a fixed @xmath281 , and @xmath297 such that @xmath298 . let @xmath299 , @xmath300 , and @xmath282 be the unique integer such that @xmath272 . then @xmath301 by hlder s inequality and the fact that @xmath234 , @xmath302 to estimate @xmath303 , we represent @xmath3 by its fourier integral ( see proposition [ proplfundamental ] ) and integrate by parts . indeed , @xmath304 where the final step follows from fubini s theorem . from ( * ? ? ? * proposition 1.3 ) , the _ mixed hilbert transform _ defined by @xmath305 is a bounded linear operator from @xmath8 to @xmath278 , and @xmath306 , where @xmath307 is a constant depending only on @xmath308 . consequently , @xmath309 the final inequality comes from boundedness of the mixed hilbert transform and theorem [ thm l3 ] . the conclusion of the theorem follows from the estimates on @xmath310 and @xmath303 . we now estimate the norms in the cases @xmath280 and @xmath279 . [ proprs38 ] suppose @xmath244 and @xmath222 are fixed . the following hold : \(i ) let @xmath311 be defined via . then @xmath312 where @xmath313 is a constant . \(ii ) @xmath314 \(iii ) @xmath315 we supply the proof for the case @xmath316 , the estimates for the negative range of @xmath24 being wholly similar . \(i ) by theorems [ thm l1 ] and [ thm l3 ] , @xmath317 since @xmath311 is 1-periodic , it suffices to check the inequality for @xmath246 $ ] . let @xmath222 and @xmath318 . then by @xmath319\\ & \leq a_\alpha\left[1+\sum_{1\leq|j|\leq n-1}|j|^{-1}+\sum_{|j|\geq n}c^{5\alpha+2}|j|^{-2}\right]\\ & \leq a_\alpha\left[1+\ln(n)+c^{5\alpha+2}n^{-1}\right],\\\end{aligned}\ ] ] whence the result . \(ii ) simply note that if @xmath320 , then @xmath321 , @xmath235 , and apply ( i ) . \(iii ) similarly , if @xmath322 , then @xmath323 whereby ( i ) provides the desired bound . we make note that the case @xmath324 provides an interesting special case . [ thmrsthm311 ] for any @xmath244 and @xmath222 , @xmath325 first , note that by plancherel s identity and a standard periodization argument , @xmath326 consequently , @xmath327}\max { \underset{k\in { \mathbb{z } } } { \displaystyle\sum}}\left(\widehat{l_{\alpha , c}}(\xi+2\pi k)\right)^2.\ ] ] taking the supremum over @xmath328 in the unit ball of @xmath329 yields @xmath330}\max { \underset{k\in { \mathbb{z } } } { \displaystyle\sum}}\left(\widehat{l_{\alpha , c}}(\xi+2\pi k)\right)^2.\ ] ] that the maximum on the right hand side is at most 1 is the content of , but the maximum is attained at @xmath331 by theorem [ thm h3 ] . having established some information about the interpolation operators and their norms for different values of @xmath308 , we now shift our gaze to some convergence properties when the shape parameter @xmath6 for a fixed @xmath24 . as one might expect , for large ( in absolute value ) @xmath24 , we obtain better convergence results owing to the more rapid decay of the fundamental functions . to begin our discussion , consider the _ whittaker operator _ defined via @xmath332 this operator is bounded from @xmath8 to @xmath278 for @xmath281 , and the following holds . [ thmconvergencetowhittaker ] if @xmath333 , @xmath281 , then for a fixed @xmath244 , @xmath334 boundedness of @xmath335 , theorem [ thmrsthm36 ] , and the uniform boundedness principle imply that it is sufficient to check convergence on the coordinate basis for @xmath8 , @xmath336 . that is , if suffices that @xmath337 as @xmath338 by theorem [ thmbandlimitedconvergence ] , @xmath339 converges to 0 uniformly as @xmath6 ; moreover , proposition [ thm l3 ] implies that @xmath340 , whence an application of the dominated convergence theorem yields the statement of the theorem . suppose a function @xmath10 has sufficiently slow growth so that @xmath261 , with @xmath341 , is well - defined ( see theorem [ thmrsthm34 ] ) . in this case we let @xmath342 , and call this the cardinal interpolant of @xmath10 due to the identity @xmath343 we will consider pointwise and uniform convergence of @xmath344 to @xmath10 as @xmath6 , but first we must make some preliminary arrangements . consider @xmath24 to be fixed . then define @xmath345 and @xmath346 [ lemrslem42 ] if @xmath253 and @xmath347 , then @xmath348 , is well - defined , continuous in each of its variables , and @xmath96-periodic in the second variable . if @xmath258 and @xmath349 , then @xmath348 is well - defined , continuous in each of its variables , and @xmath96-periodic in the second variable . continuity of @xmath350 is provided by corollaries [ cor l3 ] and [ cor l4 ] , while proposition [ prop l2 ] shows that the series @xmath351 is uniformly convergent on @xmath114 $ ] . thus @xmath352 is well - defined and uniformly continuous for @xmath353 $ ] , and moreover , we may differentiate the series term by term . finally , @xmath96-periodicity is evident . now let @xmath354 be the space of continuous , @xmath96-periodic functions on the real line , and let @xmath355 denote its dual , which is the space of all @xmath96-periodic complex borel measures on the real line , with the total variation norm given by @xmath356 . following @xcite , given @xmath357 , define the @xmath358-th fourier - stieltjes coefficient of @xmath359 by @xmath360 and the @xmath361-th fejr mean of the fourier - stieltjes series of @xmath359 by @xmath362 we begin by showing that for certain functions , the interpolant exhibits a special form . [ thmrsthm43 ] let @xmath24 be fixed , and suppose @xmath10 is given by @xmath363 for some @xmath357 and some @xmath364 in the case @xmath253 or @xmath365 in the case @xmath258 . let @xmath348 be defined by . then @xmath366 by definition , @xmath367 , thus theorem [ thmrsthm34 ] implies that @xmath344 is well - defined and continuous . by property of the fejr means , @xmath368 therefore , by the inversion formula and a standard periodization argument , @xmath369 { \widehat{l_{\alpha , c}}}(t)e^{-ixt}dt\nonumber\\ & = \dfrac{1}{2\pi } { \displaystyle\int}_{-\pi}^{\pi}\left [ { \underset{j =- n}{\overset{n}\sum}}\left(1-\frac{|j|}{n+1}\right)f(j)e^{ijt}\right]\phi_{\alpha , c}(x , t)dt.\end{aligned}\ ] ] by definition , @xmath370 for every @xmath371 . so if @xmath372 , @xmath373 , and if @xmath374 , @xmath375 . consequently , if @xmath372 , then @xmath376\phi_{\alpha , c}(x , t)dt = { \dfrac{1}{2\pi } } { \displaystyle\int}_{-\pi}^\pi\sigma_n(\mu , t)\phi_{\alpha , c}(x , t)dt.\ ] ] if @xmath374 , then we integrate by parts @xmath377 times . the boundary terms cancel due to periodicity , so @xmath378\phi_{\alpha , c}(x , t)dt \nonumber\\ & = ( -1)^k { \dfrac{1}{2\pi } } { \displaystyle\int}_{-\pi}^\pi\left [ { \displaystyle\sum}_{1\leq|j|\leq n}\left(1-\dfrac{|j|}{n+1}\right)\dfrac{f(j)}{(ij)^k}e^{ijt}\right]\phi^{(k)}_{\alpha , c}(x , t)dt \nonumber\\ & = { \dfrac{1}{2\pi } } { \displaystyle\int}_{-\pi}^\pi\left [ { \displaystyle\sum}_{1\leq|j|\leq n}\left(1-\dfrac{|j|}{n+1}\right)\widehat{\mu}(j)e^{ijt}\right]\phi_{\alpha , c}^{(k)}(x , t)dt\nonumber\\ & = { \dfrac{1}{2\pi } } { \displaystyle\int}_{-\pi}^\pi\left [ { \underset{j =- n}{\overset{n}\sum}}\left(1-\dfrac{|j|}{n+1}\right)\widehat{\mu}(j)e^{ijt}\right]\phi_{\alpha , c}^{(k)}(x , t)dt , \end{aligned}\ ] ] where the final equality comes from the fact that @xmath379 for @xmath380 . combining , , and , we see that @xmath381 using theorem [ thmrsthm43 ] , we show the following result on uniform convergence . [ thmrsthm44 ] suppose @xmath24 is as above , and @xmath10 is given by @xmath382 for some @xmath377 as in theorem [ thmrsthm43 ] and @xmath357 such that @xmath383,\ ] ] for some fixed @xmath384 . then @xmath385 with convergence being uniform on compact subsets of @xmath249 . in the case @xmath372 , convergence is uniform on @xmath249 . suppose first that @xmath380 . let @xmath79 be fixed . we will show that @xmath386 uniformly for @xmath387 . by theorem [ thmrsthm43 ] and the definitions of @xmath10 and @xmath348 , @xmath388 by applying the leibniz rule , @xmath389 the first term above converges to 0 uniformly for @xmath390 , which can be seen from the last portion of the proof of proposition [ propconvergencetosinc ] . additionally , by proposition [ prop l2](i ) the second term above is bounded above by @xmath391 and hence converges uniformly to 0 as @xmath6 . similarly , write @xmath392.\ ] ] here , if @xmath390 , then whenever @xmath393 and @xmath394 , @xmath395 falls into the respective ranges for proposition [ prop l2](ii ) and ( iii ) . applying the estimates of that proposition demonstrates that @xmath396 uniformly as @xmath6 for @xmath387 . if @xmath372 , then we again split the integral into two pieces , and analyze the corresponding integrands @xmath397 and @xmath398 however , notice that there are no terms of the form @xmath399 , and so the inequalities from proposition [ prop l2 ] give upper bounds on the integrands that do not depend on @xmath165 at all , and so the convergence of @xmath7 to @xmath10 is uniform on @xmath249 . the condition on the support of the measure @xmath359 in the previous theorem was essential because of the use of proposition [ prop l2 ] in the proof . heuristically , the condition on the support should be of no surprise due to the fact that we have no uniform control ( in @xmath25 ) of the derivatives of @xmath109 at the boundary points @xmath400 . in fact , it is likely that the derivatives get much larger near the boundary as @xmath25 grows , especially given the fact that @xmath109 converges to the characteristic function of @xmath401 . nevertheless , we may make a weaker assumption on @xmath359 which still yields a convergence result . [ thmrsthm46 ] suppose @xmath24 is as above , and @xmath10 is given by @xmath402 for some @xmath357 which is absolutely continuous with respect to the lebesgue measure . then @xmath385 with convergence being uniform on @xmath249 . the proof is quite similar to that of theorem [ thmrsthm44 ] . let @xmath403 be arbitrary , and choose @xmath404 such that @xmath405+|\mu|[\pi(1- { \varepsilon}),\pi]<\gamma$ ] since @xmath406 is absolutely continuous with respect to the lebesgue measure . then , as before , @xmath407 split each integral into one over @xmath408 $ ] and one near the endpoints . the integral over the interior segment can be handled exactly as in the proof of theorem [ thmrsthm44 ] . for the integrals near the endpoints , notice that @xmath409 , and so the first integral is at most @xmath410+|\mu|[\pi(1- { \varepsilon}),\pi])$ ] , which is at most @xmath411 by the choice of @xmath412 . meanwhile , by proposition [ prop l2](iii ) , the second integrand is at most @xmath413 which can be bounded by a constant depending only on @xmath24 . thus @xmath414 for some constant @xmath415 independent of @xmath25 , and so the conclusion follows . we conclude this section with the special consideration of the case @xmath220 , where @xmath4 is called the poisson kernel . as mentioned above , the fundamental function for the poisson kernel , @xmath416 , decays faster than any polynomial . thus any results in this section that depend on the decay rate of the fundamental function hold true for the poisson kernel . in particular , propositions [ proprsprop33 ] and [ proprs38 ] , and theorems [ thmrsthm35 ] , [ thmrsthm36 ] , [ thmrsthm311 ] , [ thmconvergencetowhittaker ] , and [ thmrsthm46 ] are all valid for the poisson kernel , since existence of the interpolant primarily depends on the decay of the fundamental function . moreover , because the fundamental function for the poisson kernel decays so rapidly , we find much stronger versions of the remaining results in this section . we begin by stating the analogue of theorem [ thmrsthm34 ] . [ thmrsthm34poisson ] suppose that @xmath417 for some @xmath223 and constant @xmath62 . then the _ cardinal poisson interpolant _ , @xmath418 is well - defined , continuous on @xmath249 , and satisfies @xmath419 for any @xmath420 . as a consequence of the preceding theorem , the function @xmath421 defined by has well - defined derivatives of all orders . in particular , lemma [ lemrslem42 ] holds , and consequently we find the following analogue of theorem [ thmrsthm43 ] . suppose @xmath10 is given by @xmath422 for some @xmath357 and some @xmath223 . then @xmath423 finally , theorem [ thmrsthm44 ] holds for any @xmath223 for the poisson kernel . in this section , we illustrate the convergence phenomena discussed in the previous section . the examples are of a similar flavor to those found in @xcite . [ example1 ] let @xmath253 and @xmath347 , or @xmath258 and @xmath424 , or @xmath220 and @xmath223 . let @xmath425 be the @xmath96-periodic extension of the measure @xmath426 , where @xmath427 is the usual dirac measure at 0 . if @xmath428 then theorem [ thmrsthm44 ] implies that @xmath429 uniformly on compact subsets of @xmath249 . however , theorem [ thm h3 ] allows us to say more given this information . if @xmath372 , then we find the following identity on account of theorem [ thmrsthm43 ] : @xmath430 for higher order polynomials , we may use theorems [ thmrsthm43 ] and [ thmrsthm44 ] to show that @xmath431 whereby one can obtain an error bound in terms of @xmath25 via proposition [ prop l2 ] . this also demonstrates that @xmath432 uniformly on compact subsets of @xmath249 . let @xmath433 be fixed , and let @xmath359 be the @xmath96-periodic extension of @xmath434}dt$ ] , where @xmath435}$ ] takes value 1 on @xmath436 $ ] and 0 elsewhere . define @xmath437 since @xmath359 is absolutely continuous with respect to the lebesgue measure , theorem [ thmrsthm46 ] implies that @xmath438 uniformly on @xmath249 . note that this fact also follows from theorem [ thmbandlimitedconvergence ] , albeit from substantially different reasoning . however , the hypothesis of theorem [ thmbandlimitedconvergence ] requires the function being interpolated to be bandlimited , and so the following result can not be obtained simply by appealing to that theorem . let @xmath24 and @xmath377 be as in example [ example1 ] . let @xmath425 be the periodic extension of @xmath439}dt$ ] for some @xmath440 . then if @xmath441 theorem [ thmrsthm44 ] implies that @xmath442 , uniformly on compact subsets of @xmath249 . in this section we prove the various results listed in section [ secpropfundamental ] . our methods closely resemble those found in @xcite . to reduce the clutter in our calculations , we will henceforth drop explicit dependence upon @xmath24 in our calculations that follow . we begin by rewriting in terms of a laplace transform to exhibit the singularity at the origin . to do this , we require an integral representation for the bessel function , which we find from @xcite : @xmath443 where @xmath415 is a constant . we relabel the product of the exponential and the integral in the above expression @xmath447 . that is , @xmath448 hence may be abbreviated as @xmath449 for @xmath450 , we have @xmath451 we turn our attention to the estimates involving @xmath447 and its derivatives , and begin by noting that @xmath452(c),\ ] ] where @xmath453 denotes the usual laplace transform , @xmath454(s)=\int_{0}^{\infty}f(t)e^{-st}dt$ ] . our estimates for @xmath447 will rely on estimates for the laplace transform . the first result in this direction is a lower bound . we obtain this bound by combining the inequalities @xmath457(c ) \geq \int_{0}^{\infty } e^{-ct } ( 2|\xi|)^\alpha t^{\alpha } dt = \dfrac{(2|\xi|)^\alpha \gamma(\alpha+1)}{c^{\alpha+1}}\ ] ] and @xmath457(c ) \geq \int_{0}^{\infty } e^{-ct } t^{2\alpha } dt = \dfrac{\gamma(2\alpha+1)}{c^{2\alpha+1}}.\ ] ] 1 . if @xmath461 , @xmath462(c)\right|\leq \dfrac{(4|\xi|)^{\alpha - l}\gamma(\alpha+1)}{c^{\alpha+1 } } + \dfrac{2^{\alpha - l}\gamma(2\alpha+1-l)}{c^{2\alpha+1-l}};\ ] ] 2 . if @xmath463,@xmath464(c)\right|\leq \dfrac{2^{3\alpha-2l}|\xi|^{2\alpha - l}}{c } + \dfrac{2^{\alpha - l}\gamma(2\alpha+1-l)}{c^{2\alpha+1-l}};\ ] ] 3 . if @xmath465 ( thus @xmath24 is an odd half integer ) , @xmath464(c)\right|\leq \ln\left ( 1+|\xi|^{-1 } \right ) + \dfrac{e^{-c}}{c(1+|\xi|)}.\ ] ] for ( i ) , we have the following @xmath466(c ) = & \int_{0}^{2|\xi|}e^{-ct } t^\alpha(t+2|\xi|)^{\alpha - l } dt + \int_{2|\xi|}^{\infty}e^{-ct } t^\alpha ( t+2|\xi|)^{\alpha - l } dt \\ \leq & \int_{0}^{2|\xi|}e^{-ct } t^\alpha(4|\xi|)^{\alpha - l } dt + \int_{2|\xi|}^{\infty}e^{-ct } t^\alpha(2t)^{\alpha - l } dt \\ \leq & \dfrac { ( 4|\xi|)^{\alpha - l}\gamma(\alpha+1)}{c^{\alpha+1 } } + \dfrac{2^{\alpha - l}\gamma(2\alpha+1-l)}{c^{2\alpha+1-l}}.\end{aligned}\ ] ] inequality ( ii ) follows from a similar calculation , replacing @xmath467 with @xmath468 in the first integral below : @xmath466(c ) = & \int_{0}^{2|\xi| } e^{-ct}t^\alpha(t+2|\xi|)^{\alpha - l } dt + \int_{2|\xi|}^{\infty}e^{-ct } t^\alpha ( t+2|\xi|)^{\alpha - l } dt \\ \leq & \int_{0}^{2|\xi|}e^{-ct}2^{3\alpha -2l}|\xi|^{2\alpha -l } dt + \int_{2|\xi|}^{\infty } e^{-ct}2^{\alpha - l}t^{2\alpha - l } dt \\ \leq & \dfrac{2^{3\alpha -2l}|\xi|^{2\alpha -l } } { c } + \dfrac{2^{\alpha - l } \gamma(2\alpha+1-l ) } { c^{2\alpha+1-l}}.\end{aligned}\ ] ] finally , for inequality ( iii ) , we have @xmath469(c ) = & \int_{0}^{\infty}e^{-ct}\dfrac{t^\alpha}{(t+2|\xi|)^{\alpha+1}}dt = \int_{0}^{\infty}e^{-c|\xi|t}\dfrac{t^\alpha}{(t+2)^{\alpha+1 } } dt\\ = & \int_{0}^{|\xi|^{-1}}e^{-c|\xi|t}\dfrac{t^\alpha}{(t+2)^{\alpha+1}}dt + \int_{|\xi|^{-1}}^{\infty}e^{-c|\xi|t}\dfrac{t^\alpha}{(t+2)^{\alpha+1}}dt \\ \leq & \int_{0}^{|\xi|^{-1 } } ( 1+t)^{-1}dt + \int_{|\xi|^{-1}}^{\infty } e^{-c|\xi|t } ( 1+t)^{-1}dt\\ \leq & \ln\left(1+|\xi|^{-1}\right ) + \dfrac{e^{-c}}{c(1+|\xi| ) } \leq \ln\left(1+|\xi|^{-1}\right ) + \dfrac{e^{-c}}{c}.\end{aligned}\ ] ] applying the leibniz rule to , we see that @xmath473(c ) \right)(\xi).\end{aligned}\ ] ] differentiating under the integral sign , which is justified by the exponential decay of the integrand , and using the triangle inequality yield . to see ( i ) , apply the leibniz rule , triangle inequality , and lemma [ lem jl3 ] to . as a prelude to ( ii ) , we remark that combining ( i ) with lemma [ lem jl2 ] ( i ) and ( ii ) reveals that if @xmath485 , latexmath:[\[\label{eq jl8 } if @xmath476 , we pick up an extra singularity from lemma [ lem jl2 ] ( iii ) , in which case we split the double sum in ( i ) into three parts , the first of which corresponds to @xmath487 , the second to @xmath488 , and the third to @xmath489 : @xmath490(c)\\ \nonumber = & a_\alpha c^{2\alpha+1}e^{-c|\xi|}|\xi|^{-4\alpha-2 } \left ( |\xi|^{2\alpha+1}\ln(1+|\xi|^{-1 } ) \right.\\ \nonumber + & \left . \sum_{l'=0}^{k-1}a_{k , l'}c^{2\alpha+1-l'}|\xi|^{2\alpha+1}\mathcal{l}[t^\alpha(t+2|\xi|)^{\alpha - l'}](c)\right . \\ \nonumber + & \left.\sum_{l=0}^{k-1}\sum_{l'=0}^{l}a_{k , l , l'}c^{l - l'}|\xi|^{l}\mathcal{l}[t^\alpha(t+2|\xi|)^{\alpha - l'}](c ) \right)\\ \label{eq jl9}\leq & a_\alpha c^{2\alpha+1}e^{-c|\xi|}|\xi|^{-2\alpha-1}\ln(1+|\xi|^{-1})\nonumber\\ + & a_\alpha c^{4\alpha+1-\lfloor\alpha\rfloor}e^{-c|\xi|}|\xi|^{-4\alpha-2}[1+o(1 ) ] , \quad |\xi|\to 0.\end{aligned}\ ] ] to track the largest powers of @xmath25 , we note that there is nothing to do in the logarithmic term ; in the second term we use lemma [ lem jl2 ] and take @xmath491 to obtain @xmath492 ; for the third term we let @xmath493 and @xmath491 to obtain @xmath494 . to prove ( ii ) , we again apply the leibniz rule to obtain @xmath495 where @xmath496 is the set of increasing non - negative integer partitions of @xmath377 , that is , @xmath497 . for @xmath485 , we may plug , , and into to obtain : @xmath498 , \quad|\xi|\to 0.\end{aligned}\ ] ] recall that for @xmath503 , @xmath504 . from the leibniz rule , we have @xmath505 thus we may use and to obtain @xmath506 where @xmath507 $ ] . since @xmath198 , @xmath508 , and @xmath509 , we have @xmath510 this is the desired estimate for @xmath511 . if @xmath476 , we must use and to handle the logarithmic term . this changes @xmath512 ; however , it is still true that @xmath513 as @xmath202 . an immediate consequence of proposition [ prop l1 ] is that the function @xmath514 defined in converges for @xmath198 and we can differentiate the series term by term . in fact , by applying the estimates from , we have latexmath:[\[\label{eq jl14 } analogous to the argument given in @xcite , we find that if @xmath189 and @xmath521 $ ] , then @xmath522 where @xmath523 . this means that @xmath524 $ ] , so we may use the leibniz rule together with proposition [ prop l1 ] and lemma [ lem jl5 ] ( letting @xmath525 ) to obtain our next result . in the special case of @xmath220 , we get the poisson kernel , whose fourier transform is given by @xmath531 using this much simpler formula allow us to simplify our earlier work and get stronger results . in fact , we see that the analogue of proposition 4.1 is given by @xmath532 where there is no longer a restriction on @xmath197 . the proof of theorem [ thm kpoissonftdecay ] now follows the same line of reasoning as that used to prove theorem [ thm l1 ] . we begin by noting that @xmath533 is even , so we need only consider the integral @xmath534 for @xmath32 , we use lemma [ lem jl5 ] with @xmath535 and find that @xmath536 . the quantity @xmath537 may be estimated similarly using lemmas [ lem jl6 ] and [ lem jl7 ] ( and again letting @xmath538 ) . we have @xmath539 the last inequality follows from the fact that both terms in parentheses have a global maximum which depends only on @xmath24 . finally , we show that @xmath533 is monotone on @xmath540 $ ] , which implies that @xmath541 . to that end , we use the quotient rule to write @xmath542 we recall that @xmath543 and by the formula found in @xcite , @xmath544 . these allow us to rewrite the numerator , @xmath545 where @xmath546 from these expressions , we see that @xmath547 for @xmath548 $ ] and @xmath549 on @xmath540 $ ] . the latter is true since @xmath550 is decreasing . this shows that @xmath533 is decreasing on @xmath540 $ ] as desired . recall that @xmath533 is non - negative as commented in section [ secbaxteranalogue ] , and theorems [ thm l1 ] and [ thm l2 ] show that @xmath551 is continuous and integrable for the given range of @xmath24 . consequently , @xmath551 is positive definite , and thus for every @xmath235 , @xmath552 the final equality coming from . we first consider the case @xmath372 . as in section 3 , write @xmath553 . proposition [ prop l1 ] with @xmath554 implies that the series converges uniformly for @xmath555 . consequently , @xmath556 . to see that this limit is 1 requires estimating @xmath557 in a slightly different manner than we have so far . combining and , we estimate @xmath558(c)}{\mathcal{l}[t^\alpha(t+2|\xi|)^\alpha](c)}\right|,\ ] ] which by lemma [ lem jl1 ] and proposition [ lem jl2](i ) is at most @xmath559 the series is summable , and uniformly bounded for @xmath555 , and it follows that @xmath560 . now for @xmath561 , let @xmath562 with @xmath555 . then we may write @xmath563}\\ & = \dfrac{|\xi|^{2\alpha+1}\widehat{\phi_c}(\xi+2\pi k)}{a_\alpha c^{2\alpha+1}f_\alpha(\xi)\left[1+\dfrac{c^{-2\alpha-1}|\xi|^{2\alpha+1}}{f_\alpha(\xi ) } { \underset{j\neq0 } { \displaystyle\sum}}\widehat{\phi_c}(\xi+2\pi j)\right]},\end{aligned}\ ] ] the second equality following from . by definition , @xmath564 , and a similar argument to the case @xmath372 above shows that we may allow @xmath81 to tend to 0 , and conclude that @xmath565 . s. d. riemenschneider and n. sivakumar , on cardinal interpolation by gaussian radial - basis functions : properties of fundamental functions and estimates for lebesgue constants , _ j. anal . _ * 79 * ( 1999 ) , 33 - 61 .

the development of multi - functional codes capable of predicting quantum transport properties of complex systems is an increasingly active field of research@xcite . this is driven in part by the top - down scaling of the active elements within cmos devices , for which quantum effects are becoming important . it is also driven by the bottom - up demands of communities working on single - molecule electronics and low - dimensional systems , where structures and molecules of increasing size and complexity are of interest . in particular , multi - functional codes are needed to describe the fundamental properties of quasi - two - dimensional materials such as graphene , silicene , germanene and their integration into workable devices . the need to understand the interplay between all of the above structures and their surrounding environments creates further demands for such codes . at a more fundamental level , over the past forty years , a standard model of electron transport has been developed , based on computing the scattering matrix of quantum systems connected to external sources and there is a need for a universal code which describes the many realizations of such systems under a common umbrella . a key task of any quantum transport code is to start from the hamiltonian describing a system and calculate the quantum - mechanical scattering matrix @xmath0 , from which a wide range of measurable quantities can be predicted . unfortunately for most nanoscale systems of interest , the full many - body atomistic hamiltonian is too complex to allow this task to be completed and therefore one usually resorts to a description based on a mean - field hamiltonian . the system of interest is then composed of a scattering region , connected to external crystalline leads , which are in turn connected to external reservoirs . the problem of computing the scattering matrix for such a system described by an arbitrary mean - field hamiltonian is solved in reference @xcite . the main question therefore is how to obtain the correct mean - field hamiltonian . the simplest mean - field approach to describing quantum transport through nanostructures is to build a tight - binding hamiltonian , which reproduces key electronic properties near the fermi energy . this approach has been available for more than half a century and is still popular today when describing generic properties of materials such as graphene@xcite . tight binding parameters can be obtained by fitting to known band structures and then varied spatially to describe external fields and other perturbations . however such an approach does not easily capture the effects of interfaces between different materials or edge terminations of finite - size systems , whose properties are distinct from those of bulk materials . nor does it easily describe finite - voltage effects . to capture these additional features of inhomogeneous systems , a more material - specific approach is needed . this problem was solved in part by the non - equilibrium green s function technique@xcite , that combines with density functional theory ( dft)@xcite to obtain the self - consistent mean - field hamiltonian of the system subject to a finite bias voltage and from it , the lesser green s functions providing the non - equilibrium electronic density and current . this approach is utilized within the smeagol code@xcite , which was the first to describe spin - dependent and finite - voltage transport properties of systems with inhomogeneous magnetic moments and in the presence of spin - orbit scattering . it is almost 10 years since the release of smeagol and during this period , we have developed a new code with increased speed , versatility and functionality , which is particularly suited to the modeling of larger - scale nanostructures , interacting with their environments . this new code is called gollum and will be freely available from _ http://www.physics.lancs.ac.uk/gollum _ within the coming weeks . our previous experience in the development of smeagol@xcite allowed us to understand that non - equilibrium transport codes are quite difficult to handle , in part because of their complex input data structures , which can create a steep learning curve , and also because they carry very heavy computational demands . as a consequence , we have devised the new code gollum to be more user friendly , with simple and easy to understand input and output structures , and having no accuracy parameters to tune . we present now a short summary of the features and functionalities of the two programs to better appreciate its differences . smeagol is a negf program that computes the charge and spin transport properties of two - terminal junctions subject to a finite voltage bias . smeagol can not read a user - defined tight - binding hamiltonian . instead , it reads the mean - field hamiltonian from the program siesta@xcite and is tightly bound to the old versions of it . smeagol can read from siesta hamiltonians carrying non - collinear spin arrangements as well as the spin - orbit interaction . siesta and smeagol have indeed been used successfully to simulate the magnetic anisotropies of atomic clusters@xcite and the spin transport functionalities of several atomic chains and molecular junctions subjected to strong spin - orbit interaction@xcite . however , smeagol does not profit from other recent density functionals . examples are the van der waals family of functionals or those based on the lda+u approach . gollum is a program that computes the charge and spin , and the electronic contribution to the thermal transport properties of multi - terminal junctions . in contrast to negf codes , gollum is based on equilibrium transport theory , which means that it has a simpler structure , it is faster and consumes less memory . the program has been designed for user - friendliness and takes a considerable leap towards the realization of ab initio multi - scale simulations of conventional and more sophisticated transport functionalities . the simpler interface of gollum allows it to read model tight - binding hamiltonians . furthermore , gollum has been designed to interface easily with any dft code that uses a localized basis set . it currently reads information from all the latest public flavors of the codes siesta@xcite and fireball@xcite . these include functionals that handle the spin - orbit or the van der waals interactions , or that include strong correlations in the spirit of the lda+u approach . plans to generate interfaces to other codes are underway . two- and three - dimensional topological materials display fascinating spin transport properties . gollum can simulate junctions made of these materials either using parametrized tight - binding hamiltonians@xcite , or dft@xcite . dft does not handle correctly strong electronic correlation effects , that are inherent many nano - scale electrical junctions . as a consequence , a number of negf programs like smeagol underestimate such effects . gollum includes several tools to handle strong correlations . these include the above - mentioned interface to the versions of siesta containing the lda+u functional . a second tool uses a phenomenological but effective approach called the scissors correction scheme . a third tool maps the dft hamiltonian into an anderson - like hamiltonian that is handled with an impurity solver in the spirit of dynamical mean field theory . the lighter computational demands required by gollum make it possible to construct conductance statistics relevant to break - junction and stm measurements of single - molecule conductances , therefore making closer contact with experiments . gollum also incorporates an interface with some classical molecular dynamics programs , which enables it to handle interactions with the environment . gollum makes use of the concept of _ virtual leads _ , that allows it to integrate easily a wide range of phenomena by the use of tight - binding hamiltonians . these include spintronics , superconductivity , kondo physics and topological phases . in contrast with smeagol , which only computes the magnitudes of transmission coefficients of two - terminal junctions , gollum has access to the full scattering matrix of a multi - terminal junction , enabling it to compute scattering amplitudes , phases and wigner delay times and thereby describe the properties of quantum pumps . even though gollum is based on equilibrium transport theory , our experience with the use of the negf code smeagol has enabled us to incorporate non - equilibrium physics into the mean - field hamiltonian . gollum has therefore the ability to compute non - equilibrium current - voltage curves . in this article , our aim is to describe the structure of the code and then present a set of demonstrator calculations . the latter will illustrate the additional functionality and versatility of gollum and at the same time constitute a set of new results for the transport properties of selected structures . all of the functionalities that will be discussed below are available either in the current public version of the code , or in the current development version , that will be made public in the autumn of 2014 . the layout of this article is as follows . in section ii , we describe the theoretical approach behind the program and outline the theoretical and practical details of the current implementation . the section starts with a detailed description of the generic two - probe and multi - probe junction setups available within gollum and introduces the terminology that will be used throughout the article . this is followed by subsections describing the determination of the surface green s function of each current - carrying lead and the full scattering matrix . we then introduce a convenient method that allows us to describe finite - voltage non - equilibrium effects . a subsection explaining the concept of virtual leads enables us to describe hybrid structures containing non - collinear magnetism or superconductivity within the scattering region . two additional subsections explain two facilities included in gollum that enable us to describe electronic correlation effects beyond dft , including coulomb blockade and kondo physics . we then show how to include a gauge field in the gollum hamiltonian . a final subsection explains the multi - scale methodology used to describe large - scale junctions and environmental effects , using a combination of classical molecular dynamics for the environment and quantum transport for the central scattering region . in section iii , we present the details and results of the simulations of a series of sixteen different demonstrator systems . the purpose of each demonstrator is to present one or more of the functionalities of gollum . we start with a few model junctions , described by tight - binding hamiltonians , which show basic capabilities and demonstrate how easily and flexibly the program can analyze non - trivial physical effects . these include two and four - terminal normal - metal junctions , a two - dimensional system showing the quantum hall effect , and two hybrid structures , the first containing two superconducting islands sandwiched by two normal metal electrodes , and the second containing a non - collinear spin structure . there follow a series of dft - based calculations , that describe spin - active junctions ; graphene junctions , where the use of van der waals functionals is crucial and junctions enclosing metallo - organic molecules , where a treatment of strong - correlations beyond dft is mandatory . these are handled using three different approaches . in the first , the properties of metalloporphyrin junctions are described using the lda+u methodology ; in the second , the lda spectrum of an ope derivative is adjusted to improve the agreement with experimental data ; in the third , we describe coulomb blockade and kondo features of model or simple gold junctions . the next demonstrator shows how gollum can compute the thermoelectric transport properties of a junction.this is followed by a discussion of the transport properties of a carbon - nanotube - based four - probe junction . the next three examples require the use of multi - scale techniques , where we use a three - step methodology described later in the article . the first demonstrator describes how liquid environmental effects modify the transport properties of a single - molecule junction . the second example demonstrates that single strands of dna can be trans - located though graphene nanopores , where the strands effectively gate the nanopore structure yielding a highly sensitive dna sensor based on field - effect - transistor concepts . a final demonstrator shows how gollum can compute full sequences of pulling and pushing cycles in single - molecule junctions resembling the opening and closing cycles of mechanically controllable break junction ( mcbj ) experiments , enabling the construction of theoretical conductance histograms . the final demonstrator shows how gollum has access to the phase of the full scattering matrix , and describes non - trivial quantum pumping effects related to the phase evolution of the scattered wave function . a concluding section summarizes the features delivered by the program and an appendix illustrates our method to decimate hamiltonians and overlap matrices . gollum describes open systems comprising an extended scattering region ( colored dark blue in figs . 1 and 2 ) connected to external crystalline leads ( colored light blue in figs . 1 and 2 ) . depending on the problem of interest and the language used to describe the system , the material ( m ) of interest forming the central part of the scattering region could comprise a single molecule , a quantum dot , a mesoscopic cavity , a carbon nanotube , a two - dimensional mono- or multi - layered material , a magneto - resistive element or a region containing one or more superconductors . figure 1 shows an example of a 4-lead system whose central scattering region ( generically labelled m throughout the paper ) is a molecule . it is important to note that in an accurate ab initio description of such a structure , the properties of the leads closest to the molecule ( or more generally the central scattering material ) will be modified by the presence of the central scattering region ( m ) and by the fact that the leads terminate . in what follows , we refer to those affected portions of the leads closest to the central scatterer as branches and include them as part of the extended scatterer ( denoted em throughout this article ) . consequently within gollum , a typical structure consists of an extended scatterer ( em ) , formed from both the central scatterer ( m ) and the branches . the extended scattering region is connected to crystalline current - carrying leads of constant cross - section , shown in light blue in the figs . 1 and 2 . for an accurate description of a given system , the branches are chosen to be long enough such that they join smoothly with the ( light blue ) crystalline leads . crucially , the properties of this interface region between the central scatterer m and the leads are determined by their mutual interaction and are not properties of either m or the electrodes alone . , where @xmath1 is an applied bias . ( bottom ) each lead is composed of an infinite chain of identical pl with hamiltonian @xmath2 coupled with each other via coupling hamiltonians @xmath3 . the extended scattering region comprises the actual central scattering region and several pl in each branch up to the tpl . the tpl connect the em region to the leads . the central scattering region consists in this example of the electrodes surfaces and a molecule . ] 2 shows a two - terminal device in more detail and introduces further terminology to be used throughout the paper . the regions in light blue are called electrodes or leads and are described by perfect periodic hamiltonians subject to chosen chemical potentials . each lead @xmath4 is formed by a semi - infinite series of identical layers of constant cross section , which we refer to as principal layers ( pls ) . 2 shows only two pls per lead ( colored white ) , although an infinite number is implied . furthermore in the figure , the leads are identical and therefore the lead index @xmath4 has been dropped . these pls are described mathematically by intra - layer hamiltonians @xmath5 . pls must be chosen so that they are coupled only to their nearest neighbors by the hamiltonians @xmath6 , which means that in the presence of long - range couplings , a pl may contain more than one longitudinal unit cell of the lead . then , if each pl contains @xmath7 orbitals , then @xmath5 and @xmath6 are square @xmath8 matrices . the extended scatterer ( em ) in dark blue is composed of a central scattering region ( m ) and branches . each branch contain several pls . these pls have an identical atomic arrangement as the pls in the leads . however , their hamiltonians differ from @xmath5 and @xmath6 due to the presence of the central scattering region . pl numbering at each branch starts at the pl beside the central scattering region . the outermost pl at each branch of the em region is called the terminating principal layer ( tpl ) and must be described by hamiltonians @xmath9 and @xmath10 which are close enough to @xmath5 and @xmath6 , to match smoothly with the corresponding lead hamiltonian . for this reason , gollum requires that the em contain at the very least one pl . the central scatterer ( m ) itself is described by an intra - scatterer hamiltonian @xmath11 and coupling matrices to the closest pls of the branches . in the example in fig . 2 , the central scattering region m comprises a molecule and the atoms forming the electrode surfaces . the surfaces in gollum include all atoms belonging to the electrodes whose atomic arrangements can not be cast exactly as a pl , due to surface reconstructions , etc . for simplicity , fig . 2 shows the case of a symmetric system , although no such symmetries are imposed by gollum . all hamiltonians are spin- dependent , but again for notational simplicity , the spin index @xmath12 will not be written explicitly here . this means that the hamiltonian @xmath13 for a given lead @xmath4 can be written as the semi - infinite matrix : @xmath14 when using a non - orthogonal basis set , overlap matrices must be defined with the same structure as the hamiltonian matrices : @xmath15 and @xmath16 . it is convenient to introduce the notation @xmath17 for notational simplicity , from now we will consider the case where all leads are equal so that the super - index @xmath4 can be omitted , although gollum imposes no such restrictions . the program can assume either open or periodic boundary conditions in the plane perpendicular to the transport direction . in this last case , unit cells are chosen in the plane perpendicular to the transport direction and the hamiltonians and overlap matrices acquire a specific dependence on the transverse k - vector , @xmath18 , @xmath19 where @xmath20 and @xmath21 label the @xmath22 orbitals in the unit cell ( ie pl ) at the origin of @xmath23 , while @xmath24 denote orbitals equivalent to @xmath21 , but placed in adjacent unit cells located at transverse positions @xmath23 . for @xmath25 , @xmath20 and @xmath21 must belong to the same pl @xmath26 , while for @xmath27 , @xmath24 must belong to the pl @xmath28 . finally @xmath23 are vectors in the two - dimensional bravais lattice , joining the unit cell taken as origin with its neighboring unit cells . to illustrate how an em is connected to leads , we now consider the 4-lead example of fig.1 , where we assume that the tpl is the third pl in each branch . to describe such a multi - terminal setup , the hamiltonian matrix @xmath29 of the em is arranged in a non - conventional way . the first matrix block corresponds to the central scattering region @xmath30 , the second matrix block corresponds to the pls in the em branch connecting to lead 1 , the third matrix block to the pls in the em branch connecting to lead 2 , and so on . @xmath31 finally , the em described by @xmath29 and the leads described by @xmath32 are coupled via matrices @xmath33 to yield a four - terminal junction described by the infinite matrix : @xmath34 where , @xmath35 couple @xmath29 and for example , @xmath36 this arrangement of the hamiltonian enables the straightforward generalization of the approach to an arbitrary number of leads . the transport properties of the junction are encapsulated in its scattering matrix @xmath37 , which can be obtained by computing the green s function of the whole junction @xmath38 by solving the infinite system of equations @xmath39 this equation can be simplified by replacing the semi - infinite lead greens functions @xmath40 by their surface green s functions @xmath41 , whose dimensions are @xmath42 . the remaining system of equations takes the form @xmath43 where the coupling hamiltonians are now @xmath44 and the surface green s functions of the isolated leads @xmath45 can be obtained as described in section ii.a.3 below . the equation for the full green s function can be written in a more compact form as @xmath46 the scattering matrix can be computed using the green s functions matrix elements @xmath47 connecting the different leads , which can be obtained by inverting only the upper matrix box in eq . ( [ gf ] ) @xmath48 in contrast , access to the local electronic and current densities at the em region is obtained from @xmath49 the above expressions for the hamiltonians are very general . any appropriate tight - binding hamiltonian could be introduced by hand to allow computation of the transport properties of a parametrized model . alternatively , any dft code using localized basis sets can provide them . in this case the dft program produces the hamiltonians and fermi energy of the em region @xmath50 , @xmath51 and @xmath52 , and of each lead @xmath53 , @xmath15 and @xmath54 in separate runs . gollum has an interface to the latest versions of the dft program siesta ( siesta 3.1 , siesta vdw and siesta lda+u ) and of fireball and more interfaces will be developed in the future . spin degrees of freedom in spin - active systems are handled as follows : if the spins are all collinear , then we compute separate hamiltonians and perform separate transport calculations for the spin - up and -down degrees of freedom . however , if the junction has non - collinear spins or is subject to spin - orbit interactions@xcite , then the spin - up and down - degrees of freedom are regarded as two distinct sets of orbitals in the hamiltonian , whose distinct labels allow the computation of spin - currents or magneto - resistive behaviors . ( color online ) infinite system used to generate the leads hamiltonians @xmath55 and @xmath56 . a positive direction is defined to be towards the scatterer . ] each of the lead green s functions @xmath45 is determined following the procedures described in refs . [ ] with some minor modifications . to do so , we start by associating to each semi - infinite lead @xmath4 a periodic infinite system , whose unit cell contains a single pl , as sketched in fig . ( [ fig : lead - generation ] ) . @xmath55 and @xmath56 can then be created for this infinite system by hand as model hamiltonians , or can be generated by a dft program in a dedicated simulation . notice that we will drop the @xmath4 super - index until the end of the section for simplicity . by expanding the bloch eigenstates of the infinite system in a localized basis set @xmath57 where @xmath26 , @xmath20 are indices for its unit cells and the orbitals within them , and @xmath58 is a dimensionless , longitudinal bloch wave - vector , the following secular @xmath42 equation can be deduced @xmath59 where the column vector @xmath60 contains the wave - function coefficients @xmath61 . the above equation is usually solved by choosing a wave vector @xmath58 and solving for the eigen - energies and the corresponding eigenvectors . however , in the present transport problem , we do the opposite : we choose the energy @xmath62 and solve for the allowed wave vectors and corresponding eigenstates . for a given energy @xmath62 , the above equation has @xmath63 solutions with either real or complex wave vectors @xmath64 , @xmath65 . to obtain the latter , the equation can be recast as @xmath66 where @xmath67 and @xmath68 are the @xmath42 identity and zero matrices , @xmath69 and @xmath70 is a n - component column vector . gollum solves equation ( [ gep ] ) as it is superior in numerical terms to equation ( [ gep1 ] ) . we compute the group velocities of the states corresponding to real wave vectors as @xmath71 note that @xmath72 has units of energy and therefore the real , fully - dimensioned group velocity is @xmath73 , where @xmath74 is the spacing between neighboring pls in a given lead . we now divide the @xmath63 wave vectors obtained from eq . ( [ gep ] ) into two sets , each containing n values.the first set denoted \{@xmath64 } are real ( complex ) wave vectors that have @xmath75 ( @xmath76 ) and therefore propagate propagate ( decay ) to the right of the figure . they are consequently called positive open ( closed ) channels . the second set denoted \{@xmath77 } are real ( complex ) wave - vectors that have @xmath78 ( @xmath79 ) propagate ( decay ) to the left of the figure . they are called negative open ( closed ) channels . we introduce the dual vectors @xmath80 , @xmath81 , which satisfy @xmath82 and @xmath83 . these can be found by inverting the @xmath42 matrices @xmath84 the above eigenvectors can be used to construct the following transfer matrices @xmath85 these transfer matrices allow us to build the coupling matrix @xmath1 , the self - energies @xmath86 , and the surface green s functions @xmath45 : @xmath87 the procedure described above to compute the lead green s functions can fail , because of the singular behavior of the hamiltonians matrices @xmath27 , which lead to numerical inaccuracies in the solution of eq . ( [ gep ] ) , and is usually manifested in the program producing a number of positive and negative channels different from @xmath22 . notice that if the number of positive and negative channels is different from @xmath22 , then the dual vectors can not be found by inverting @xmath88 and @xmath89 . there exist several schemes to regularize @xmath27 , based on decimating out the offending degrees of freedom . these procedures are explained in detail in refs . [ ] . gollum uses a suitable adaptation of these methods , which is described in the appendix . ( color online ) infinite system whose unit cell is the em region . this is linked to neighboring em cells by periodic boundary conditions . ] ( color online ) super cell containing the extended scattering region em . the em region is surrounded by buffer vacuum regions to its left and right . ] as noted above , @xmath90 can be provided as a model hamiltonian , or generated by a dft or other material - specific program . one of the strengths of gollum is an ability to treat interfaces with high accuracy . in a tight - binding description , tight - binding parameters of a particular material are often chosen by fitting to a band structure . however this does not solve the problem of choosing parameters to describe the interface between two materials . often this problem is finessed by choosing interface parameters to be a combination of pure - material parameters such as an arithmetic or geometric mean , but there is no fundamental justification for such approximations . therefore we describe here methods to generate @xmath90 using a dft program , where the inclusion of branches as part of the extended scatterer occurs naturally . ( [ fig : em1 ] ) shows an example of a junction where the electrodes are identical . the system is composed of super cells formed from a central scatterer and pls . there are periodic boundary conditions in the longitudinal direction , such that the tpl of one branch of a super - cell is linked smoothly to the tpl of a neighboring super - cell . running a dft program for such a super - cell then automatically generates @xmath90 . provided the super cells contain sufficient pls , the hamiltonians @xmath91 and @xmath92 associated with the tpls will be almost identical to those generated from a calculation involving an infinite periodic lead . if the hamiltonians @xmath25 and @xmath27 associated with the pls are generated from a calculation involving an infinite periodic lead , then provided the super cells contain sufficient pls , these will be almost identical to @xmath91 and @xmath92 respectively . in this case then there will be minimal scattering caused by the junction between the tpl and the lead . clearly there is a trade - off between accuracy and cpu time , because inserting more pls increases the size and cost of the calculation . in practice , the number of pls retained in such a super - cell is increased in stages until the results do not change significantly as the number of pls is increased further . there exist situations where the electrodes are dissimilar , either chemically , or because of their different crystalline structure , or because their magnetic moments are not aligned . in these cases , there can not be a smooth matching between tpls of neighboring super cells in fig . 4 . to address this situation , we use a setup similar to that displayed in fig . ( [ fig : em2 ] ) , where additional pls are appended to the branches in the em region . these additional pls are terminated by artificial surfaces and surrounded by vacuum . the tpls are then chosen to be one of the pls near the middle of each branch and should be surrounded by enough pls both towards its artificial surface and towards the central scattering region . then the pls placed between the tpl and the artificial surface are discarded . these sacrificial pls ensure that the chosen tpl is unaffected by the presence of the artificial vacuum boundary . clearly , calculations of this sort are more expensive in numerical terms than those performed with super cells generated as in fig . ( [ fig : em1 ] ) , because they contain many more atoms . in an ab initio calculation of the transport properties of a junction , the dft program produces the hamiltonians and fermi energy of the em region @xmath50 , @xmath51 and @xmath93 , and of each lead @xmath53 , @xmath15 and @xmath54 in separate runs . notice that the hartree potential is defined up to a constant , which is usually different for the em and for each lead this usually means that the energy origin of the em and of the corresponding lead pls , as well as their fermi energies do not agree with each other , so eq . ( 2 ) must be rewritten as follows : @xmath94 where we have referred the energy of each lead to its own fermi energy . to fix the hamiltonian mismatch we define a realignment variable for each lead as follows : @xmath95 where @xmath20 indicates a relevant orbital or group of orbitals . then , the hamiltonian of each lead is realigned with that of the em @xmath96 it turns out that the renormalized @xmath97 and bare @xmath54 fermi energies of each lead do not match perfectly with each other if the number of pls in the em region is not sufficiently large . this is the case when for efficiency reasons , it is desirable to artificially minimize the size of the em . sometimes it is advisable to choose the fermi energy of one of the leads @xmath98 as the reference energy . in this case , a second overall shift can be performed using either @xmath99 or the quantity @xmath100 . we note that the most general scattering state in a given lead @xmath4 at a given energy @xmath62 can be written as a linear combination of open and closed channels as follows @xmath101 where @xmath102 and @xmath103 denote here open positive and negative channels and @xmath104 and @xmath105 are their normalized kets . here , the contribution of all the closed channels in lead @xmath4 is described by the ket @xmath106 . consequently , the number of electrons per unit time flowing between two adjacent pls within the lead is @xmath107 we pick in this section the convention that positive direction in the lead means flow towards the em region and vice versa . so positive ( negative ) open channels are also called incoming ( outgoing ) channels of lead @xmath4 . with this notation . the wave - function coefficients of the incoming open channels of a given lead are determined by the properties of the reservoir connected to the lead . the wave function coefficients of the open outgoing channels of lead @xmath4 are obtained from the amplitudes of all incoming channels by @xmath108 where @xmath103 ( @xmath109 ) is an outgoing ( incoming ) dimensionless wave - vector of lead @xmath4 ( @xmath110 ) . it is therefore convenient to assemble the wave - functions of the @xmath111 outgoing and @xmath111 incoming open channels of a given lead @xmath4 in the column vectors @xmath112 , @xmath113 , and all the scattering matrix elements connecting leads @xmath4 and @xmath110 into the matrix block @xmath114 . notice that the dimensions of @xmath114 are @xmath115 . then the above equation can be written more compactly as @xmath116 by normalizing the bloch eigenvectors @xmath60 , @xmath117 and their duals to unit flux , @xmath118 the matrix elements of the scattering matrix block connecting leads @xmath4 and @xmath110 can be written as . @xmath119 here @xmath120 is the off - diagonal block of the surface green s function defined in eqs . ( 11 ) and ( 12 ) , that connects leads @xmath4 and @xmath110 and @xmath121 is the matrix defined in eq . ( 20 ) . with the above notation , if the incoming channel @xmath102 of lead @xmath4 is occupied with probability @xmath122 ( ie if in eq . ( [ ok ] ) , @xmath123 with probability @xmath122 ) then the number of electrons per unit time , entering the scattering region from reservoir @xmath4 along channel @xmath102 with energy between @xmath62 and @xmath124 is @xmath125 and the number per unit time , per unit energy leaving the scatterer and entering reservoir @xmath4 along channel @xmath126 with energy between @xmath62 and @xmath124 is @xmath127 in many cases , the incoming and outgoing channels of each lead @xmath4 can be grouped into channels possessing particular attributes ( ie quantum numbers ) labeled @xmath128 etc . this occurs when all incoming channels of a particular type @xmath129 in lead @xmath4 possess the same occupation probability @xmath130 . for example , all quasi - particles of type @xmath129 in reservoir @xmath4 may possess a common chemical potential @xmath131 and @xmath132 may take the form @xmath133 , where @xmath134 is the fermi function . in this case , if the incoming and outgoing channels of type @xmath129 belonging to lead @xmath4 possess wave - vectors @xmath135 , @xmath136 , then the number of quasi - particles per unit time of type @xmath129 leaving reservoir @xmath4 with energy between @xmath62 and @xmath124 is @xmath137 where @xmath138 and @xmath139 is the number of open incoming channels of type @xmath140 , energy @xmath62 in lead @xmath4 . note that in the above summation , @xmath141 runs over all outgoing wave - vectors of energy @xmath62 and type @xmath129 of lead @xmath4 and @xmath142 runs over all incoming wave - vectors of energy @xmath62 and type @xmath143 in lead @xmath110 . if @xmath4 and @xmath110 are different leads , then @xmath144 is often called the transmission amplitude and denoted @xmath145 , while if they are the same lead , then @xmath146 is called the reflection amplitude @xmath147 . similarly , for @xmath148 , it is common to define the transmission coefficient @xmath149 as @xmath150 and for @xmath151 , we define the reflection coefficient as @xmath152 so that @xmath153f^i_{\beta_i}(e ) \nonumber \\ & - & \sum _ { { j\ne i},{\beta_j}}t^{i , j}_{\alpha_i,\beta_j } f^j_{\beta_j}(e)\}\label{x1}\end{aligned}\ ] ] note that unitarity of the scattering matrix requires @xmath154 hence the sum of the elements of each row and column of the matrix @xmath155 is zero : @xmath156 or equivalently , @xmath157 and @xmath158 from eqs . ( [ x1 ] ) and ( [ x3 ] ) , if @xmath159 is independent of @xmath110 and @xmath143 then @xmath160 for all @xmath4 and @xmath129 , as expected . for this reason , in the above equations , @xmath159 can be replaced by @xmath161 , where @xmath134 is an arbitrary function of energy , which in practice is usually chosen to be a fermi function , evaluated at a convenient reference temperature and chemical potential . when comparing theory with experiment , we are usually interested in computing the flux of some quantity @xmath88 from a particular reservoir . from eq . ( [ xxx ] ) , if the amount of @xmath88 carried by quasi - particles of type @xmath129 is @xmath162 , then the flux of @xmath88 from reservoir @xmath4 is @xmath163 in the simplest case of a normal conductor , choosing @xmath164 , independent of @xmath129 , the above equation yields the electrical current from lead @xmath4 . within gollum @xmath129 may represent spin and in the presence of superconductivity may represent hole ( @xmath165 ) or particle ( @xmath166 ) degrees of freedom . in the latter case , the charge @xmath167 carried by particles is -e , whereas the charge @xmath168 carried by holes is + e. gollum starts from a mean - field hamiltonian provided either by the user or by an outside material - specific dft code . it then computes the scattering matrix and its related transport properties . when finite voltages are applied to the electrodes , they change the distribution of incoming and outgoing electrons and therefore the underlying hamiltonian . for example , a finite voltage in a two - terminal device may introduce an electrostatic potential , which should be included in the hamiltonian . a key feature of many negf codes including smeagol is that such effects can be treated self - consistently , albeit at the cost of a greatly increased computing overhead . to avoid this overhead , gollum assumes that the user is able to provide a modified hamiltonian at finite voltages . based on our experience on the development and usage of negf programs and as demonstrated in in section iii.j below , we have found that in many cases , the following intuitive modification of the initial zero - voltage hamiltonian yields reasonable - accurate voltage - dependent transmission coefficients @xmath169 connecting leads @xmath4 and @xmath110 . the scheme enables the simulation of non - trivial @xmath170 curves which compare favorably to those obtained using negf techniques and enables the modeling of generic non - equilibrium transport phenomena such as negative differential resistance ( ndr ) and current rectification in close agreement with negf codes@xcite . consider the case where each lead has a different voltage @xmath121 . then the finite - voltage hamiltonian takes the form @xmath171 we find that @xmath29 needs only be computed at zero voltage in most cases ; the effect of a finite bias can be accounted for by a suitable re - alignment of the energies of the orbitals in the em region with the shifted energy levels of the electrodes . mathematically , we apply a simple shift to the hamiltonian matrix elements at each orbital @xmath26 in the em region @xmath172 where these local shifts @xmath173 depend on the junction electrostatics , which in many cases are known . for example , in the case of a highly - transparent junction , the shifts can be modeled by a linear voltage ramp connecting the matrix elements of the orbitals at the tpls of the em region . in contrast , when the central scattering region hamiltonian @xmath174 is connected to the pl hamiltonians @xmath175 in each branch of the em region by weaker links @xmath176 , the voltage drop and therefore the resistance is concentrated at these spots . in this case , we take @xmath177 for all orbitals in branch @xmath4 , starting at the tpl and up to the linker atoms , and @xmath178 for all the orbitals inside the m region itself . this scheme performs specially well for systems where the states around the fermi level ( homo or lumo ) are localized at or close to the contact atoms . it enables us to mimic accurately junctions displaying non - trivial negative differential resistance , as well as rectification effects for asymmetric molecules@xcite . what is the difference between a lead and a channel ? from a mathematical viewpoint , channels connect an extended scattering region to a reservoir and the role of lead @xmath4 is simply to label those channels @xmath179 , which connect to a particular reservoir @xmath4 . conceptually , this means that from the point of view of solving a scattering problem at energy @xmath62 , a single lead with @xmath180 incoming channels can be regarded as @xmath180 virtual leads , each with a single channel . gollum takes advantage of this equivalence by regarding the above groups of channels with wave - vectors @xmath181 as virtual leads and treating them on the same footing as physical leads . from this viewpoint , ( [ xxx ] ) and ( [ x1 ] ) yield the number of quasi - particles per unit time `` from virtual lead @xmath129 '' entering the scattering region with energy between @xmath62 and @xmath124 . this viewpoint is particularly useful when the hamiltonians @xmath182 , @xmath183 describing the pls of the physical lead @xmath4 are block diagonal with respect to the quantum numbers associated with @xmath181 . for example , this occurs when the leads possess a uniform magnetization , in which case the lead hamiltonian is block diagonal with respect to the local magnetization axis of the lead and @xmath140 represents the spin degree of freedom @xmath12 . this occurs also when the leads are normal metals , but the scattering region contains one or more superconductors , in which case the lead hamiltonian is block diagonal with respect to particle and hole degrees of freedom and @xmath140 represents either particles @xmath184 or holes @xmath185 . more generally , in the presence of both magnetism and superconductivity , or combinations of singlet and triplet superconductivity , @xmath140 would represent combinations of spin and particles and holes degrees of freedom . in all of these cases , @xmath182 , @xmath183 are block diagonal and it is convenient to identify virtual leads @xmath129 with each block , because gollum will compute the channels @xmath181 belonging to each block in separate calculations and therefore guarantees that all such channels can be separately identified . this is advantageous , because if all channels of @xmath182 , @xmath183 were calculated simultaneously , then in the case of degeneracies , arbitrary superpositions of channels with different quantum numbers could result and therefore it would be necessary to implement a separate unitary transformation to sort channels into the chosen quantum numbers . by treating each block as a virtual lead , this problem is avoided . examples of this approach are presented below , when describing the scattering properties of magnetic or normal - superconducting - normal systems . in the presence of non - collinear magnetic moments , provided the lead hamiltonians are block diagonal in spin indices ( in general relative to lead - dependent magnetization axes ) choosing @xmath186 and @xmath164 in eq . ( [ xxxx ] ) yields for the total electrical current @xmath187 note that in general it is necessary to retain the subscripts @xmath188 associated with @xmath189 or @xmath190 , because the leads may possess different magnetic axes . similarly the thermal energy from reservoir @xmath4 per unit time is @xmath191 for the special case of a normal multi - terminal junction having collinear magnetic moments , @xmath192 for all @xmath4 and since there is no spin - flip scattering , @xmath193 . in this case , the total hamiltonian of the whole system is block diagonal in spin indices and the scattering matrix can be obtained from separate calculations for each spin . we assume that initially the junction is in thermodynamic equilibrium , so that all reservoirs possess the same chemical potential @xmath194 . subsequently . we apply to each reservoir @xmath4 a different voltage @xmath195 , so that its chemical potential is @xmath196 . then from equation ( [ xxx ] ) , the charge per unit time per spin entering the scatterer from each lead can be written as @xmath197 and the thermal energy per spin per unit time is @xmath198 where @xmath199 and @xmath200 is the deviation in fermi distribution of lead @xmath4 from the reference distribution @xmath201 . in the limit of small potential differences or small differences in reservoir temperatures , the deviations in the distributions from the reference distribution @xmath202 can be approximated by differentials and therefore to evaluate currents , in the presence of collinear magnetism , gollum provides the following spin - dependent integrals @xmath203 in the presence of two leads labeled @xmath204 , the spin - dependent low - voltage electrical conductance @xmath205 , the thermopower ( seebeck coefficient ) @xmath206 , the peltier coefficient @xmath207 and the thermal conductance @xmath208 can be obtained as @xmath209 so that the equivalent spin - summed magnitudes are @xmath210 note that the thermal conductance is guaranteed to be positive , because the expectation value of the square of a variable is greater than or equal to the square of the expectation value . for a two - terminal system , the above expressions allow us to obtain the electronic contribution to the thermoelectric figure of merit@xcite : @xmath211 a phenomenological scheme that improves the agreement between theoretical simulations and experiments in , for example , single - molecule electronics consists of shifting the occupied and unoccupied levels of the m region downwards and upwards respectively to increase the energy gap@xcite of the m region . the procedure is conveniently called spectral adjustment in nanoscale transport ( saint ) . at the request of a user , gollum modifies the hamiltonian operator of the m region as follows : @xmath212 where @xmath213 are energy shifts and ( @xmath214 , @xmath215 ) denote the occupied and unoccupied states , respectively . by using the definition of the density matrix operator of the m region @xmath216 the above hamiltonian can be rewritten as @xmath217 the equation can also be written in matrix form as @xmath218 to find the density matrix , we first solve the generalized eigenvalue problem : @xmath219 where @xmath220 is the @xmath221 identity matrix , and we have arranged the eigen - energies @xmath222 into a diagonal matrix @xmath223 . then @xmath224 in the simplest case , for a single - molecule junction , the shifts @xmath225 are chosen to align the highest occupied and lowest unoccupied molecular orbitals ( ie the homo and lumo ) with ( minus ) the ionization potential ( ip ) and electron affinity ( ea ) of the isolated molecule @xmath226 however the coulomb interactions in the isolated molecule are screened if the molecule is placed in close proximity to the metallic electrodes . currently , gollum takes this effect by using a simple image charge model@xcite , where the molecule is replaced by a point charge located at the middle point of the molecule and where the image planes are placed 1 above the electrodes surfaces . then the shifts are corrected by screening effects as follows : @xmath227 where @xmath74 is the distance between the image plane and the point image charge . many nanoscale - scale junctions are expected to show coulomb blockade behavior , and in specific situations also kondo features @xcite . these features can be demonstrated by gating the junction , and should appear as coulomb and kondo diamond lines in contour density plots of the low - voltage conductance as a function of bias and gate voltages . these strong correlation effects are completely missing in conventional dft . accurate parametrizations of the ground - state energy density functional of the single - channel anderson model exist that allow a correct description of those phenomena@xcite . however , most nanojunctions are better modeled in terms of a multi - channel anderson model as we have chosen to do in gollum . this model is described by the hamiltonian @xmath228 where the hamiltonian at the central scattering region is given by @xmath229 here the @xmath230-sum runs over the @xmath231 correlated degrees of freedom and includes the spin index , @xmath232 denote the on - site energies and @xmath233 is the electronic coulomb repulsion , that is assumed to be the same for all degrees of freedom . we map the central scattering region hamiltonian @xmath30 in eq . ( 4 ) into @xmath234 to extract the self - energies @xmath235 of the correlated degrees of freedom , in the spirit of dynamical mean field theory@xcite . following ref . ( ) , our correlated degrees of freedom are a subset of the eigenstates of @xmath30 , that we will call here molecular orbitals . this contrast with the approach followed in ref . ( ) where the correlated degrees of freedom where taken to be atomic orbitals of transition metal atoms . for a four - lead device , the details of our implementation are as follows . we take the em hamiltonian @xmath236 where @xmath174 has dimensions @xmath237 and therefore describes @xmath155 molecular orbitals . we first solve the generalized eigenvalue problem at the m region as in eqs . ( 54 - 57 ) in the previous section to find the rotation matrix @xmath238 that diagonalizes @xmath174 : @xmath239 where @xmath240 is the @xmath221 identity matrix . we then perform the direct product @xmath241 to enlarge the size of the matrix @xmath238 to the dimensions of the em hamiltonian matrix . we then rotate the em hamiltonian and compute the green s function of the em region @xmath242 we now choose which molecular orbitals @xmath243 with associated on - site energies @xmath244 are the correlated degrees of freedom , and use the projectors @xmath245 to find their projected green s functions and occupancies @xmath246\,f(e_f)\end{aligned}\ ] ] where @xmath247 is the em fermi level . we then shift the on - site energy of the correlated orbitals by the conventional double counting term@xcite @xmath248 as a consequence , we have to recompute again the green s functions @xmath249 this is cast in the form @xmath250 which allows us to extract the hybridization function @xmath251 . the initial ingredients in the solution of the multichannel anderson model are the set @xmath252 . they allow us to extract the self - energies @xmath253 using a impurity solver . these are added again to the on - site energies @xmath254 leading to a new em hamiltonian @xmath29 and associated green s function @xmath255 . from here we compute a new hybridization function @xmath256 with which new self - energies @xmath257 are determined . the cycle is repeated until self - consistency in @xmath258 and @xmath257 is achieved . the resulting @xmath29 is inserted back into eq . ( 12 ) and the surface green - function matrix @xmath259 is computed to extract the transport properties of the correlated junction . we have decided to include in gollum a finite-@xmath233 impurity solver . this way , we can subtract the double - counting terms and place the molecular orbitals at their correct bare energy positions by using eq . ( [ eq : double_counting ] ) . there exist a variety of finite-@xmath233 impurity solvers based on perturbation expansions on the coulomb interaction @xmath233 , on the hybridization function @xmath251 , on interpolative approaches , on monte - carlo algorithms ( see ref . ( ) for a detailed account of some of these solvers ) , or on numerical renormalization group techniques@xcite ( nrg ) . nrg techniques have superior accuracy , but they bring high computational demands . slave - boson - based expansions on @xmath251 like the oca@xcite are rather accurate and less expensive numerically . the impurity solver used in gollum is based on the interpolative perturbation theory approach@xcite , where the second - order in @xmath233 expression for the electron self - energy is interpolated to match the atomic self - energy , and adjusted to satisfy consistency equations for the high - energy moments together with luttinger s theorem . this approach is computationally very simple , but has been proven to provide reasonable results for the multi - channel finite-@xmath233 anderson model@xcite . its main shortcoming is that it overestimates the kondo temperature , as we discuss in section ( [ section_kondo ] ) . specifically , the impurity solver that we have implemented to handle the multi - channel hamiltonian ( [ eq : anderson - hamiltonian ] ) is described in ref . ( ) , although we have corrected errors in some of the equations in that reference . we note that this impurity solver handles @xmath260 spin - degenerate correlated degrees of freedom , so that @xmath231 must be an even number . in other words , these channels must come as kramers pairs . we stress that other impurity solvers can be implemented straightforwardly , due to the modular nature of gollum . to compute transport properties in the presence of a magnetic field gollum allows the user to introduce a peierls substitution by changing the phase factors of the coupling elements@xcite between atomic orbitals . for example in the case of a nearest - neighbor tight - binding hamiltonian , the inter- site matrix element @xmath261 between site i and site j is replaced with the modified element , @xmath262 where @xmath263 and @xmath264 are the positions of site i and j and @xmath265 is the vector potential . the gauge is chosen such that the principal layers of the leads remain translational invariant after the substitution . as an example , below we demonstrate how gollum describes the quantum hall effect in a disordered square lattice , with a perpendicular uniform magnetic field . and @xmath266 connected to a superconductor ] figure [ fig : superconductor](a ) shows a two - probe normal - superconductor - normal ( n - s - n ) device with left and right normal reservoirs connected to a scattering region containing one or more superconductors . if the complete hamiltonian describing a normal system of the type shown in fig . 2 is @xmath267 , then in the presence of superconductivity within the extended scattering region , the new system is described by the bogoliubov - de gennes hamiltonian @xmath268 where the elements of the matrix @xmath269 are non - zero only in the region occupied by a superconductor , as indicated in figure [ fig : superconductor](b ) . physically , @xmath267 describes particle degrees of freedom , @xmath270 describes hole degrees of freedom and @xmath269 is the superconducting order parameter . the multi - channel scattering theory for such a normal - superconducting - normal ( n - s - n ) structure was first derived by lambert in ref . [ ] , where the following current - voltage relation was presented : @xmath271 where @xmath272 ( @xmath273 ) is the current from the left ( right ) reservoir , @xmath274 ( @xmath275 ) is the difference between the chemical potential of the left ( right ) reservoir and the chemical potential @xmath20 of the superconducting condensate and the voltage difference between the left and right reservoirs is @xmath276 . this expression is the low - voltage limit of more general current - voltage relations discussed in [ ] . the generalization to multi - probe structures is described in refs . , to thermoelectric properties of superconducting nanostructures in refs . [ ] and to ferromagnetic - superconducting structures in refs . [ ] . in this equation , @xmath277 where @xmath278 ( @xmath279 ) is the number of open channels in the left ( right ) lead , @xmath280 ( @xmath281 ) are normal ( andreev ) reflection and transmission coefficients for quasi - particles emitted from the right lead , @xmath282 ( @xmath283 ) are normal ( andreev ) reflection and transmission coefficients from the left lead and all quantities are evaluated at the fermi energy @xmath284 . as a consequence of unitarity of the scattering matrix , these satisfy @xmath285 and @xmath286 . the current - voltage relation of equ . ( [ x7 ] ) is fundamentally different from that encountered for normal systems , because unitarity of the s - matrix does not imply that the sum of each row or column of the matrix @xmath74 is zero . consequently , the currents do not automatically depend solely of the applied voltage difference @xmath276 ( or more generally on the differences between incoming quasi - article distributions ) . in practice such a dependence arises only after the chemical potential of the superconductor adjusts itself self - consistently to ensure that the current from the left reservoir is equal to the current entering the right reservoir . insisting that @xmath287 , then yields @xmath288 and therefore the two - probe conductance @xmath289 takes the form of @xmath290 the above equation demonstrates why a superconductor possesses zero resistivity , because if the superconductor is disordered , then as the length @xmath291 of the superconductor increases , all transmission coefficients will vanish . in this limit , the above equation reduces to @xmath292 . in contrast with a normal scatterer , this shows that in the presence of andreev scattering , as @xmath291 tends to infinity , the resistance ( = 1/conductance ) remains finite and therefore the resistivity ( ie resistance per unit length ) vanishes . in the notation of eqs . ( [ xxx ] ) and ( [ x1 ] ) , the above current - voltage relations and their finite - temperature , finite voltage generalizations can be obtained from eq . ( [ xxxx ] ) by writing @xmath293 or @xmath294 to yield @xmath295 since @xmath296 and @xmath297 , this becomes @xmath298 \bar f^j_{\beta_j}(e ) \label{x6}\ ] ] since , in the low - bias limit , @xmath299 , where @xmath134 is the fermi distribution with chemical potential @xmath20 , this simplifies to @xmath300 where @xmath301 \label{x77}\ ] ] the total current is obtained by multiplying eq . ( [ x6 ] ) by a factor of 2 to account for spin . on the other hand , in the limit of zero temperature , @xmath302 hence in this limit , the current - voltage relation ( [ x6 ] ) reduces to eq . ( [ x7 ] ) . simulation of the transport properties of a nanoscale - scale junction involves three distinct tasks . first , model geometries must be generated . secondly , the hamiltonian for each geometry must be constructed . thirdly , the s - matrix can be calculated and transport properties of the junction calculated . gollum separates these three tasks into three different processes . an overview of the work - flow of a generic gollum calculation is shown in figure [ multiscale ] . the three consecutive stages of the work process are denoted by the three dotted rectangular boxes . the initial step consists usually of modeling the atomistic arrangement of the junction . an initial structure is usually guessed , followed by geometry optimization or molecular dynamics simulations to obtain a more realistic atomic arrangement . this task can be performed by either ab - initio or classical molecular - dynamics methods . for systems containing a few hundred atoms , a quantum - mechanical dft - based simulation is usually the method of choice . however , experiments are often performed under ambient conditions or in a liquid environment . in these cases that the microscopic model should include the atomic structure of the environment , as we show below in section iii.gollum addresses this task by using classical molecular dynamics to model the environment and in the spirit of the born - oppenheimer approximation , feeding snapshots of the associated electrostatic field into the the dft - based mean - field hamiltonian . a similar approach is used to model the evolution of mechanically - controlled break junctions upon stretching , where the atomistic arrangement of the junction evolves slowly in time . in this case , if the same experiment is repeated a number of times , the junction geometry will be slightly different each time . therefore , a proper statistical analysis of the junction geometries is mandatory and calculations of the associated distribution of transmission coefficients is required . the task of generating junction geometries is also better suited for classical molecular dynamics situations . gollum also facilitates the use of combined dft and classical molecular dynamics approaches to gain accurate , yet quicker simulation results@xcite . a non - comprehensive set of software tools is listed in figure [ multiscale ] . once the atomic arrangements are generated , these are fed into the second stage , where the hamiltonian matrix is generated . this stage is in practice independent of the previous geometry construction and can be run separately , taking only the output geometries of the first stage . the junction hamiltonian can be generated using a variety of tools , some of which are listed in box ii in figure [ multiscale ] . a popular approach is the use of dft codes that are able to write the hamiltonian in a tight - binding language . in this way , model tight - binding hamiltonians can also be easily generated . other approaches involve the use of slater - booster or semi - empirical methods . in addition , gollum has the ability to modify suitably these hamiltonian matrices as discussed above . for example , the hamiltonian matrix can be modified to include scissor corrections , coulomb - blockade physics , a gate or bias voltage , a magnetic phase factor or a superconducting order parameter . finally , stage iii is the actual quantum transport calculation . this takes the hamiltonian matrix as an input and calculates the s - matrix and associated physical quantities , such the electrical or spin current , the conductance , or the thermopower . in this section , we present a diversity of calculations , which demonstrate the broad capabilities of gollum . for simplicity , we begin with a set of calculations on model hamiltonians , which demonstrate that gollum can easily handle tight - binding models for a range of physical systems . we then move on to more material - specific calculations , in which the hamiltonian is obtained from dft . these include examples exhibiting kondo physics , coulomb blockade and non - linear , finite - voltage effects . next we present more computationally challenging calculations involving van der waals interactions , environmental effects and series of geometries associated with break - junction measurements . finally an example of a quantum pump is presented , which requires access to the phase of scattering amplitudes . we define the conductance quantum @xmath303 , that will be used frequently below . , @xmath304 , @xmath305 and @xmath306 ( taken in arbitrary units ) . ] ( a ) as a function of the energy . energies are referred to the fermi energy @xmath247 and are given in units of @xmath307 . ] ( b ) that has four one - dimensional chain leads . as a function of the energy energies are referred to the fermi energy @xmath247 and are given in units of @xmath307 . ] as a first example , we consider a simple one - dimensional tight - binding chain containing a single orbital per pl and a single impurity orbital in the em region , as shown in fig [ fig:2 - 4probetb](a ) . we take the following parameters , that are given in arbitrary units . within the leads , the site energies are @xmath308 , and the nearest neighbor couplings are @xmath309 . the impurity has a site energy @xmath304 and is coupled to the leads by a hopping element @xmath310 . results are shown for @xmath305 and @xmath306 . the transmission coefficient for this chain is shown in figure [ fig:2probetb ] . as a second example , we consider the four - probe structure of fig . [ fig:2 - 4probetb](b ) , that shares the same set of parameters as the two - probe model above . the various transmission coefficients for this structure are shown in figure [ fig:2probetb4 ] . by symmetry , these are all identical . as an example of a quantum transport calculation with a magnetic field , we demonstrate the quantum hall effect within the simple tight - binding square lattice shown in the inset of fig . [ quantumhall ] . the lattice constant is set to @xmath311 . the onsite energies of the perfect lattice are @xmath312 ev , the hopping integrals at zero magnetic field are @xmath305 ev and the fermi energy is set at zero . the red area in the figure denotes a disordered portion of the lattice . in this disordered area , the onsite energies are randomly varied as @xmath313 , where @xmath314 is a random number distributed with uniform probability in the range @xmath315 , @xmath316 and @xmath317 ev ( red , green and blue dashed curves , respectively ) . the transport direction is chosen to be the @xmath318 axis ( e.g. : from bottom to top ) while the @xmath319 axis goes along the horizontal direction . to demonstrate the quantum hall effect we introduce a homogeneous magnetic field perpendicular to the square lattice , pointing out of the paper , which is expressed in units of @xmath320 tesla . with this setup the vector potential is chosen so that the lead remains translationally invariant along the @xmath318 direction . this means that we implement a peierls substitution of the form @xmath321 where @xmath322 and @xmath323 are the coordinates of the site i. with this modified hamiltonian the conductance calculated by gollum is shown in fig . ( [ quantumhall ] ) . this clearly shows the presence of quantum hall plateaus , which are resilient to the presence of disorder . conductance @xmath324 in units of @xmath325 as a function of inverse magnetic field with various level of disorder . the magnetic field unit is set to @xmath326 tesla . the inset shows the square lattice used for the calculation . the black area denotes a perfect square lattice . the red area denotes a disordered portion of the lattice , where the inter - site distances are slight modified from @xmath327 to perturb the phase contribution . the onsite energy for the regular lattice is @xmath312 ev , the coupling with zero magnetic field is @xmath305 ev and the fermi energy is chosen as zero . in the disordered area the onsite energy is randomly varied as @xmath313 , where @xmath314 is a random number distributed with uniform probability in the range @xmath315 , @xmath316 and @xmath317 ev ( red , green and blue dashed curves , respectively ) . ] and @xmath328 . the left ( right ) physical lead consists of two virtual leads @xmath329 and @xmath330 ( @xmath331 and @xmath332 ) carrying particle and hole channels respectively . ] as an example of scattering in the presence of superconductivity , we now compute the electrical conductance of the n - s - n structure shown in fig . ( [ fig : sup-1 ] ) , which contains two superconducting regions with order parameters @xmath333 and @xmath334 . such a structure is known as an andreev interferometer and was first analyzed in refs . [ ] , where it was predicted that the electrical conductance is a periodic function of the order - parameter phase difference @xmath335 , with period @xmath336 . at that time , this effect was completely missing from the more traditional quasi - classical description of superconductivity . when the missing terms were restored , good agreement between quasi - classical theory and scattering theory was obtained @xcite . in the following calculation , the hamiltonian @xmath267 of eq . ( [ super ] ) is simply a nearest neighbor tight - binding hamiltonian on a square lattice , with diagonal elements @xmath308 and nearest - neighbor couplings with @xmath305 ( in arbitrary units ) . within the regions occupied by superconductor @xmath110 , ( where @xmath337 or @xmath338 ) the top ( particle ) sites are coupled to the bottom ( hole ) sites by @xmath339 , with @xmath340 given in units of @xmath307 . for @xmath341 , figure ( [ fig : sup-2 ] ) shows the energy dependence of the andreev refection coefficient @xmath342 and the normal and andreev transmission coefficients @xmath343 and @xmath344 respectively . the green line in figure ( [ fig : sup-2 ] ) represents the number of open channels in electron ( hole ) conducting leads . as expected , the andreev reflection coefficient is large for small energies and decreases for energies above @xmath345 . substituting the values of these coefficients at @xmath346 into eq . ( [ eq : gsup ] ) and evaluating them for all @xmath335 yields the conductance versus @xmath335 plot shown in figure [ fig : sup-3 ] . as expected , the conductance is an oscillatory function of the order - parameter phase difference @xmath335 with period @xmath336 . ( dot dashed line line ) , @xmath347 ( solid blue line ) , @xmath348 ( dashed line ) and the number of open channels in the left lead for the device shown in fig . ( [ fig : sup-1 ] ) , as a function of the energy . as a function of the energy . energies are referred to the fermi energy @xmath247 and are given in units of @xmath307 . ] in units of @xmath325 , for a n - s - n structure shown in fig . ( [ fig : sup-1 ] ) as a function of the phase difference @xmath335 between the two order parameters . ] and @xmath349 the left ( right ) physical lead consists of two virtual leads @xmath350 and @xmath351 ( @xmath352 and @xmath353 ) carrying up - spin and down - spin channels respectively . ] in this section , we compute the electrical conductance of the structure shown in figure ( [ fig : mag1 ] ) , which we again describe using a simple tight - binding model of the form @xmath354 the hamiltonian @xmath267 is simply a nearest neighbor hamiltonian on a square lattice , with diagonal elements @xmath308 and nearest - neighbor couplings with @xmath305 ( in arbitrary units ) and for simplicity we choose @xmath355 everywhere . the systems consists of two magnetic islands with magnetic moments @xmath356 and @xmath349 , connected to non - magnetic leads . choosing the fermi energy to be @xmath357 , and evaluating eq . ( [ xxxx1 ] ) at zero temperature , figure ( [ fig : mag2 ] ) shows the resulting electrical conductance as a function of the angle @xmath335 between the two magnetic moments . as expected , the conductance is an oscillatory function of the magnetic angle @xmath335 . in units of @xmath325 , for the structure shown in shown in fig . ( [ fig : mag1 ] ) as a function of the angle @xmath335 between the two magnetic moments . ] having discussed model systems described by simple tight - binding hamiltonians , we now turn to more material - specific descriptions based on dft . we will use the program siesta in most of the calculations below , and will provide many of the simulation parameters to help people to reproduce our calculations . ( color online ) spin - resolved transmission coefficients as a function of energy for nickel - chain junctions . the different curves correspond to simulation with different levels of accuracy in the @xmath358 summations : ( a ) , ( b ) , ( c ) and ( d ) correspond to 1 , 4 , 16 and 64 @xmath58-points , respectively . ( 1 ) and ( 2 ) correspond to parallel and anti - parallel configurations , respectively . ] gollum can describe voltage - dependent spin - polarized transport in spin - active junctions , made from a variety of metals , including iron @xcite , platinum or palladium @xcite . to demonstrate this , we describe here the voltage - dependent spin - filtering and magneto - resistive behavior of a two - terminal junction where ( 001 ) fcc nickel electrodes with parallel ( p ) or anti - parallel ( ap ) spin orientations are connected by a nickel atomic chain@xcite . notice that because we may have electrodes with ap spin orientations , we are forced to use em setups such as those shown in fig . we sketch in fig . ( [ em_victor1 ] ) the scattering region used in the present calculation . it comprises a 6-atom - long nickel chain , the electrodes surfaces and the two branches . the electrodes surfaces contain 2/3 atomic layers with 4 atoms each . the left and right branches contain two pls that have 4 atoms each . the second pl is followed by vacuum . we have checked that the transport results in this example are reasonably converged if we choose pl1 as the tpl , which means that pl2 is sacrificial . to find the junction hamiltonian , we use the program siesta . we use the generalized gradient approximation ( gga ) functional@xcite and take the theoretical gga lattice constant of 3.45 for the pls as well as inter - atomic distances of 2.27 along the chain . we have employed a single-@xmath359 ( sz ) basis set to span the valence states and a mesh cutoff of 400 ry to define the real - space grid where the density , potential and matrix elements are calculated . ) in the ( a ) p and ( b ) ap spin configurations as a function of the bias voltage applied to the junction . the different curves correspond to different numbers of @xmath18-points . ] to understand the spin - polarized transport properties of the junction , we will analise below the spin - dependent transmission coefficients @xmath360 , together with the spin - dependent charge currents . these are computed using the approximate expression @xmath361 the above approximation is quantitatively accurate for small enough bias voltages ( @xmath362 v ) and also shows the expected qualitative behavior at larger voltages . we have found that the transmission coefficients and currents depend sensitively on the number of transverse @xmath18-points taken along the plane perpendicular to the transport direction . as we will show below , we need to use at least 16 @xmath18-points to achieve convergence . in other words , a @xmath363-point calculation provides a poor estimate of the transport properties of these junctions . we plot @xmath360 as a function of the energy referred to the fermi energy of the scattering region for p and ap spin orientations in fig . ( [ fig : victor2 ] ) . the upper panel of the figure shows that the transmission coefficients for the p configuration are strongly spin - polarized . this polarization remains at the fermi level , which suggests that these junctions could act as spin filters . the bottom panel of the figure shows the transmission coefficients for the ap configuration . the fact that these are different from those of the p spin orientation hints that these junctions could show significant gmr ratios . to quantify these statements , we compute the charge current of the junction in the p and ap configurations @xmath220 and @xmath364 and plot them in fig . ( [ fig : victor3 ] ) . the figure shows that indeed these junctions show magnetorresistive properties . the figure also demonstrates that the currents depend on the number of @xmath18-points . to further give quantitative estimates of the spin activity of the junctions , we define the spin polarization in the p arrangement and the gmr ratio as @xmath365 where @xmath366 are the spin - dependent currents for the p orientation . we show in fig . ( [ fig : victor4 ] ) these two magnitudes as a functions of the bias voltage applied to the junction . we indeed find large spin signals for these devices . furthermore , the figure demonstrates that at least 16 @xmath18-points are needed to achieve converged results . and ( b ) gmr ratio plotted as a function of the bias voltage applied to the junction shown in fig . ( [ em_victor1 ] ) . the different curves correspond to different numbers of @xmath18-points . ] gollum can profit from the improved chemical accuracy delivered by the most advanced density functionals . as an example , we discuss here the transport properties of the junction shown in fig . ( [ fig : graphene - junction ] ) , where a single phthalocyanine trimer molecule bridges two graphene electrodes separated by a physical gap of length 17.265 . these graphene sheets are armchair - terminated and passivated by hydrogen atoms . periodic boundary conditions are applied in the two directions across the graphene plane . the phthalocyanine units are linked by butadiyne chains . the planar anchors couple to the graphene via interaction of the @xmath367-clouds and therefore an accurate description of the chemical bonding and transport properties can only be achieved by the use of a van der waals density functional . we use here the implementation of dion et al . in the siesta program @xcite . we have computed the hamiltonian and overlap matrix elements using a double - zeta basis set for all the elements in the simulation , together with a grid fineness of 200 rydberg . by minimizing the energy , we find that the molecule sits at a height of 3.4 above the sheets . we have studied the impact of the length of the electrode gap on the transport properties of the junction by attaching additional armchair layers to the edges of both sheets ; these layers are made of two carbon rows , and have a width of 2.502 . the transmission coefficients for several gap widths are shown in fig . ( [ fig : graphene - gap ] ) . we find several breit - wigner resonances associated with molecular levels of the trimer . remarkably , these do not shift in energy as the gap width varies@xcite . however , deep dips appear for several gap widths . these are associated with interference among the different paths whereby electrons can propagate between the molecule and the electrodes . we have also studied the change in the transmission curves as the molecule is displaced laterally and longitudinally across the physical gap . we show representative examples of the transmission curves for longitudinal displacements in fig . ( [ fig : graphene - y ] ) . the figures show that the energy positions of the molecular breit - wigner resonances remain almost constant . we have found the same behavior for other graphene - based junctions : the energy position of the breit - wigner resonances for a given graphene - based junction does not depend on the molecule position relative to the physical gap , provided that the bonding mechanism is by physisorption . this universality arises because physisorption carries no charge transfer between the molecule and the sheets . furthermore the electrodes are made from the same material and therefore there is no dipole moment associated with the contacts . finally , the @xmath368 hybridization between molecular orbitals and the electrode states is weaker than for the bonds present in most noble - metal / single - atom contacts , and does not have a large impact on the nature of the molecular orbitals . as an example of a gollum calculation using a lda+u density functional@xcite , we describe in this section a case where strong electronic correlations may affect the transport properties of a nanoscale - scale junction . we discuss the junction shown in fig . ( [ fig : ldau - geometry ] ) . here gold ( 001 ) electrodes bridge either a porphyrin ( p ) or a metallo - porphyrin molecule ( cup or cop ) . electron flow through any of these three porphyrin molecules is carried by molecular orbitals that hybridize strongly with the gold s - orbitals . this gives rise to broad breit - wigner resonances in the transmission coefficients that are identical for the three molecules . however , for the cup and cop junctions , additional electron paths are created whereby electrons hop into and off the localized d - orbitals of the transition metal atom . the interference between direct and d - orbital - mediated paths creates sharp fano resonances that can however be masked by the much wider breit - wigner resonances@xcite . we see below how including strong correlations in the d - orbitals of the co and cu atoms in terms of a lda+u approach produces strong shifts in the energy dependence of those resonances . we have computed the hamiltonian using the siesta code , where we have picked a single zeta basis for the gold atoms at the electrodes , a double - zeta - polarized basis set for all the atoms in the molecule and a gga functional . we have included a @xmath233 correction term for the d - orbitals of the cu and co atoms in a mean - field fashion , in the spirit of the lda+u approach@xcite . we present here our results for values of @xmath233 equal to 0 , 2.5 and 4.5 ev . of the junction shown in fig . ( [ fig : ldau - geometry ] ) , where the central molecule is cup . @xmath369 is plotted as a function of the energy @xmath62 referred to the fermi energy @xmath247 of the scattering region . the different panels correspond to the different @xmath233 corrections added to the dft hamiltonian ( see text ) . the green and gold ellipses encircle masked fano resonances . they are originated by paths hopping onto the cu 3d@xmath370 or d@xmath371 orbitals . the red ellipse circles a sharp breit - wigner resonance coming from c and n atoms . ] we find that the transmission coefficients of the three molecules display the same wide breit - wigner resonances , that correspond to molecular orbitals hybridizing strongly with the electrodes . these are shown in figs . ( [ fig : ldau - cup ] ) and ( [ fig : ldau - cop ] ) for cup and cop respectively . in addition , the three molecules show a sharp breit - wigner resonance that is marked by a red ellipse in the figures . this resonance corresponds to a molecular orbital encompassing c and n atoms that is weakly bonded to the electrodes . interestingly , this sharp breit - wigner resonance shifts in energy if we change the value of the coulomb interaction @xmath233 for the cup and cop junctions . to understand this phenomenon , we have looked at the density of states of the junction projected onto each atomic orbital , and the local density of states integrated in a narrow energy window around the red resonance . we have found that the n- and c - based molecular orbital hybridizes with the d@xmath372 orbital of the copper or cobalt atoms and is therefore affected by the @xmath233-term . we have found additional sharp peaks appearing in @xmath373 for the cup and cop junction that do not show up for the simple porphyrin junction . these are marked by green , blue and gold circles in figs . ( [ fig : ldau - cup ] ) and ( [ fig : ldau - cop ] ) . these seem to be sharp breit - wigner resonances also . however , we have demonstrated@xcite that they are actually fano resonances , where the fano dip is masked by the transmission of the neighboring wide breit - wigner resonances . by plotting the density of states of the junction projected in each orbital , we indeed find that they correspond to the copper d@xmath370 or d@xmath371 orbitals . because these fano resonances are associated with atomic d - orbitals strongly localized in the transition metal atom , we expect that adding a @xmath233-term will have a strong impact on their energy position . ( [ fig : ldau - cup ] ) shows how these resonances indeed shift in energy as @xmath233 is increased . note that one of the fano resonances coming from the d@xmath370 copper orbital is strongly pinned to the fermi energy , while other resonances rapidly move down in energy . we analyze in this section the transport properties of a series ( 111 ) gold junctions that are bridged by ope derivatives . the backbone of these molecules has a varying number of rings ranging from one to three . the molecules may be oriented fully perpendicular to the electrodes surfaces , or making a tilting angle , as we show in fig . ( [ tricene45 ] ) it is well established by now @xcite that gold junctions that contain conjugated thiol - terminated molecules like opes have a larger conductance when the molecule is tilted . this is due to the increased overlap of the @xmath374 states of the sulfur atoms when the angle between the molecule and the normal to the surface increases . we show in this section that a plain dft - based calculation predicts that the largest conductance occurs when the molecule is oriented perpendicular to the electrodes . this deficiency is remedied by the use of the saint method . this method is an efficient semi - empirical correction that allows us to obtain quantitative agreement between dft calculations and experiments@xcite , as we have already stressed in section ii . ( color online ) a ( 111 ) gold junction sandwiching a tricene - dithiol molecule . the molecule is coupled to one ad - atom on one side and to a hollow gold site at the other end , and its orientation is titled with respect to the electrodes normal line . ] the hamiltonian of the junction has been obtained with the code siesta and a local density approximation ( lda ) functional@xcite . we have picked a single - zeta basis for the gold atoms of the electrodes , and a double - zeta - polarized basis for the atoms in the molecule . the pls of the electrodes contain three atomic layers , each having @xmath375 atoms . we have chosen junction geometries where the molecular derivatives are oriented either perpendicularly to the gold surfaces or making a 45 degrees angle , as shown in fig . ( [ tricene45 ] ) . due to this tilting angle , we had to use a non - periodic scattering region , as in fig . the scattering regions consisted therefore of the molecule , the two surfaces containing two atomic layers each and 3 pls on each branch followed by vacuum . we chose pl2 as the tpl , so all hamiltonian matrix elements of pl3 were chopped off . ( color online ) transmission of the junction shown in fig . ( [ tricene45 ] ) , containing an ope molecule with 3 rings . @xmath373 has been calculated ( a ) without and ( b ) with saint corrections . continuous and dashed lines correspond to perpendicular and tilted ( 45 degrees ) molecules , respectively . ] we have computed the transmission curves @xmath373 for the referred ope derivatives using conventional dft , and have found for all of them that the conductance ( computed from @xmath376 ) is larger if the molecule is oriented perpendicular to the electrodes . the upper panel in fig . ( [ t_saint_1 ] ) demonstrates this behavior for an ope containing three rings . the figure shows that the higher conductance is originated by the position of the homo level of the molecule , that is placed only slightly below the fermi energy of the scattering region . in contrast , the homo level of the tilted molecule is shifted farther away from @xmath247 . this situation demands for the use of the saint correction scheme , that will reposition the molecular orbital levels at their correct energies . we show in the bottom panel of the figure that this is indeed the case , and that by the use of the saint scheme , the correct experimental trend is recovered , where tilted molecules show larger conductances . we have verified that the same change happens for ope molecules containing one and two rings . finally , we plot in fig . ( [ t_saint_2 ] ) @xmath373 for the three molecules ( containing one , two and three rings ) for perpendicular and tilted orientations . ( [ g.saint ] ) we show the transmission of the ope derivatives with a number of rings between 1 and 3 and two tilting angles , 0 and 45 degrees . the figures show that all those junctions where the derivative is oriented perpendicularly to the gold surface ( defined here to be 0 degrees ) show a larger transmission at the fermi level than the tilted cases in contrast with our expectations discussed above @xcite . ( color online ) transmission of the junction shown in fig . ( [ tricene45 ] ) , containing an ope molecule with ( a ) 1 , ( b ) 2 and ( c ) 3 rings . @xmath373 has been calculated with saint corrections . continuous and dashed lines correspond to perpendicular and tilted ( 45 degrees ) molecules , respectively . ] .[tabsaint.2 ] corrections entering the saint scheme for occupied and unoccupied levels given in ev . [ cols="^,^,^",options="header " , ] the physical mechanism whereby the conductance of the tilted configuration is higher , is due to the higher hybridization between the molecular orbitals and the electrodes in the tilted configuration . this increases the width of the transmission resonances and therefore decreases the effect of opening the gap with the saint correction . on the other hand , the image charge correction is also larger in the tilted configuration , since the molecule is closer to the surfaces , and therefore the reduction in the opening of the gap is also larger , which means the final gap ends up smaller in the tilted configuration . the conductance of each case is summarized in fig . ( [ g.saint ] ) . notice that the saint correction scheme changes qualitatively the physical picture in this junction . our procedure for the saint correction scheme is as follows . we first calculate the ionization potential ( ip ) and electron affinity ( ea ) of the molecules in the gas phase . these gas phase corrections open the dft homo - lumo gap . however , these bare shifts need to be corrected because of image charge effects . the final values for the correction shifts are summarized in table ( [ tabsaint.2 ] ) . notice that the corrections are very similar in magnitude and have opposite signs@xcite . as noted above , gollum has a simple and flexible input data structure so that model hamiltonians can be utilized easily . as a simple example of a simulation exhibiting kondo and coulomb blockade behavior , we show here results obtained from gollum for a tight - binding single - level anderson model coupled to two semi - infinite chains , that corresponds to taking @xmath377 in eq . ( [ eq : anderson - hamiltonian ] ) . due to the coupling to the two leads , the correlated level acquires a finite bandwidth @xmath378 where @xmath379 is the density of states in the leads . for vanishing @xmath363 the model is in the so - called atomic limit which is characterized by sharp peaks in the @xmath380-level density of states @xmath381 at @xmath382 and @xmath383 . this limit corresponds to the coulomb blockade regime in an actual junction where the conductance is strongly suppressed except at the charge degeneracy points . however , when the coupling to the leads increases ( @xmath363 becomes larger than the temperature @xmath369 , but is still smaller than @xmath233 ) , virtual processes allow the charge and spin in the molecule to fluctuate and a resonance close to the fermi energy appears due to the kondo effect . this simple model therefore captures relevant physics of molecular junctions such as the appearance of the coulomb blockade effect , and the crossover from the coulomb blockade to the kondo regime as the temperature is lowered below the kondo temperature @xmath384 where the last expression holds in the so - called symmetric limit @xmath385 . the interpolative impurity solver implemented in gollum provides a good quantitative description of the above phenomena in the weak coupling regime and is also qualitatively correct in the intermediate and strong coupling regimes . it however overestimates the width of the kondo resonance . we show in fig . [ fig : anderson - t ] ( a ) the zero - voltage transmission curve @xmath373 computed with gollum , and using the single - level anderson hamiltonian ( [ eq : anderson - hamiltonian ] ) in the symmetric limit . we take the following parameters : @xmath386 ev , @xmath387 and @xmath388 so that @xmath389 and @xmath390 , placing the junction in the strong correlation regime . these parameters yield a kondo temperature @xmath391 . the figure shows that the interpolative solution provides a transmission curve featuring the lower and upper hubbard bands placed at their correct position and having the right width , together with a sharp kondo resonance at low temperatures which progressively smoothens and eventually disappears as the temperature is raised . however , the interpolative solution provides a kondo temperature @xmath392 , e.g. : two orders of magnitude larger than the exact one . we now show the results obtained from gollum for a similar junction , shown in fig . ( [ fig : anderson - molecule ] ) , where a hydrogen atom bridges two gold ( 001 ) electrodes . in this case , the input hamiltonian is generated by the dft code siesta and the leads are repeated periodically in the plane perpendicular to the transport direction , using pl unit cells in ab stacking and @xmath393 atoms in each atomic layer . we have used a single - zeta basis set and a generalized gradient approximation functional . we have adjusted the distance @xmath380 between the hydrogen atom and the leads to reproduce a coupling similar to that set for the above anderson model which is achieved with @xmath394 . the generated transmission curve is shown in fig . [ fig : anderson - t ] ( b ) for the same three temperatures used for the anderson hamiltonian . we find that the shape of the transmission curves remains qualitatively the same . however , both the lower and upper hubbard bands and the kondo resonance are now sharper at their tips . using gollum , we subsequently apply a gate voltage @xmath395 to the gold - hydrogen - gold junction and compute the low - voltage conductance @xmath324 as a function of @xmath395 to compare the results using plain dft versus dft in combination with the interpolative method . the results are shown in fig . ( [ fig : anderson - gate ] ) . we have not included the double - counting term to make the comparison between both approaches more explicit . the figure nicely shows how the interpolative method splits the single dft peak into two coulomb blockade peaks and also a kondo peak . the figure also shows how the kondo peak disappears at temperatures above @xmath396 leaving only the coulomb blockage features . finally , we subject the gold - hydrogen - gold junction to the combined effect of finite bias @xmath1 and gate @xmath395 voltages . fig . ( [ fig : anderson - diamond ] ) shows density - contour plots of the low - voltage conductance as a function of @xmath1 and @xmath395 . this figure demonstrates that gollum can nicely reproduce coulomb blockade diamonds , as well as the kondo line , that disappears as the temperature is raised above @xmath396 . we demonstrate with two examples how gollum incorporates finite - voltage effects . in the first , we have computed the current - voltage characteristics of a gold ( 001 ) junction sandwiching an alkane molecule , that we show in fig . ( [ alkane4 ] ) . we have computed the zero - bias hamiltonian of the junction using the siesta code , with a double - zeta - polarized basis for all the atoms , and a gga functional . the pls contain two atomic layers , with @xmath393 atoms each . we have applied periodic boundary conditions across the plane perpendicular to the transport direction and have computed @xmath397 at the @xmath363 point . for every given voltage @xmath1 , we modify the em hamiltonian as described in eq . ( [ voltage ] ) . all the orbitals @xmath26 at the left branch in the em region in fig . ( [ alkane4 ] ) are shifted by @xmath398 , starting at the tpl and stopping at the linking sulfur atom . similarly , the orbitals at the right branch in the em region are shifted by @xmath399 all , starting at the linking sulfur atom and including those at the tpl . the current @xmath400 is then computed from the modified hamiltonian @xmath401 . the resulting @xmath170 curve is shown as a red dashed line in fig . ( [ alkane5 ] ) . for comparison , the black line in the same figure shows the @xmath170 curves obtained with a full negf simulation using the code smeagol . finally , the dot - dashed green line shows the current - voltage curve obtained from gollum by integrating the zero - voltage transmissions obtained from @xmath402 . the figure demonstrates that our proposed method reproduces rather accurately the features found in the full negf calculation , including the ndr feature at @xmath403 volt , while the plain equilibrium calculation fails to reproduce the gross features of the current - voltage characteristics . it underestimates the low - voltage conductance by a factor of two . as a second example , we have calculated the current - voltage characteristics of a ( 111 ) gold junction sandwiching a porphyrin molecule . the junction geometry is similar to that shown in fig . ( [ fig : ldau - geometry ] ) . the sulfur atoms attach to the electrodes at a hollow site . the porphyrin molecule does not have in the present case a metallic atom at the center , but has two saturating hydrogen atoms instead . we have performed two series of calculations . in the first , the geometry and physical gap distance has been relaxed and the calculations have been done at the most stable configuration . in the second , we have pulled the electrodes and sulfur atoms away . to do so , we have increased the sulfur - molecule distance by 0.3 . the current - voltage curves are shown in fig . ( [ porphirin ] ) . these characteristics do not show non - trivial features . we note again that the plain calculation using zero - voltage transmissions fails to reproduce the negf curve , underestimating the conductance by a factor close to 2 . in contrast , our prescription provides characteristics that reproduce accurately the negf results . to understand the difference between the dot - dashed green lines and the finite - voltage results of figs . ( [ alkane5 ] ) and ( [ porphirin ] ) , we have analyzed the evolution of the transmission coefficients @xmath404 as the voltage bias @xmath1 is ramped . we have found running negf simulations that the main non - equilibrium effect that affects the current for the two junctions above is an energy shift of the molecular homo resonance , that moves up as the voltage is increased . our prescription not only captures the effect , but also follows accurately the evolution of the resonance shifts dictated by the negf calculation , at least at low voltages . by shifting the homo resonance upwards in energy , a larger weight of the resonance enters into the energy integration window used to compute the current integral , hence increasing the current . in contrast , this effect can not be captured at all if one uses the plain @xmath405 transmission coefficients . ( color online ) a junction where a butane - dithiol molecule is sandwiched by ( 001 ) gold electrodes and subjected to a finite bias potential . ] in this section , we show how gollum can compute the thermoelectric properties of complex junctions formed by trapping a c@xmath406 molecule between gold electrodes . in a previous paper @xcite we have demonstrated both experimentally and theoretically that @xmath407-based nanojunctions show promisingly - high values for the thermopower and figure of merit . however the temperature dependence of these values and the fluctuations caused by the exact geometrical details have not been thoroughly investigated , partly due to the computationally - expensive nature of the calculations . here we show that the fast and efficient implementation of gollum allows us to undertake a more complete exploration of the transport properties of these c@xmath406-based junctions . the systems of interest consist of two ( 111 ) gold leads that can be tilted an angle @xmath21 relative to each other . each lead is terminated using either a flat surface or a pyramid , as shown in fig . ( [ fig : c60 ] ) . we have performed dft calculations using the code siesta , with a double - zeta - polarized basis set , and the lda functional@xcite . we have relaxed the molecular geometries using a force tolerance of 0.02 ev / and have found the equilibrium distance between the leads and the c@xmath406 molecule to be of about 0.22 nm , depending on the exact orientation of the molecule . this result is in good agreement with other previously reported distances @xcite . we have kept this distance fixed in all subsequent transport calculations . however , we have taken for completeness five possible orientations of the @xmath407 molecule relative to the electrodes . these are : ( a ) a c - c bond between a hexagon and a pentagon facing the au surface ; ( b ) a hexagon facing the au surface ; ( c ) a pentagon facing the au surface ; ( d ) a bond between two hexagons facing the au surface ; and ( e ) a single atom facing the au surface . we have also tilted one of the electrodes in steps of 15 degrees between 0 and 60 degrees to see the interference effects caused by the exact position of the tip on the surface of the fullerene , recalculating the thermoelectric properties at each step . the starting position ( for @xmath408 ) for the c@xmath406 against the electrodes is such that one of its pentagons is facing the au surfaces . molecule . the electrodes are terminated as either a flat surface ( upper panel ) or a pyramid ( lower panel ) , and can be tilted an angle @xmath21 relative to each other . ] ) having flat or pyramid - terminated electrodes ; @xmath21 is the rotation angle . figs . ( a)-(b)-(c)-(d ) show two - dimensional contour plots of the conductance @xmath324 ( measured in units of @xmath325 ) , the thermopower @xmath409 ( measured in @xmath410 ) , the thermal conductance @xmath411 ( measured in nw / k ) , and the figure of merit @xmath412 ( dimensionless ) . the vertical and horizontal axes are the temperature t measured in kelvin and the tilting angle @xmath21 for flat - electrode junctions ( e)-(f)-(g)-(h ) show the same magnitudes for pyramid - terminated electrodes . note that the color code is different for each figure . ] the results of our calculations are shown in figure ( [ fig : c60 - 1 ] ) . by taking horizontal cuts through these surfaces , we can see clear evidence of quantum interference as the angle changes and the tip is repositioned around the fullerene surface . for the conductance @xmath324 , these oscillations are almost temperature independent , whereas in the case of the thermopower @xmath409 , the thermal conductance @xmath411 and the electronic figure of merit zt , these oscillations are almost negligible at low temperatures and then grow with @xmath369 . comparing the conductance obtained with flat electrodes against that with pyramid - terminated electrodes , we can clearly see that @xmath324 decreases substantially when using the pyramid - like electrodes , in agreement with ref . furthermore in the case of flat electrodes the conductance fluctuates with the angle @xmath21 by about half an order of magnitude , whereas for the pyramid - terminated electrodes we find a larger change of almost one order of magnitude . this again demonstrates that the pyramid - terminated electrode scans the molecular surface like an stm tip , with improved detail , while part of these features are blurred when using a flat electrode . the flat - electrode junctions possess conductances values of about 0.6 - 1.2 @xmath413 , while junctions with pyramid tips have conductances of order 0.015 - 0.15 @xmath413 . we overestimate the experimental values for @xmath324 @xcite , by about one order of magnitude , a known problem associated with the underestimation of the homo - lumo gap inherent to the plain dft approach . as expected , the thermopower @xmath409 is more sensitive to the angle @xmath21 when using pyramid - terminated junctions compared with the case of flat - surfaced junctions . interestingly , @xmath409 is quite high especially at the higher temperatures , achieving values of of about @xmath414 to @xmath415 . ab - initio force - relaxation simulations show that it is possible to sculpt complex three - dimensional structures of nanoscale - scale dimensions by cutting shapes into a graphene bilayer @xcite . we find that the edges of the two graphene sheets coalesce in order to saturate dangling bonds and to maximize the degree of @xmath416 hybridization . for example , by cutting a cross shape in a bilayer graphene sheet , the resulting sculpturene is the three - dimensional crossbar carbon nanotube ( cnt ) shown in fig . ( [ fig : four - probe - cnt ] ) , which is an example of a four - terminal electronic device . this four - terminal device is composed of two armchair and two zigzag cnt electrodes . these become perfectly periodic cnt leads ( shown with blue ) far enough from the junction . to perform a gollum - based four - terminal calculation , we obtain the mean - field hamiltonian of this structure using the siesta code , with a double - zeta - polarized basis set and a gga functional@xcite . we start with the referred cross - shaped bilayer graphene sheet and after relaxing the inter - atomic forces to tolerances below 0.02 ev / , we find the crossbar shaped device shown in figure ( [ fig : four - probe - cnt ] ) . we feed the resulting ab - initio hamiltonian into gollum , and compute the transmission coefficients @xmath417 between every possible combination of pairs of leads . notice that the armchair cnt leads are semiconducting , while the zigzag cnts are metallic . we therefore expect different qualitative behaviors for the transmission properties among the different arms . this is shown in fig . ( [ fig : t-12 - 34 ] ) , where the transmission coefficients between the two armchair arms @xmath418 are much smaller than those connecting the zigzag arms @xmath419 . in addition , the figure indicates that the central cross area , where the two arms join together is not transparent , but introduces strong scattering . similarly , the transmission from lead 1 to lead 3 also shows a reduced transmission at low energies due to the semiconducting behavior of the armchair lead 1 , see fig ( [ fig : t-12 - 34 ] ) . this figure shows that @xmath420 due to the junction symmetry . and leads 3 and 4 @xmath419 ( solid black and dashed red lines respectively ) . energies are measured in ev and referred to the fermi energy of the scattering region . ] , solid black line ) and between lead 1 and lead 3 ( @xmath421 , dashed red line ) . energies are measured in ev and referred to the fermi energy of the scattering region . ] in the literature , most theoretical analyses of phase - coherent transport properties assume that the junction is immersed in vacuum and therefore ignore the effects of the surrounding environment . in contrast , many experiments are carried out under ambient conditions , which can have a marked effect on transport properties@xcite . if surrounding environmental molecules possess a dipole moment , then the scattering region will be subject to a fluctuating electrostatic field . previous work to investigate the impact of environmental water on the transport properties of a a single - molecule junction@xcite also took into account the effect of a solvation shell of water molecules surrounding the junction . gollum describes these effects systematically , by noting that for nanostructures such as single - molecule junctions , the timescale for such fluctuations is typically longer than the time taken for an electron to pass through the device and therefore one can adopt the born - oppenheimer approximation and freeze the environment during each electron transit . however , successive electrons experience different environmental snapshots and therefore are subjected to different instantaneous mean field hamiltonians leading to different instantaneous conductances . the measured ( time - averaged ) electrical conductance will hence be an ensemble average over these snapshots . to obtain a series of environmental snapshots , gollum uses classical molecular dynamics to describe the environmental molecules and for each snapshot , feeds the resulting geometries into a dft code to compute the corresponding self - consistent hamiltonian . the resulting mean field hamiltonian is then used to compute the scattering matrix and related transport properties . ) for 100 snapshots . ] to illustrate this approach , we compute here the ensemble averaged conductance of the junction shown in fig . ( [ fig : solvent - molecule ] ) , where two pi - stacked monothiol terminated oligophenyleneethynylenes form a bridge between two gold ( 001 ) electrodes . the bridging molecule is surrounded by two different solvents : decane and 1,4-dioxane,1,2,4-trichlorobenzene ( tcb ) . an example of a junction surrounded by a shell of tcb molecules is shown in the lower panel of fig . ( [ fig : solvent - molecule ] ) . we have tested here the classical molecular dynamics packages lammps@xcite and dlpoly@xcite , but gollum is flexible enough to accept coordinates from other classical molecular dynamics packages . in what follows , we show results obtained with lammps , where we have used the dreiding force field to describe the intra- and inter - molecular interactions and have employed the reaxff forcefield to obtain the initial charges . to create the environment , we place two hundred solvent molecules surrounding the backbone molecule . we perform the simulations using a constant temperature and volume ( nvt ) ensemble and subsequently a constant temperature and pressure ( npt ) thermostat . we equilibrate the junction for 150 ps with 0.1 fs time steps continuously raising the temperature to 290 k. we do not include the gold electrodes in the molecular dynamics simulation , so to simulate the binding of the anchor groups , we hold the positions of the two terminating atoms which connect to the electrodes fixed . we record between 350 and 500 snapshots of the junction , that have been taken every 2ps . for each snapshot , we feed the atomic coordinates into the dft code siesta and generate the dft hamiltonian . we then feed the hamiltonians into the transport code to compute the electrical conductance . some example transmission curves for 100 snapshots of the junction with a tcb solvent are shown in the right panel in fig . ( [ fig : solvent - molecule ] ) . we note that the the room - temperature dynamics of the atoms at the junction lead to a large spread in the transmission curves , and therefore to many different values of the low - voltage conductance . we therefore assemble conductance histograms to help identify the most probable conductance values . the resulting histograms in the presence of decane and tcb solvents are shown in fig . ( [ fig : solvent - histogram ] ) . the fact that the most probable conductance values are different shows that ambient - conditions or liquid - immersed molecular electronics experiments are affected by the surrounding solvent . notice that in these simulations we kept fixed the molecule - electrode geometry , so the spread in @xmath324 is due entirely to environmental effects . ) for two different solvents : decane ( left ) and tcb ( right ) . ] the fact that the transport properties of nanoscale junctions depends on the surrounding environment leads to a wide range of possible sensing applications . in this section , we demonstrate the versatility of gollum by showing how it can be used to predict the change in conductance of a nanopore , when a single dna strand is trans - located through it . deoxyribonucleic acid ( dna ) is a molecule that encodes the genetic instructions used in the development and functioning of all known living organisms and many viruses . dna molecules are double - stranded helices , consisting of two long biopolymers composed of simpler units called nucleotides . each nucleotide is composed of one of the four nucleobases guanine ( g ) , adenine ( a ) , thymine ( t ) and cytosine ( c ) , which are attached to a backbone made of alternating sugar and phosphate groups . the two polymer strands are bound together by non - covalent bonds that link base pairs and are easily separated to form two single - stranded dna molecules ( ssdna ) molecules . dna sequencing aims at identifying the sequence of the dna bases in a sample of ssdna . many researchers are actively seeking new methods to sequence dna with improved reliability and scalability and that are economically viable . biological nanopores made from protein such as a - hemolysin have been shown experimentally to sense the presence of dna @xcite , but are also very sensitive to temperature and ph , and can only be used within a limited voltage bias window@xcite . as an alternative , solid state devices which can be integrated into existing semiconducting circuitry technology and that are robust to the chemical environment have been proposed as sensors @xcite . to demonstrate the versatility of gollum , we examine here the potential for dna nucleobase sensing of the sculpturene device shown in fig . ( [ fig : ttht ] ) , which comprises a torus - like nanopore connected to two cnt electrodes@xcite . the torus in the figure has an inner pore with a diameter of 1.6 nm , whereas the leads are two ( 6,6 ) armchair nanotubes having a diameter of about 5 . we have selected for our study six short strands of ssdna containing three bases , that are shown in fig . ( [ fig : base - sequences ] ) . we have first relaxed the coordinates of the nucleotides that are threading the pore , using the dft code siesta with a double - zeta basis set and a lda functional@xcite . since the pore diameter is slightly larger than the strand width , the strand and its nucleotides can adopt different conformations and orientations inside the pore . we accumulate snapshots of these different conformations and orientations for each of the six ssdna strands as they trans - locate the pore . for each snapshot , we compute the current - voltage curve and subtract the current for the empty pore @xmath422 . the current averaged over snapshots @xmath423 for each ssdna strand is plotted in fig . ( [ fig : ivcurves - sequences ] ) . the sizable height of the curves demonstrate that the conductance of the pore is sensitive to the gating effect produced by the presence of ssdna strands inside the pore . furthermore , the different behavior of the curves means that , armed with a proper statistical analysis , the sculpturene device can distinguish different nucleotide sequences , so that this kind of device could be utilized potentially as a discriminating dna sensor . curves for each of the six sequences shown in fig . ( [ fig : base - sequences ] ) , where the current has been averaged over different pore - nucleotide relative angles and the current of the empty pore has been subtracted . ] a large body of experiments in single - molecule electronics is performed using the mechanically - controlled break junction ( mcbj ) technique , in which a metallic strip is pulled slowly until it breaks into two separate pieces . this process enables the formation of electrodes with molecular - scale gaps , which can be bridged by a single molecule . experimentally , these two electrodes are repeatedly pulled away or pushed towards each other . by applying a small bias voltage and recording the current passing through the junction , the low - voltage conductance can be measured as a function of the distance between the electrodes . when the distance is small enough , a single molecule can bridge the gap between the electrodes and its conductance can be measured . in the literature , most theoretical studies are confined to small numbers of ideal geometries and binding configurations . in this section , our aim is to demonstrate that the versatility of gollum allows us to compute whole pulling curves of conductance versus electrode separation . the single - molecule junction that we discuss here is shown in fig . ( [ fig : pulling - example ] ) and consists of gold ( 111 ) electrodes . the electrodes in the simulation are terminated by pyramids and bridged by a bipyridine molecule . to simulate a stretching process we have created one hundred geometries of the junction , each with a different distance @xmath380 between the center of the end atoms of the two leads . to optimize the atomic arrangement of each of the hundred geometries , we start from an idealized setup consisting of the two pyramids surrounded by vacuum ( e.g. : not attached to the gold leads ) . we place the molecule slightly shifted to one side to break the symmetry and keep an au - n bond - length of about 2 . we then relax the inter - atomic forces with the siesta code using a gga functional@xcite and a double - zeta - polarized basis set until each individual force is smaller than 0.02 ev / . we keep fixed the atomic positions of the bottom two layers of the pyramids during this geometry optimization . ( [ fig : pulling - geometries ] ) shows four of the hundred relaxed configurations achieved . ) for four different distances @xmath380 . ] we then reattach the crystalline gold leads , and impose periodic boundary conditions along the plane perpendicular to the transport direction . we use a sz basis for the gold atoms of the leads , together with a simplified pseudo - potential , where only the 6s channel is included to speed up the simulations . however , we use a double - zeta - polarized basis set for the atoms at the gold pyramids and in the molecule . siesta then creates the hamiltonian of each of the hundred junctions that we feed into gollum . figure ( [ fig : pulling - curve ] ) shows the low - voltage conductance @xmath324 versus the electrode separation @xmath380 . this pulling curve shows that during the pulling process the conductance possesses a plateau , in agreement with many experiments using mcbjs . this simulation also reveals that the aromatic rings contact directly the gold surface , therefore increasing the molecule - gold coupling and the molecular conductance . ) . @xmath324 is measured in units of @xmath325 and plotted as a function of the distance @xmath380 . each of the conductance points displayed correspond to one of the hundred relaxed configurations of the mcbj simulation . the figure also shows the four different distances @xmath380 corresponding to the relaxed geometries displayed in fig . ( [ fig : pulling - geometries ] ) . ] so far , all calculated quantities have been obtained from the modulus squared of the scattering matrix elements . to demonstrate that gollum also provides information about transport properties associated with the phases of the scattering matrix , we now examine an example of a quantum pump . quantum pumps are time - dependent electron scatterers , which are able to transport electrons between two external reservoirs subjected to the same chemical potential . the pump process is adiabatic if the frequency of the pump cycle is smaller than the inverse of the characteristic timescale of the scatterer , the wigner delay time@xcite . experimental @xcite and theoretical @xcite studies of adiabatic quantum pumps have examined the conditions for optimal pumping and the effects of noise and dissipation . adiabatic pumping can be understood in terms of the parametric derivative of the full scattering matrix @xmath409 at fixed chemical potential @xcite . an adiabatically - slowly time - varying scatterer connected by ideal channels to external reservoirs , produces a current @xmath424 pumped into the @xmath110th channel , where @xmath425 is the energy shift matrix defined by @xmath426 with @xmath409 being the full scattering matrix and @xmath427 the fermi energy . pumping can occur if the @xmath428-matrix depends on time through a parameter @xmath429 . hence currents can be expressed in terms of parametric derivatives @xmath430 where @xmath431 is the parametric current entering channel @xmath110 . since gollum gives us access to the full scattering matrix @xmath432 , it offers the possibility of investigating adiabatic pumping in nanostructures . we demonstrate this capability by calculating the charge pumped in a double - walled carbon nanotube nano - electromechanical device shown in fig . ( [ fig : shuttle ] ) @xcite , that mimics the experimental setup of ref . ( ) . since an electron current travelling along the inner tube can cause a chiral outer tube to rotate @xcite , the quantum pump shown in fig . ( [ fig : shuttle ] ) represents the inverse effect , in which rotation of the outer tube causes a current to flow along the inner tube . the position and orientation of the inner tube is kept fixed , while the shorter outer tube rotates slowly . the angle @xmath433 describes the real space rotation angle of the outer tube and also plays the role of the pumping parameter in this system . surrounds concentrically an inner tube , with an inter - layer spacing @xmath434 corresponding to the van der waals distance . the inner wall remains fixed , while the outer tube is rotated about the tube axis by the angle @xmath433 . a slow variation of @xmath433 results in a parametric current @xmath435 . finite charge can be pumped in one rotational cycle depending on the chiralities of the constituent tubes . ] to reveal the rich behavior of this family of quantum pumps , fig . ( [ fig : paremplot ] ) shows the parametric current , as a function of the rotational angle @xmath433 for a typical device . depending on the particular angle , charge may be pumped either from left to right or vice versa . the integral of this parametric emissivity within a full parametric cycle of @xmath436 is the number of electrons pumped per cycle . for a device consisting of a ( 5,5 ) carbon nanotube surrounded by a ( 14,6 ) outer nanotube . the outer tube rotates slowly around the inner one . the contour plot shows the current as a function of the rotational angle @xmath433 measured in degrees and the fermi energy @xmath427 measured in units of the hopping integral @xmath307 between carbon atoms . the orange color indicates charge being pumped from left to right and vice versa . ] in fig . ( [ fig:14 - 6plot ] ) we show the charge pumped in a ( 5,5 ) carbon nanotube with a ( 14,6 ) outer nanotube rotating slowly about it . the average pumped charge clearly drops by several orders of magnitude as the fermi energy is increased from zero . therefore for a most efficient pumping , the fermi energy should be close to the dirac point . note however , that the pumped charge could again increase if the fermi level is large enough to open another channel . beyond this average behavior , there exist numerous sharp peaks in the pumped charge . the location of these peaks correlates with fabry - perot resonances in the reflection coefficient . this suggests that the largest pumping occurs at those resonances . in other words , when the transmission is high , pumping is low , and vice versa . through the device discussed in fig . ( [ fig : paremplot ] ) as a function of the fermi energy measured in units of the hopping integral @xmath307 between carbon atoms at the cnts . the green solid line shows the charge pumped towards the left ; the red solid line shown the charge pumped towards the right . at certain energies , the pumped charge is very high . these peaks correlate with the fabry - perot resonances in the reflection coefficient @xmath238 . @xmath238 is shown as a solid blue line . ] finally , to demonstrate that gollum can handle the disordered systems , we calculate the ensemble - averaged conductivity @xmath12 of a two - terminal system on a square lattice , with leads attached to a disordered scattering region as shown in fig . ( [ fig : bd1 ] ) . the width of the system is @xmath437 unit cells and the length is varied between @xmath438 to @xmath439 unit cells . the conductivity is defined as @xmath440 , where @xmath441 is the transmission from the left lead to right lead evaluated at the fermi energy , @xmath442 ev . the tight - binding hamiltonian of the system has a single orbital per site , with nearest neighbour couplings . @xmath443 the site energies within the leads are @xmath444 ev , while the random site energies within the scattering region are uniformly distributed over the interval @xmath445 $ ] ev . figure ( [ fig : bd2 ] ) shows the the ensemble - averaged conductivity ( @xmath12 ) and transmission coefficient ( @xmath376 ) for the system shown in fig . ( [ fig : bd1 ] ) . it contains three regions . within the ballistic regime between @xmath446 and approximately @xmath447 , the conductivity increases linearly with length . in the diffusive region ( @xmath448 ) , the conductivity exhibits ohmic behavior and is almost independent of length . finally for @xmath291 greater than 100 , there is a cross over to the anderson localized regime . and transmission coefficient @xmath376 as a function of the number @xmath291 of unit cells in the transport direction , for the two probe square lattice shown in the previous figure . ] we end this article by showing how the spin - orbit interaction induces gaps at certain band crossings in the one - dimensional electronic structure of infinite nickel chains . these gaps may appear or not depending on the orientation of the atomic spins relative to the axis of the chain . they lead to dips in the transmission coefficient @xmath373 of the chain at the gap energies . we have simulated linear nickel chains using the dft program siesta . the chains have a single atom per unit cell and are oriented along the z - axis . we have used a standard set of pseudopotential parameters , a double - zeta basis set and simple lda for the exchange - correlation potential . we have checked that the electronic structure and @xmath373 are the same for any spin orientation if the spin - orbit interaction is set to zero , as is should due rotational invariance . however , if the spin - orbit interaction is switched on , then a finite yet small magnetic anisotropy barrier appears . we have found that if we choose the atomic spins to lie along the chain axis , then there are no spin - orbit gaps close to the fermi energy . as a consequence , the transmission coefficients with and without the spin - orbit interaction are indistinguishable from each other ( as shown in fig . ( [ fig : ni - chain - so ] ) ) . in contrast , when the atomic spins are oriented in a plane perpendicular to the chain axis , then several small gaps open around the fermi energy . these gaps are seen as dips in fig . ( [ fig : ni - chain - so ] ) . we have developed a new quantum transport code , which is fast , easy to use and versatile . this flexibility has been demonstrated by presenting a wide range of example calculations , encompassing charge , spin and thermal transport , corrections to density functional theory such as lda+u and spectral adjustments , transport in the presence of non - collinear magnetism , the quantum - hall effect , kondo and coulomb blockade effects , finite - voltage transport , multi - terminal transport , quantum pumps , superconducting nanostructures , environmental effects and pulling curves and conductance histograms for mechanically - controlled - break - junction experiments . further developments are in the pipeline , including the incorporation of phonon transport . gollum will soon be freely available from the following web site _ http://www.physics.lancs.ac.uk/gollum _ and the authors of this article are available to help potential users access the code . the research presented here was funded by the spanish ministerio de economa y competitividad through the grant fis2012 - 34858 , by the uk epsrc and by the european commission fp7 itn `` molesco '' project no . 606728 vmgs thanks the spanish ministerio de economa y competitividad for a ramn y cajal fellowship ( ryc-2010 - 06053 ) . lo has been supported by the hungarian scientic research fund no . jf wishes to thank useful discussions with n. lorente and a. kormanyos . we have described in section ii.a.2 the method employed by gollum to find the surface green s function @xmath259 of each lead . however , the solution of eq . ( [ gep ] ) gives with some frequency numerical inaccuracies which render the method useless as it stands . these inaccuracies are caused by the highly non - singular behavior of the hamiltonian matrix @xmath27 connecting adjacent pls . we discuss here the adaptation of the method described in refs . ( ) that gollum uses to regularize @xmath27 . mathematically , we perform an svd decomposition of this @xmath42 matrix , @xmath449 where @xmath233 and @xmath1 are unitary matrix and @xmath409 is a diagonal matrix containing the eigenvalues @xmath450 of @xmath27 . numerical algorithms usually arrange them in descending order . the condition number of @xmath27 is defined as the ratio @xmath451 between the maximum and minimum eigenvalues of @xmath27 . @xmath411 determines how singular is @xmath452 and therefore the propensity to suffer inaccuracies when handling @xmath27 . small eigenvalues @xmath450 appear whenever @xmath27 is very sparse . physically , this is originated for example if the pl are very long so that a large fraction of hopping integrals ( matrix elements of @xmath27 ) is zero . our first procedure to regularize @xmath27 consists in adding a real or complex random matrix to @xmath27 . we have found that this is frequently enough to render a regular @xmath27 matrix . if this first procedure fails this is because the orbitals involved do not play a role in the transport properties of the lead and should be decimated out , so that the dimensions of the @xmath453 matrix are reduced . notice that reducing the dimensions of @xmath27 has the advantage that the computation of the surface green s function @xmath259 is much lighter . however , the procedure must be performed with some care as finally @xmath259 must connect with the corresponding tpl of the em branch , whose matrices have dimensions @xmath42 . explicitly , we write the schroedinger equation of the infinite chain of the corresponding lead : @xmath454 where @xmath455 , @xmath26 labels the pl and runs from @xmath456 to @xmath457 and we assume that the chain will be chopped off at the @xmath458 pl and then connected to the tpl of the em . we first perform the svd decomposition of @xmath27 described in eq . ( [ svd ] ) . we then set to zero all eigenvalues smaller than a given tolerance @xmath459 . we have checked that setting @xmath460 provides unproblematic @xmath27 and @xmath461 . let us assume that we set to zero @xmath462 eigenvalues of @xmath27 , so that @xmath463 remain non - zero . we now construct the matrices @xmath464 we transform eq . ( [ lead - equation ] ) for all sites up to site @xmath465 as follows : this set of equations generates the new hamiltonians of each lead , and connects it with the extended molecule . overlap matrices must also be decimated as they enter into the expression for the group velocity , eq . ( [ group - velocity ] ) . the denominator in this equation looks like @xmath471 where we have applied the svd transformation to the overlap matrices @xmath472

soon after the invention of inflationary cosmology it was realized that inflation may divide our universe into many exponentially large domains corresponding to different metastable vacuum states @xcite . the total number of such vacuum states in string theory can be enormously large @xcite . a combination of these two facts with the kklt mechanism of vacuum stabilization @xcite gave rise to what is now called the string landscape scenario @xcite . the purpose of this paper is to clarify some features of the landscape . in particular , one may wonder what are the properties of the domain walls separating different vacua , how these domains may form , and how large will be the fraction of the volume of the universe inside different domains . to this end , we will construct a broad class of static bps domain wall solutions interpolating between different ads or minkowski vacua with unbroken supersymmetry . quite generally , these bps domain walls may interpolate between various minima , maxima , or saddle points of the scalar field potential . bps domain wall solutions in 4-dimensional @xmath3 supergravity were first studied in @xcite . the methods used for that analysis have been later refined and generalized to a variety of models , for different number of dimensions and supersymmetries , see e.g. @xcite . our present aim is to further extend these methods to make them suitable for investigation of the string landscape scenario . following the logic of the recent developments , we will begin with an investigation of the supersymmetric vacua arising from flux compactifications prior to the kklt uplifting , and later introduce the more physically relevant metastable de sitter vacua . we start in section 2 with an overview of some known features of domain walls in effective d=4 , n=1 supergravity . the new results here include the derivation of bps domain wall gradient flow equations in generic models with f - term potentials with arbitrary khler potential , superpotential and number of moduli . we rewrite the gravitational action , including the boundary @xmath4-terms in a bps form : when the flow equations are satisfied , the integrand of the action defines the bps domain wall tension . afterward , in the spirit of the analogous work in five dimensions in the context of the ads / cft correspondence , where domain walls are viewed as dual to renormalization group flows @xcite , we present the bps flow equations in the form of rg flows and provide the relevant `` c - theorem '' . in section 3 we give examples of bps domain walls where the potential has more than one critical point but there is no barrier between them . our examples include an ads minimum and an asymptotically minkowski vacuum . other examples have critical points at finite distance in the moduli space : an ads saddle point and an ads minimum , or an ads maximum and an ads saddle point . a particularly interesting case of flux vacua area codes in a calabi yau hypersurface required the study of the flow of four moduli . they flow from one critical point of the potential , the landau ginsburg fixed point , towards the second critical point with larger absolute value of the cosmological constant near the conifold . in section 4 we study a class of models in stringy landscape with a barrier between critical points . we have examples of a local minkowski or ads minimum and another absolute ads minimum of larger absolute value of the cosmological constant . in case of ads - ads wall , we display the domain walls with critical and super - critical tension . the last case has the superpotential crossing zero between two ads vacua . the results of our investigation in sections 2 - 4 directly apply to the transitions between supersymmetric ads and minkowski vacua . however , the most interesting parts of the string theory landscape correspond to ds vacua . these vacua can be obtained from supersymmetric ads or minkowski vacua by uplifting , which can be achieved , e.g. , by adding some anti - d3 branes @xcite . this procedure breaks supersymmetry and may destabilize some of the vacua , in particular those which corresponded to maxima and saddle points . as a result , some of the bps wall solutions may disappear after the uplifting , whereas some others will remain . however , instead of describing the static bps domain walls , they will describe domain walls moving with acceleration . such domain walls can be formed during the tunneling and bubble formation in the process of decay of metastable vacua , as discussed by coleman and de luccia ( cdl ) in @xcite . we will discuss this process in section 5 . if the uplifting is relatively small , which may happen when we uplift a supersymmetric minkowski vacuum , or a shallow ads vacuum , as in the version of the kklt scenario developed in ref . @xcite , then the properties of the domain walls do not differ much form the properties of the bps walls prior to the uplifting . in this case we will be talking about near - bps , or near - extremal walls and bubbles . the concept of near - bps walls and bubbles is closely related to the concept of near - extremal black holes . whereas at the extremal limit the bps black holes are absolutely stable and have zero temperature , the near - extremal black holes have a non - vanishing temperature , they can evaporate and eventually approach a stable extremal limit . the physics of near - extremal black holes has been studied intensively over the years : a remarkable agreement with the physics of d - branes was established under condition of a small deviation from extremality @xcite . in our new endeavor we will try to understand the rate of tunneling into a collapsing vacuum with a negative cosmological constant from uplifted slightly non - bps vacua using the proximity of the large size bubbles to the bps domain walls whose tension we can evaluate exactly . in the bps limit the corresponding bubbles have an infinite radius , but the near - bps ones have a large but finite radius @xcite . the description of these bubbles can be performed using the deviation from the bps limit as an expansion parameter . this simplifies investigation of bubble formation and vacuum decay in such models . in section 5 we will find , in particular , that the vacuum decay rate of the uplifted supersymmetric minkowski vacuum does not depend on the details of the model and is given by the universal expression @xmath5 , where @xmath6 is the entropy of the ds state . this decay rate is much higher than the typical rate of decay of a ds vacuum to a minkowski vacuum , which in many cases does not differ much from @xmath7 @xcite . in application to our vacuum with @xmath8 , we are talking about the difference in the decay rate by a factor @xmath9 . the situation is especially interesting in a more general case when we uplift the models with several different ads minima of the moduli potential . the ads vacuum with a smaller value of the volume modulus experiences a bigger uplifting . the scale of supersymmetry breaking in the uplifted vacuum is related to the magnitude of the uplifting . as a result , the decay of the uplifted ads vacuum to the collapsing space with a negative vacuum energy usually occurs much faster than the typical ds rate of the uplifted minkowski vacuum @xmath10 . under some conditions , the decay rate can be quite large , with the suppression factor @xmath11 being replaced by @xmath12 . here @xmath13 and @xmath14 is the depth of the ads minimum prior to the uplifting , which may be a hundred orders of magnitude greater than @xmath8 . in the kklt - type models , the depth of the ads vacuum prior to the uplifting is of the same order as @xmath15 , where @xmath16 is the gravitino mass after the uplifting @xcite . therefore in the class of models studied in section 5 , the vacuum decay rate due to the tunneling to a collapsing universe is related to the gravitino mass , @xmath17 . after the tunneling , the universe within each bubble with negative vacuum energy collapses within a microscopically small time @xmath18 . the condition that the lifetime of the uplifted vacuum is greater than the age of our part of the universe may serve as a vacuum superselection rule . an important feature of the decay of ds space to a minkowski vacuum or to a collapsing universe with a negative vacuum energy density is that the part of space that experiences such a decay never returns back to ds space . this is quite different from the back - and - forth transitions between various ds vacua , which may keep the system in a kind of thermal equilibrium . in this respect , the channels of irreversible decay of an uplifted ds vacuum to a minkowski vacuum or to a collapsing universe serve as sinks for the probability flow in the landscape . in section 6 we will show that the existence of such sinks has important implications for the calculations of the probabilities in the landscape . in particular , if one ignores the different rate of expansion of the universe in different vacua , i.e. uses a comoving volume distribution , then in the absence of the sinks , the fraction of the comoving volume of the universe in the state with a given vacuum energy density @xmath19 asymptotically approaches @xmath20 . this distribution has a sharp maximum at the smallest possible @xmath21 . meanwhile if the vacuum with the smallest possible @xmath21 is not the ground state but a sink , and the probability of tunneling to the sink is sufficiently large , then the fraction of the volume in the vacuum with the smallest @xmath21 will be exponentially suppressed , and the maximum of the probability distribution will be at larger values of @xmath21 . walls and bubbles considered in our paper interpolate between domains with different values of the scalar fields , but with the same values of fluxes . in section 7 we will briefly review our results and discuss a possibility of their generalization for the transitions between vacua with different fluxes . 1 cm in this paper we are going to analyze the domain wall solutions interpolating between different supersymmetric vacua of the landscape of string theory flux compactifications to 4 dimensions . we will focus on the case of @xmath3 supergravity . bps domain wall solutions with 1/2 of unbroken @xmath3 supersymmetry in the minimal 4-dimensional supergravity have been discussed in @xcite . the equations presented there refer to the case of a single scalar field . here we are going to extend this analysis to a more general setup allowing for the possibility of many moduli fields with generic metric in the moduli space . of course , in any sufficiently small patch , there is a possible reparametrization of the moduli space so that there is a single scalar field supporting the domain wall . in a rather generic setup , however , this is not possible globally , and therefore the equations we will propose in the following are more useful in this respect . the analysis presented here is mainly based on the approach presented in @xcite for the domain wall solutions of the minimal 5-dimensional supergravity . we briefly recall the main ingredients needed in order to discuss supersymmetric bosonic configurations of n=1 supergravity coupled to chiral multiplets . the bosonic sector of the action is @xmath22 where the potential can be expressed in terms of a superpotential @xmath23 as @xmath24 or , in terms of the covariantly holomorphic superpotential @xmath25 , as @xmath26 the covariant derivative is defined as @xmath27 on holomorphic quantities ( like @xmath28 ) and as @xmath29 on covariantly holomorphic quantities ( like @xmath30 ) . we will not consider vector multiplets and possible d - term extensions here . the relevant supersymmetry transformations ( of the gravitino and modulini ) are @xmath31 \delta\chi^{i}_{l } & = & \frac12 \gamma^{\mu } \partial_{\mu } z^{i}\epsilon_{r } - \frac12 g^{i\bar\jmath}{\rm e}^{k/2 } \overline{d_{j}w}\epsilon_{l } , \label{deltachi } \end{aligned}\ ] ] where @xmath32 is the @xmath33 khler connection . in the following we will consider flat space - like domain wall solutions of the supersymmetry conditions and of the supergravity equations of motion . for this reason we are going to use the following metric ansatz @xmath34 given the domain wall structure , we split the indices @xmath35 , @xmath36 . for the metric ( [ typei ] ) there is only one non - trivial component of the spin connection @xmath37 where @xmath38 , leading also to the ricci scalar ( which will become useful later ) @xmath39 . \label{scalari}\ ] ] plugging ( [ typei ] ) into the supersymmetry equations ( [ deltapsi])([deltachi ] ) and assuming that also the scalars @xmath40 depend only on the radial coordinate @xmath41 , one obtains @xmath42 \delta \psi_{i l } & = & \frac{a^\prime}{2 } \gamma_{\underline i}\gamma_{\underline r } \epsilon_l + \frac12 { \rm e}^{k/2 } w \gamma_i \epsilon_r \label{delpsiii } \\[2 mm ] \delta \chi_l & = & \frac12 \gamma^r z^{i\prime}(r ) \epsilon_r - \frac12 { \rm e}^{k/2 } g^{i\bar \jmath } \overline{d_j w } \epsilon_l , \label{delchii } \end{aligned}\ ] ] where @xmath43 was assumed . using also @xmath44 , one can start solving the various equations . from ( [ delpsiii ] ) , splitting the complex superpotential into norm and phase @xmath45 we get the projector condition @xmath46 from which the bps equation follows @xmath47 the condition ( [ proj0 ] ) can be assembled into a real projector @xmath48 ( satisfying @xmath49 ) acting on the full spinor as @xmath50 : @xmath51 . \label{projectori}\ ] ] the radial equation on the gravitino gives a differential equation on the spinor parameter @xmath52 here , the projector condition was used . moreover , one can use the projector condition inside ( [ diffeps ] ) , or take the charge conjugate ( using the majorana condition ) . the two resulting equations are @xmath53 \epsilon_r^{\prime } & = & \frac{i}{2}a_r^b \epsilon_r \pm \frac12 { \rm e}^{k/2 } |w| \epsilon_r , \label{majorana } \end{aligned}\ ] ] whose consistency implies @xmath54 the solution is given in terms of an arbitrary constant left - handed spinor parameter @xmath55 ( hence the 1/2 supersymmetry preserved ) : @xmath56 the bps equation ( [ bpsi1 ] ) has been used to express the integration over the superpotential in terms of the warp factor and ( [ bpsextra ] ) for the integration over @xmath57 . in the last formula we have used the definition of the charged conjugate spinor @xmath58 that follows from choosing the charge conjugation matrix @xmath59 , with @xmath60 . in this way the spinor @xmath61 satisfies the relation ( [ proj0 ] ) . finally , using again the projector condition in the modulini equations , one finds the bps equation for the scalar fields @xmath62 we can now also remove the dependence on the phase @xmath63 , by using the holomorphicity of the superpotential : @xmath64 this implies that the bps equations for the scalars can take the form : @xmath65 summarizing , the supersymmetry conditions that have to be satisfied in order to specify the domain wall solutions are @xmath66 z^{i \prime } & = & \mp 2 g^{i\bar \jmath } \partial_{\bar \jmath } |{\cal z}| . \end{array } $ } \label{summary}\ ] ] the other consistency condition we derived from the solution of the supersymmetry equations , namely ( [ bpsextra ] ) , is identically satisfied . this can be seen by using the fact that @xmath63 is the phase of the superpotential , and by equating with ( [ bpsextra ] ) the condition coming from its definition , @xmath67 . the resulting equation is an identity upon using ( [ summary ] ) . domain walls in @xmath3 supergravity were studied before in @xcite in a more restrictive setting , so we compare our results . first there is a change in the signature . this implies that one should change the sign for every upper index and put an @xmath68 for each gamma matrix in the supersymmetry equations . this can be re - absorbed in sending @xmath69 . then there is a difference in the metric . on the solution they have @xmath70 , but also their radial coordinate @xmath71 is related to ours by @xmath72 . also , the decomposition of the superpotential in norm and phase implies that @xmath73 . altogether the equations in @xcite become @xmath74 t^{\prime } & = & -\zeta { \rm e}^{k/2+i \theta } g^{t\overline t } \overline{d_t w } , \label{cv2 } \\[2 mm ] \theta^{\prime } & = & \frac{i}{2 } \left(t^{\prime } \partial_{t } k - \overline t^{\prime } \partial_{\bar t } k\right ) . \label{cv3 } \end{aligned}\ ] ] it is easy to see that these equations are equivalent to the ones presented above , when restricted to a single scalar field flowing . we find it useful to present the domain wall equations in the form of renormalization group equations . for this purpose we will describe the flow not as a function of @xmath41 coordinate but as a function of the warp factor . we multiply eq . ( [ bpsi2a ] ) on @xmath75 and find , using the eq . ( [ bpsi1 ] ) @xmath76 and @xmath77 we will find below that in certain cases the domain wall bps equations in the form ( [ summary ] ) are useful . sometimes we will use equations in the form ( [ renorm1 ] ) , particularly in the case of many moduli participating in the flow and @xmath28 not crossing zero . a careful inspection of the bps conditions ( [ summary ] ) shows that we can associate to the domain wall solutions a monotonic function , which therefore specifies the direction of the flow . as it is known since the analysis of the 5-dimensional domain walls and the interpretation of such solutions as renormalization group flows in the context of the ads / cft correspondence , there is a c - function that is given by the first derivative of the logarithm of the warp factor @xmath78 @xcite . by using the bps equations we easily see that @xmath79 where the last inequality follows from the properties of the scalar metric . there is , however , another interesting quantity that in most cases works as a c - function . this is the square of the covariantly holomorphic superpotential , which is related to the square of the mass of the gravitino at a generic point of the moduli space @xmath80 we can now look at the variations of @xmath81 along the flow . in order to compute it , we actually evaluate @xmath82 and use the bps equation ( [ bpsi2 ] ) . after some easy algebra we get that @xmath83 due to the positive definiteness of the metric of the scalar fields , this expression is clearly either always increasing , or always decreasing along the flow , depending on the sign of the bps equation , unless the flow crosses points where the superpotential vanishes . for these special cases the bps equations change sign as we cross the zeros of the superpotential and therefore @xmath81 is first decreasing until it reaches zero and then starts increasing again . the supersymmetric domain walls studied above are static 1/2 bps configurations of @xmath84 supergravity . in an analogous situation for static 1/2 bps cosmic strings in n=1 supergravity @xcite it was possible to rewrite the gravitational action , including the boundary terms , in the bps form . the action therefore was given by some squares of the bps conditions and total derivative terms . a saturation of the bps bound is equivalent to solving first order differential equations for unbroken supersymmetry . the on shell action therefore is equal , up to a sign , to the total energy of the string . therefore the tension of the string , the energy per unit area , can easily be derived from the bps action . the metric of the cosmic string is @xmath85 and the relation between the @xmath86 on shell and a tension @xmath87 is @xmath88 where @xmath89 and @xmath90 is a fi term in the action . for flat domain walls the metric is @xmath91 . therefore one would expect the following relation between the action and the tension ( the energy per unit of area ) @xmath92 in the case of cosmic strings @xmath93 and @xmath71 have flat metric , therefore the volume of the area is just @xmath94 . however , for our domain walls we have to distinguish the physical volume from coordinate volume and we have to ignore ( divide by ) the physical volume which is @xmath95 . in @xcite the tension of the bps domain walls was calculated using a rather complicated nester construction . the formula for the tension / energy density of the wall between two ads critical points is given for two cases : in one case the superpotential does not vanish between critical points , in the other case it vanishes at some intermediate point : * 2 ads vacua , @xmath96 in between @xmath97 * 2 ads vacua , @xmath98 in between @xmath99 here we show that for domain walls we can put the supergravity action in the standard bps form , namely as the integral of the square of the bps conditions plus boundary terms . we start with the gravitational action supplemented by the gaussian curvature at the boundaries . @xmath100 here @xmath101 is the metric at the boundaries . in our case the boundaries are at @xmath102 surface . therefore @xmath103 and @xmath104 where @xmath105 and @xmath106 and @xmath107 . one starts by plugging ( [ scalari ] ) in the einstein term and using that all the quantities depend only on the radial coordinate . then , we try to re - combine the various terms in the form of the bps equations ( [ bpsi1])([bpsi2a ] ) plus some boundary term . the action ( [ action1 ] ) becomes @xmath108\nonumber \\ & & + \int d^3 x \ , ( a^3)'| _ { r=+\infty } - ( a^3)'| _ { r=-\infty } ) \ .\end{aligned}\ ] ] this can now be recast in the form of bps equations squared plus additional boundary terms as @xmath109 \nonumber \\[2 mm ] & \mp & \int d^3 x \ , dr \left[2(a^3)^\prime |{\rm e}^{k/2}w| + a^3 z^{i\prime } \partial_i ( e^{k/2}w){\rm e}^{-i\theta}+a^3 \bar z^{\bar \imath}\bar \partial_i ( e^{k/2}\bar w){\rm e}^{i\theta}+ a^3 ( e^{k/2})^{\prime } |w|\right ] \label{bpsform0 } \end{aligned}\ ] ] finally , the second line becomes a total derivative by using ( [ holow ] ) , and the total action reads @xmath110 \nonumber \\[2 mm ] & \mp & \int d^3 x \ , dr \frac{d}{dr}\left[a^3\left(2 |{\cal z}|\right)\right ] , \label{bpsformi } \end{aligned}\ ] ] where we recovered the bps equations in the first line and we have a boundary term contribution in the second line . one can define the energy density functional @xmath111 , since we have a static solution , so that @xmath112 . if the superpotential does not vanish between the critical points we find @xmath113\nonumber\\ & & + \int d^3 x ( a^3 2 |{\cal z}| _ { r=+\infty}- a^3 2 |{\cal z}| _ { r=-\infty } ) . \label{bpsform2 } \end{aligned}\ ] ] in case that @xmath114 we have to make a choice of the sign in bps equations such that the tension is positive , as shown in eq . ( [ bpsform2 ] ) . thus when @xmath115 does not vanish between fixed points and the bps equation of motion are satisfied with the upper sign the energy density of the wall is given by @xmath116 note that @xmath117 . so our answer is exactly as in eqs . ( [ 1 ] ) , in agreement with @xcite . it was explained in @xcite that the bps domain walls with this tension are related to cdl bubbles with infinite radius . now we assume that there is a solution of the bps equations where the superpotential vanishes at some point @xmath118 between the ads critical points . the boundary integral in eq . ( [ bpsformi ] ) must be split into two terms since one part of the flow proceeds with one sign in the bps equations and for the other part , after @xmath119 point , the sign is flipped . therefore the answer for the energy density in this case is @xmath120 noting that @xmath117 , we have derived the correct tension as in eq . ( [ 2 ] ) in agreement with @xcite . this case , according to @xcite is not related to cdl tunneling . we would like to conclude this subsection by the observation that our new derivation of the bps domain wall tension follows the pattern discovered for the bps cosmic string in @xcite . we find the derivation using the bps form of the action significantly more elegant than using the nester construction . in this section we are going to give some examples of supersymmetric domain wall solutions for some of the generic models which capture the most important features of the landscape of flux vacua . we start with the simple scenario with just one khler modulus and all the complex structure moduli frozen : the kklt model @xcite . this model has one supersymmetric ads vacuum at a finite point of the volume modulus and a minkowski vacuum at infinite value of the volume modulus . we will solve the bps equations numerically and display the corresponding domain wall . after this simple example we show a more involved setup where also the complex structure moduli and the axion - dilaton are allowed to flow : the @xmath121 models . these include a wide variety of models ( and also the kklt model ) as a special case . the case which we will solve numerically has two different ads critical points at finite distance in the moduli space and three scalars are flowing between the critical points , two of them proportional to each other . finally we focus on a model where the effect of the khler deformations is neglected , but the form of the complex - structure deformations is more involved . the flux vacua in such models are described by a simple set of `` new attractor equations '' @xcite relating the points in the moduli space where the moduli are stabilized by fluxes . these flux vacua attractors have some resemblance as well as some differences with the black hole attractor equations @xcite . we will focus here on a model of a calabi yau threefold obtained as a hypersurface in @xmath122 . this model has two different ads critical points @xcite at finite distance in the moduli space . we will find out how four scalars in this model are flowing from the landau ginsburg ads critical point towards the near conifold region of the moduli space . the kklt model @xcite suggests a way to generate meta - stable de sitter vacua in string theory starting from supersymmetric anti de sitter minima with all moduli stabilized due to fluxes and using non - perturbative effects , such as gaugino condensation and/or d3-instantons . the uplifting of these stabilized ads vacua to ds vacua is possible under certain conditions by the addition of positive energy density sources , like anti - d3 branes . the simplest realization has a single volume modulus @xmath123 . after all the complex structure moduli have been fixed by the fluxes , the superpotential for the volume modulus ( whose khler potential is @xmath124 ) is @xmath125 with @xmath126 , @xmath127 and @xmath128 parameters depending on the details of the setup . the corresponding potential , before uplifting , has one ads supersymmetric critical point at finite volume and an asymptotic minkowski vacuum at infinity of the volume . it is easy to check that we can consistently set the axion @xmath129 to zero , since this is a fixed point for any value of the volume @xmath130 . for the sake of simplicity we will also take the parameters @xmath127 , @xmath128 and @xmath126 to be real ( and positive the first two @xmath131 , @xmath132 , negative the last one @xmath133 ) . the potential for this case and the covariantly holomorphic superpotential @xmath30 are given by fig . [ kkltpot ] as functions of the volume modulus . [ kkltpot ] ( blue upper curve ) and @xmath30 ( red curve below ) as functions of the modulus . one can see that the potential has an ads minimum at @xmath134 and tends to minkowski limit at large @xmath130 . @xmath30 is always negative in the interval between ads and minkowski vacua . potential energy density @xmath21 is shown in units @xmath135 , whereas the covariantly holomorphic superpotential @xmath30 is shown in units @xmath136.,title="fig : " ] one finds that for these parameters during the total flow the covariantly holomorphic superpotential @xmath30 is negative and approaches zero at infinity . we can reduce the problem to a one - dimensional gradient - flow , where the bps equations follow by using @xmath137 we solve numerically the flow equations in the opposite direction ( for technical reasons ) , starting with the ads minimum towards the infinitely far minkowski . for this purpose we have to use ( [ summary ] ) : @xmath138 \phi^\prime ( r ) & = & \partial_\phi |{\cal z}(r)|.\end{aligned}\ ] ] the ads critical point is at @xmath134 , and for values slightly above @xmath139 we get a domain wall solution that drives @xmath140 and @xmath141 . in this case we obviously have just one basin of attraction and the attractor point is the supersymmetric ads vacuum . asymptotically , at large @xmath41 , the value of @xmath130 grows , and @xmath115 becomes @xmath142 we can therefore solve the bps equations above in this limit by @xmath143 it should be noted that @xmath144 which means that the curvature of the space - time ( [ scalari ] ) vanishes asymptotically , as it should be in minkowski space . , the covariantly holomorphic superpotential @xmath30 ( in units @xmath145 ) , the warp factor @xmath128 and the curvature @xmath146 ( in units @xmath147 ) , all as functions of @xmath41 . here @xmath41 is the coordinate of the domain wall configuration given in eq . ( [ typei ] ) ; we show @xmath41 in units @xmath148 . , title="fig:"]0.5 cm , the covariantly holomorphic superpotential @xmath30 ( in units @xmath145 ) , the warp factor @xmath128 and the curvature @xmath146 ( in units @xmath147 ) , all as functions of @xmath41 . here @xmath41 is the coordinate of the domain wall configuration given in eq . ( [ typei ] ) ; we show @xmath41 in units @xmath148 . , title="fig : " ] 0.5 cm , the covariantly holomorphic superpotential @xmath30 ( in units @xmath145 ) , the warp factor @xmath128 and the curvature @xmath146 ( in units @xmath147 ) , all as functions of @xmath41 . here @xmath41 is the coordinate of the domain wall configuration given in eq . ( [ typei ] ) ; we show @xmath41 in units @xmath148 . , title="fig:"]0.5 cm , the covariantly holomorphic superpotential @xmath30 ( in units @xmath145 ) , the warp factor @xmath128 and the curvature @xmath146 ( in units @xmath147 ) , all as functions of @xmath41 . here @xmath41 is the coordinate of the domain wall configuration given in eq . ( [ typei ] ) ; we show @xmath41 in units @xmath148 . , title="fig : " ] in fig . [ kkltabcd ] we plot the wall : first on the top right we have a scalar interpolating between the ads value @xmath134 and infinity . on the top right we plot @xmath30 as the function of wall coordinate @xmath41 . it is always negative but here we see clearly that it takes a constant negative value at ads and starts interpolating towards zero at minkowski vacuum . the next plot on the bottom left shows the warp factor @xmath149 . finally on the right bottom we show the curvature of the domain wall : it is vanishing at large values of modulus in the flat space and grows towards the decreasing values of modulus and sharply drops down to its negative ads value upon reaching the ads vacuum . from this example we learn that in the absence of a barrier between the ads and minkowski vacua there exists a static domain wall where the flow of the volume modulus from the infinite value at minkowski towards the finite value at ads is controlled by 1/2 of unbroken supersymmetry . the tension of this wall is equal to @xmath150 where @xmath151 is the cosmological constant of the ads minimum . the previous example had only one ads vacuum . however , in the landscape of flux compactifications there are also examples with multiple ads vacua and therefore it is interesting to display an example with multiple basins of attraction to different ads vacua and the supersymmetric domain wall solutions interpolating between them . in order to find a workable example , we need an analytic superpotential admitting multiple ads vacua . such an example is provided by various models of toroidal orbifolds analyzed in @xcite . the generic moduli content is given by the axio / dilaton @xmath152 and by the complex - structure moduli @xmath153 and 3 khler moduli @xmath154 , with khler potential @xmath155 for the sake of simplicity we focus on the case of one untwisted complex structure modulus , as for instance in the @xmath156 , @xmath157 orbifolds or in the @xmath158 orbifold with @xmath159 , @xmath160 , @xmath161 or @xmath162 lattice . the superpotential is a polynomial in the complex - structure and dilaton moduli that comes from the fluxes , added by some non - perturbative exponential terms involving the khler - structure moduli . if we consider the cases with just one complex structure modulus , the full untwisted superpotential reads @xcite @xmath163 here the parameters @xmath164 and @xmath165 are related to the fluxes ( they are the coefficients of the real sections of the 3-cycles ) , while the parameters @xmath166 and @xmath167 are to be associated with non - perturbative effects . this potential includes the one - modulus kklt scenario as a special example . we can then look for supersymmetric vacua , satisfying @xmath168 , @xmath169 and @xmath170 . for simplicity , let us consider the case that @xmath171 , @xmath172 and look for vacua where @xmath173 . then , the supersymmetry conditions become @xmath174 b+ d\ , s & = & \frac{1}{u+\bar u } w , \label{duw } \\[2 mm ] - gh\ , { \rm e}^{-h t } & = & \frac{1}{t+\bar t } w. \label{dtw } \end{aligned}\ ] ] a generic solution for @xmath175 of these equations requires that @xmath176 and if we further look for the case where im@xmath177 = im@xmath152 = 0 , then we also obtain that @xmath178 , \label{cond1 } \\[2 mm ] { \rm re}\,t & = & \frac{3c{\rm re}\,s(b + d\ , { \rm re}\,s)}{h(a b- cd \,{\rm re}\,s^2)}. \label{label3 } \end{aligned}\ ] ] when the values of the fluxes are chosen appropriately , one can obtain 2 distinct vacua at values of @xmath179 and @xmath180 . one instance is the choice @xmath181 . the critical points are at @xmath182 and @xmath183 . ( the value of the size of the manifold is tuned by @xmath184 . the smaller @xmath101 , the bigger @xmath93 . increasing @xmath185 on the other hand allows for smaller values of @xmath186 . ) we can then use the bps equations to obtain the interpolating solution between them . having seen how the critical points are obtained , we would like to reduce the problem to a two - dimensional system depending only on @xmath187 , @xmath188 . both critical points lie on the section of the moduli space where all the axions are set to zero . this condition can be kept consistently all along the flow . when the axions are vanishing , @xmath189 and therefore @xmath190 . this condition can be preserved along the flow as @xmath191 if the axions are vanishing . this also implies , since the metric is diagonal , that the real and imaginary parts decouple consistently . given the symmetry of the potential , we can also assume that @xmath173 all along the flow . the bps equations therefore reduce to @xmath192 t^{\prime } & = & \mp 8t^2 \partial_t |{\cal z}|(s , t , u ) , \label{boh } \\[2 mm ] u^{\prime } & = & \mp 8u^2 \partial_u |{\cal z}|(s , t , u ) , \nonumber \end{aligned}\ ] ] and , by using ( [ relus ] ) , one can prove that the differential equations for @xmath193 and @xmath186 are proportional to each other . this means that we can further reduce the problem to the description of the gradient flows for the flow potential @xmath194 , where @xmath195 and @xmath196 . in these variables , the bps equations become @xmath197 \displaystyle \frac{a^{\prime}}{a } & = & \pm |{\cal z}|(\phi_i ) \end{array } \right.\label{bpseff}\ ] ] where the absolute value of the central charge reads @xmath198 notice the double exponentials due to the non - perturbative terms in the khler moduli . given this flow potential , we can look for the special case where we have two distinct critical points , for instance when the parameters are chosen as above : @xmath181 . for these parameter values ( [ z ] ) shows two supersymmetric ads critical points with different values of the cosmological constant and an asymptotically minkowski vacuum . although the bps equations can be written in a closed form , the integration can only be done numerically . in fig . [ fig : contplotw ] we show the contour plot of @xmath30 , where the @xmath199 modulus ( proportional to the complex structure and to the dilaton @xmath195 ) is on the horizontal axis and @xmath200 , proportional to the khler modulus @xmath196 is on the vertical one . and the gradient flow between the two critical points . @xmath201 is everywhere positive between ads critical points , it vanishes only at the asymptotically minkowski vacuum . ] we plot the total potential on fig . [ fig : stupot ] . . ] we can clearly see from the pictures that there are two distinct critical points , and we have plotted the numerical solution interpolating between them . as it is clear from the picture , the two critical points have different behavior . using the nomenclature coming from 5d , one is approached by the flow in an ultra - violet direction ( i.e. the warp factor goes as @xmath202 , for @xmath203 ) , the other in an infra - red ( i.e. the warp factor goes as @xmath204 for @xmath205 ) . an ultra - violet critical point is a minimum of the absolute value of the superpotential ( minimum with positive @xmath28 or maximum with negative @xmath28 ) , an infra - red critical point is a maximum . this implies that there is no need for the superpotential to cross zero . the warp factor indeed shows no critical points , as it always increases along the flow . the superpotential will necessarily cross zero along the flow and therefore the bps equations will change sign only when the interpolation must be done between two infra - red ( two ultra - violet critical points are excluded as the wall should be singular to interpolate between them ) . this is the only way to obtain a smooth domain wall interpolating between them . a similar phenomenon occurred in the realization of the randall sundrum scenario with a thick brane in @xcite . we will have examples of such domain walls in sec . [ sec : subcritical ] . from fig . [ fig : contplotw ] we can also clearly see the two basins of attraction for the ads and dine - seiberg minkowski critical points and the ads repeller ( with a marginal line of attraction separating the two basins ) . this shows that there is one main basin of attraction for the ads vacuum with the larger absolute value of the cosmological constant . there is a line of attraction for the other critical point , which is really a saddle point of the flow potential . finally there is another basin of attraction for the values of the moduli that leads to the decompactification of the space . the size modulus @xmath93 runs to infinity at the same time as the dilaton goes to zero , so we get a decompactification of the space in the weak coupling regime . this example shows that the landscape attractors are different from the black - hole attractors . for the latter , supersymmetry vacua are always minima of the superpotential @xmath28 and therefore one always has clear distinct basins of attractions and there are no repeller - like vacua . moreover , if one is looking at the 5-dimensional case there is always only one basin of attraction for the vector - multiplet moduli . the panorama of the landscape can be further refined by considering a representative example in the class of type ii compactifications on calabi yau threefolds , defined as hypersurfaces in a weighted projective space . when fluxes are active for the rr and ns three - form field strengths on the orientifold of these hypersurfaces , a non - trivial superpotential is induced for the dilaton and complex structure moduli , leading to the many interesting physical properties that have driven a massive research activity ( for a review and bibliography see @xcite ) . these models are particularly relevant even for a single deformation parameter because of the results obtained in @xcite showing that , quite unexpectedly , the density of flux vacua is not uniform in moduli space , but rather it peaks around the conifold singularity , @xmath206 . indeed , upon extensive numerical computations , it has been shown that the conifold region plays the role of attractor " basin and it serves as accumulation region for flux vacua . this supports the analytic predictions made in @xcite . on the other hand , also the landau ginzburg singularity at @xmath207 enjoys very peculiar properties , deriving from it being a fixed point of the modular group . in general , it is the special point in moduli space where the number of vacua with vanishing superpotential @xmath208 and discrete symmetries can be of the same order as the total number of vacua ( when non - zero ) @xcite . for our purposes , it will be convenient for computations because of our need to use perturbative expansions for the cy periods , that are known around this point @xcite . thus , in spite of neglecting the khler structure moduli that in principle would be present in these systems , this case is particularly interesting as a representative of the large amount of possible flux vacua that are making the landscape so rich . another important related feature is that one may encounter multiple attraction basins @xcite . these consist of distinct supersymmetric vacua that can be found within a finite region in moduli space , or `` area code '' , for fixed values of the @xmath209 and @xmath210 fluxes . we thus proceed to identify such a situation and to construct the domain wall solution interpolating between different area codes . as a start , we quickly recall some necessary formulas , while referring for instance to @xcite for the relevant lore and more conventions . one examines in general the @xmath211 complex structure deformations . as usual , one chooses a symplectic basis @xmath212 for the @xmath213 homology cycles ( @xmath214 ) , with dual cohomology elements @xmath215 such that @xmath216 and one assembles into a symplectic vector the periods of the unique holomorphic three - form @xmath217 , @xmath218 the @xmath219 are projective coordinates on the complex structure moduli space and @xmath220 , where @xmath221 is the holomorphic prepotential . the khler potential for the complex - structure and axion - dilaton moduli space is given by @xmath222 where @xmath223 is the axion - dilaton and @xmath224 . the superpotential induced by non - trivial 3-form fluxes reads @xmath225 where @xmath226 and @xmath227 are the integer valued flux components @xmath228 we will take for convenience @xmath229 . let us consider in the following @xmath230 to be the calabi yau locus defined by the equation @xmath231the supersymmetric vacua are found by solving the f flatness " condition , that is a ( khler covariantized ) extremization with respect to @xmath223 and the single complex - structure modulus @xmath232 , @xmath233 leading to the equivalent conditions @xmath234 alternatively , one can solve the flux vacua attractor equations @xcite . one finds that the moduli space of this model presents landau ginzburg , conifold and large complex structure singular points respectively at @xmath207 , @xmath206 and @xmath235 @xcite . the monodromy group , @xmath236 , of the complex structure moduli space has two generators : @xmath127 , which generates phase rotations @xmath237 with @xmath238 around the landau ginzburg point at @xmath239 , and @xmath177 which corresponds to the logarithmic monodromy @xmath240 around the conifold singularity @xmath241 . by itself , @xmath127 generates a @xmath242 subgroup , with an associated fixed point at @xmath239 ; @xmath177 , on the other hand , is of infinite order . the monodromy matrix @xmath127 generates rotations by a root of unity around @xmath243 : @xmath244 and is explicitly given by @xmath245 the periods admit , for @xmath246 , an expansion in a picard fuchs basis @xmath247 where @xmath129 is the @xmath248 root of unity @xmath249 the usual choice for the periods in this basis appropriate for the polynomial m is @xmath250 . then , the periods in the symplectic basis @xmath251 are related to the picard fuchs periods by a linear transformation @xmath252 the matrix of transformation from the picard fuchs to the symplectic basis has rational elements that are given by @xmath253 quite generally , in the symplectic basis the periods have the expansion @xmath254 where the @xmath255 are complex constants and @xmath256 are constant four - vectors with components that are rational linear combinations of powers of @xmath129 . as a further computational tools , we note in passing that for this model the meijer periods ( we find them following method developed in @xcite for quintic in the context of black hole attractors ) are given by the following expression in mathematica notations @xmath257,\\ & u^{-}_{1}(z ) & = { \rm meijerg}\left[\left\{\left\{\frac16,\frac26,\frac46,\frac56\right\},\{\}\right\},\{\{0,0\},\{0,0\}\},z\right],\\ & u^{-}_{2}(z ) & = { \rm meijerg}\left[\left\{\left\{\frac16,\frac26,\frac46,\frac56\right\},\{\}\right\},\{\{0,0,0\},\{0\}\},-z\right],\\ & u^{-}_{3}(z ) & = { \rm meijerg}\left[\left\{\left\{\frac16,\frac26,\frac46,\frac56\right\},\{\}\right\},\{\{0,0,0,0\},\{\}\},z\right ] , \end{aligned}\ ] ] where @xmath258 are valid for im@xmath259 . the transformation to the picard fuchs basis is given by the matrix @xmath260 so that @xmath261 with @xmath262 for im@xmath263 . we now proceed to specifying the vacua for given values of the fluxes . let us take , as a convenient choice among the solutions of the f - flatness conditions , the flux components to be @xmath264 and @xmath265 the virtue of this set of numbers is that the consequent vacua fall well within the convergence region of the picard fuchs periods , thus making the numerical computations for the flow somewhat easier . it turns out that these fluxes give rise to two vacua : the first vacuum is a maximum at the landau ginzburg point , with @xmath266 the critical value of the potential is @xmath267 . we find the following diagonalized mass matrix for this vacuum @xmath268 the second vacuum is a saddle point close to the conifold point , with @xmath269 and with potential equal to @xmath270 . in this case the diagonalized mass matrix is @xmath271 the interpolating solution going from one vacuum to the other is found by solving the flow equations ( [ renorm1 ] ) , that become @xmath272 where we have used for convenience the variable @xmath273 . the resulting flows are displayed below on the @xmath232 complex plane ( figure [ plot1a ] ) and on the @xmath223 complex plane ( figure [ plot2 ] ) . complex plane . ] complex plane . ] the corresponding flow of the potential @xmath274 as a function of @xmath275 is shown in fig . [ plot3 ] . it shows clearly that there is no barrier . . ] these domain walls are associated to flows that go from a maximum in the landau ginzburg point , towards the conifold which is a saddle point . these are explicit examples of critical points which may be sometimes repellers ( for saddle points or maxima of the potential ) and sometimes attractors , in particular stable attractors in case of minima of the potential . in this case , the flow tends to escape from the lg point towards the conifold . we are confident that these results on interpolating domain walls are common to other one - parametric calabi yau models that have been studied in @xcite and generically have similar distribution of vacua on moduli space . when the vacua are separated by a barrier , one can expect to find an absolute minimum as well as a meta - stable local minimum . this class of models in the stringy landscape was suggested in @xcite : in addition to the ads minimum and to the asymptotically minkowski vacuum , there is also a supersymmetric minkowski minimum at some finite value of the volume modulus . the main motivation for the new class of models with such a minimum was to have more flexibility with the gravitino mass , and to avoid certain cosmological problems specific to the standard kklt scenario . here we will show that the same class of models ( which we will call kl models ) allows two ads minima , both at finite value of the volume modulus . moreover , we will find two types of flows here : depending on the choice of parameters the superpotential may never vanish between the two ads vacua , or it may vanish at some point and therefore change the sign along the flow . this results in several different types of bps domain walls in the kl models . the model has one modulus as in the kklt case , but a more involved superpotential of the racetrack type @xcite . the superpotential is a modification of the kklt model by an additional exponential term @xmath276 this model is a simple instance of multiple basins of attraction . it has one supersymmetric minkowski vacuum at fixed volume , one supersymmetric ads vacuum also at fixed volume and one supersymmetric minkowski asymptotic vacuum at infinite volume as shown in an upper curve in figure [ klpot ] . the plot of the potential clearly shows the barrier between the left minkowski minimum and the ads minimum , the superpotential is negative everywhere away from minkowski vacua . [ h ! ] , blue upper curve , and the covariantly holomorphic superpotential @xmath30 , the red curve below , as the functions of the modulus @xmath130 for the kl model . the potential is shown in units @xmath277 , and @xmath30 is in units @xmath278 . both @xmath21 and @xmath30 have extrema at the same point , which is a reflection of the fact that supersymmetry is unbroken in the minima of @xmath21 . this plot is for @xmath131 , @xmath279 , @xmath280 , @xmath281 and @xmath282.,title="fig : " ] for the choice @xmath131 , @xmath279 , @xmath280 , @xmath281 and @xmath282 , also using once more the definition @xmath283 , the vacua are at @xmath284 , @xmath285 and @xmath286 , with @xmath287 . therefore , one can construct two interpolating domain walls . one is between the ads minimum and the local minkowski minimum , which we will call m@xmath288 . another wall is between the ads minimum and the minkowski extremum at infinity , m@xmath289 . here we study only the domain walls interpolating between m@xmath288 and ads , see figure [ plotx ] . the features of the domain wall between m@xmath289 and ads minimum in this model will be the same as the ones studied in kklt case in section [ sec : kkltmodel ] . and ads . we plot the scalar @xmath130 , @xmath30 ( in units @xmath290 ) , warp factor @xmath128 and the curvature scalar @xmath146 ( in units @xmath277 ) as functions of the coordinate @xmath41 . the curvature @xmath146 is equal to zero in the minkowski vacuum , has a peak at the domain wall , and approaches a constant negative value in the ads vacuum.,title="fig:"]0.5 cm and ads . we plot the scalar @xmath130 , @xmath30 ( in units @xmath290 ) , warp factor @xmath128 and the curvature scalar @xmath146 ( in units @xmath277 ) as functions of the coordinate @xmath41 . the curvature @xmath146 is equal to zero in the minkowski vacuum , has a peak at the domain wall , and approaches a constant negative value in the ads vacuum.,title="fig : " ] 0.7 cm and ads . we plot the scalar @xmath130 , @xmath30 ( in units @xmath290 ) , warp factor @xmath128 and the curvature scalar @xmath146 ( in units @xmath277 ) as functions of the coordinate @xmath41 . the curvature @xmath146 is equal to zero in the minkowski vacuum , has a peak at the domain wall , and approaches a constant negative value in the ads vacuum.,title="fig:"]0.5 cm and ads . we plot the scalar @xmath130 , @xmath30 ( in units @xmath290 ) , warp factor @xmath128 and the curvature scalar @xmath146 ( in units @xmath277 ) as functions of the coordinate @xmath41 . the curvature @xmath146 is equal to zero in the minkowski vacuum , has a peak at the domain wall , and approaches a constant negative value in the ads vacuum.,title="fig : " ] in figure [ plotx ] we plot on the top left the scalar potential @xmath21 , and on the right the covariantly holomorphic superpotential @xmath30 , which vanishes at m@xmath288 and stays negative all the way towards ads . the warp factor on the bottom left starts as a constant at m@xmath288 and interpolates into an exponent of @xmath41 near the ads . finally , the curvature jumps from zero at m@xmath288 to some positive value and falls down to a negative constant at ads . it is interesting that the curvature of the wall grows significantly between zero at m@xmath288 before becoming negative at ads . now we are going to consider the models with two ads minima . this can be done , e.g. , by taking the kl model with a supersymmetric minkowski minimum m@xmath291 , and change any of the parameters of this model . in particular , as one can see from fig.[fig : klppota ] , one can obtain an ads minimum from the minkowski minimum by a slight modification of a single parameter in the previous model : we changed @xmath292 to @xmath293 . using once more the definition @xmath283 , the ads vacua are at @xmath284 , @xmath294 and @xmath295 , with @xmath296 , and @xmath297 for @xmath284 . the potential @xmath21 ( upper curve ) and the covariantly holomorphic superpotential @xmath30 ( lower curve ) for the kl model as function of the modulus are shown in figure [ fig : klppota ] , for the choice @xmath298 , @xmath279 , @xmath280 , @xmath281 and @xmath282 . @xmath30 is always negative . ( upper curve , in units @xmath277 ) and the covariantly holomorphic superpotential @xmath30 ( lower curve , in units @xmath278 ) for the kl model , as function of the modulus . notice that we took @xmath298 instead of @xmath292 , all other parameters being the same as for the potential in fig . [ klpot].,title="fig : " ] we show various features of the kl wall between the two ads vacua in fig . [ fig : plota ] . we plot the scalar and the covariantly holomorphic superpotential , which is a negative constant near the first ads vacuum and interpolates towards another negative value when it approaches another ads vacuum . the warp factor @xmath128 starts as a decreasing exponent and then changes into a faster decreasing exponent of @xmath41 near the second ads . the curvature @xmath146 jumps from some constant negative value near the first ads vacuum , becomes positive near the domain wall , and then falls down to a negative constant near the second ads vacuum . and we plot the scalar @xmath130 and the covariantly holomorphic superpotential @xmath30 ( in units @xmath290 ) , which is always negative , warp factor @xmath128 , and the curvature scalar @xmath146 ( in units @xmath277 ) , as a function of the coordinate @xmath41 . the curvature @xmath146 has a peak at the domain wall , and approaches different constant negative values in two different ads minima.,title="fig:"]0.5 cm and @xmath299 . we plot the scalar @xmath130 and the covariantly holomorphic superpotential @xmath30 ( in units @xmath290 ) , which is always negative , warp factor @xmath128 , and the curvature scalar @xmath146 ( in units @xmath277 ) , as a function of the coordinate @xmath41 . the curvature @xmath146 has a peak at the domain wall , and approaches different constant negative values in two different ads minima.,title="fig : " ] 0.7 cm and @xmath299 . we plot the scalar @xmath130 and the covariantly holomorphic superpotential @xmath30 ( in units @xmath290 ) , which is always negative , warp factor @xmath128 , and the curvature scalar @xmath146 ( in units @xmath277 ) , as a function of the coordinate @xmath41 . the curvature @xmath146 has a peak at the domain wall , and approaches different constant negative values in two different ads minima.,title="fig:"]0.5 cm and @xmath299 . we plot the scalar @xmath130 and the covariantly holomorphic superpotential @xmath30 ( in units @xmath290 ) , which is always negative , warp factor @xmath128 , and the curvature scalar @xmath146 ( in units @xmath277 ) , as a function of the coordinate @xmath41 . the curvature @xmath146 has a peak at the domain wall , and approaches different constant negative values in two different ads minima.,title="fig : " ] the tension of this bps wall is given by eq . ( [ tension1 ] ) . @xmath300 it is critical in a sense explained in @xcite : the tension of such bps brane corresponds to a limit of the cdl bubble solution @xcite with infinite radius . when such bps wall exist there is no tunneling since these supersymmetric walls saturate the cdl bound . we will discuss this issue later when we will study tunneling , see section [ decay ] . in the previous subsection we considered a kl model with two ads minima and the superpotential which did not change sign on the way from one minimum to another . now we will construct a kl model with two ads minima and superpotential vanishing at some point between them ( fig . this can be done by a small deviation of one of the parameters : instead of taking @xmath292 as in original kl model , or @xmath301 , as in the previous subsection with two ads vacua , it is sufficient to take @xmath302 . for example , for the choice @xmath303 , @xmath279 , @xmath280 , @xmath281 and @xmath282 , also using once more the definition @xmath283 , one finds that the ads vacua are at @xmath284 , @xmath304 , with @xmath305 , and at @xmath284 , @xmath306 , with @xmath307 . the covariantly holomorphic superpotential @xmath30 is positive near the ads minimum at @xmath308 . it crosses zero at @xmath309 and becomes negative when approaching the second ads minimum at @xmath306 . it approaches zero at infinite values of the volume modulus . ( upper curve , in units @xmath310 ) and the covariantly holomorphic superpotential @xmath30 ( lower curve , in units @xmath311 ) for the kl model as function of the modulus . @xmath30 is positive at @xmath308 and negative at @xmath312 . here we took @xmath303 instead of @xmath292 , all other parameters being the same as for the potential in fig . [ klpot].,title="fig : " ] the corresponding bps solution is shown in fig . [ plotb ] . we plot on the top left the scalar field @xmath130 , and on the right the covariantly holomorphic superpotential , which changes the sign on the solution . the warp factor on the bottom left starts as a growing exponent near the first ads minimum and changes into a decreasing exponent of @xmath41 near the second minimum . finally , the curvature jumps from some constant negative value near the first ads vacuum , becomes positive near the domain wall , and then falls down to a negative constant near the second ads vacuum . as far as we know , this solution provides the first explicit example of a smooth bps domain walls in @xmath313 with the superpotential vanishing between the critical points . changes the sign on the domain wall solution . we plot the scalar @xmath130 , the covariantly holomorphic superpotential @xmath30 ( in units @xmath290 ) , the warp factor @xmath128 , and the curvature @xmath146 ( in units @xmath277 ) . notice the characteristic peak of the scale factor at the point where @xmath30 changes sign.,title="fig:"]0.5 cm changes the sign on the domain wall solution . we plot the scalar @xmath130 , the covariantly holomorphic superpotential @xmath30 ( in units @xmath290 ) , the warp factor @xmath128 , and the curvature @xmath146 ( in units @xmath277 ) . notice the characteristic peak of the scale factor at the point where @xmath30 changes sign.,title="fig : " ] 0.3 cm changes the sign on the domain wall solution . we plot the scalar @xmath130 , the covariantly holomorphic superpotential @xmath30 ( in units @xmath290 ) , the warp factor @xmath128 , and the curvature @xmath146 ( in units @xmath277 ) . notice the characteristic peak of the scale factor at the point where @xmath30 changes sign.,title="fig : " ] changes the sign on the domain wall solution . we plot the scalar @xmath130 , the covariantly holomorphic superpotential @xmath30 ( in units @xmath290 ) , the warp factor @xmath128 , and the curvature @xmath146 ( in units @xmath277 ) . notice the characteristic peak of the scale factor at the point where @xmath30 changes sign.,title="fig : " ] the tension of this bps wall is given by eq . ( [ tension2 ] ) . @xmath314 it is super - critical in the sense explained in @xcite . the tension of such bps walls exceeds the cdl bound and therefore the tunneling is super - suppressed . in previous sections we studied domains with unbroken supersymmetry separated by static bps domain walls . in our world supersymmetry is broken , but if supersymmetry breaking is not very large as compared to other parameters , one may hope that our domain wall solutions will provide an approximate description of slowly moving near - bps domain walls . just as in the situation with the near - extremal black holes , one may hope the physics of the near - bps walls can be under a much better theoretical control than the physics of generic domain walls . it would be especially interesting to see whether the bps domain walls may play any role in the cosmological context , dividing the universe into different regions separated by static domain walls . however , from the point of view of the friedmann cosmology , the universe with a negative cosmological constant @xmath315 should collapse to a cosmological singularity within a very short time @xmath316 . therefore in the usual cosmological context one can hardly describe coexisting ads domains separated by static domain walls . one of the most interesting applications of the results obtained in the last section is the theory of vacuum decay and bubble formation . it is well known that ds vacua obtained in the kklt construction by uplifting of ads vacua are unstable with respect to the decay to the minkowski vacuum corresponding to infinitely large values of the volume modulus @xmath130 , i.e. to a supersymmetric 10d vacuum @xcite . according to @xcite , the decay probability is given by @xmath317 where @xmath318 is the euclidean action for the tunneling trajectory , and @xmath319 is the euclidean action for the initial metastable configuration @xmath320 . the euclidean action calculated for the metastable vacuum ds solution @xmath321 is given by @xmath322 this action has a simple sign - reversal relation to the entropy of de sitter space @xmath323 : @xmath324 therefore the decay time of the metastable ds vacuum @xmath325 can be represented in the following way : @xmath326 here @xmath327 is the so - called recurrence time . it was shown in @xcite that if the decay of the metastable ds vacua occurs due to the tunneling through the barrier with a positive scalar potential , then @xmath318 is always negative , and therefore the decay always happens during the time shorter than the recurrence time @xmath328 . this result directly applied to the simplest kklt model where the tunneling occurs through the positive barrier separating the metastable ds vacuum and the supersymmetric 10d vacuum . however , the situation with the tunneling to the minima with a negative vacuum density remained less clear ; see refs . @xcite for the recent discussion . in what follows we will examine this issue using the results obtained in the previous section . our consideration will be based on the observation of ref . @xcite that in all cases when @xmath329 does not vanish across the domain wall , the domain wall solutions can be represented as a limiting configuration describing the wall of the cdl bubbles of an infinitely large radius . for such bubbles , the tunneling action is also infinitely large , and the vacuum decay is impossible . in our context , this fact is related to the supersymmetry of the different vacua @xcite , and of the interpolating bps wall solution . however , in realistic situations supersymmetry is broken . for example , in the kklt construction the metastable ds vacuum state appears after the uplifting of a supersymmetric ads vacuum . uplifting can be achieved , e.g. , by adding a @xmath330 brane contribution , which results in a supersymmetry breaking in the ds vacuum . similarly , in the kl model @xcite one must uplift a minkowski vacuum or an ads vacuum to a ds vacuum , thus breaking the supersymmetry . as we will see , this may lead to a relatively rapid decay of the uplifted ds vacuum due to creation of the bubbles describing a collapsing universe with negative vacuum energy corresponding to the ads vacuum . note that in terms of a canonically normalized volume modulus @xmath130 , which we will use hereafter , the @xmath330 brane contribution looks as @xmath331 , where @xmath332 is some constant . ( the canonically normalized field , @xmath333 , is obtained by dividing the field @xmath130 used in the previous section by the factor @xmath334 . ) this contribution rapidly decreases at large @xmath130 . therefore this term may significantly uplift the ads ( or minkowski ) minimum at small values of @xmath130 , but it makes a much smaller effect on the ads minima at large @xmath130 . this suggests that the transitions should typically occur towards larger values of the volume of the compactified space . first of all , consider the kl potential with the supersymmetric minkowski vacuum , and uplift it by a tiny amount to make the vacuum energy there equal to the present cosmological constant , @xmath335 , in planck energy density units . this should make this ds vacuum metastable , and lead to a vacuum decay due to creation of bubbles with negative vacuum energy . . ] to simplify the investigation of its decay , we will consider potentials such that the difference between the depth of the two minima is much smaller than the height of the barrier . an example of such a potential is provided by the kl model with parameters @xmath336 , @xmath337 , @xmath338 , @xmath339 and @xmath340 . for these values of parameters the potential has a supersymmetric minkowski minimum at @xmath341 and an ads minimum at @xmath342 . if one slightly changes the parameters , e.g. takes @xmath343 , then the minkowski minimum becomes an ads minimum . if one uplifts the potential by adding a term @xmath331 , the potential acquires a ds minimum , as shown in fig . [ klthin ] . during all of these changes , the second ads minimum changes only slightly . one expects that the thin wall approximation should be valid for the description of tunneling in such potentials . one part of this condition is automatically satisfied for the near - bps bubbles . indeed , by considering sufficiently small uplifting one can always make such bubbles arbitrarily large without changing the thickness of the wall . however , in order to use the thin - wall approximation as formulated in ref . @xcite , one should also check that the difference between the depth of the two minima is much smaller than the height of the barrier , and that the width of the wall of the cdl bubble remains much smaller than the radius of the bubble even if one turns off the gravitational interaction . we checked numerically that this condition is also satisfied for our potential . having in mind potentials of this type , we will use the thin - wall approximation to investigate the decay of the ds vacuum with the field @xmath308 and vacuum energy density @xmath344 to the state with the field @xmath312 with a negative vacuum energy density @xmath345 . we will be interested in the realistic situation @xmath346 , as illustrated by fig . [ klthin ] . examples of the vacuum decay considered in @xcite described transitions between ds and minkowski space , and decay of the minkowski vacuum . we will use the results obtained by parke @xcite , who generalized the results of @xcite for the tunneling between the states with arbitrary values of @xmath347 in the thin - wall approximation . according to @xcite , the bubble size is given by @xmath348 where @xmath349 is the bubble radius in the thin wall approximation ignoring gravity , @xmath111 is the bubble wall tension , and @xmath350 ( it can be positive or negative ; in our case it is negative ) , and @xmath351 in the thin - wall approximation , the tension of the wall can be approximated by the integral @xmath352^{1/2 } \ .\ ] ] here @xmath353 is the point near @xmath312 where @xmath354 . @xcite uses a slightly different prescription , which is equivalent to ours in the limit when the height of the barrier is much greater than the difference between @xmath347 . ) in the limit when the radius of the cdl bubble becomes infinitely large , the wall tension is known exactly : @xmath355 see @xcite , and also our derivation of this result in sect . 2 , eq . ( [ tension1 ] ) . using the variables @xmath356 one can represent eq . ( [ rad ] ) in the following convenient form : @xmath357 for a particular case of tunneling from minkowski space ( @xmath358 ) , the infinite bubble size regime ( the threshold regime for the tunneling ) corresponds to @xmath359 , @xmath360 , when the denominator vanishes . the tunneling probability is given by @xmath361 , where @xmath362 here the first term is the no - gravity result , and the gravitational correction factor @xmath363 is given by @xcite @xmath364}{x^2(y^2 - 1)(1 + 2xy+x^2)^{\frac{1}{2 } } } \ . \label{gravfactor}\ ] ] this factor also blows up at @xmath359 , @xmath360 , i.e. the tunneling becomes forbidden . now let us uplift the minkowski vacuum , assuming that the ads vacuum at @xmath312 remains approximately at the same height @xmath365 . as we already argued , this is indeed the case if the ads vacuum is at large values of the volume modulus @xmath130 . uplifting changes not only @xmath366 , but also the bubble wall tension @xmath111 . typically uplifting decreases the wall tension , which increases the probability of the tunneling ; a sufficiently large uplifting may remove the wall altogether , see @xcite . any small uplifting makes the tunneling from @xmath308 to @xmath312 possible . in order to study the size of the bubbles and the probability of the tunneling , let us change @xmath367 and @xmath368 by @xmath369 and @xmath370 . in the linear approximation @xmath371 and both depend on the small uplifting @xmath372 . in general , one would expect that the result should depend both on @xmath369 and @xmath370 . however , we find that in the leading approximation in @xmath369 and @xmath370 , the results of the expansion near @xmath359 , @xmath360 depend only on @xmath370 , i.e. they are not sensitive to the small changes of the bubble wall tension . in particular , the denominator in ( [ size ] ) is equal to @xmath373 , where @xmath374 , for @xmath346 . this yields @xmath375 here @xmath376 is the hubble constant in ds space with vacuum energy density @xmath377 . thus , instead of an infinite size bubble , we have a bubble with the size equal to the size of the ds horizon ! now we will perform a similar investigation for the tunneling rate . by considering the function @xmath378 in eq . ( [ gravfactor ] ) and expanding it with respect to @xmath369 and @xmath370 , one finds that in the limit @xmath346 @xmath379 where @xmath380 is the entropy of ds space , @xmath381 this means that the tunneling probability is given by the universal model - independent equation which has a simple geometric interpretation in terms of ds entropy : @xmath382 several comments are in order . first of all , we have found that the decay rate of the vacuum obtained by uplifting of a supersymmetric minkowski minimum does not depend on the tension of the wall , or on the depth of the supersymmetric ads minimum . the only thing we needed to know is that prior to the uplifting these two minima were supersymmetric , which is a generic property of all known flux vacua . an unexpected feature of this result is that the tunneling is suppressed not by a factor @xmath383 , but by a factor @xmath384 . this means that the decay time of an uplifted minkowski vacuum will be shorter than the recurrence time @xmath385 by a huge factor @xmath386 . we will return to the discussion of this fact in section [ sink ] . as we have seen in the previous subsection , a tiny supersymmetry breaking achieved by uplifting of the supersymmetric minkowski vacuum to the ds vacuum with @xmath387 renders this vacuum unstable . now we will extend our results for a more general case . suppose that initially we have two supersymmetric ads vacua , with depth @xmath388 , such that @xmath389 . one can achieve it e.g. by considering the kl potential discussed in the previous section , but taking @xmath390 instead of @xmath391 . tunneling between the two supersymmetric ads minima is forbidden . let us also assume that the deeper minimum is at larger values of the volume modulus , so the kklt uplifting of the minimum @xmath366 to ds space with @xmath387 leaves the depth of the second ads minimum almost unaffected . this is indeed the case in the model discussed above . we will also assume that prior to the uplifting @xmath392 was much greater than the present value of the vacuum energy @xmath393 , so for all practical purposes it is not important whether we uplifted it to a ds state @xmath387 or to the minkowski state with @xmath358 . this subtle difference was important in our previous investigation because the degree of supersymmetry breaking was proportional to @xmath387 . now we are considering a much greater supersymmetry breaking , which is achieved due to a much more significant uplifting , from the initial value of @xmath366 to ( approximately ) zero . in order to simplify notation and avoid mixing the values of @xmath366 before and after the uplifting , we will assume that after the uplifting @xmath358 , and in all subsequent equations we will understand by @xmath366 its value in the ads minimum prior to the uplifting . in order to study the decay rate of the ( nearly ) minkowski vacuum formed by uplifting , one should use , as before , eqs . ( [ coeffb ] ) and ( [ gravfactor ] ) . in minkowski vacuum , one has @xmath360 , so the expression ( [ gravfactor ] ) looks singular , but in fact this singularity is fictitious , which can be easily seen by expanding it near @xmath360 . the outcome of the calculations can be represented in the following form : @xmath394 here @xmath111 is the bubble wall tension for the tunneling in the uplifted potential . the first term in this expression is the probability of the decay of the uplifted minkowski vacuum ignoring gravitational corrections . the second term represents the gravitational suppression of the probability of the tunneling due to the factor @xmath363 . as we see , the probability of the tunneling depends on the wall tension @xmath111 . however , the only part of the expression ( [ tunnads ] ) where this dependence is crucially important is the gravitational suppression term @xmath395 . if after uplifting we would still have two supersymmetric vacua ( minkowski and ads ) , then we would have @xmath396 , the last term in ( [ tunnads ] ) would blow up , and the tunneling would be forbidden . however , prior to the uplifting we had a smaller value of tension , @xmath397 . as we argued before , uplifting tends to decrease the barrier between the two vacua , and therefore it tends to decrease the tension . ( our numerical investigation of the tunneling in the kl model with the potential shown in fig . [ klthin ] is consistent with this expectation . ) if this is the case , the gravitational suppression gradually disappears , and the probability of the tunneling becomes very large . to illustrate various options , let us assume , in the first approximation , that @xmath111 does not change during the uplifting . in this case we find that @xmath398 where @xmath399 . for @xmath13 ( i.e. for @xmath365 of the same order as @xmath366 , prior to the uplifting ) , this expression coincides , up to a numerical coefficient @xmath400 , with the entropy of ds space with the vacuum energy @xmath392 , see eq . ( [ action2a ] ) . the main difference is that for the ds space in which we live now we have @xmath387 , and the factor @xmath401 is incredibly small , @xmath402 . meanwhile the depth of the ads vacuum prior to the uplifting can be a hundred orders of magnitude greater than @xmath393 , and therefore the decay rate of this vacuum after the uplifting can be very large ( see below for a numerical estimate ) . note that if , as we expect , @xmath111 slightly decreases during the uplifting in our model , then @xmath403 will be even smaller , and the decay rate will be even higher . it would be nice to confirm our conclusions by other methods , or to calculate the tunneling probability directly , going beyond the thin - wall approximation . however , our main goal here was to show that generically there is no reason to expect a strong gravitational suppression of the decay of ds vacua if they are obtained by the kklt uplifting of the supersymmetric ads minima . qualitatively , the main argument is as follows : if there are many ads minima in the stringy landscape , and if the tunneling between them is suppressed by supersymmetry , then this suppression disappears after the uplifting and supersymmetry breaking . let us check whether these intuitive expectations match our estimates , eq . ( [ tunndsads ] ) . according to ( [ potential2 ] ) , the scalar potential prior to the uplifting is given by @xmath404 . in the supersymmetric minimum at @xmath308 we have @xmath405 , and @xmath406 . uplifting practically does not change the position of the minimum @xmath308 , and therefore it practically does not change @xmath407 . meanwhile the gravitino mass squared after the uplifting is given by @xmath408 . this means that the depth of the ads minimum before the uplifting is equal to @xmath409 , where @xmath16 is the present value of the gravitino mass @xcite . therefore our result for the tunneling rate of the uplifted ads state ( [ tunndsads ] ) can be represented in the form directly related to the gravitino mass : @xmath410 the scale of supersymmetry breaking in our world is much greater than the energy scale corresponding to the present value of the cosmological constant . for example , if the gravitino mass is @xmath411 gev @xmath412 in planck units , then @xmath413 . therefore eq . ( [ tunndsadsgrav ] ) suggests that a typical decay rate of the ads vacua after their uplifting to the ds state with @xmath387 is greater than @xmath414 . most importantly , this decay rate is greater than @xmath402 by an enormously large factor @xmath415 . this conclusion may have important cosmological implications which we are going to discuss in the next section . but before doing so , let us point out a possible caveat in our discussion . as we already mentioned , if the tunneling between the two vacua should go through the region where @xmath30 changes the sign , then @xmath416 . in this case , if we ignore the decrease of the tension during the uplifting , the tunneling between the two vacua does not occur . one can clearly see it by looking at the gravitational suppression term @xmath395 . this term for @xmath416 is negative , which corresponds to super - suppression of the tunneling @xcite . on the other hand , we expect that in the realistic theories with hundreds of moduli fields there should be many ads minima , some of which will remain ads after the uplifting . therefore it is natural to expect that the decay channels at least to some of these vacua will be open , and will be described by eqs . ( [ tunndsads ] ) , ( [ tunndsadsgrav ] ) . this is the only thing that matters to us , since the total rate of the vacuum decay is determined by the path of the least resistance , i.e. by the channels which maximize the tunneling probability . one of the main reasons to study vacuum transitions which occur on an incredibly large timescale is to find a probability distribution to live in one of the many vacuum states . this issue has a long history involving quantum cosmology , eternal inflation , and statistical properties of the stringy landscape . for a long time , vacuum transitions were studied in the context of inflationary cosmology . the main subject was the possibility of transitions between ds vacuum and minkowski vacuum , or between two ds vacua . one of the most surprising and controversial results was obtained by hawking and moss @xcite . they found that a phase transition from a ds minimum with energy density @xmath366 may occur via a quantum jump to the top of the barrier separating the two vacua , with the height @xmath417 , and by a subsequent classical rolling to the second ds minimum with energy density @xmath418 . the probability of this jump , according to @xcite , was given by @xmath419 initial interpretation of this result was rather obscure because the corresponding instanton seemed to describe a homogeneous tunneling , which was certainly impossible in an infinite ( or exponentially large ) inflationary universe . moreover , from the derivation of this result it was not clear why the tunneling should occur to the top of the potential instead of going directly to the second ds minimum . a proper interpretation of the hawking moss tunneling was achieved only after the development of the stochastic approach to inflation @xcite . one may consider quantum fluctuations of a light scalar field @xmath130 with @xmath420 . during each time interval @xmath421 this scalar field experiences quantum jumps with the wavelength @xmath422 and with a typical amplitude @xmath423 . then the wavelength of these fluctuations grows exponentially . as a result , quantum fluctuations lead to a local change of the amplitude of the field @xmath130 , which looks homogeneous on the horizon scale @xmath424 . from the point of view of a local observer , this process looks like a brownian motion of the homogeneous scalar field . if the potential has a ds minimum at @xmath425 with @xmath426 , then eventually the probability distribution to find the field with the value @xmath130 at a given point becomes time - independent , @xmath427 that is why the probability to gradually climb to the local maximum of the potential at @xmath428 and then fall to another ds minimum is given by eq . ( [ hm ] ) @xcite . note that the distribution @xmath429 is a fraction of the _ comoving _ volume of the universe corresponding to each of the ds vacua . this probability distribution does not take into account different rate of growth of different part of the universe . its interpretation can be understood as follows : at some initial moment one divides the universe into many domains of the same size , assigns one point to each domain , and follows the subsequent distribution @xmath429 of the points where the scalar field takes the value @xmath130 . a necessary condition for the derivation of this result in @xcite was the requirement that @xmath420 . this requirement is violated for all known scalar fields at the present ( post - inflationary ) stage of the evolution of the universe . however , one may try to generalize it for the situations with @xmath430 . following @xcite ( see also @xcite ) , we will look for a probability distribution @xmath431 to find a given _ point _ in a state with the vacuum energy @xmath347 . the main idea is to consider cdl tunneling between two ds vacua , with vacuum energies @xmath366 and @xmath365 , such that @xmath432 , and to study the possibility to tunneling in both directions , from @xmath366 to @xmath365 , or vice versa . the action on the tunneling trajectory , @xmath318 , does not depend on the direction in which the tunneling occurs , but the tunneling probability does depend on it . it is given by @xmath433 on the way up , and by @xmath434 on the way down @xcite . let us assume that the universe is in a stationary state , such that the comoving volume of the parts of the universe going upwards is balanced by the comoving volume of the parts going down . this can be expressed by the detailed balance equation @xmath435 which yields ( compare with eq . ( [ hm ] ) ) @xmath436 independently of the tunneling action @xmath318 . equations ( [ e38a ] ) and ( [ weinb ] ) imply that the fraction of the comoving volume of the universe in a state @xmath130 ( or @xmath312 ) different from the ground state @xmath308 ( which is the state with the lowest , but positive , vacuum energy density ) is proportional to @xmath437 , with the normalization coefficient @xmath438 . the probability distribution @xmath439 coincides with the square of the hartle hawking wave function describing the ground state of the universe @xcite . it has a simple physical meaning : the universe wants to be in the ground state @xmath308 with the lowest possible value of @xmath440 , and the probability of the deviations from the ground state is exponentially suppressed . this probability distribution also has a nice thermodynamic interpretation in terms of ds entropy @xmath441 @xcite : @xmath442 here , as before , @xmath443 . this result and its thermodynamic interpretation played a substantial role in the discussion of the string theory landscape @xcite . unfortunately , there are some problems associated with this result . it suggests that the universe is similar to a pond filled with still water , where all things accumulate at the bottom : the fraction of the comoving volume with the smallest possible vacuum energy must be overwhelmingly large , @xmath444 . for a discussion of paradoxes associated with this conclusion and their possible resolution in the kklt scenario see , e.g. @xcite . investigation of the stationary probability distribution alone does not give us a full picture . for example , the probability distribution ( [ e38a ] ) tells us about the fraction of the comoving volume of the universe in a given state , but it tells us nothing about the evolution towards this state . a partial answer to this question can be given by investigation of the stochastic diffusion equations describing the evolution of the scalar field in the inflationary universe . but now , instead of looking for the most probable outcome of the evolution , one should follow the evolution backwards and look for the initial condition @xmath308 for the trajectories which bring the field to its final destination @xmath130 . in the stationary regime considered above , the corresponding solution looks very similar to ( [ e38a ] ) @xcite : @xmath445 in this equation , however , @xmath308 is not the position of the ground state , but a position of an arbitrary initial point for the diffusion process which eventually brings us to the point @xmath130 . as we see , the probability is maximized by the largest possible value of @xmath446 . interestingly , expression @xmath447 describing the probability of initial conditions coincides with the expression for the square of the tunneling wave function describing creation of a closed ds universe `` from nothing '' @xcite , whereas the second term looks like the square of the hartle - hawking wave function describing the ground state of the universe . in the stationary regime the squares of these two wave functions coexist in the same equation , but they provide answers to different questions . here we should note that it is far from being clear whether the simplest probability distribution which we studied so far should be used in anthropic considerations . indeed , the probability that something happen at a given _ point _ , which is described by the probability distribution in the _ comoving volume _ as discussed above , does not take into account the exponential growth of the _ physical volume _ of the universe , which is the main property of inflationary models . for the discussion of the volume - weighted probability measure , see e.g. @xcite . despite this reservation , it is quite important to study the probability distribution in the comoving volume since its investigation is an essential step in several different attempts to calculate the probabilities in the landscape . here we would like to discuss another problem with the stationary probability distribution described above : this stationary distribution is not very useful during inflation when one can ignore the existence of the lowest ds space with @xmath335 ; in order to obtain a stationary distribution one should take into account the growth of the physical volume of the universe @xcite . moreover , as we are going to show now , this distribution does not necessarily apply to the string theory landscape , simply because in the kklt scenario there are no stable ds vacua that could serve as a ground state of the universe . metastability of ds space in the kklt construction was emphasized in @xcite and in many subsequent papers . here we would like to look at this issue in a more detailed way . stationarity of the probability distribution ( [ weinb2 ] ) was achieved because the lowest ds state did not have any further way to fall . meanwhile , in string theory it is always possible for the metastable ds state to decay . it is important that if it decays by production of the bubbles of 10d minkowski space @xcite , or by production of bubbles containing a collapsing open universe with a negative cosmological constant @xcite , the standard mechanism of return back to the original ds state does not operate any longer . these processes work like sinks for the flow of probability in the landscape . the fraction of the comoving volume in ds vacua will decrease in time . to get a visual understanding of this process , assume that we are looking at the universe in the comoving coordinates , and paint black all of its parts corresponding to one of the two ds states , and paint white the parts in another ds state . then the stationary regime which we discussed so far would imply that eventually the whole universe in average becomes gray , and the level of gray asymptotically will remain constant . suppose now that some parts of the universe may tunnel to a state with a negative cosmological constant . these parts will collapse , so they will not return to the initial ds vacua . if we paint such parts red , then the universe , instead of reaching the constant gray level , eventually will look completely red . this does not mean that eventually all parts of our universe will end up in the sink . due to the exponential expansion of the universe in various ds vacua , the universe eternally rejuvenates , and the total volume of the universe in different ds states continues to grow exponentially @xcite . however , if one wants to find the ratio of the comoving volume of the universe in different states , then instead of solving the detailed balance equation ( [ balance ] ) , one should solve the `` vacuum dynamics '' equations @xmath448 @xmath449 here @xmath450 is the probability current from the lower ds vacuum to the sink , i.e. to a collapsing universe , or to a minkowski vacuum , @xmath451 is the probability current from the upper ds vacuum to the sink , @xmath452 is the probability current from the lower ds vacuum to the upper ds vacuum , and @xmath453 is the probability current from the upper ds vacuum to the lower ds vacuum . combining this all together , gives us the following set of equations for the probability distributions : @xmath454 @xmath455 since the entropy is inversely proportional to the energy density , the entropy of the lower level is higher , @xmath456 . since the tunneling is exponentially suppressed , we have @xmath457 , so we have a hierarchy @xmath458 . we will associate the lower vacuum with our present vacuum state , with @xmath459 , and therefore in what follows we will often have in mind the regime @xmath460 . for simplicity , we will study here the possibility that only the lower vacuum can tunnel to the sink , i.e. we will take the limit @xmath461 and drop the term @xmath462 in eq . . we will analyze more general solutions of equations ( [ v1 ] ) , ( [ v2 ] ) in a separate publication @xcite . on the other hand , we will keep in mind the results of the previous section , where we have found that typically the probability of the decay of a metastable ds vacuum to a sink can be quite high , @xmath463 . therefore we expect that @xmath464 . other mechanisms of a relatively fast vacuum decay have been discussed e.g. in @xcite . the main conclusions to be obtained in this section will be valid for any of these mechanisms . by solving equations ( [ v1 ] ) , ( [ v2 ] ) , one can show that @xmath465 and @xmath466 exponentially decrease in the course of time due to the existence of the sink ( the term @xmath467 ) , but the corresponding physical volume of the universe exponentially grows , and the ratio @xmath468 approaches a stationary regime @xmath469 . in order to find @xmath470 , one can add to each other our equations ( without the term @xmath471 ) . this yields @xmath472 the solution is @xmath473 as for the ( asymptotically ) constant ratio @xmath474 , from eqs . ( [ v1 ] ) , ( [ v3 ] ) one finds @xmath475 one may consider two interesting regimes . suppose first that @xmath476 , i.e. the probability to fall to the sink from the lower vacuum is smaller than the probability of the decay of the upper vacuum . in this case one recovers the previous result , eq . ( [ weinb2 ] ) , which is related to the square of the hartle - hawking wave function : @xmath477 now let us consider an opposite regime , and assume that the decay rate of the uplifted ds vacuum to the sink is relatively large , @xmath478 ( which automatically means that @xmath479 ) . in this case the solution of eq . ( [ ttt ] ) is @xmath480 i.e. one has an inverted probability distribution . this result has a simple interpretation : if the `` thermal exchange '' between the two ds vacua occurs very slowly as compared to the rate of the decay of the lower ds vacuum , then the main fraction of the volume of the ds vacua will be in the state with higher energy density . note that the total fraction of the comoving volume remaining in both of the ds spaces exponentially decreases in time due to the existence of the sink in the lower ds vacuum : @xmath481 this result may have interesting methodological implications . in many recent discussions of the probabilities in the landscape the authors consider the world as seen by an imaginary eternal observer . in the context of the stringy landscape such considerations are problematic , because every would - be eternal observer living in the landscape ( unlike an observer in an eternally existing ds space ) begins his life in a cosmological singularity and ends up in a sink , within the typical time determined by eq . ( [ sinkpower ] ) . moreover , if the sinks occur in the higher ds vacua ( see below ) , then the `` eternal observers '' , or markers , will have a good chance to die well before reaching the thermalized regions produced after inflation . introduction of a finite lifetime of such observers may result in the modification of the probability measure as suggested in @xcite . here we should make an important clarification . the probability distribution @xmath431 describes the fraction of the volume of the universe in a particular ds state . however , when the bubbles of a new phase expand , their interior becomes an empty ds space . if we are usual observers which are born after reheating of the universe , then one may argue that the probability _ to live _ in the bubble @xmath482 is proportional not to the relative volume of the bubbles @xmath483 but to the frequency of their production , i.e. to the ratio of the probability currents @xmath484 @xcite . in the absence of the sink , the fraction of the comoving volume which flows to the lower ds vacuum due to the tunneling from the upper ds vacuum is equal to the fraction of the volume jumping upwards from the lowest vacuum to the higher vacuum . in other words , the two probability currents are exactly equal to each other , j_01 = j_10 , which is the essence of the detailed balance equation ( [ balance ] ) . interestingly , our new results imply that this regime remains approximately valid even in the presence of the sink , under the condition @xmath485 . on the other hand , in the regime described by eq . ( [ ooo ] ) , which occurs if the decay rate to the sink is large enough , @xmath486 , one has a completely different result : @xmath487 thus we have a crucial regime change at the moment when the decay rate of the lower vacuum to the sink starts competing with the decay rate of the upper ds vacuum . now we will compare the probability current to the sink , @xmath488 , and the probability current from the lower ds upwards , @xmath489 . in the absence of the sink , @xmath488 was zero , so every point falling to the lower ds vacuum eventually jumps up to the higher ds vacua . this leads to eternal recycling of ds vacua @xcite , and to the problems discussed in @xcite : in the absence of the sinks ( i.e. in the models with eternal ds space ) each point in the comoving volume would indefinitely wonder between different ds spaces . in this scenario , it would be more probable that our part of the universe was formed as a result of a large quantum fluctuation upwards from an empty ds space , instead of being formed at the stage of inflation . this would leave the homogeneity of our universe , and the origin of perturbations of metric unexplained @xcite . invention of the kklt mechanism demonstrated that all ds vacua in string theory are metastable @xcite , with the lifetime smaller than the recurrence time @xmath490 . this means that we do not have much time for the incredibly improbable jumps upwards to happen @xcite . however , the time necessary for the improbable fluctuations upwards , which could re - create conditions for the existence of life , is also somewhat smaller than the recurrence time @xcite . therefore in order to make sure that the universe does not recycle back from our ds state we would need to double - check that the probability to jump upwards is indeed strongly suppressed as compared to the probability to tunnel to the sink . the estimates of the decay probability of ds vacuum contained in our paper confirm that this is indeed the case . indeed , investigation of the tunneling in several models discussed in our paper suggests that the probability that any given part of ds space with @xmath491 will jump upwards and participate in the process of recycling of the universe is suppressed by a factor of @xmath492 . this can be expressed in terms of the ratio of the probability current from the lower ds vacuum to the upper vacuum @xmath489 with the probability current from the lower ds vacuum to the sink @xmath488 : @xmath493 the last part of this equation follows from the condition @xmath460 , and from the estimate of the decay rate obtained in the previous section , which suggests that in many cases this rate is quite high , @xmath464 . despite this fact , recycling may still be possible , since for @xmath485 one has @xmath494 . on the other hand , for @xmath486 one has @xmath495 . this suggests that in this regime the probability that our universe was created due to the jumps up @xcite becomes exponentially small as compared with the probability that it was formed as a result of inflation . using our results , one can easily check that this is indeed the case if the rate of spontaneous formation of the parts of the universe of our type in the vacuum @xmath496 is smaller than the decay rate of the vacuum @xmath496 to the sink . for wide sinks , this condition can be easily satisfied . this conclusion is closely related to the inversion of the probability distribution for @xmath486 ( wide sink regime ) , see eq . ( [ ooo ] ) . note that inverted probability distributions in the presence of the sink represented by minkowski vacuum naturally appear in the simplest chaotic inflation scenario with the volume - weighted probability measures proposed in @xcite ; see figs . 10 and 12 in @xcite , where the basic theory of the volume - weighted distributions in the inflationary cosmology was developed . with the first of the two volume - weighted probability measures proposed in @xcite the problem of non - inflationary jumps up does not appear at all ; the growth of the total volume of the universe ( and the corresponding fluxes ) is predominantly determined by the exponential expansion of domains with larger vacuum energy and by the subsequent inflation on the way down . in our investigation so far we considered the simplest model with two ds vacua , with the probability leak in the lower ds vacuum . on the other hand , if the probability leak occurs only in the upper ds vacuum , one may ignore the term @xmath497 in eq . ( [ v1 ] ) but keep the term @xmath498 in eq . ( [ v2 ] ) . in this case , for @xmath499 one again recovers the distribution ( [ hheq ] ) , whereas for @xmath500 one finds @xmath501 note that in this regime we have a flat probability distribution , which does not depend on @xmath502 . more generally , one should consider a dynamical equilibrium of a system of many ds , minkowski and ads vacua , where the probability leaks may occur at each level . the fraction of the comoving volume occupied by different metastable vacua in such a universe will be exponentially sensitive to the decay modes for each of these vacua . this makes the investigation of the probabilities in the landscape much more involved , but also much more interesting . previously we were trying to construct the wave function of the ground state of the universe . now we should learn how to study the universe with the holes in the ground . there are two possible approaches to string theory landscape . the first one is based on the calculation of the wave function of the universe , which may describe all branches of the wave function corresponding to all possible vacua . in fact , the first two proposals for the anthropic solution of the cosmological problem in the context of inflationary cosmology were based on the investigation of quantum creation of the universe with various values of scalar fields and fluxes @xcite , or with different types of compactification @xcite . another approach is to consider an inflationary universe consisting of many parts corresponding to all possible vacua @xcite , which allows , in particular , to consider a single universe divided into parts corresponding to all possible values of scalar fields , fluxes and compactifications , and look for the anthropic solution of the cosmological constant problem in this context , see e.g. @xcite . conceptually , this approach is much simpler than the approach based on quantum cosmology , but in order to make it sufficiently general one must find a way to describe the universe consisting of many different parts and describe possible transitions between them . this was the goal of our paper . one of the problems here is that a fully consistent `` landscape action '' with the bulk and brane sources is not known @xcite . however , one can hope that such action can be constructed at least in the case of unbroken supersymmetry following the `` supersymmetry in singular spaces '' proposal @xcite . this strategy was already applied in examples of fully consistent supersymmetric bulk and brane actions in @xcite which include the d8 domain walls with the piece - wise constant @xmath503 flux . in a more general situation with moduli stabilization when various fluxes may jump due to brane sources , the corresponding construction has still to be developed . therefore in the current project we restricted ourselves to studies of the regions of the landscape where stabilization of the moduli is under control and fluxes are fixed . even in this restricted and simplified situation we were able to find multiple vacua and bps domain walls separating them . in the first part of this paper we concentrated on the stringy landscape prior to the uplifting and found many bps domain wall solutions separating different parts of the stringy landscape from each other . we did it using the methods developed earlier in @xcite , and also more recent methods based on the `` new attractor equations '' @xcite . in the second part of the paper we studied what happens with the domain walls after the uplifting . in general , potentials of many moduli before and after uplifting may have many different ads minima of the moduli potential , for the same values of fluxes . before the uplifting , the tunneling between these minima was forbidden by supersymmetry . one could expect , therefore , that after the uplifting the rate of tunneling may be related to the degree of supersymmetry breaking . indeed , in the particular examples that we have studied the probability of the decay of the uplifted vacuum to the remaining ads vacuum was suppressed by a factor @xmath504 , where @xmath505 is the gravitino mass . this decay probability is much greater than @xmath506 . one of the interesting features of the decay of ds vacua to minkowski or ads vacua is that this decay is irreversible . this process is substantially different from the stationary `` thermal exchange '' between various ds vacua . the existence of the channels of the irreversible vacuum decay , which serve as sinks for the flow of probability , is a distinguishing new feature of the stringy landscape which deserves further investigation . as we show in section 6 of our paper , the existence of the sinks strongly affects the probability distributions in string cosmology . it is a pleasure to thank l. alvarez gaum , r. bousso , t. banks , m. cveti , f. denef , b. freivogel , g. horowitz , s. kachru , m. mario , d. martelli , e. silverstein , s. shenker , l. susskind and w. taylor for stimulating conversations . the work of a.c . is supported in part by the european community s marie curie actions , project no . mrtn - ct-2004 - 005104 ( forces universe ) . the work of r.k . , a.l . and a.g . was supported by nsf grant phy-0244728 . the work of a.g . was also supported by nsf grant phy-0097915 , and by doe under contracts de - ac03 - 76sf00515 and rfbr 05 - 01 - 00758 . linde , `` nonsingular regenerating inflationary universe , '' print-82 - 0554 , cambridge university preprint , 1982 , see http://www.stanford.edu/@xmath507alinde/1982.pdf ; a.d . linde , `` the new inflationary universe scenario , '' in : _ the very early universe _ gibbons , s.w . hawking and s.siklos , cambridge university press ( 1983 ) , pp . 205 - 249 , see http://www.stanford.edu/@xmath507alinde/1983.pdf ; a. d. linde , `` eternally existing self - reproducing chaotic inflationary universe , '' phys . lett . b * 175 * , 395 ( 1986 ) . w. lerche , d. lst and a. n. schellekens , `` chiral four - dimensional heterotic strings from selfdual lattices , '' nucl . b * 287 * , 477 ( 1987 ) . r. bousso and j. polchinski , `` quantization of four - form fluxes and dynamical neutralization of the cosmological constant , '' jhep * 0006 * , 006 ( 2000 ) [ arxiv : hep - th/0004134 ] . m. r. douglas , `` the statistics of string / m theory vacua , '' jhep * 0305 * 046 ( 2003 ) [ arxiv : hep - th/0303194 ] ; f. denef and m. r. douglas , `` distributions of flux vacua , '' jhep * 0405 * , 072 ( 2004 ) [ arxiv : hep - th/0404116 ] . m. cvetic and h. h. soleng , `` supergravity domain walls , '' phys . rept . * 282 * , 159 ( 1997 ) [ arxiv : hep - th/9604090 ] ; m. cvetic , s. griffies and s. j. rey , `` nonperturbative stability of supergravity and superstring vacua , '' nucl . b * 389 * , 3 ( 1993 ) [ arxiv : hep - th/9206004 ] ; m. cvetic , s. griffies and s. j. rey , `` static domain walls in n=1 supergravity , '' nucl . b * 381 * , 301 ( 1992 ) [ arxiv : hep - th/9201007 ] ; m. cvetic , s. griffies and h. h. soleng , `` local and global gravitational aspects of domain wall space - times , '' phys . rev . d * 48 * , 2613 ( 1993 ) [ arxiv : gr - qc/9306005 ] . a. ceresole , g. dallagata , r. kallosh and a. van proeyen , `` hypermultiplets , domain walls and supersymmetric attractors , '' phys . d * 64 * , 104006 ( 2001 ) [ arxiv : hep - th/0104056 ] . k. behrndt , g. lopes cardoso and d. lst , `` curved bps domain wall solutions in four - dimensional n = 2 supergravity , '' nucl . b * 607 * , 391 ( 2001 ) [ arxiv : hep - th/0102128 ] ; g. lopes cardoso , g. dallagata and d. lst , `` curved bps domain wall solutions in five - dimensional gauged supergravity , '' jhep * 0107 * ( 2001 ) 026 [ arxiv : hep - th/0104156 ] ; g. lopes cardoso , g. dallagata and d. lst , `` curved bps domain walls and rg flow in five dimensions , '' jhep * 0203 * ( 2002 ) 044 [ arxiv : hep - th/0201270 ] ; k. behrndt and m. cvetic , `` bent bps domain walls of d = 5 n = 2 gauged supergravity coupled to hypermultiplets , '' phys . d * 65 * , 126007 ( 2002 ) [ arxiv : hep - th/0201272 ] . j. louis and s. vaula , `` n=1 domain wall solutions of massive type ii supergravity as generalized geometries , '' arxiv : hep - th/0605063 . d. z. freedman , s. s. gubser , k. pilch and n. p. warner , `` renormalization group flows from holography supersymmetry and a c - theorem , '' adv . * 3 * , 363 ( 1999 ) [ arxiv : hep - th/9904017 ] . r. kallosh and a. linde , `` landscape , the scale of susy breaking , and inflation , '' jhep * 0412 * , 004 ( 2004 ) [ arxiv : hep - th/0411011 ] ; j. j. blanco - pillado , r. kallosh and a. linde , `` supersymmetry and stability of flux vacua , '' jhep * 0605 * , 053 ( 2006 ) [ arxiv : hep - th/0511042 ] . g. t. horowitz and a. strominger , `` counting states of near - extremal black holes , '' phys . lett . * 77 * , 2368 ( 1996 ) [ arxiv : hep - th/9602051 ] ; c. g. . callan and j. m. maldacena , `` d - brane approach to black hole quantum mechanics , '' nucl . b * 472 * , 591 ( 1996 ) [ arxiv : hep - th/9602043 ] ; a. w. peet , `` tasi lectures on black holes in string theory , '' arxiv : hep - th/0008241 . d. z. freedman , c. nunez , m. schnabl and k. skenderis , `` fake supergravity and domain wall stability , '' phys . d * 69 * , 104027 ( 2004 ) [ arxiv : hep - th/0312055 ] ; a. celi , a. ceresole , g. dallagata , a. van proeyen and m. zagermann , `` on the fakeness of fake supergravity , '' phys . d * 71 * , 045009 ( 2005 ) [ arxiv : hep - th/0410126 ] . r. kallosh , `` new attractors , '' jhep * 0512 * , 022 ( 2005 ) [ arxiv : hep - th/0510024 ] . g. dallagata , `` non - kaehler attracting manifolds , '' jhep * 0604 * ( 2006 ) 001 [ arxiv : hep - th/0602045 ] . j. p. hsu , a. maloney and a. tomasiello , `` black hole attractors and pure spinors , '' arxiv : hep - th/0602142 . a. giryavets , `` new attractors and area codes , '' jhep * 0603 * , 020 ( 2006 ) [ arxiv : hep - th/0511215 ] . d. lst , s. reffert , w. schulgin and s. stieberger , `` moduli stabilization in type iib orientifolds . i : orbifold limits , '' arxiv : hep - th/0506090 . k. behrndt and g. dallagata , `` vacua of n = 2 gauged supergravity derived from non - homogenous quaternionic spaces , '' nucl . b * 627 * , 357 ( 2002 ) [ arxiv : hep - th/0112136 ] . m. graa , `` flux compactifications in string theory : a comprehensive review , '' phys . rept . * 423 * , 91 ( 2006 ) [ arxiv : hep - th/0509003 ] . a. giryavets , s. kachru and p. k. tripathy , `` on the taxonomy of flux vacua , '' jhep * 0408 * , 002 ( 2004 ) [ arxiv : hep - th/0404243 ] . s. ashok and m. r. douglas , `` counting flux vacua , '' jhep * 0401 * , 060 ( 2004 ) [ arxiv : hep - th/0307049 ] . o. dewolfe , a. giryavets , s. kachru and w. taylor , `` enumerating flux vacua with enhanced symmetries , '' jhep * 0502 * , 037 ( 2005 ) [ arxiv : hep - th/0411061 ] . a. klemm and s. theisen , `` considerations of one modulus calabi - yau compactifications : picard - fuchs equations , kahler potentials and mirror maps , '' nucl . phys . b * 389 * , 153 ( 1993 ) [ arxiv : hep - th/9205041 ] . f. denef , b. r. greene and m. raugas , `` split attractor flows and the spectrum of bps d - branes on the quintic , '' jhep * 0105 * , 012 ( 2001 ) [ arxiv : hep - th/0101135 ] . r. bousso , b. freivogel and m. lippert , `` probabilities in the landscape : the decay of nearly flat space , '' arxiv : hep - th/0603105 . s. weinberg , `` does gravitation resolve the ambiguity among supersymmetry vacua ? , '' phys . * 48 * ( 1982 ) 1776 . s. j. parke , `` gravity , the decay of the false vacuum and the new inflationary universe scenario , '' phys . b * 121 * , 313 ( 1983 ) . s. w. hawking and i. g. moss , `` supercooled phase transitions in the very early universe , '' phys . b * 110 * , 35 ( 1982 ) . a. a. starobinsky , `` stochastic de sitter ( inflationary ) stage in the early universe , '' in : _ current topics in field theory , quantum gravity and strings _ , lecture notes in physics , eds . de vega and n. sanchez ( springer , heidelberg 1986 ) * 206 * , p. 107 . a. d. linde , `` hard art of the universe creation ( stochastic approach to tunneling and baby universe formation ) , '' nucl . b * 372 * , 421 ( 1992 ) [ arxiv : hep - th/9110037 ] . a. d. linde , d. a. linde and a. mezhlumian , `` from the big bang theory to the theory of a stationary universe , '' phys . d * 49 * , 1783 ( 1994 ) [ arxiv : gr - qc/9306035 ] . k. m. lee and e. j. weinberg , `` decay of the true vacuum in curved space - time , '' phys . d * 36 * , 1088 ( 1987 ) . j. garriga and a. vilenkin , `` recycling universe , '' phys . d * 57 * , 2230 ( 1998 ) [ arxiv : astro - ph/9707292 ] . n. goheer , m. kleban and l. susskind , `` the trouble with de sitter space , '' jhep * 0307 * , 056 ( 2003 ) [ arxiv : hep - th/0212209 ] . a. d. linde , `` quantum creation of the inflationary universe , '' lett . nuovo cim . * 39 * , 401 ( 1984 ) ; a. vilenkin , `` quantum creation of universes , '' phys . d * 30 * , 509 ( 1984 ) . j. garcia - bellido , a. d. linde and d. a. linde , `` fluctuations of the gravitational constant in the inflationary brans - dicke cosmology , '' phys . d * 50 * , 730 ( 1994 ) [ arxiv : astro - ph/9312039 ] ; a. vilenkin , `` predictions from quantum cosmology , '' phys . lett . * 74 * , 846 ( 1995 ) [ arxiv : gr - qc/9406010 ] ; j. garcia - bellido and a. d. linde , `` stationarity of inflation and predictions of quantum cosmology , '' phys . d * 51 * , 429 ( 1995 ) [ arxiv : hep - th/9408023 ] ; j. garcia - bellido and a. d. linde , `` stationary solutions in brans - dicke stochastic inflationary cosmology , '' phys . rev . d * 52 * , 6730 ( 1995 ) [ arxiv : gr - qc/9504022 ] . a. vilenkin , `` making predictions in eternally inflating universe , '' phys . d * 52 * , 3365 ( 1995 ) [ arxiv : gr - qc/9505031 ] ; s. winitzki and a. vilenkin , `` uncertainties of predictions in models of eternal inflation , '' phys . d * 53 * , 4298 ( 1996 ) [ arxiv : gr - qc/9510054 ] ; j. garriga and a. vilenkin , `` a prescription for probabilities in eternal inflation , '' phys . d * 64 * , 023507 ( 2001 ) [ arxiv : gr - qc/0102090 ] ; j. garriga , d. schwartz - perlov , a. vilenkin and s. winitzki , `` probabilities in the inflationary multiverse , '' jcap * 0601 * , 017 ( 2006 ) [ arxiv : hep - th/0509184 ] ; r. easther , e. a. lim and m. r. martin , `` counting pockets with world lines in eternal inflation , '' jcap * 0603 * , 016 ( 2006 ) [ arxiv : astro - ph/0511233 ] ; a. vilenkin , `` probabilities in the landscape , '' arxiv : hep - th/0602264 . a. h. guth , `` the inflationary universe : a possible solution to the horizon and flatness problems , '' phys . d * 23 * , 347 ( 1981 ) . p. j. steinhardt , `` natural inflation , '' in : _ the very early universe _ gibbons , s.w . hawking and s.siklos , cambridge university press , ( 1983 ) . a. r. frey , m. lippert and b. williams , `` the fall of stringy de sitter , '' phys . d * 68 * , 046008 ( 2003 ) [ arxiv : hep - th/0305018 ] . d. green , e. silverstein and d. starr , `` attractor explosions and catalyzed vacuum decay , '' arxiv : hep - th/0605047 . a. d. linde , `` the inflationary universe , '' rept . phys . * 47 * 925 ( 1984 ) . a. d. sakharov , `` cosmological transitions with a change in metric signature , '' sov . phys . jetp * 60 * , 214 ( 1984 ) [ zh . . fiz . * 87 * , 375 ( 1984 ) ] . j. d. brown and c. teitelboim , `` neutralization of the cosmological constant by membrane creation , '' nucl . phys . b * 297 * , 787 ( 1988 ) ; j. d. brown and c. teitelboim , `` dynamical neutralization of the cosmological constant , '' phys . b * 195 * , 177 ( 1987 ) . j. l. feng , j. march - russell , s. sethi and f. wilczek , `` saltatory relaxation of the cosmological constant , '' nucl . b * 602 * , 307 ( 2001 ) [ arxiv : hep - th/0005276 ] . t. banks , m. dine and e. gorbatov , `` is there a string theory landscape ? , '' jhep * 0408 * , 058 ( 2004 ) [ arxiv : hep - th/0309170 ] . f. denef , private discussion . de alwis , `` transitions between flux vacua , '' arxiv : hep - th/0605184 . e. bergshoeff , r. kallosh and a. van proeyen , `` supersymmetry in singular spaces , '' jhep * 0010 * , 033 ( 2000 ) [ arxiv : hep - th/0007044 ] . e. bergshoeff , r. kallosh , t. ortin , d. roest and a. van proeyen , `` new formulations of d = 10 supersymmetry and d8 - o8 domain walls , '' class . grav . * 18 * , 3359 ( 2001 ) [ arxiv : hep - th/0103233 ] ; r. kallosh , s. prokushkin and m. shmakova , `` domain walls with strings attached , '' jhep * 0107 * , 023 ( 2001 ) [ arxiv : hep - th/0107097 ] .

random tessellations * st*able with respect to * it*erations ( stit ) have recently been introduced in @xcite and @xcite as a new model for random tessellation in @xmath3 and have quickly attracted considerable interest in modern stochastic geometry as well fitting the growing demand for non - trivial and flexible but mathematically tractable tessellation models . the stit tessellations may be interpreted as outcome of a random cell division process , which makes them very attractive for applications , see for example @xcite . other potential applications include mathematical modelling of systems of cracks , joints or fissures in rock or the so - called craquele of thin layers . a general approach to random cell division processes has recently appeared in @xcite and the construction of stit tessellations can roughly be described as follows . at first , we fix a compact and convex window @xmath1 in which the construction is carried out . for simplicity we assume @xmath4 to be a polytope , as this implies that the resulting tessellation has polytopal cells with probability one . next , an exponentially random life time is assigned to @xmath4 , whereby the parameter of the distribution is given as a constant multiple of the integral - geometric mean width of @xmath4 , see section [ secstit ] for details . upon expiry of this life time a @xmath4 hitting uniform random hyperplane is chosen , is introduced in @xmath4 and divides the window into two polytopal sub - cells @xmath5 and @xmath6 . the construction continues now recursively and independently in both of these sub - cells , where the newly introduced hyperplanes are always chopped - off by the boundaries of their mother - cells . the whole construction is continued until some deterministic time threshold @xmath0 is reached . regarded in time , the construction can by interpreted as a pure - jump markov process on the space of tessellations of the window @xmath4 . + our construction shares some common features with random fragmentation processes or branching markov chains in the sense of @xcite . the cells of the resulting tessellation within @xmath4 can be regarded as particles in a suitable polish space and the dynamics of the particles is non - interacting in a sense that different particles ( cells of the tessellation ) have independent evolutions , which is indeed the case in our construction . moreover , whenever a particle dies it is replaced by exactly two new particles , namely the two newly generated sub - cells . in addition , the life times of the particles are exponentially distributed , as assumed in @xcite , but in general not independent in contrast to the fragmentation theory . + in the recent paper @xcite the authors have introduced a new technique for studying the geometric properties of stit tessellations based on martingales and the general theory of martingale problems for pure - jump type markov processes . in particular , with these new developments , the variance of the total surface area of a stationary and isotropic iteration stable random tessellation @xmath7 in @xmath1 ( a random stit tessellation ) has been determined by integral - geometric means and the corresponding central limit theory has been established . strikingly , as already signalled by results in the special two dimensional case @xcite , and as confirmed by the results of the present paper , it turns out that in a certain rather strong sense the asymptotic behaviour of the surface area process dominates and fully determines the asymptotic geometry of the stit tessellation . in particular the surface area variance is _ the _ basic second - order parameter of the tessellation and all second - order characteristics of functionals considered in our work can be reduced to it , likewise non - trivial functional limits in law exist for the stit surface area process whereas the limits in law for other natural related processes arise as deterministic functionals of the corresponding surface process . another crucial phenomenon arising in this context , as first noted in @xcite , is that the asymptotic theories for dimension @xmath8 and @xmath9 differ strongly in many important aspects . we have studied the planar case @xmath8 in the recent separate paper @xcite and thus only consider @xmath9 in the present one . + the purpose of this paper is to establish a second - order and limit theory for integral - geometric characteristics of stationary and isotropic stit tessellations in dimensions higher than two . more precisely , the characteristics studied in this work are the cumulative intrinsic volumes of all orders for the collections of so - called maximal polytopes of @xmath7 . these are in codimension @xmath10 the cell - separating facets introduced during the random cell division process described above , constituting the basic building blocks of a stit tessellation as discussed in detail below . we shall provide explicit as well as asymptotic variance expressions for these parameters of the random tessellation @xmath11 as @xmath12 tends to infinity for a sequence @xmath13 of expanding convex windows . further , we will find the covariance measures for random lower - dimensional face measures generated by @xmath14 finally , we shall also give the corresponding convergence in law statements , obtaining non - gaussian limits for the studied case @xmath15 as opposed to the classical gaussian limits arising for @xmath8 , see @xcite . + + the paper is organized as follows : in section [ secstit ] we define stit tessellations , specializing to the stationary and isotropic set - up in the focus of this paper , and we discuss their basic properties that are needed in our arguments . next , in section [ sec2ndord ] we calculate the variances of the cumulative intrinsic volumes of all orders for stit tessellations . this includes both , exact formulae and asymptotic analysis upon letting the window size grow to infinity . further , in section [ sec2ndmeas ] we extend the second - order analysis to the level of lower - dimensional face measures induced by stit tessellations , thus taking into account not only the numeric characteristics but also the spatial profile of the stit face processes . finally , in section [ secclt ] we develop the corresponding functional limit theory with non - gaussian limit processes for @xmath16 in order to keep the paper self - contained , we will recall important faces from @xcite and sketch some of their proofs for the readers convenience . the purpose of this section is to provide a short self - contained discussion of stit tessellations in @xmath3 as studied in this paper and to summarize their basic properties for easy reference . we will restrict to the stationary and isotropic case in subsection [ subsecmnwc ] below , specializing to the scope of the paper . the general reference for these by now classical properties throughout this section is @xcite . next , in subsection [ subsecmart ] we will discuss the martingale tools developed in @xcite and underlying our present theory . finally , in subsection [ subsec1stord ] we provide certain useful mean value relationships for intrinsic volumes in context of stit tessellations . we start with a compact and convex polytope @xmath1 in which our construction is carried out and denote by @xmath17 the standard isometry - invariant measure on the space @xmath18 of ( affine ) hyperplanes in @xmath3 normalized so as to induce unit surface intensity on @xmath19 ( note that in our earlier papers @xcite and @xcite this measure has been denoted by @xmath20 ) . we call @xmath17 the _ driving measure _ of the construction . assign now to @xmath4 an exponentially random lifetime with parameter @xmath21)$ ] , where @xmath22:=\{h\in{\cal h}:\ h\cap w\neq\emptyset\}\ ] ] is the set of hyperplanes hitting @xmath4 . upon expiry of this random life time , a random hyperplane is chosen according to the distribution @xmath21)^{-1}\lambda(\cdot\cap[w])$ ] , is introduced in @xmath4 and is chopped off by its boundary . this is , the window @xmath4 splits into the two polyhedral sub - cells @xmath5 and @xmath6 that are separated by the introduced hyperplane piece . the construction continues now recursively and independently in @xmath5 and @xmath6 and is stopped if some previously fixed deterministic time threshold @xmath23 is reached . our assumptions ensure that the cells of the tessellation constructed until time @xmath23 are convex polyhedra in @xmath4 with probability one . they are denoted by @xmath24 and we denote by @xmath7 the random closed set in @xmath4 that consists of the union of cell - boundaries of cells constructed until time @xmath0 , see figure [ figure ] . the construction of @xmath7 is referred to as the _ mnw - construction _ after the names of its inventors mecke , nagel and weiss and the tessellation @xmath7 itself is called a random _ stit tessellation_. the abbreviation stit comes from the crucial property enjoyed by @xmath25 namely that of being able under eration , for which we refer to @xcite or @xcite . we call the cell separating @xmath26-dimensional faces the _ @xmath26-dimensional maximal polytopes _ and denote the collection of all such polytopes of @xmath7 by @xmath27 . moreover , we introduce the set @xmath28 of @xmath29-dimensional faces of @xmath26-dimensional maximal polytopes of @xmath7 by @xmath30 where by @xmath31 we mean the set of all @xmath29-dimensional faces of the @xmath26-dimensional polytope @xmath32 , @xmath33 . the maximal polytopes are often referred to as _ i - polytopes _ this terminology originates from the historically first considered particular case @xmath8 where the maximal polytopes ( which are just line segments in this case ) assumed shapes similar to the literal @xmath34 + it is an important observation that the spatio - temporal construction of @xmath7 satisfies the markov property in time parameter @xmath0 , which means that @xmath35 where @xmath36 denotes the operation of iteration of tessellations , see @xcite , @xcite or @xcite . to make the paper more self - contained , we recall now the definition of the meaning of the operation @xmath36 . to this end , we regard the tessellation @xmath7 as a frame or primary tessellation and associate with each cell @xmath37 an i.i.d . copy @xmath38 of @xmath39 . another tessellation @xmath40 of @xmath4 is now defined as @xmath41 i.e. we consider the union of @xmath7 and the cut - outs of @xmath38 s within the cells @xmath42 of the primary tessellation ( note that in the definition we have used the interpretation of a tessellation as a random closed subset of @xmath4 ) . we say that @xmath40 is the _ iteration _ of @xmath7 with @xmath39 . the remarkable property of our tessellations constructed by the mnw - process is that the outcome @xmath43 coincides in law with @xmath44 , i.e. with the continuation of the mnw - construction until time @xmath45 . thus , it is the same either to continue the mnw - construction from @xmath0 until time @xmath45 or to perform at time @xmath0 an iteration of @xmath7 with @xmath39 . + the local properties established so far , can be extended to the whole space , since the random tessellations @xmath7 are _ consistent _ in @xmath4 , by which we mean that for any fixed @xmath23 and @xmath46 it holds @xmath47 where @xmath48 stands for equality in distribution . this implies in view of the consistency theorem ( * ? ? ? 2.3.1 ) the existence of the whole - space tessellation @xmath49 such that @xmath50 for each compact convex @xmath1 . + it directly follows from the markov property that the random tessellations @xmath49 are stable with respect to the operation @xmath36 , i.e. @xmath51 where @xmath52 means the dilation with a factor @xmath53 . this property explains also the abbreviation stit , because the last equation is a classical probabilistic stability relation . the random tessellations @xmath49 share another important property , namely that the _ intersection _ of @xmath49 with a @xmath29-dimensional plane @xmath54 , @xmath55 , is again an iteration stable random tessellation . more precisely , we have @xmath56 and the integral - geometric constant @xmath57 is given by @xmath58 see eq . ( 3.29 t ) in @xcite . moreover , it is easy to see from the properties of the capacity functional of @xmath49 , compare with ( * ? ? ? 5(ii ) ) , that stit tessellations have the following _ scaling property _ : @xmath59 i.e. the tessellation @xmath49 has , upon rescaling with the factor @xmath60 the same distribution as @xmath61 , that is the stit tessellation with surface intensity @xmath10 . + we close this section by mentioning that the random tessellations @xmath49 have another important property , namely that @xmath49 has _ poisson typical cells _ or poisson cells for short , see lemma 3 in @xcite or part ( 1 ) of proposition [ propxx ] below . this is to say , the typical cell of @xmath49 has the same distribution as the typical cell of a stationary and isotropic poisson hyperplane tessellation with surface density @xmath0 ( see chap . 10.3 in @xcite for references about the classical poisson hyperplane model ) . in particular , this fact combined with the intersection property from above shows that the typical cell of the lower dimensional stit tessellation @xmath62 , with @xmath54 as above , has the same distribution as the typical cell of a poisson hyperplane tessellation in @xmath63 with surface density @xmath64 . more formally , we define the cell intensity measure @xmath65 of the stit tessellation @xmath7 by @xmath66 and its @xmath26-dimensional maximal polytope intensity measure @xmath67 by @xmath68 where @xmath69 stands for the uni mass dirac measure concentrated at @xmath70 . moreover , we let @xmath71 be the cell intensity measure and @xmath72 be the @xmath26-face intensity measure of a poisson hyperplane tessellation @xmath73 within @xmath1 having intensity measure @xmath74 , which are defined similarly to @xmath65 and @xmath67 above ( clearly , the @xmath26-dimensional maximal polytopes have to be replaced by the set of @xmath26-faces of @xmath73 ) . these definitions bring us in the position to reformulate special cases of theorems 1 and 2 from @xcite adapted to our later purposes : [ propxx ] we have @xmath75 as already noted above , the mnw - construction of iteration stable random tessellations @xmath7 in finite volumes @xmath1 enjoys a markov property in the continuous time parameter @xmath76 in our previous work @xcite we have used this fact combined with the classical theory of martingale problems for pure jump markov processes to construct a class of natural martingales associated to the mnw - process . in this paper we only need a part of that theory . to formulate it we let @xmath77 be a bounded and measurable functional on the space of @xmath26-polytopes in @xmath3 and we denote by @xmath78 the collection of @xmath26-dimensional maximal polytopes of a tessellation @xmath79 standing for a generic realization of @xmath7 for some @xmath80 write @xmath81 any hyperplane @xmath82 $ ] hitting the window @xmath4 is tessellated by the intersection with @xmath79 and we denote by @xmath83 the set of all @xmath26-dimensional cells of this tessellation and introduce @xmath84}\sum_{f\in{\operatorname{cells}}(y\cap h)}\phi(f)\lambda(dh).\ ] ] it is also convenient to introduce the bar notation for centered versions of these quantities with @xmath85 that is to say @xmath86 and likewise @xmath87 with this notation , in view of the results developed in @xcite , we have [ martprop ] for bounded and measurable functionals @xmath77 and @xmath88 on the space of @xmath26-polytopes in @xmath3 , the stochastic processes @xmath89 and @xmath90ds\label{mart2}\end{aligned}\ ] ] are martingales with respect to the filtration @xmath91 induced by @xmath92 . [ [ sketch - of - a - proof ] ] sketch of a proof : + + + + + + + + + + + + + + + + + + in order to make the paper self - contained , we give here the main idea of the proof of proposition [ martprop ] , although there is some overlap with @xcite . + at first , it is a direct consequence of the mnw - construction that the generator @xmath93 of the pure - jump markov process @xmath7 , @xmath23 , is given by @xmath94}\sum_{f\in{\operatorname{cells}}(y\cap h)}[f(y\cup f)-f(y)]\lambda(dh),\ ] ] where @xmath79 stands for some instant of @xmath7 and @xmath95 is a bounded and measurable function on the space of tessellations of @xmath4 . applying now the classical dynkin formula ( see for example lemma 19.21 in @xcite ) with @xmath96 gives the martingale property of the random process ( [ mart1 ] ) . ( it should be mentioned that so defined @xmath95 need not to be bounded in general . however , this technical difficulty can be overcome with a suitable localization argument . ) + for the second statement we consider the time - augmented random process @xmath97 , @xmath23 , which has generator @xmath98 given by @xmath99(y)+\left[{\partial\over\partial t}g(y,\cdot)\right](t)\ ] ] for appropriate functions @xmath100 . now , using dynkin s formula now for the product process @xmath97 and with @xmath101 , @xmath102 , gives ( again after a suitable localization argument ) the martingale property of the random process @xmath103 for another functional @xmath88 we can apply the latter property for @xmath104 and @xmath105 and subtract the two results , which shows the martingale property of the random process ( [ mart2]).@xmath106 in this subsection we discuss certain basic first - order consequences of proposition [ martprop ] to be of use for our future reference . we will write @xmath107 for the intrinsic volume of order @xmath29 with @xmath108 ( for the standard definition of these functionals we refer to @xcite and the references cited therein ) . further , write @xmath109 note now that whenever a new facet @xmath32 splits a cell @xmath42 into @xmath110 and @xmath111 of @xmath7 giving rise to a new tessellation @xmath112 , we have @xmath113 since @xmath107 has the valuation property . consequently , constructing the tessellation @xmath102 by successive cell splits we easily get @xmath114 it is our aim to relate @xmath115 and @xmath116 . for this purpose we use a special case of _ crofton s formula _ from classical integral geometry , which reads @xmath117}v_j(k\cap h)\lambda(dh)=\gamma_{j+1}v_{j+1}(k),\label{eqcrofton}\ ] ] where @xmath118 is a convex body and where the constant @xmath57 is given by ( [ gammaj ] ) , see ( * ? ? ? applying this formula to @xmath115 yields @xmath119}\sum_{f\in{\operatorname{cells}}(y\cap h)}v_j(f)\lambda(dh)=\int_{[w]}\sum_{c\in{\operatorname{cells}}(y)}v_j(c\cap h)\lambda(dh)\\ & = & \sum_{c\in{\operatorname{cells}}(y)}\gamma_{j+1}v_{j+1}(c)=\gamma_{j+1 } f_{j+1}(y).\label{avjtechnicaltool}\end{aligned}\ ] ] to streamline the notation , we shall write @xmath120 and @xmath121 below . note that , upon combining ( [ fjsj ] ) and ( [ avjtechnicaltool ] ) , it follows from ( [ mart1 ] ) with @xmath122 that @xmath123 are all @xmath91-martingales for @xmath124 in particular , bearing in mind that @xmath125 we conclude that @xmath126 is a @xmath91-martingale , in particular @xmath127 , where we have used @xmath128 . the latter equation is extended by [ propofirstorder ] for @xmath129 we have @xmath130 with the convention that @xmath131 and @xmath132 . [ [ proof ] ] proof : + + + + + + taking expectations in ( [ martexpl1ord ] ) we get @xmath133 continuing recursively by applying ( [ fjsj ] ) and ( [ avjtechnicaltool ] ) we end up with @xmath134 however , the definitions of @xmath39 and @xmath135 imply that @xmath136 for any @xmath137 . thus , @xmath138 and the result follows immediately.@xmath106 + + it is interesting to compare the mean value formula from the last proposition with that from @xcite , section 3.2 . we denote by @xmath139 the density of the @xmath29-th intrinsic volume of the collection of @xmath26-dimensional maximal polytopes of @xmath49 , i.e. @xmath140 for arbitrary compact windows @xmath1 with @xmath141 , see ( * ? ? ? using now proposition [ propofirstorder ] we get @xmath142 because of the homogeneity of intrinsic volumes . on the other hand , we have shown in @xcite that @xmath143 where here and in the sequel we denote by @xmath144 , @xmath145 , the volume of the @xmath146-dimensional unit ball . indeed , these two values are identical , since @xmath147 [ rema1 ] our first - order formula from the last proposition contains the extra boundary correction term @xmath148 in contrast to our mean value formula ( [ xxxeqmean ] ) from @xcite . these additional terms come from the fact that we consider maximal polytopes possibly chopped off by the boundary of the domain @xmath4 rather than the points of an associated center function for full facets in @xmath149 thus , it may happen for instance that in two neighbouring regions one observes two facets which can coalesce when putting these regions together into one volume , whence the lower - order finite volume corrections arise . in this section we develop a full second - order theory for intrinsic volumes of stationary and isotropic random stit tessellations . this is first done exactly in subsection [ subsecexact ] for @xmath150 inside a bounded convex observation window @xmath1 . then , in subsection [ subsecasympt ] we derive the corresponding asymptotic expressions for @xmath151 to proceed with second - order calculations we shall use the notation already introduced in subsection [ subsec1stord ] above . observe first that in view of ( [ fjsj ] ) the relation ( [ avjtechnicaltool ] ) simplifies and we obtain @xmath152 recalling that @xmath120 and @xmath153 putting likewise @xmath154 and using ( [ mart2 ] ) in proposition [ martprop ] with @xmath155 and @xmath156 , we see that @xmath157 ds\ ] ] is a martingale with respect to the filtration @xmath91 induced by @xmath158 . substituting ( [ baraj ] ) , this implies that @xmath159 ds\ ] ] is an @xmath91-martingale as well . this is a crucial formula because , upon taking expectations , it allows us to express @xmath160 in terms of corresponding covariances with indices @xmath161 and @xmath162 in other words , we get @xmath163 ds.\ ] ] it is important to observe that this recursion terminates because @xmath164 , which allows us to provide an explicit expression for the covariances @xmath160 . to obtain the desired formula , denote @xmath165 for any @xmath166 \to { \bbb r}$ ] and @xmath167 for which the iterated integral exists . then we claim that for @xmath168 it holds @xmath169 @xmath170 with the convention that @xmath171 and @xmath172 whereas this can be readily verified by a straightforward induction in view of ( [ principalcov ] ) , there is a more natural way to see it . in fact , in the course of recursive applications of ( [ principalcov ] ) the covariance @xmath173 is represented in terms of @xmath174 and @xmath175 which can be interpreted in terms of lattice walks on pairs of indices from @xmath176 to @xmath177 where only steps @xmath178 and @xmath179 are allowed , each step @xmath178 involving multiplication by @xmath180 plus integral iteration and each step @xmath179 resulting in multiplication by @xmath181 plus integral iteration . there are @xmath182 such walks , whence the first line in ( [ auxeq ] ) follows . the second line of ( [ auxeq ] ) is due to additional @xmath183 terms in ( [ principalcov ] ) , which are born at all times @xmath184 of the lattice walk discussed above , with @xmath185 standing for the number of @xmath178 steps and @xmath186 for the number of @xmath179 steps . note that no additional term is born at @xmath187 though , as it corresponds to having reached the ( co)variance @xmath188 , which does not get expanded any further . in ( [ auxeq ] ) we have substituted @xmath189 and @xmath190 for notational convenience . + applying now ( [ principalcov ] ) for @xmath191 and using that @xmath164 we see that @xmath192 whence we finally get from ( [ auxeq ] ) the following exact variance expression : [ thmconvexact ] the covariance between the intrinsic volume processes @xmath193 and @xmath194 for @xmath168 of a stationary and isotropic random stit tessellation @xmath7 is given by @xmath195 @xmath196 with the usual convention that @xmath171 and @xmath172 unfortunately , we are not able to make the covariance in the above theorem any more explicit . this is due to the presence of mixed moments @xmath197 whose evaluation is technically related to the problem of providing a general formula for joint moments of lower - order intrinsic volumes of @xmath26-dimensional poisson cells ( arising as hyper - planar sections of @xmath7 , which are in addition possibly chopped off by the boundary of the window @xmath4 ) which is not currently available in required generality up to the best of our knowledge . fortunately though , the offending mixed moments turn out to be of negligible order in large window size asymptotics and we are able to provide fully explicit asymptotic formulae in subsection [ subsecasympt ] below . + we would like to point out that in the special case @xmath8 our theorem [ thmconvexact ] reduces to theorem 1 and corollary 1 in @xcite where we have studied variances and central limit theory for maximal edge ( i - segment ) lengths and vertex counts and where we could give fully explicit exact variance formulae exploiting the particular features of the planar setting . in corollary 2 ibidem we also have provided asymptotic expressions for these variances for sequences of growing observation windows . for this reason , in our asymptotic considerations below we will restrict to the case @xmath9 where the general asymptotic expressions are essentially different from those arising in the exceptional planar case . this is due to the variance dichotomy established in @xcite and mentioned in section [ secintro ] above . let @xmath1 be a compact convex window and consider the sequence @xmath198 . we will write from now on @xmath199 instead of @xmath200 and likewise @xmath201 for @xmath202 in order to emphasize the dependence on @xmath12 . our main interest in this section is to derive from theorem [ thmconvexact ] an asymptotic expression for the covariances @xmath203 as @xmath204 . we start with the following [ propaux1 ] for @xmath205 with @xmath0 fixed we have @xmath206 [ [ proof-1 ] ] proof : + + + + + + observe first that , in view of ( [ itint ] ) , the integral @xmath207 only involves integration with respect to the variable @xmath208 and does not affect the order in @xmath12 . thus , it is sufficient to prove the proposition for @xmath209 and @xmath210 which we shall henceforth assume without loss of generality . we have @xmath211 where @xmath212 and @xmath213 to provide a bound for @xmath214 we need some additional notation . write @xmath215 for the common value of @xmath216 with @xmath217 ranging through hyperplanes in @xmath3 and where @xmath218 the typical cell of the sectional stit tessellation @xmath219 . using the scaling property ( [ stitscaling ] ) and the homogeneity of the intrinsic volumes we readily get @xmath220 indeed , @xmath221 to proceed , write , recalling ( [ adef ] ) , @xmath222}\sum_{f\in{\operatorname{cells}}(y(s , w_r)\cap h)}v_{d-1-m}(f)v_{d-1-n}(f)\lambda(dh)\\ \nonumber & \leq & { \bbb e}\int_{[w_r]}\sum_{f\in{\operatorname{cells}}(y(s)\cap h),\ ; f \cap w_r \neq \emptyset } v_{d-1-m}(f)v_{d-1-n}(f)\lambda(dh)\\ \nonumber & = & \int_{[w_r ] } \frac{\varsigma_{m , n;s}}{\varsigma_{0,d-1;s } } { \bbb e}{\operatorname{vol}}_{d-1}({\operatorname{typicalcell}}(y(s ) \cap h ) \oplus ( -(w_r \cap h ) ) ) \lambda(dh),\end{aligned}\ ] ] where @xmath223 stands for the usual minkowski addition . recalling that @xmath224 is poisson , see section [ subsecmnwc ] , we readily get for @xmath225 , @xmath226 and thus we conclude in view of ( [ bound1 ] ) that , for @xmath227 @xmath228 consequently , @xmath229 next , we find a bound for @xmath230 to this end , we note that during the time interval @xmath231 $ ] there are @xmath232 cell splits within @xmath233 in the course of the mnw - construction and hence the expectations of sums @xmath234 are of order @xmath235 uniformly in hyperplanes @xmath236 $ ] and in @xmath237.$ ] thus , using that @xmath238 ) = o(r)$ ] and recalling the definition ( [ adef ] ) we obtain @xmath239 putting ( [ bounds1 ] ) and ( [ bounds2 ] ) together completes the proof of proposition . @xmath106 + + this implies that asymptotically , as @xmath204 , the terms appearing in theorem [ thmconvexact ] with @xmath240 are of order at most @xmath241 and thus negligible compared with the leading @xmath242-term . for this reason , the asymptotic behaviour of @xmath243 is dominated by the term with @xmath244 , which is of order @xmath245 where by @xmath246 we mean quantities bounded both from below and above by multiplicities of the argument in the usual landau notation . for this dominating case proposition [ propaux1 ] is refined by [ propasympt ] we have for @xmath168 @xmath247 [ [ proof-2 ] ] proof : + + + + + + first , recall from thm . 4 in @xcite that the variance of the total surface area of @xmath7 equals @xmath248 the main argument for ( [ totalvar ] ) reads as follows : at first , we use the fact that stit tessellations have poisson typical cells and find @xmath249}\sum_{f\in{\operatorname{cells}}(y(s , w)\cap h)}v_{d-1}^2(f)\lambda(dh)\\ \nonumber & = & \int_{[w]}\int_{w\cap h}\int_{w\cap h}{\bf 1}[x , y\ \text{are in the same cell of}\ y(s , w)\cap h]dxdy\lambda(dh)\\ \nonumber & = & \int_{[w]}\int_{w\cap h}\int_{w\cap h}e^{-s\lambda([\overline{xy}])}dxdy\lambda(dh),\end{aligned}\ ] ] where @xmath250 $ ] stands for the usual indicator function and @xmath251 for the line segment connecting @xmath252 and @xmath253 . using now ( [ mart2 ] ) with @xmath254 , taking expectations and noting that the mixed term vanishes leads to @xmath255}\int_{w\cap h}\int_{w\cap h}e^{-s\lambda([\overline{xy}])}dxdy\lambda(dh)\\ \nonumber & = & \int_{[w]}\int_{w\cap h}\int_{w\cap h}{1-e^{-t\lambda([\overline{xy}])}\over\lambda([\overline{xy}])}dxdy\lambda(dh).\end{aligned}\ ] ] the latter integral can be transformed into ( [ totalvar ] ) by using an integral - geometric formula of blaschke - petkantschin - type and the crofton formula ( [ eqcrofton ] ) see @xcite for details . + recall now ( [ vard1expr ] ) and write @xmath256 where the last equality follows by the scaling property ( [ stitscaling ] ) of stit tessellations . thus , @xmath257 now , the binomial theorem implies that there exists some constant @xmath258 such that @xmath259 with @xmath260 . in view of ( [ eeee1 ] ) this proves the desired result.@xmath106 + + consequently , by combining propositions [ propaux1 ] and [ propasympt ] with theorem [ thmconvexact ] we get [ corasycov ] asymptotically , as @xmath204 we have for @xmath168 @xmath261 @xmath262 where @xmath263 denotes the @xmath264-energy of @xmath4 , i.e. @xmath265 the affine blaschke - petkantschin formula ( * ? ? ? 7.2.7 ) can be used to provide an integral - geometric expression for the measure - geometric energy functional @xmath266 . in fact , we have @xmath267 where @xmath268 denotes the @xmath26-st chord power integral of the convex body @xmath4 in the sense of @xcite , page 363 , i.e. @xmath269 where @xmath270 denotes the space of lines in @xmath3 with invariant measure @xmath271 and @xmath272 is the volume of the @xmath273-dimensional unit ball . this means that in the asymptotic covariance formula from corollary [ corasycov ] the dependence on the geometry of @xmath4 is encoded by the non - additive @xmath263 or equivalently by @xmath268 . we will from now on use the representation in terms of chord power integrals , as it allows an easier comparison with other tessellation models . summarizing , this yields the following asymptotically , as @xmath204 , we have for @xmath168 @xmath274 and for @xmath275 @xmath276 especially for the practically relevant case @xmath277 , we have for @xmath278 and @xmath279 @xmath280 in general , @xmath268 or equivalently @xmath263 is rather difficult to evaluate explicitly . however , for the unit ball @xmath281 in @xmath3 we have by applying @xcite , theorem 8.6.6 ( with a corrected constant ) @xmath282 for example for @xmath277 this gives us the value @xmath283 . for the interpretation of computer simulations it is of particular interest to evaluate the chord power integral @xmath284 for a @xmath285-dimensional cube @xmath286 with side length @xmath287 . it can be shown that the numerical value of @xmath284 is given by @xmath288 . in this section we focus our attention on the spatial pair - correlation structure for the processes of maximal polytopes of arbitrary dimensionalities induced by @xmath289 arising for @xmath108 as random @xmath29-volume measures concentrated on the union of all @xmath29-faces of @xmath26-dimensional maximal polytopes of @xmath49 . the nature of the so defined face measures is somewhat different than that of cumulative intrinsic volume processes considered above , in particular the face measures keep track not just of the cumulative numeric characteristics of stit tessellations but also of their spatial profile , moreover it should be emphasized that in general the total mass of a face measure may be quite unrelated to the cumulative intrinsic volume of the corresponding order , as for example @xmath290 , the number of @xmath26-dimensional maximal polytopes of @xmath49 in @xmath1 , is not deterministically related to the number of vertices of @xmath7 as soon as @xmath15 etc . however , we decided to consider the face measures in this paper as they are of interest in their own right and supplement our results from the other sections , showing the power and versatility of the methods developed in this paper . + for @xmath291 we consider the ( random ) @xmath29-th order face measure @xmath292 where @xmath293 is the @xmath294-truncated @xmath29-volume measure @xmath295(a ) = { \operatorname{vol}}_j(a \cap e),\ ; a \subseteq { \bbb r}^d , $ ] whereas @xmath296 where by @xmath31 we mean the collection of all @xmath29-faces of the @xmath26-dimensional polytope @xmath32 . we also abuse the notation by putting @xmath297 we shall be interested in the covariance measures @xmath298 , on @xmath299 given by @xmath300 for all bounded measurable @xmath301 with bounded support , where the standard duality notation @xmath302 is used for integration @xmath303 denote by @xmath304 the law of the typical @xmath26-dimensional maximal polytope , of the stit tessellation @xmath49 and let @xmath305 be the corresponding facet density . the following proposition is crucial for this section : [ covrecursionlemma ] for @xmath306 we have @xmath307 @xmath308 ds\ ] ] with @xmath309 and @xmath310 given as in ( [ gammaj ] ) . [ [ proof-3 ] ] proof : + + + + + + for bounded measurable and boundedly supported @xmath311 and @xmath312 and for @xmath313 consider the facet functionals @xmath314 and @xmath315 defined analogously . then , choosing some compact convex @xmath4 containing the supports of @xmath311 and @xmath316 in its interior we see that , recalling ( [ sigmadef ] ) and ( [ adef ] ) , @xmath317 } \sum_{f \in { \operatorname{cells}}(y \cap h ) } j^g_i(f ) \lambda(dh ) = \sum_{e \in { \operatorname{maxpolytopes}}_{i+1}(y ) } \left\langle g , \int_{[w ] } { \operatorname{vol}}_i(\cdot \cap ( e \cap h ) ) \lambda(dh ) \right\rangle\ ] ] @xmath318 where we have used the crofton formula ( [ eqcrofton ] ) . applying now ( [ mart2 ] ) with @xmath319 and @xmath320 we get upon taking expectations @xmath321 @xmath322 ds\ ] ] and thus , in view of ( [ aformeas ] ) , @xmath323 @xmath324 ds.\ ] ] using that @xmath325 and similar relationships , we end up with @xmath326 @xmath327 ds \right\rangle.\ ] ] thus , to establish ( [ covmeasrecursion ] ) it is enough to show that @xmath328 to prove ( [ covaux2 ] ) use ( [ stitsection ] ) and write , recalling ( [ adef ] ) , @xmath329 } { \bbb e } \sum_{f \in { \operatorname{cells}}(y(s , w ) \cap h ) } j^g_i j^h_j(f)\lambda(dh)\\ & = & \int_{[w ] } \int j^g_i j^h_j(f ) { \bbb m}^{y(s , w ) \cap h}(df ) \lambda(dh),\label{eqxxx } \end{aligned}\ ] ] where @xmath330 is the cell intensity measure for the sectional stit tessellation @xmath331 , compare with ( [ cellintensitymeasure ] ) . using proposition [ propxx ] ( a ) we are led to @xmath332 } \int j^g_i j^h_j(f ) { \bbb m}^{y(s , w ) \cap h}(df ) \lambda(dh)=\int_{[w ] } \int j^g_i j^h_j(f ) { \bbb m}^{{\operatorname{pht}}(s , w ) \cap h}(df ) \lambda(dh).\ ] ] now , using slivnyak s theorem ( cf . 3.2.5 ) ) we obtain for any bounded measurable function @xmath77 on the space of @xmath26-dimensional polytopes @xmath333}\int\phi(f){\bbb m}^{{\operatorname{pht}}(1,w)\cap h}(df)\lambda(dh ) & = & \int\int_{[c]}\phi(c\cap h)\lambda(dh){\bbb m}^{{\operatorname{pht}}(1,w)}(dc)\\ \nonumber & = & \int\phi(f){\bbb f}_{d-1}^{{\operatorname{pht}}(1,w)},\end{aligned}\ ] ] where @xmath334 is the cell intensity measure of @xmath335 , whereas @xmath336 is the cell intensity measure and @xmath337 is the @xmath26-face intensity measure of a poisson hyperplane tessellation @xmath338 within @xmath4 having intensity measure @xmath17 . thus , replacing @xmath17 by @xmath339 , @xmath340 , we arrive at @xmath332}\int\phi(f){\bbb m}^{{\operatorname{pht}}(s , w)\cap h}(df)\lambda(dh)={1\over s}\int\phi(f){\bbb f}_{d-1}^{{\operatorname{pht}}(s , w)}(df),\ ] ] whence continuing ( [ eqxxx ] ) we find @xmath341 hence , by applying proposition [ propxx ] ( b ) we get from ( [ aaux1 ] ) , @xmath342 where , recall , @xmath343 is the @xmath26-dimensional maximal polytope intensity measure for @xmath7 , see ( [ faceintensitymeasure ] ) . in view of ( [ aaux2 ] ) the required relation ( [ covaux2 ] ) follows now directly by the definition of typical @xmath26-dimensional maximal polytope ( * ? ? ? * chap . 4.1 , chap . 10 ) and the definition of @xmath344 upon taking into account that the supports of both @xmath311 and @xmath316 are contained in the interior of @xmath345 putting ( [ covaux1 ] ) and ( [ covaux2 ] ) together yields ( [ covmeasrecursion ] ) and thus completes the proof . @xmath106 + + to proceed note that @xmath346 is a constant measure and therefore any covariances involving it vanish . thus , using ( [ covmeasrecursion ] ) in proposition [ covrecursionlemma ] and arguing as in the derivation of theorem [ thmconvexact ] from the crucial relation ( [ principalcov ] ) we obtain [ thmcovmeas ] for @xmath347 we have @xmath348 @xmath349 { \cal i}^{k+l - m - n}\left(\lambda^{(d-1)}_{y(\cdot ) } { \bbb q}^{y(\cdot)}_{d-1};t \right)(df ) dx\ ] ] with the usual convention that @xmath171 and @xmath172 it is interesting to note that for the particular case @xmath350 we get [ covcor1 ] we have @xmath351{\bbb q}^{y(t)}_{d-1}(df ) dx.\ ] ] this means that for @xmath350 the covariance measure of the surface area process @xmath352 coincides with that of the surface area process induced by the homogeneous and isotropic boolean model with grain distribution @xmath304 and with grain density @xmath353 a result first established by weiss , ohser and nagel @xcite in the special planar case by completely different methods . recall further from theorem 3 in @xcite that @xmath354 where @xmath355 stands for the distribution of the typical facet of the poisson hyperplane tessellation @xmath356 with surface intensity @xmath208 ( this is to say , the hyperplane process has intensity measure @xmath339 ) . but this random polytope has the same distribution @xmath357 as the typical cell of a poisson hyperplane tessellation in @xmath358 with surface intensity @xmath359 with @xmath360 given by ( [ gammaj ] ) . thus , @xmath361{\bbb q}_{{\operatorname{typicalcell}}}^{{\operatorname{pht}}(\gamma_{d-1}s,{\bbb r}^{d-1})}(dc)dxds\ ] ] and the covariance measure is reduced to known quantities . + for @xmath362 the situation becomes more complicated and mixtures of typical cell distributions corresponding to different time moments arise in the right hand side of ( [ covmeasfullexpr ] ) . as in section [ sec2ndord ] above , also here explicit calculations are precluded for @xmath362 due to the lack of known formulae for mixed moments of general order intrinsic volumes of poisson cells . in this section we present a central limit theory for the suitably rescaled intrinsic volume processes @xmath363 in strong contrast to the two - dimensional case considered in @xcite , where a classical gaussian limit behaviour is observed , for the case @xmath364 in focus of this paper the situation is very different and non - gaussian limits arise . in this context , to proceed with a full discussion below , we recapitulate first some facts already known from @xcite in a way specialized for our present purposes . define the rescaled intrinsic volume processes @xmath365}$ ] for @xmath366 by @xmath367.\ ] ] note that the shift by @xmath368 in time argument here is of technical importance as placing the time origin just after the very initial _ big bang _ phase of the mnw - construction , where the dominating fluctuations of @xmath369 arise , but where nothing of asymptotic significance happens for @xmath370 since the big bang phase evolution for @xmath369 has been considered in full detail in ( * ? ? ? 5 ) , in this paper we only focus on the later phase @xmath371.$ ] putting together the present section with ( * ? ? ? 5 ) reveals remarkable richness of the complete asymptotic picture for stit tessellations in large windows . + we begin with the observation that follows by the theory developed in subsection 5.3 of @xcite : [ vd1 ] the process @xmath372}$ ] converges in law , as @xmath373 in the space @xmath374 $ ] of right - continuous functions with left - hand limits ( cdlg ) on @xmath375 $ ] endowed with the usual skorokhod @xmath376-topology ( * ? ? ? vi.1 ) to the constant process @xmath377 where @xmath378 is a certain non - gaussian square - integrable random variable with variance @xmath379 the relation ( 85 ) in @xcite implies that @xmath380 converges in law to @xmath381 as @xmath382 on the other hand , recall that @xmath383 is a martingale in view of ( [ martexpl1ord ] ) . moreover , by ( [ totalvar ] ) and the definition of @xmath383 combined with the scaling property of stit tessellations we have @xmath384 ||x - y||}}{||x - y||^2 } dx dy \end{aligned}\ ] ] and hence @xmath385 tends to @xmath386 as @xmath382 consequently , the asymptotic constancy of the limit process follows now by doob s @xmath387-maximal inequality ( * ? ? ? * thm 3.8(iv ) ) which completes our argument . @xmath106 + + [ nongaussexpl ] the convergence in law and non - gaussianity results referred to above come from the paper @xcite . in order to keep the present work formally self - contained we give here a very brief sketch of the main arguments . + the convergence is readily guaranteed by the martingale convergence theory as ibidem . the crucial point is the non - gaussianity . here , the idea , as described in detail in section 5.3 of @xcite , relies on showing that @xmath381 has its tails much heavier than normal random variables . we proceed , roughly speaking , by constructing a suitable class of _ initial frame tessellations _ of the body @xmath4 with the properties that * the number of frame hyperplanes hitting @xmath4 equals @xmath388 , * the frame hyperplanes intersect within @xmath4 only very seldom , which implies that @xmath4 gets subdivided into @xmath389 cells ( recall the landau notation ) . this can be achieved by choosing @xmath273 principal directions and keeping all hyperplanes in the collection under construction approximately aligned to these directions . importantly , it turns out that this way we can get the lower bound @xmath390 for the probability that the real initial frame arising in the very initial phase for @xmath391 usually called _ big bang _ , see below upon suitable spatial re - scaling does fall into the desired class . now , denoting the last event by @xmath392 , we see that on @xmath392 the total deviation @xmath393 gets decomposed into * a sum of independent contributions @xmath394 coming from all respective cells @xmath395 of the frame tessellation , * a correction term of order @xmath389 due to the centred joint contribution of all hyperplanes of the frame tessellation ( observe that each such hyperplane necessarily contributes @xmath396 , whereas @xmath397 ) . finally we can use the knowledge of @xmath398 and @xmath399 and properly exploit the independence of the random variables @xmath394 , to conclude that on @xmath392 we can find a further sub - event @xmath400 with @xmath401 such that @xmath402 and hence @xmath403 , which could not hold if @xmath381 were gaussian . we refer the reader to @xcite for further details . [ ivlimtheo ] the vector @xmath404}$ ] converges in law , as @xmath373 in the space @xmath405;{\bbb r}^d)$ ] of @xmath3-valued cdlg functions on @xmath375 $ ] endowed with the usual skorokhod @xmath376-topology ( * ? ? ? * chap.vi.1 ) , to the stochastic process @xmath406.\ ] ] before proceeding with the proof we discuss certain striking features of the phenomenon described in theorem [ ivlimtheo ] . namely , although all intrinsic volume processes @xmath407 exhibit fluctuations of order @xmath408 the nature of these fluctuations differs very much between @xmath409 and @xmath410 * as shown in subsection 5.3 in @xcite and in proposition [ vd1 ] above , the leading - order deviations of @xmath411 in large @xmath12 asymptotics arise very early in the course of the mnw - construction , in its initial stages usually referred to as the _ big bang phase_. here , this is the time period @xmath412.$ ] during the later stages of the construction , i.e. say the time interval @xmath413,$ ] the variance increase is of lower order and any newly arising fluctuations are negligible compared to those originating from the _ big bang_. in the asymptotic picture this means that the initial fluctuation remains _ frozen _ throughout the rest of the dynamics , whence the constant limit for @xmath372}$ ] as @xmath382 * in contrast , the leading - order deviations of @xmath414 for @xmath415 arise and cumulate constantly in @xmath0 with deterministic polynomial rates depending on @xmath416 times the initial _ big bang _ fluctuation of the process @xmath417 which , in this sense , _ stores the entire randomness _ of the intrinsic volume vector . the mechanism determining the dependence of fluctuations of intrinsic volume processes of orders @xmath415 given those for @xmath418 and the resulting form of the limit process ( [ limproc ] ) will be discussed in the sequel . the crucial step of the proof relies on considering for each @xmath108 the auxiliary process @xmath419 which is a centred @xmath91-martingale by ( [ mart1 ] ) and which is the same as @xmath420 by ( [ baraj ] ) . the idea is to show that for @xmath421 the processes @xmath422 are of negligible order in large @xmath12 asymptotics . indeed , upon squaring and taking expectations we get @xmath423 using that @xmath424 as follows by the martingale property of @xmath425 and inserting this to ( [ varsigmaaux ] ) we are led to @xmath426 @xmath427 with the last two terms cancelling . thus , @xmath428 whence , by ( [ principalcov ] ) , @xmath429 in view of proposition [ propaux1 ] for @xmath421 we have @xmath430\ ] ] and thus the relation ( [ varsigmaaux2 ] ) implies @xmath431\ ] ] for @xmath432 hence finally @xmath433 } ( \hat\sigma_{v_j;t}^{w_r})^2 = o(r^{2(d-1)})\ ] ] for @xmath421 by doob s @xmath387-maximal inequality ( * ? ? ? * thm 3.8(iv ) ) applied to the martingale @xmath434 defined as in ( [ sigmaaux ] , [ sigmaaux2 ] ) with @xmath4 replaced by @xmath233 there according to our usual notational convention . + with ( [ finalnegl ] ) it is now easy to complete the proof . indeed , since the normalization in the definition of intrinsic volume processes @xmath435 involves a prefactor @xmath436 the relation ( [ finalnegl ] ) allows us to recursively substitute @xmath437 for @xmath438 , as soon as @xmath421 , without affecting the large @xmath439 asymptotics in law ( note that the technically motivated shift by @xmath440 in time argument of the rescaled process @xmath435 is asymptotically negligible as inducing only a negligible @xmath387-difference precisely calculated in theorem [ thmconvexact ] ) . the application of such recursive substitutions combined with proposition [ vd1 ] completes the proof of theorem [ ivlimtheo ] . @xmath106 + + it is crucial to emphasize at this point that , in intuitive terms , under the normalization of @xmath441 @xmath442 the mechanism governing the rise of fluctuations of intrinsic volume processes of orders @xmath443 given those for @xmath135 reduces effectively to the simple approximation @xmath444 and its recursive application . of course this _ simple approximation _ follows itself by rather non - trivial arguments above . clearly , this description only characterizes the leading order fluctuations as considered in theorem [ ivlimtheo ] although our tools should allow a more delicate characterization of non - leading lower order fluctuations as well , see e.g. subsection 5.1 in @xcite for the particular case of @xmath445 we are grateful to claudia redenbach and joachim ohser who provided the two pictures of stit tessellations . we would also like to thank the anonymous referees for their comments and suggestions as well as the editor - in - chief for his help . + the first author was supported by the polish minister of science and higher education grant n n201 385234 ( 2008 - 2010 ) . the second author was supported by the swiss national science foundation , grant pp002 - 114715/1 . t. schreiber , c. thle , second - order properties and central limit theory for the vertex process of iteration infinitely divisible and iteration stable random tessellations in the plane , adv * 42 * , 913935 ( 2010 ) .

the internet is a primary example of a large - scale , self organized , complex system . understanding the processes that shape its topology would provide tools for engineering its future . the internet is assembled out of multiple autonomous systems ( ass ) , which are contracted by different economic agreements . these , in turn , compose the routing pathways among the ass . with some simplifications , we can represent the resulting network as a graph , where two nodes ( ass ) are connected by a link if traffic is allowed to traverse through them . the statistical properties of this `` internet graph '' , such as the clustering properties , degree distribution etc . , have been thoroughly investigated @xcite . however , without proper understanding of the mechanism that led to this structure , the statistical analysis alone lacks the ability to either predict the future evolution of the internet nor to shape its evolution . a large class of models , a primary example is the `` preferential attachment '' model @xcite , use probabilistic rules in order to simulate the network evolution and recover some of its statistical properties . yet , these models fail to account for many other features of the network @xcite . possibly , one of the main reasons for that is they treat the ass as passive elements rather than economic , profit - maximizing entities . therefore , an agent - based approach is a promising alternative . game theory is one of the main tools of the trade in estimating the performance of distributed algorithms @xcite . it describes the behavior of interacting rational agents and the resulting equilibria . game theory has been applied extensively to fundamental control tasks in communication networks , such as flow control @xcite , network security @xcite , routing @xcite and wireless network design @xcite . recently , there has been increased activity in the field of _ network formation games_. these studies aim to understand the network structure that results from interactions between rational agents @xcite . different authors emphasized different contexts , such as wireless networks @xcite or the inter - as topology @xcite . the theme of such research is to investigate the equilibria s properties , e.g. , establishing their existence and obtaining bounds on the `` price of anarchy '' and price of stability . these metrics measure from above and below , correspondingly , the social cost deterioration at an equilibrium compared with a ( socially ) optimal solution . alternatively , agent - based simulations are used in order to obtain statistical characteristics of the resulting topology @xcite . nonetheless , the vast majority of these studies assume that the players are identical , whereas the internet is a heterogeneous mixture of various entities , such as cdns , minor isps , tier-1 ass etc . there are only a few studies that have explicitly considered the effects of heterogeneity on the network structure . some examples include @xcite , which extends a previous model of formation games for directed networks @xcite , and in the context of social networks , @xcite . the latter describes a network formation game in which the link costs are heterogeneous and the benefit depends only on a player s nearest neighbors ( i.e. , no spillovers ) . most of the studies on the application of game theory to networks , with very few exceptions , e.g. , @xcite , focused on static properties of the game . this is particularly true for network formation games . however , it is not clear that the internet has reached an equilibrium . indeed , ass continuously draw new contracts , some merge with others while other quit business . in fact , a dynamic inspection of the inter - as network suggests that the system may be far from equilibrium . therefore , a _ dynamic _ study of an inter - as network formation game is needed . in addition , previous work ignores an important requirement that autonomous systems has - reliability . indeed , failures occur , and an as must face such events . while some game theoretic works addressed reliability in other contexts @xcite , to the best of our knowledge , there are no works that considered the _ topological properties _ that emerge in a _ heterogeneous , dynamic , _ network formation game with _ reliability constraints . _ we establish an analytically - tractable model , which explicitly accounts for the heterogeneity of players as well as reliability requirements . we base our model on the heralded fabrikant model @xcite , which was recently extended to include heterogeneous players @xcite . we model the inter - as connectivity as a network formation game with _ heterogeneous _ players that may share costs by monetary transfers . we account for the inherent bilateral nature of the agreements between players , by noting that the establishment of a link requires the agreement of both nodes at its ends , while removing a link can be done unilaterally . as reliability comes into play , agents may require to be connected to other agents , or to all the other agents in the network , by at least two disjoint paths . we investigate both the static properties of the resulting game as well as its dynamic evolution . game theoretic analysis is dominantly employed as a `` toy model '' for contemplating about real - world phenomena . it is rarely confronted with real - world data . in this study we go a step further from traditional formal analysis , and we do consider real inter - as topology data analysis to support our theoretical findings . the main contributions of our study are as follows : * in the context of network formation games , we provide a theoretical framework that introduces reliability constraints . we discuss both the case of frequent failures , where the fall - back pathways are as frequently used as the main pathways , as well as the case of rare crashes . * we introduce the concept of _ price of reliability _ , which is defined as the ratio of the social cost with reliability constraints to the social cost with no such additional constraints . surprisingly , we show that this price can be smaller than one , namely , that the additional reliability requirements may increase the social utility . * we provide dynamical analysis of topological quantities , and explain the prevalence of some `` _ _ network motifs _ _ '' , i.e. , sub - graphs that appear frequently in the network . through real - world data , we provide encouraging support to our predictions . in the next section , we describe our model . we discuss alternative variants that address different failure frequencies or whether utility transfers ( e.g. , monetary transfers ) are allowed or not ; we address both the case of allowing utility ( i.e. , monetary ) transfers as well as the case where this is not possible . next , in section 3 , we provide static analysis . dynamic analysis is presented in section 4 . in section 5 we compare our theoretical predictions with real - world data on inter - as topologies . finally , conclusions are presented in section 6 . we assume that each as is a player . while there are many types of players , following @xcite , we aggregate them into two types : _ major league _ ( or t__ype - a _ _ ) players , such as major isps , central search engines and the likes , and _ minor league _ ( or _ type - b _ ) players , such as local isp or small enterprises . each player , regardless of its type , may form contracts with other players , and should they reach a mutual understanding , a link between them is formed . a player s strategy is set by specifying which links it is interested in establishing , and , if permissible , the price it will be willing to pay for each . in order to maintain reliable routing pathways , players may be required to sustain at least _ two disjoint paths _ to other players or a subset of players . we denote the set of type - a ( type b ) player by @xmath0 ( @xmath1 ) . a link connecting node @xmath2 to node @xmath3 is denoted as either @xmath4 or @xmath5 . the total number of players is @xmath6 , and we assume @xmath7 . the _ shortest distance _ between nodes @xmath2 and @xmath3 is the minimal number of hops along a path connecting them and is denoted by @xmath8 . finally , the degree of node @xmath2 is denoted by @xmath9 . our cost function is based on the cost structure in @xcite and @xcite . players are penalized for their distance from other players . first and foremost , players require a good , fast connection to the major players , while they may relax their connection requirements to minor players . bandwidth usage and delay depends heavily on the hop distance , and connection quality is represented by this metric . similarly to @xcite , we weight the relative importance of a major player by a factor @xmath10 in the cost function in the corresponding distance term . the link prices represent factors such as the links maintenance costs , bandwidth allocation costs etc . different player types may incur different link costs , @xmath11 , due to varying financial resources or infrastructure . all ass must maintain access to the internet in case of a single link failure . this is tantamount to the requirement that all the players must have at least two _ disjoint _ paths to each other node . nevertheless , if either link prices are high , crash frequencies are low or the content of a minor as is of little value , players may relax their reliability requirements and demand the establishment of disjoint paths only to the major players . this is represented in the cost function by a control parameter @xmath12 , which is set to one if two disjoint paths are required to all nodes , and zero if the requirement holds for ( other ) nodes of major players only . conversely , if failures are often , then the regular and backup paths ( in the corresponding pair of disjoint paths ) are used almost as frequently . as such , they must be weighted the same in the cost function . therefore , the distance cost is composed of two terms , one represents the distance along the primary path and the other represents the distance along the backup path . the relative weight of these two terms is set by a parameter @xmath13 if failures are frequent and the likelihood of using either route is the same , we have @xmath14 . however , if failures are rare , traffic will be mostly carried across the shorter path . therefore its length should carry more weight in the cost than the length of the backup route , hence @xmath15 . this motivates the following cost function . two paths , @xmath16 and @xmath17 are disjoint if they have no node in common , namely if the unordered sets satisfy @xmath18 the cost function @xmath19 of node @xmath2 of type @xmath20 , is defined as:[cost - definition ] @xmath21 where @xmath8 and @xmath22 are the lengths of a pair of disjoint paths between @xmath23 that minimizes the cost function . @xmath8 denotes the length of the shorter path . formally , denote a pair of disjoint paths connecting player @xmath2 and player @xmath3 as @xmath24 , where @xmath25 @xmath26 is the length of shorter ( correspondingly , longer ) path . set @xmath27 then @xmath28 and @xmath29 . if there is not pair of disjoint path player @xmath2 and player @xmath3 , then @xmath30 , with @xmath31if there is not a path connecting players @xmath2 and @xmath3 we also have @xmath32 for convenience , we set @xmath33 . we assume @xmath34 the social cost is the sum of individual costs , @xmath35 . we denote the optimal ( minimal ) social cost as @xmath36 _ _ , _ _ and the social cost at the optimal stable solution is @xmath37 . the _ price of stability _ is the ratio between the social cost at the best stable solution and its value at the optimal solution , namely @xmath38 . similarly , denote by @xmath39 the highest social cost in an equilibrium . then , the _ price of anarchy _ is the ratio between the social cost at the worst stable solution and its value at the optimal solution , namely @xmath40 . note that the requirement of disjoint _ node _ paths generalizes the requirement of disjoint _ link _ paths and protects against link failures within an autonomous system . all of our results apply to both notions of disjoint path ( except a refinement of theorem [ thm : symmetric dynamic ] ; see the discussion there ) . for simplicity and generality we shall use the notion of a disjoint node paths . we denote the change in cost of player @xmath2 as after the addition ( removal ) of a link @xmath41 by @xmath42 ( correspondingly , @xmath43 ) . the abbreviation @xmath44 is often used . if @xmath14 , then the two routing pathways are used the same . in this case , the shortest cycle length @xmath45 is the relevant quantity that appears in the cost function . this can be found in polynomial time by using suurballe s algorithm @xcite . however , if @xmath15 , routing will occurs along two disjoint paths , such that the length of the shortest between the two is shortest ( among all pairs of disjoint paths ) . although the complexity of finding this pair is np - hard , first finding the shortest path and then finding the next shortest path is a heuristic that works remarkably well , both in the real - world data analysis and on the networks obtained in the theoretical discussion . the reason behind this is that , when failures are rare , information is predominantly routed along the shortest path . when players are required to establish a fall - back route , they will establish a path that is disjoint from the _ current _ routing path , namely the shortest one . the establishment of a link requires the bilateral agreement of the two parties at its ends , while removing a link can be done unilaterally . this is known as a _ pairwise - stable _ equilibrium@xcite . the players strategies are _ pairwise - stable _ if for all @xmath46 , the following hold : \a ) if @xmath47 , then @xmath48 ; \b ) if @xmath49 , then either @xmath50 or @xmath51 . the resulting graph is referred to as a _ stabilizable _ graph . the additional reliability requirements result in additional link expenses , as for example , the degree of every node needs to be at least two . the _ price of reliability _ is the ratio between the optimal social cost under the additional survivability constraint to the optimal social cost when the additional constraints are removed . the cost function,@xmath52 , of node @xmath2 of type @xmath20 , is obtained by setting @xmath53 in definition [ cost - definition ] and requiring the existence of only a _ path from player @xmath2 to any other players in the network . denote the optimal social cost without the additional survivability requirement in a pairwise stable equilibrium as @xmath54 . the _ price of reliability _ ( _ por ) _ is the ratio between the optimal value of the social costs among the set of stable equilibria , @xmath55 . surprisingly , we shall show that there exist scenarios in which reliability requirements _ increase _ the social utility , so that the price of reliability can be smaller than one . thus far , it was implicitly assumed that utility transfer is not feasible . nevertheless , often players are able to transfer utility , for example via monetary transactions . an extended model that incorporates such transfers is introduced by allowing for a monetary transaction in which player @xmath2 pays player @xmath3 some amount @xmath56 iff the link @xmath57 is established @xcite . player @xmath3 sets some minimal price @xmath58 and should @xmath59 the link is formed . [ monetary cost definition]the cost function of player @xmath2 when monetary transfers are allowed is @xmath60 . we recall the observation in @xcite that , without transfers , a link will be established only if _ both _ parties , @xmath2 and @xmath3 , reduce their costs , @xmath61 and @xmath62 . but , when monetary transfers are allowed , an edge will be established if ( and only if ) the relaxed condition @xmath63 holds . in game theoretic terms , this condition is equivalent to the requirement that the _ core _ of the two players game is non - empty . _ _ in the remainder of the paper , whenever monetary transfers are feasible , we will state it explicitly , otherwise the basic model ( without transfers ) is assumed . we shall now analyze the properties of stable equilibria , such as the _ price of anarchy _ , which is the ratio between the social cost at the worst stable solution and its value at the optimal solution , and the _ price of stability _ , which is the ratio between the social cost at the best stable solution and its value at the optimal solution . we shall further discuss topological properties that emerge from our analysis . it was shown in @xcite that if @xmath64 the only stable solution is a clique and in @xcite it was shown that if @xmath65 then the major players form a clique . one may have guessed that reliability requirements , which generally induce the creation of additional , backup edges , would ease the formation of the clique . the next proposition shows that this naive assumption is wrong , and in fact , as the frequency of failure increases , it becomes increasingly difficult to maintain the major player s clique . consider a dense set , in which every player may access all the other players within two hops by a at least two disjoint paths . a direct link between two players only reduces their mutual distance by one , and does not affect any other distance . if this link fails often , it may be used only partially , and it may not be worthy to pay its cost . hence , in this setting , counter intuitively , frequent failures end up with a sparser network . assume the frequency of failures is high , namely @xmath66 then , the type - a players form a clique if and only if @xmath67 . allowing monetary transfers does not change the result . we consider a major player s clique and ask under which conditions the removal of a link is a worthy move . consider an edge @xmath4 in this clique . since only the shortest distance between players @xmath2 and @xmath3 is affected , and is increased by one , the type - a players clique is stable if and only if @xmath68 . as the major players ( tier-1 as ) form a densely connected set , a clique - like subgraph , in the rest of the paper we shall only consider the case where @xmath67 . we also assume , trivially , that @xmath69 , as otherwise the only stabilizable network is a clique . the next proposition describes a scenario in which , surprisingly , the additional reliability constraints _ reduce _ the social cost . the optimal solution when @xmath70 and @xmath71 , according to prop . [ lem : reliable optimal 1 ] . the type - b players are in red circles . the type - a clique is in blue squares . ] [ lem : reliable optimal 1 ] assume @xmath70 and symmetric reliability requirements , namely @xmath71 . then , the optimal network is composed of a type - a clique , where all the type @xmath72 nodes are connected to all members of the type - a clique , as depicted in fig . [ fig : the - optimal - solution ] . this network is not stabilizable , and @xmath73 . nevertheless , for @xmath74 , we have @xmath75 . in addition , the price of reliability is smaller than one . the main idea behind this result is that in the optimal , yet unstable solution , every minor player establishes a link with all the major players ( fig . [ fig : the - optimal - solution ] ) . this configuration is unstable as it is over - saturated with links , and the optimal stable solution is obtained by diluting this network so that every minor player will connect to just two major players . if the reliability requirements are further removed , then additional dilution occurs , increasing the social cost . in other words , the stable configuration is under - saturated with edges , and the additional survivability requirements facilitate the formation of additional links . we shall now prove the first part of this theorem , namely that the network described in fig . [ fig : the - optimal - solution ] , is optimal in terms of social cost . in this network , every type - b player is connected to every type - a player . we denote this network by @xmath76 first , consider a network in which the type - a players are not in a clique . then , there exists two players @xmath77 such that @xmath78 . by establishing a link @xmath4 the change in social cost is @xmath79 therefore a network which includes this link has a lower social cost . hence , in the optimal network , all the type - a players are in clique . a similar calculation for a missing link between player @xmath80 and @xmath81 shows that @xmath82 and establishing this link reduces the social cost as well . in conclusion , in an optimal network a type - a player is connected to all the other players . consider node @xmath83 its distance from every other node @xmath81 is @xmath84 and @xmath85 . the minimal distances between two minor players @xmath86 are @xmath87 and @xmath85 whereas in @xmath88 we have @xmath89 . assume that a network with a lower social cost exists , and denote it by @xmath90 . in @xmath90 there must exists a link between two type - b players , @xmath86 . however , removing this link reduces the social cost , as @xmath91 therefore , @xmath90 is not optimal , in contradiction to our assumption . this proves that @xmath88 is optimal . next , we are going to show @xmath88 is unstable . consider @xmath92 and @xmath93 . by removing @xmath94 we cost change of player @xmath95 is @xmath96 and therefore it is beneficial for player @xmath95 to remove this link . hence , this network in not stabilizable . finally , we are going to show that the @xmath97 . the outline is as follows . first , we are going to show that the network depicted in fig . @xmath98 is stable . the optimal stable network ( without reliability requirements ) in this parameter regime was explicitly derived in @xcite . we are going to show that the cost in the latter network is higher than the former , which bound the @xmath99 from above by a value smaller than one . consider a network in which all the minor players are connected to two major players , @xmath100 . we are going to show that this network is stable . clearly , as shown before , the type - a clique is stable . in addition , for any player @xmath92 neither @xmath101 nor @xmath2 has the incentive to remove either @xmath102 or @xmath103 as it would violate the reliability requirement and would lead to unbounded cost . hence , this network is stable . denote this network by @xmath104 in @xcite it was shown that the optimal stable network ( without reliability requirements ) in this configuration is the network in which all type - a players form a clique , and all the type - b players are connected to a single type - a player . we denote the latter network by @xmath105we have @xmath106 but , @xmath107 this concludes the proof . in conclusion , if the failure frequency is high , the survivability requirements will induce dilution of the clique of major players . however , the opposite effect occurs along the graph cut - set between the set of minor players and the set of major players , where the additional constraints lead to an increased number of links connecting major and minor players . so far we have assumed that the reliability requirements are symmetric . as explained in the introduction , in some cases it is reasonable to assume that players will require a backup route only to the major players , i.e. , non - symmetric reliability constraints . we shall now show that in this case , the social cost may deteriorate considerably . hence , from a system designer point of view , it is much more important to incentivize a configuration where the reliability requirements are symmetric , than to reduce failure frequency globally . this result stems from the following lemma . [ lem : finite distances]if @xmath108 then for every two nodes i , j , @xmath109 . if @xmath110 there exists stable equiliberia such that there are no two disjoint paths between some nodes @xmath23 . first , we are going to show that @xmath22 is finite ( step 1 ) . then , in the second step , we shall bound @xmath22 from above . note , that if there are nt two disjoint paths between players @xmath2 and @xmath3 , it is beneficial for both players to add the link @xmath4 and form a path . step 1 : if @xmath2 and @xmath3 are disconnected , then it is worthy for both of them to establish the link @xmath4 . therefore , there exists a path from @xmath2 to @xmath3 and @xmath8 is finite . assume @xmath111 . if there are nt two disjoint paths between nodes @xmath2 and @xmath112 it is worthy for both nodes to establish a direct link . hence if @xmath111 the distance @xmath22 is finite . we now discuss the case @xmath87 and show there must exists an additional disjoint path , so @xmath22 is finite . we prove by negation . take node @xmath113 . as before , both @xmath114 and @xmath115 are finite . consider the following two cases : \a ) @xmath116 , @xmath117 . then , the trajectory @xmath118 is disjoint from the path @xmath4 , and it is therefore worthy for both player @xmath2 and player @xmath3 to establish the link @xmath119 hence @xmath22 is finite . \b ) @xmath116 , @xmath120 or @xmath121 , @xmath117 . without loss of generality , we assume the first case . according to the previous discussion , there are at least two disjoint paths from @xmath95 to @xmath122 since they are disjoint , only one of them may contain the edge @xmath4 . therefore , there exist a path @xmath123 that is disjoint from the edge @xmath4 and therefore @xmath22 is finite . \c ) @xmath121 , @xmath120 . applying the same reasoning as in b ) , there exists a path from @xmath2 to @xmath95 that is disjoint from the edge @xmath4 . likewise , there exists a path from @xmath3 to @xmath95 that is disjoint from the edge @xmath119 therefore there exist a path from @xmath2 to @xmath3 that is disjoint from @xmath119 step 2 : step 1 has shown that every two nodes are connected by a cycle . let us assume the longest shortest cycle connecting two nodes is of length @xmath124 , namely there exists a cycle @xmath125 . by establishing the link @xmath126 the cost of player @xmath127 due to distances from other players is reduced by at least @xmath128 , hence the link will be established as @xmath129 and similarly @xmath130 . since @xmath131 the proof is complete . if there exist players with no two disjoint paths in an equilibrium , then the social cost becomes unbounded . this immediately results in an _ unbounded _ price of anarchy , as indicated by the following theorem . [ thm : price of anarchy]consider a network such that @xmath132 . the price of anarchy is as follows : \a ) in a setting with asymmetric reliability requirements @xmath133 : unbounded . \b ) in a setting with symmetric reliability requirements @xmath134 : bounded by @xmath135 . \a ) we shall prove this by showing that there exists a stable equilibrium with unbounded social cost . consider a network in which the major players ( type - a players ) form a clique , while every minor player is connected only to a single type - a player @xmath136 . we shall show that this network is stabilizable . as discusses before , the type - a clique is stable . consider @xmath83 by forming the link @xmath94 the change of cost of player @xmath93 , @xmath137 is @xmath138 hence , this link will be formed . no additional links between minor players will be established , since such a link only reduces the distance between the participating parties by one , @xmath139 and does not provide an additional disjoint path to the type - a clique . therefore , the social cost is @xmath140 . \b ) the total cost due to the inter - connectivity of the type - a clique is identical for all link stable equilibria and is @xmath141 . this cost is composed of @xmath142 links per node , and the distance cost to every other major nodes , @xmath143 next , we evaluate the cost due to the type - b nodes inter - distances . according to lemma [ lem : finite distances ] , both @xmath144 and @xmath145 , so the cost due to the inter - distances between a type - b player and every other player is bounded from above by @xmath146 . when summed up over all minor players , this contributes a term @xmath147 to the social cost . likewise , the cost of links that at least one of their ends is a type b player is at most @xmath148 . therefore , the maximal cost in all link stable equilibria is bounded from above by @xmath149 the optimal network configuration is described in proposition [ lem : reliable optimal 1 ] . it is straightforward to evaluate the social cost in this configuration . the distance cost due to inter - distances of the type - b players is @xmath150 while the cost of all links which connect type - b players to type - a players is @xmath151 finally , the social cost due to the type - a clique remains the same , so finally we obtain that the minimal social cost is @xmath152 by taking the limits @xmath153 we obtain that @xmath154 . a stable equilibrium with infinite social cost can be easily achieved by considering a network where all minor players are connected to a single , designated , major player . there exists a single path of at most two hops between every minor player to every major player . however , as the stability requirements are asymmetric , the major players have no incentive to establish additional routes to any minor player , and the reliability requirements of the minor players remain unsatisfied . the previous discussion assumed that a player can not compensate other players for an increase in their costs . yet , contracts between ass often do involve monetary transactions . accordingly , in this subsection we shall highlight the additional insights that are obtained when utility transfers are permissible . our first result indicates that , in this setting , in contrast to the previous setting , there always exists a fallback route between every two players , regardless of the symmetric or asymmetric nature of the additional survivability constraints . if monetary transfers are feasible , players may compensate other players for the cost of additional links such that all the additional constraints are satisfied . hence , symmetry is less important than in the previous scenario . furthermore , this result suggests that every player is connected to every other player by a cycle . the following proposition shows that the maximal cycle length decays with the number of major players . as the number of ass increases in time , this predicts that this length should decrease in time . we shall verify this prediction in section [ sec : data - analysis ] . [ prop : max distance]assume @xmath155 . then , every two players are connected by a cycle , and the maximal cycle length connecting a major player to a minor player is bounded by @xmath156 _ _ lemma [ lem : finite distances ] showed that the maximal distance between players is bounded . we are going to tighten this result in the regime where monetary transfers are feasible . denote the maximal distance between type - a player and a type_b player by @xmath157 . _ _ _ _ first , we are going to show that the maximal cycle length connecting a major player and a minor player is @xmath158 this follows from a simple geometric argument . consider two players , @xmath80 and @xmath81 . assume @xmath159 . if the cycle length is @xmath160 or greater , there exists a type - b node that its distance @xmath161 in contradiction to the assumption that the maximal distance between a major player and and a minor player is @xmath157 . denote maximal distance between two minor players by @xmath162 a similar argument shows that maximal cycle length between two minor players is @xmath163 . _ _ _ _ next , are going to show that the maximal distance connecting a major player and a minor player is @xmath164 . we prove by negation . assume that the distance between player @xmath136 and @xmath92 is @xmath165 . denote the nodes on the path as @xmath166 . then , by establishing a link between them , the distance between @xmath3 and @xmath167 ( similarly , and distance between player @xmath2 and players @xmath168 ) is reduced . in addition , player @xmath2 reduces its distance to every node of the type - a clique by @xmath165 . lemma 25 of @xcite shows that the total reduction in distance is @xmath128 . then , by establishing the link @xmath4 we have @xmath169 _ _ _ _ and as @xmath170 this expression is negative . therefore , the link will be established , and the maximal shortest distance between a major player and a minor player is smaller than @xmath165 . this concludes the proof . _ _ our second result is based on the first one , and shows that the price of anarchy is bounded . in fact , as the network grows , proposition [ prop : max distance ] also indicates that its diameter _ shrinks_. therefore , in the large network limit , the price of anarchy is bounded by a constant . the price of anarchy is bounded by @xmath135 . furthermore , if @xmath171 then the price of anarchy is upper bounded by 2 . in proposition [ prop : max distance ] s proof we showed that the maximal distance between a major player a minor player is bounded by @xmath172 it immediately follows that in the large network limit , as @xmath173 , every player is directly connected to a player that , in turn is connected to the type - a players . the latter can be either a major player , as part of the clique , or a minor player that is connected to every player in the major player s clique . note that a link between any two players @xmath174 and @xmath175 reduces the social cost , since it at least lowers the costs of @xmath174 of @xmath175 and can only reduce the costs of other players . therefore , we can bound the social cost by a configuration in which every type - b player has the minimal number of links , namely two , is at distance two from every major player . therefore , the worst social cost in bounded by @xmath176 comparing this bound with @xmath36 , as derived in proposition [ lem : reliable optimal 1 ] , in the limit @xmath153 completes the proof . the internet undergoes continuous transformations , such as the emergence of new ass , or formation of new traffic contracts . in fact , it may very well be out of equilibrium . therefore , a static analysis of the equilibrium points must be accompanied by dynamic analysis . accordingly , our main focus in this section is to identify prevalent network motifs @xcite , i.e. , small sub - graphs that emerge during the natural evolution of the network . in section [ sec : data - analysis ] we shall show that these motifs are indeed ubiquitous in the real as topology , and the frequency of their occurrences is few folds _ _ more than expected in a random network . while there are many possible equiliberia , we shall show that convergence occurs only to _ just a few_. we shall also show that the convergence time is short , namely _ linear _ in the number of players . we start the discussion by setting up the dynamic framework , as first formulated in @xcite . we split the game into _ turns _ , where at each turn only a single player is allowed to remove or initiate the formation of links . at each point in time , or turn , the players that already joined the game form a subset @xmath177 . we shall implicitly assume that the cost function is calculated with respect to the set @xmath178 of players that are already present in the network . each turn is divided into _ moves , _ at each of which a player either forms or removes a single link . a player s turn is over when it has no incentive to perform additional moves . note that disconnections of several links can be done unilaterally and hence iteratively . dynamic rule # 1 : in player @xmath2 s turn it may choose to move @xmath179 times . in each move , it may remove a link @xmath180 or , if player @xmath3 agrees , it may establish the link @xmath4 . player @xmath3 would agree to establish @xmath4 iff @xmath181 . according to this definition , during player s @xmath2 turn , all the other players will act in a greedy , rather than strategic , manner . for example , although it may be that player @xmath3 prefers that a link @xmath182 would be established for some @xmath183 , if we adopt dynamic rule # 1 it will accept the establishment of the less favorable link @xmath119 in other words , the active player has the advantage of initiation and the other players react to its offers . there are numerous scenarios in which players can not fully forecast other players moves and offers , e.g. , when information is asymmetric or when only partial information is available @xcite . in these settings , it is likely that a greedy strategy will become the modus operandi of many players . this is a prevalent strategy also when the system evolves rapidly and it is difficult to assess the current network state and dynamics . in a dynamic network formation game , a key question is : can a player temporarily disconnects itself from the graph , only to reconnect after getting to a better bargaining position ? or must a player stay connected ? if the timescale in which the costs are evaluated is comparable to the timescale in which the dynamics occur , then , clearly , a player will not disconnect from the network voluntarily . however , if the latter is much shorter , it may , for a very brief time , disconnect itself from the graph in order to perform some strategic move . the following rules address the two alternative limits . dynamic rule # 2a : let the set of links at the current move @xmath184 be denoted as @xmath185 . a link @xmath4 will be added if @xmath2 asks to form this link and @xmath186 . in addition , any link @xmath4 can be removed in move @xmath187 dynamic rule # 2b : in addition to dynamic rule # 2a , player @xmath2 would only remove a link @xmath4 if @xmath188 and would establish a link if both @xmath186 and @xmath189 according to dynamic rule # 2a , a player is allowed to perform a strategic plan in which the first few steps will increase its cost , as long as when the plan is completed its cost will be reduced . on the other hand , if the game follows dynamic rule # 2b , then a player s cost must be reduced _ at each move _ , hence such multi - move plan is not possible . our first result in this section shows that , during the natural evolution of the network , a `` double star '' sub - graph , or network motif , often emerges . in the `` double star '' motif , as depicted in fig . [ fig : generalized star 2 ] , there exists a primary and a secondary star . all the minor players are connected to the primary star s center . part of the players are also connected to the other star s center , forming the secondary star . consider a region where it is immensely difficult to establish a link to a major player , either due to geographical distance , link prices or perhaps additional physical links are simply not accessible . nevertheless , in order to maintain a reliable connection , there must be at least two links that connect this region to the internet backbone via some major players . in order to provide a stable , fault tolerant service , every player in this region will form links with the players hosting the endpoints of these links , forming the double star sub - graph . assume now that link prices reduce over time , or that the importance of a fast connection to the internet core increases in time . in this case , players may decide to establish direct links with the major players , and remove either one of both links connecting them to the star centers . note though , that players will be reluctant to disconnect from the star center if the number of nodes in the star is large . a network configuration which includes a `` double - star '' structure of minor players . every node in the primary star ( encircled in yellow ) is linked to a major player , node @xmath190 ( in green ) . a direct link connects the two star centers , denoted by 1 and 2 ( in pink ) . the members in the secondary star ( in purple ) are connected to both star centers . in addition , there secondary star center is also connected to the major player @xmath190 . there might be additional minor players outside the stars ( in red ) . needless to say , the type - a clique ( square boxes ) is also present . ] moreover , the next theorem also shows that , eventually , and fairly quickly , the system will converge to either the optimal stable state , or to a state in which the social cost is a low multiple of the optimal social cost . [ thm : symmetric dynamic]assume symmetric reliability requirements , i.e. , @xmath71 . if the players follow dynamic rules # 1 and # 2a , then , in any playing order : \a ) the system converges to either the optimal stable state , depicted in fig . [ fig : optimal and stable ] , or to the network depicted in fig . [ fig : generalized star 2 ] . \b ) in the large network limit , namely , when @xmath74 , the social costs ratio satisfy @xmath191 , with @xmath192 . \c ) if players play in a uniformly random order , the probability that the system has not converged by turn @xmath193 decays exponentially with @xmath193 . otherwise , if every player plays at least once in o(n ) turns , convergence occurs after o(n ) steps . we claim that the network , at any time , has the following structure ( fig . [ fig : generalized star 2 ] ) : a type - a players clique , and _ at most _ two type - b star centers @xmath194 . the larger ball s center is labelled as @xmath195 , and the smaller is @xmath196 . we denote the set of type - b players which are members of the star centered about node 1 ( 2 ) as @xmath197 and ( correspondingly , @xmath198 ) . we have @xmath199 in addition , some type - b players might be linked to two nodes @xmath200 , the set of these type - b players is denoted by @xmath201 . the set of type - a players that have a direct link with the star center @xmath195 ( star center @xmath196 ) is denoted by @xmath202 ( correspondingly @xmath203 ) . players @xmath204 form a clique . assuming the second star exists , it is connected to either players @xmath195 or @xmath205 ( or both ) . finally , there may be additional links between players is @xmath201 and @xmath203 . an example of this network structure is presented in fig . [ fig : generalized star 2 ] . .[tab : d(i , k)]the shortest distance @xmath8 between two nodes , as discussed in prop . [ thm : symmetric dynamic ] . note that if @xmath206 , than column 9 applies instead of column 8 for @xmath207 . [ cols="^,^,^,^",options="header " , ] we prove by induction . after the first three players played the induction base case is true . we first assume that the link @xmath208 and consider the case @xmath209 later . denote the active player by @xmath210 . consider the following cases : \1 . @xmath211 : as @xmath212 , @xmath210 will form ( or maintain ) links with every type - a player . if additional links to minor players are to be formed , it is clearly better for player @xmath210 to establish links first with either player @xmath195 or player @xmath196 than to any player in @xmath213 or @xmath207 . we first consider the case where @xmath214 and split into two cases : 1a ) if no two disjoint paths exists between player @xmath210 to players @xmath195 or @xmath196 , namely , if @xmath215 and either @xmath216 or @xmath217 , then it is beneficial for both @xmath210 and the star centers to link such that reliability requirements will be satisfied . 1b ) if @xmath218 then by establishing the edge @xmath219 we have @xmath220 while by establishing @xmath221 the change of cost is : @xmath222 as @xmath223 , player @xmath210 will prefer to establish first a link with player 1 rather than with player 2 . the link will be formed if @xmath224 , which is true if @xmath225 if condition [ eq:2r edge ] holds , then player @xmath210 will attempt to form a link with the star center @xmath196 , and succeed if condition [ eq : a - b link condition ] holds as well . establishing a link to any other type - b node is clearly an inferior option , as if a player had decided to establish links with either node @xmath195 or node @xmath226 an additional link to one of their leafs will only reduce the distance to it by 1 and hence is not a worthy course of action . therefore , @xmath227 . \2 . @xmath228 and @xmath229 . a player may disconnect itself and then choose its two optimal links . clearly , among the type - a clique , player @xmath230 best candidates are players @xmath231 , while among the type - b players , the cost reduction is maximal by linking to nodes @xmath232 . in addition , a link to node @xmath190 is at least as preferred a link to @xmath205 , while a link to node 1 is always preferred over a link to node @xmath196 as long as @xmath233 . therefore , we only need to consider three possible moves by player @xmath210 - linking to @xmath231 ( the red players in fig . [ fig : generalized star 2 ] ) , linking to @xmath234 or linking to @xmath232 . using tables [ tab : d(i , k ) ] and [ tab : d(i , k ) ] we can compare the cost of connecting to nodes @xmath232 versus the costs of connecting to nodes @xmath195 and @xmath190 . a simple calculation shows that if the expression @xmath235 is positive , than linking to players @xmath195 and @xmath190 is preferred over linking to player @xmath195 and @xmath196 . this expression is positive , as @xmath236 and @xmath223 . in a similar fashion , we can compare between the choices of linking to nodes @xmath195 and @xmath190 or forming links with nodes @xmath190 and @xmath237 if the expression @xmath238 is positive , then establishing links to @xmath195 and @xmath190 is preferred , otherwise the alternative will be chosen . two special cases are of interest : if @xmath239 and @xmath240 , then every player will prefer to link to @xmath190 and @xmath205 . that is , we ll get the optimal configuration . if @xmath14 , @xmath241 , i.e , node 1 is a member of the clique , @xmath242 and @xmath243 , then a minor player will connect to players @xmath195 and @xmath190 . note that if there are no type - a players present when @xmath210 plays , then it must connect to the star centers @xmath195 and @xmath196 . in addition , after @xmath210 formed two links , no player @xmath244 will agree to form the link @xmath245 , as the induced change of cost is @xmath246 if @xmath247 and @xmath248 then player @xmath249 and player @xmath250 will establish the link @xmath245 . by symmetry , this also happens when @xmath251 and @xmath252 . if @xmath253 then it must maintain links to all nodes in @xmath254 ( if there are any ) in order to satisfy the reliability criteria . additional links to type - a players will be formed according to the discussion in case 1 , while no other links to type - b player will be formed according to the discussion in case 2 . \4 . if @xmath255 , then it must maintain links to all nodes in @xmath198 ( if there are any ) in order to satisfy the reliability criteria . a similar calculation to the one in case @xmath196 shows that player @xmath196 s best course of action is to remain connected to player in order to complete the induction proof all that is left is to address the case @xmath209 . in this case , player @xmath196 must be connected to at least one additional player @xmath256 in order to have two disjoint path to every player in the network . clearly , its optimal choice is player @xmath205 . in order to maintain reliable path to every @xmath257 player @xmath205 must agree to form this link . player @xmath196 will disconnect @xmath208 if either @xmath258 or @xmath259 . however , if either of this conditions hold , tables [ tab : d(i , k ) ] and [ tab : d(i , k ) ] shows that for every @xmath260 , it would prefer to link to player @xmath205 rather than player @xmath196 . in particular , this is also true for every node @xmath207 . therefore , as soon as every member in @xmath198 played once , the star will be empty and no additional star will raise again . therefore , there will be at most one star left in the network . this completes the proof on the momentarily structure of the network . note that if @xmath261 we obtain the optimal stable solution . \b ) since every minor player has either two or three links , the contribution to the social cost due to minor players links is @xmath262 while the contribution to the social costs due to the inter - distances between minor players is @xmath263 . in the limit @xmath153 , the dominant term in the cost function is the term proportional to @xmath264 . both @xmath8 and @xmath22 for @xmath86 are bounded by @xmath265 , and comparing this results with the optimal social cost ( proposition [ lem : reliable optimal 1 ] ) we have @xmath266 with @xmath192 in the limit @xmath153 . \c ) the discussion of case 2 shows that in a large network , it is suboptimal for a minor player to be a member of @xmath198 . given the the opportunity , it will prefer to become a member in either @xmath197 or @xmath201 . therefore , after every player has played at least twice , @xmath267 , and after every player has played four times the system will reach equilibrium . lemma 13 in @xcite then shows that the probability that the system has not converged by turn @xmath193 decays exponentially with @xmath193 . the optimal stable network , as described in theorem [ thm : symmetric dynamic ] . ] in section [ sec : static - analysis ] we emphasized the importance of symmetry in the reliability requirements for reducing the price of anarchy . the next theorem affirms this assertion , and shows that if the constraints are asymmetric , the system converges to a state with an unbounded social cost on a large set of possible dynamics and initial conditions . [ thm : dynamic asymetric]assume asymmetric reliability requirements , namely @xmath110 . if the players follow dynamic rules # 1 and either dynamic rule # 2a or # 2b , then the system converges to a state with an unbounded social cost . the idea behind the proof is to show that the reliability requirements of at least one minor player remain unsatisfied , the therefore the social costs , as a sum over individual costs , is unbounded . first assume that @xmath268 . we shall now show that at any given turn , the network is composed of a type - a ( possibly empty ) clique , a set of type - b players @xmath269 linked to player @xmath95 , acting as a star center , and an additional ( possibly empty ) set of type - b players @xmath201 connected to the type - a player @xmath190 . player @xmath95 is also connected to player @xmath190 . some major player may establish links with the star center @xmath270 the set of these major players is denoted by @xmath271 . note that the cost of every player in either @xmath269 or @xmath201 is @xmath272 , since every path from each of these players to the type - a clique crosses player @xmath190 . we prove by induction . at turn @xmath273 , this is certainly true . denote the active player at time @xmath193 as @xmath274 consider the following cases : \1 . @xmath211 : since @xmath275 all links to the other type - a nodes will be established or maintained , if @xmath210 is already connected to the network . clearly , the optimal link in @xmath210 s concern is the link with the star center @xmath95 . therefore , player @xmath210 will attempt to establish the link @xmath276 if @xmath277 is negative . if @xmath278 or that @xmath279 then @xmath95 will accept this link . regardless , if player @xmath210 has formed the link it will not establish a link with @xmath280 , as it only reduces its distance to player @xmath2 by a single hop , and @xmath281 if eq . [ eq : term1 ] is positive , then , player @xmath210 has no incentive to establish a link with any @xmath280 , as @xmath282 and the by establishing @xmath283 the sole change is the reduction of @xmath284 from @xmath265 to @xmath195 , @xmath285 eq . [ eq : no single hop ] also shows that if @xmath286 , it will not form a link with @xmath287 , while if @xmath288 it may not remove the link @xmath289 as otherwise @xmath2 is disconnected . @xmath290 : first , assume that @xmath210 is a newly arrived player , hence it is disconnected . obviously , in its concern , a link to the star s center , player @xmath95 , is preferred over a link to any other type - b player . similarly , a link to a player @xmath190 is preferred over a link to any other type - a player ( or if @xmath291 and @xmath292 , equivalent to a link to any other type - a player ) . therefore , @xmath210 first link choice would be either @xmath293 or @xmath276 . in other words , @xmath249 or @xmath294 . if @xmath249 , than no @xmath80 will agree to establish a link with @xmath210 , as it only reduces its distance from @xmath210 by one hop , and does nt alter @xmath114 for any other @xmath95 similarly , if @xmath294 than @xmath295 and the link would not be formed . likewise , no link between @xmath287 and @xmath296 may be formed , as @xmath297 \3 . @xmath298 , the star s center : @xmath210 may not remove any edge connected to a type - b player and render the graph disconnected . on the other hand , the previous discussion shows it will not establish additional links to nodes in @xmath201 . note that the reliability requirements of players in @xmath299 are invalidated . therefore , the cost of every player @xmath300 is at least @xmath301 , and we have @xmath302 . hence , at any given turn , as soon as either @xmath303 and @xmath282 , the social cost is unbounded . if @xmath304 a link between player @xmath80 and player @xmath228 will be formed . however , as soon as player @xmath2 becomes the active player , it is beneficial for it to remove the link @xmath283 and establish a link to the star center @xmath94 instead . therefore , after player @xmath2 s turn , the cost of player @xmath210 is again @xmath305 . in a similar fashion , if @xmath306 than player @xmath280 and player @xmath307 may establish the link @xmath4 . this does not affect any other player and while it may reduce their costs , it does not provide them with two disjoint paths to the type - a clique , as they both must traverse player @xmath190 in order to access it . therefore , their costs is still @xmath305 . although , at first sight , it might seem that this result is due to the greedy and myopic choice of edges , it is possible to show that there are cases in which this result holds even when players may pick their links strategically rather than in a greedy manner . for example , if @xmath268 and the number of minor players is odd , than a similar analysis shows that a player in @xmath308 will establish a link to a single player in @xmath201 . however , as the number of minor players is odd , there will always be an unmatched player in either @xmath201 or @xmath269 and its cost will be @xmath305 . recall that under the presence of monetary transfers , players @xmath2 and @xmath3 will agree to establish an edge if @xmath309 . nevertheless , it may be that during the active player turn there are a few links that satisfy this condition , and the player _ _ must prioritize them . each player s decision is myopic , and is based solely on the current state of the network . hence , the order of establishing links is potentially important . needless to say , a player s preference order will depend on the link prices . while there are several alternatives , we adopt the following preference order @xcite : denote the active player as player @xmath2 . each link @xmath4 carries different utility in player @xmath310 respect . it is reasonable to assume that a link with a lower connection value will be priced lower , so that the link with the least connection utility will be marked with the lowest price . in fact , one may assume its price will be as much as the implied cost of the other party of this link . we will denote this price as @xmath311 . every other player @xmath95 will use this value and demand an additional payment from player @xmath2 for the link @xmath94 , as it is more beneficial for player @xmath2 . formally : `` strategic '' pricing mechanism : set @xmath312 as the node that maximizes @xmath313 . set @xmath314 . finally , set @xmath315 the price that player @xmath3 requires in order to establish _ @xmath4 _ is _ _ @xmath316 . under this pricing mechanism , there could be many links that carry the same utility . some of these links have a better connection value , but they come at a higher price . since all the links carry the same utility , we need to decide on some preference mechanism for player @xmath2 . the simplest one is the `` cheap '' choice , in which , if there are a few equivalent links , the player will choose the cheapest one . this can be a reasonable choice , as new players can not spend too much resources , and therefore they will choose the `` cheapest '' option that belongs to the set of links with maximal utility . preference order : player @xmath2 will establish links with player @xmath3 if player @xmath3 minimizes @xmath317 and @xmath318 . if there are several players that minimize @xmath319 , then player @xmath2 will establish a link with a player that minimizes @xmath56 . if there are several players that satisfy the previous condition , then one out of them is chosen randomly . we are now at a position to identify an additional network motif , namely , the `` entangled cycles '' . this network motif is composed of a line ( i.e. , interconnected sequence ) of minor players nodes , with some cross - links between the nodes along this line , breaking the hierarchy ( fig . [ fig : the - entangled - loops ] ) . the `` entangled cycle '' of length three is the `` feedback loop '' motif , which was previously found to exist in a higher frequency than expected in the internet graph @xcite . the `` entangled cycles '' motif . a ) six minor player are connected in an `` entangled cycles '' subgraph . the first two nodes have direct connection to some major player , and access the rest of the network by the major player s additional links , represented by the dotted line . b ) if at some point , a link between a player in this subgraph and some external player is formed ( in this example , a major player ) , some links may be removed without violating the reliability requirement and without increasing the distance cost appreciably . the removed links marked by a red x. ] when a new minor player arrives , it will choose the two cheapest links and will connect to the corresponding players . clearly , its costs due to the distance from the rest of the network will be the highest . as such , when the next player arrives , it will offer the lowest link price . the new arrival will link to it and to one of its provider . the process will repeat , until , at some point , an existing player will decide that this growing branch is too far from it , and will connect to one of the nodes along this `` entangled cycles '' . at this point , this subgraph will be over - saturated with links as players may utilize this link to access the internet core . hence , some links will be removed ( see , for example fig . [ fig : the - entangled - loops](b ) ) . the set of the links that will be removed depends heavily on the playing order and the temporary network structure . nevertheless , some cross - links may remain in order to satisfy reliability constrains . this explains the following result . [ thm : entangled cycles motif]assume the number of major players is at least 10 . denote the distance cost of player @xmath320 @xmath20 , as @xmath321 , namely @xmath322 assume that a subset @xmath323 of minor player first join the game and play consecutively and that the two players with the maximal distance cost are adjacent . then : \a ) these players will form an `` entangled cycles '' structure of length @xmath165 , as depicted in fig . [ fig : the - entangled - loops ] , and @xmath324 \b ) the entangled cycles structure is semi - stable , in the following sense : if , at some later turn , there exist players @xmath325 , @xmath326 such that the link @xmath4 is formed ( fig . [ fig : the - entangled - loops](b ) ) , then some links in the `` entangled cycles '' structure may be removed in subsequent turns . \a ) assume that at time @xmath193 a subset of @xmath327 minor players first join the game . denote the set of players that are currently connected to the network by @xmath178 . we denote by @xmath328 ( @xmath329 the number of type a players ( correspondingly , type - b player ) at that moment . let us denote the players with the highest distance cost term as @xmath127 and @xmath330 , namely , @xmath331 where for simplicity we assumed that @xmath332 are minor player ( type - b players ) . according to the `` strategic '' pricing mechanism a new player will establish links with these players , as they will offer the cheapest links . note that by connecting to these players , its survivability requirements are satisfied , as each of these players maintain two disjoint path to every major player ( if the reliability requirements are asymmetric ) or to all other players ( in case of symmetric reliability requirements ) . we are now going to prove by induction that player @xmath333 will first connect to players @xmath334 and @xmath335 . we shall prove by this by proving that the distance costs of @xmath334 and @xmath335 are maximal . first , note that for every player @xmath336 , we have @xmath337 and @xmath338 , since the path from @xmath3 to every player in @xmath178 pass through @xmath339 or @xmath340 , and @xmath341 . therefore , in order to show that players @xmath334 and @xmath335 have the highest distance cost , it is sufficient to show that @xmath342 for every @xmath343 . for every player @xmath344 we have @xmath345 since the path that connects player @xmath334 to the any player @xmath346 crosses either player @xmath127 or @xmath347 , which are adjacent . therefore , the distance cost of player @xmath334 due to its distance from every player @xmath346 , denoted by @xmath348 is at least grater than the corresponding distance cost of player @xmath127 , @xmath349 by at least @xmath350 . note that @xmath351 for every @xmath343 according to the definition of @xmath127 . however , player @xmath339 may be closer to players @xmath352 than player @xmath353 . denote the maximal distance between any two players by @xmath274 we have , @xmath354 where @xmath210 is the maximal distance between two players . in proposition [ prop : max distance ] it was shown that @xmath355 hence , @xmath354 as @xmath356 this expression is positive . hence , links will be established as stated . this shows that at the time player @xmath333 joins the game , the distance cost of player @xmath334 is maximal . a similar calculation shows that at this turn , the distance cost @xmath357 is maximal in the set @xmath358 . we have thus far shown that the first two links of a new player @xmath333 are to players @xmath334 and @xmath335 . nevertheless , as the `` entangled cycles '' motif grows , the incentive of other players to connect to players in it increases . at some point , player @xmath3 may be able to form links with players not in @xmath323 , or an active player @xmath359 may decide to connect to some player in @xmath323 ( or vice versa for @xmath360 ) . in either of these cases , players in the `` entangled cycles '' motif may have three disjoint paths to other players , and may therefore remove a link , should they deemed to do so . in this case , the `` entangled cycles '' motif will be diluted . this theorem shows that reliability is a major factor in breaking up tree hierarchy in the internet . in addition , it also hints that the hierarchical structure does not break frequently in the top levels of the internet , but rather mostly in the intermediate and lower tiers . note that according to theorem [ thm : entangled cycles motif](b ) , in large networks the length of the `` entangled motifs '' is short . therefore , we do not expect to see excessively long structures , but rather small ones , having just a few ass . in this section we compare our theoretical predictions with the real - world inter - as topology graph @xcite . we classified ass to major players and minor players according the popular caida ranking @xcite . for the sake of comparison with previous work @xcite , we classified the major players as the top 100 ass according to this ranking . the mean length of the shortest cycle connecting a major player and a minor player as function of time , from january 2006 to october 2008 . the length decrease as time passes and the network grows , in agreement with our model . ] in section [ sec : monetary - transfers ] , we showed that , by allowing monetary transfers , the maximal cycle length connecting a major player to a minor player depends inversely on the number of major players . as the number of ass increases in time , it is reasonable that the number of major player grows as well . hence , we expect that the length of the shortest cycle connecting a major player to a minor player will decrease in time . [ fig : the - mean cycle - length ] shows the mean cycle length connecting one of the secondary leading 2000 ass , ranked 101 - 2100 in caida ranking , and one of the top 100 nodes . the steady decline of the cycle s length in time is predicted by our model . the ratio of the mean number of disjoint paths connecting a minor player to the core to the mean degree of a minor player . this ratio is above @xmath361 , and increases in time , showing that additional links are likely to part of disjoint paths to the core . this ratio is presented as a function of time from january 2006 to october 2008 . ] our analysis showed that in most of the generated topologies , the minor players are organized in small subgraphs that have direct connection to the internet core , namely the major players clique , or the tier-1 subgraph , in agreement with @xcite . in order to maintain a reliable connection , in each subgraph there must be at least two links that connect minor players to the core . indeed , we have found out that the ratio between the mean number of disjoint paths from a minor player to the core and the mean degree of minor players is more than @xmath361 , and it increases in time ( fig . [ fig : disjoint path ratio ] ) . that is , almost every outgoing link of a minor player is used to provide it with an additional , disjoint path to the core . in other words , a player is more likely to establish an additional link , hence increase its degree , if it supplies it with a new path to the core that does not intersect its current paths . in section [ sec : dynamic - analysis ] we predicted the ubiquity of two network motifs , the `` double star '' motif ( fig . [ fig : generalized star 2 ] ) and the `` entangled cycles '' motif ( fig . [ fig : the - entangled - loops ] ) . we define the occurrence of a `` double star '' motif as the existence of a a connected pair of nodes , each with degree greater than @xmath184 , designated as the centers , such that at least @xmath184 neighbors of one center are also neighbors of the other center . we generated random networks according to the configuration model ( cm ) , in which each node is given a number of stubs according to its degree , and stubs are connected uniformly . then , we evaluated the mean number of occurrences of this motif in a random cm network with the same number of nodes and the same degree distribution as the real inter - as topology . for @xmath362 , we have found @xmath363 occurrences in the real - world inter - as topology , whereas the mean number of occurrences in the random cm network was only * @xmath364 * instances ( the @xmath365 indicates standard deviation ) . namely , there are more than four times displays of this subgraph in the internet than in a random network with the same degree distribution . chebyshev s inequality provides a bound on the _ p - value , @xmath366 . _ this low value indicates that it is highly unlikely that a random cm network explains the frequent appearance of this network motif . we have tested the prevalence of this motif with other values of @xmath184 and the number of occurrences is consistently a few times more than expected in a random cm network . our analysis suggests that reliability considerations is one of the factors leading to the increased number of incidents . the unexpected prevalence of the `` feedback loop '' , which is a special case of the `` entangled cycles '' motif , was first reported in @xcite . the `` feedback loop '' motif coincides with the entangled cycles motif of length three , and in order to further assess our results we tested for the occurrence frequency of the `` entangled cycles '' motif of length four . we compared the number of occurrences of this motif in the real - world internet graph to the expected number of occurrences in a random configuration model network . while the number of instances of this motif in the internet graph was @xmath367 , the expected number of occurrences in the random network was only @xmath368 . as before , the abundance of this network motif , an order of magnitude greater than expected ( _ @xmath369 _ , a relaxed bound based on chebyshev s inequality ) provides a positive indication to the implications of survivability requirements . in summary , we have provided both static and dynamic empirical evidence that conform with our predictions , suggesting the importance of reliability considerations on the structure and dynamics of the inter - as topology . while many studies have tried to model the internet structure , the list of works the explicitly address reliability considerations is much shorter . furthermore , most of the theoretical models of the internet structure and dynamics assume homogenous agents , while the internet is inherently heterogeneous , composed of a wide variety of entities with different business models . in this work , we found that constructing a model that includes these two factors , namely reliability and heterogeneity , may provide important insights on the internet structure and dynamics . we first rigorously formulated a model of a network formation game in this context . our model is flexible , and may be used in a wide variety of settings . it allows for many variations and schemes , for example situations in which failures are frequent or rare , or to account for varying centrality of different types of players . the homogeneity of the model also allows us to describe scenarios in which a fallback route is required only to a subset of players . indeed , a reasonable as policy is to require a backup routing path only to the internet backbone , rather than establishing fault - tolerant routing pathways to every particular autonomous system . we established the _ price of reliability , _ which measures the excess social cost that is required in order to maintain network survivability in an optimal stable equilibrium . surprisingly , we showed that it can be smaller than one , that is , that the additional survivability constraints _ add _ to the social utility . we have also showed that reliability requirements have disparate effects on different parts of the network . while it may support dilution in dense areas , it facilitates edges formation in sparse areas , and in particular it supports the formation of edges connecting minor players and major players . in our dynamic analysis we have found the repetitive appearance of small sub - graphs , or network motifs , namely the `` entangled cycles '' motif and the `` double star '' motif . indeed , the number of appearance of these motifs in the real inter - as topology surpassed the expected number by a few folds , indicating that additional factors support their formation , and as our analysis shows , survivability is one of them . we have also predicted that the length of the minimal cycle connecting a major player to a minor player should decrease in time . this prediction , too , was verified by a dynamic data analysis . finally , while our analysis focuses on the inter - as topology , it may be applied to other networks as well , that are composed of heterogeneous , rational agents that are required to maintain some reliability aspects . primary examples are trade networks and mvno operators in the cellular market . in this work we have shown that a game theoretic analysis of network formation , which encompasses heterogeneous agents and explicitly addresses survivability concerns , holds promising results . nevertheless , there are many open questions , such as the following . how does the emerging network handle more than a single failure ? which incentive mechanism will promote increased reliability of the future internet ? what can a comparative analysis of experimental results on different networks tell us about the players strategies in each network ? these , and many more , indicate that there is yet a lot to uncover in this intersection between network formation , heterogeneity and reliability .

the time delay between two dynamical systems can provide information on conduction velocity , and the nature and origin of coupling , between the processes . so it is necessary to use a well validated method for this purpose . literature on time delay is vast and well documented in @xcite . often , in physiological time series analysis , a single method can not be made unique to be applicable for a wide class of data stemming from the processes seemingly operated by similar mechanisms . methods used for time delay estimation are no exception from this fact . here , we use a spectral based method , maximising coherence , for the time delay estimation @xcite . first , we apply this method to uni- and bi - directionally coupled rssler attractors with a known delay between the systems . in both cases , the results obtained are in good agreement within a narrow range of the delay used in the simulation . then , we apply this method to simultaneously recorded traces of electroencephalogram ( eeg ) and electromyogram ( emg ) of subjects with essential tremor a well known pathological form of hand tremor . we obtain a delay of 11 - 27 milli seconds ( @xmath0 ) between the tremor correlated cortical activity ( eeg ) and emg and the results are reasonably within the range of experimentally observed value for cortico - muscular transmission @xcite . this result , further confirms the involvement of cortex in the generation of essential tremor @xcite . the paper is organised as follows : in section 2 , we discuss in detail , the methodology of coherence analysis . then , we extend the coherence analysis for time delay estimation . due to a time delay there will be a time misalignment between the two time series thereby causing a reduction in the coherence estimated between them . in order to compensate for the reduction in coherence due to delay , we shift one of the time series keeping the other constant and estimate the coherence as a function of the shift . this method has been successfully applied to estimate the time delay between the acoustic source and the receiving signals @xcite . on similar lines , phase synchronisation is used to estimate the time delay between and among the atmospheric variables observed at different meteorological sites @xcite . in section 3 , we apply this method to afore mentioned theoretical models . after validating the method by the results of standard models , in section 4 , we apply this method to estimate the time delay between the simultaneously recorded eeg and emg data of essential tremor subjects . we discuss the results and conclude in section 5 . let @xmath1 and @xmath2 be two simultaneously recorded data sets of length @xmath3 . the mean and standard deviation of the two data sets are , respectively , set to zero and one . we divide the data sets into @xmath4 disjoint segments of length @xmath5 , such that @xmath6 . we calculate power spectra , @xmath7 , @xmath8 and cross spectrum @xmath9 , which is the fourier transform of the cross - correlation function of the signals @xmath1 and @xmath2 @xcite , in each segment . over cap in all the quantities indicates that it is an estimate of that quantity . finally , we average the power spectra and the cross - spectrum across all the segments and calculate coherence as follows @xcite , @xmath10 the coherence spectrum provides the strength of correlation between the two signals , @xmath1 and @xmath2 . the confidence limit for coherence at the @xmath11 is given by @xmath12 . thus , the coherence spectrum is always considered with this line . in all of our analysis , @xmath13 is set to 0.99 , and hence the confidence limit is @xmath14 . the estimated value of coherence , at a frequency , above this line indicates ( a ) significant coherence between the two time series at this frequency ; ( b ) the magnitude ( deviation of coherence from this line ) determines the degree of linear correlation between the two time series at this frequency . the estimated value of coherence at a frequency below this line is considered as the lack of correlation between the two time series at this frequency . if the sampling frequency of the signals is @xmath15 @xmath16 ( i.e. @xmath17 number of data points are sampled per second ) , then the frequency resolution of the quantities in eq . 1 , is @xmath18 . thus , one should optimally choose the value of @xmath5 depending on the purpose of analysis , to compromise between the sensitivity and reliability . usage of the fixed segment length @xmath5 is questioned and a variable segment length is suggested in @xcite , to make quantitative assessment of the signal . but , for all practical purposes , a fixed segment length is easily implementable and hence , it is used in the forthcoming analysis . while the coherence spectrum provides the strength of correlation between the two signals @xmath1 and @xmath2 , time ( delay ) information between the two signals can be obtained from the phase spectrum , which is the argument of the cross - spectrum @xcite , @xmath19 following @xcite , eq . 2 can be further simplified to see the explicit appearance of the time delay in it as follows , @xmath20 where @xmath21 is the time delay . the phase estimate @xmath22 and its upper and lower 95% confidence interval are given by @xcite @xmath23^\frac12 . \eqno 4\ ] ] thus , the confidence interval of the phase estimate is inversely related to coherence . in the rest of the paper we use estimates of the coherence and phase without the over hat for the sake of convenience . one of the conventional ways to estimate a time delay , in frequency domain , is to fit a straight line to the phase spectrum ( eq . 3 ) in the frequency band of significant coherence , as the phase can be reliably estimated only in the frequency band of significant coherence ( see eq . this method of estimating delay is possible when we have a broad band coherence and limits its applicability to the narrow band coherent signals . in some cases , coherence extends to first harmonic and hence the phase values at the harmonics are used to increase the reliability of the delay estimate @xcite . further , if two time series show significant coherence over a wide range of frequency band , but have a minimal phase relation , then the estimation of time delay from the phase estimate is not straight forward . under such conditions , phase estimate , in addition to the @xmath24 term ( see eq . 3 ) , will also contain a frequency dependent factor , namely , the argument of the transfer function , @xmath25 . in such cases , @xmath26 is estimated by hilbert transform , and then delay is estimated from the phase estimate after subtracting the estimated transfer function from the phase estimate @xcite . from the above discussion it is clear that phase estimates can not be used to estimate the delay from narrow band coherent signals which are often observed in human physiological data @xcite . for this purpose , we use the method of maximising coherence . as discussed in the last part of the introduction , a delay @xmath21 between two time series , will cause a reduction in the estimated coherence between them . in order to compensate for the reduction in coherence due to delay and thereby to estimate the delay , we realign the time series by artificially shifting them . for this , we shift one of the time series by a lag @xmath27 keeping the other constant and consider coherence at the selected frequency band @xmath28 as a function of @xmath29 , @xmath30 and repeat the same for the other time series . if there is a delay between the two time series , coherence in the selected frequency band @xmath30 will increase from its value at @xmath31 and reach a maximum at the @xmath29 corresponding to @xmath21 . after shifting one of the time series by a lag @xmath29 , the length of the shifted time series will be less than the un - shifted time series by @xmath32 ( @xmath29 in sample units ) data points . we discard the extra data points in the un - shifted time series to have the same length for both time series . further , coherence is a relative measure and changes with the length of the data . it is reflected in the 99% confidence level ( see above ) . for each shift , we discard the points corresponding to the maximum lag from the entire length of the time series , and consider only the length of the data which is integer multiple of @xmath5 . by doing so , coherence @xmath30 calculated for all the lags will have the same confidence level , thereby confirming that the maximum value of coherence is reached only because of the delay and not due to the spurious effects caused in estimating coherence for different lengths of data . as we have confidence level as a check for the reliability of @xmath21 , we do nt need any additional tests like , surrogate tests @xcite to assess the reliability of the results . however , in order to get the variability in the estimated delay , we use surrogate analysis . surrogate analysis is introduced in the context of nonlinear time series analysis to check whether or not the time series under consideration has got a nonlinear structure @xcite . making the ( null ) hypothesis that the time series has come from a linear process , several linear realizations of the time series namely , surrogates , are synthesised . then , the original time series and the surrogates are quantified by a suitable discriminating statistic . any deviation in the discriminating statistic calculated for the original time series and surrogates indicates the presence of nonlinear structure in the original time series and thereby rejecting the null hypothesis @xcite . there are different ways of preparing the surrogates @xcite . the misconception of surrogate analysis is discussed in @xcite . as mentioned above , the objective of the methods to generate surrogates @xcite is to synthesise a new data set by destroying the nonlinear structure present in the original data set . but our aim is not to show the presence of nonlinear structure in our data . for the purpose of calculating the error - bars of delay , surrogates are generated by exploiting one of the basic assumptions of spectral analysis that distinct parts of the time series are independent @xcite . instead of shuffling the time series as a whole ( which is done in one of the methods to synthesise surrogates , amplitude adjusted surrogates ) we shuffle the disjoint data segments from which the original spectrum is estimated ( see above ) . we shuffle only the un - shifted time series ( see above ) from which the information is assumed to flow to the other time series ( which is time shifted in advance in maximising coherence analysis ) with a delay . by doing so , the spectra of both the time series , @xmath33 and @xmath34 in eq . 1 will remain the same but the cross spectrum @xmath35 in eq . 1 will be different . this type of surrogate is similar to the one proposed in @xcite where the whole time series is shuffled but by preserving two point correlation ( i.e. power spectrum ) of the original time series . for all the analyses reported in this paper we synthesise 19 different realizations of surrogates . we make a null hypothesis that @xmath30 obtained is due to spurious correlations between the two time series . for each realization of surrogate we calculate time delayed coherence @xmath36 . we calculate the significance of difference @xmath37 between the @xmath36 calculated for surrogates and @xmath30 where @xmath38}$ ] , where @xmath39 indicates the average over different realizations of surrogates and @xmath40 $ ] indicates the standard deviation between different realizations of surrogates . any value of @xmath41 indicates that the @xmath30 obtained is not due to spurious correlations @xcite and hence the null hypothesis is rejected . if the null hypothesis is rejected we consider @xmath30 for further analysis otherwise we discard it as spurious correlation . for the @xmath30 qualified in the surrogate analysis ( for which null hypothesis is rejected ) we calculate the error in the delay in the following way : we subtract @xmath36 calculated for each realization from @xmath30 and calculate delay for each subtracted function ( i.e. ) @xmath42 . we report here the delay of the system as the mean value of the delays ( calculated for the 19 surrogate subtracted realizations ) and their standard deviation as error - bar . as the increase in the @xmath30 when compensated for the delay is very small , for the sake of clarity we plot @xmath43 -\left[c(\omega_0)-<c(\omega_0)^{surr}>\right]$ ] . note that @xmath44 is the value of @xmath30 at @xmath31 and @xmath45 is the average of @xmath36 for different realizations of the surrogates at @xmath31 . by this definition @xmath46 will pass through a zero value at @xmath31 and show a maximum value ( above zero ) at @xmath47 ( for @xmath48 ) between the processes . though we calculate @xmath37 for different @xmath29 values at which @xmath30 is evaluated , we report here @xmath49 as it is the most relevant one in determining the significance of the @xmath50 ( and hence @xmath51 ) . based on the above arguments the delay between the two time series is given by : @xmath52 in this section , we apply the method of maximising coherence , to coupled rssler systems . the dynamics of the @xmath53-th attractor is given as follows : @xmath54~+~\epsilon_{j , i}~x_j(t-\delta ) \nonumber \\ \dot{y}_i(t)&=&x_i(t)~+~a~y_i(t ) \nonumber \\ ~\dot{z}_i(t ) & = & b~+~z_i(t)~[x_i(t)~+~c ] \nonumber , \end{aligned}\ ] ] where @xmath55 , are the parameters . coupling is established through the @xmath56 component between the attractors . @xmath57 , is the coupling constant and @xmath58 is the @xmath59 component of the @xmath60th attractor . in our study , we consider the dynamics of two coupled rssler attractors , and hence @xmath61 and @xmath62 and @xmath21 is the delay . further , we do nt consider the self coupling of the oscillators . and @xmath21 is set to 2 sec.,width=192 ] first , we consider the coupling scheme as shown in fig . 1 , where two rssler attractors are coupled and the information flow is from attractor 2 to attractor 1 ( @xmath64 and @xmath63 ) as governed by the above set of equations . @xmath21 is set to 2 sec . the above equations are simulated by euler scheme with a step size of 0.01 sec and for the subsequent analysis the data are down sampled to 0.1 sec . as the coupling is established through the @xmath59 component between the attractors , we use the @xmath59 time series to perform the delay estimation . we have used 30000 data points for further analysis . power spectra of the two attractors are shown in fig 2a . solid line represents the power spectrum of attractor 1 and solid line with dots represents the power spectrum of attractor 2 . for the sake of clarity the spectrum of attractor 1 is shifted vertically upwards by a factor of 5 from its original position . we have used a segment length of @xmath65 and hence the frequency resolution ( of the quantities in eq . 1 ) is 0.01 @xmath16 . spectra show dominant activities in the frequency range of 0.1 to 0.2 @xmath16 , which can be related to the mean orbital period of @xmath66 sec of the attractor @xcite . since the dynamics of the attractor 1 is perturbed by coupling ( see fig . 1 ) , its spectrum looks slightly different from that of the attractor 2 . the coherence spectrum of the coupled attractors is shown in fig . horizontal line around the coherence of 0.15 indicates the confidence level of @xmath67 . in the frequency band between 0.1 - 0.2 @xmath16 ( see fig . 2a ) , where the individual attractors show dominant activities , the coupled systems show significant coherence . estimated phase ( eq . 2 ) between the coupled attractors is shown in fig . 2c . error - bars represent the 95% confidence interval of the phase estimate given by eq . 4 . thus the phase can be reliably estimated in the frequency band ( 0.1 - 0.2 @xmath16 ) , where the coupled attractors show a significant coherence ( see fig . the result of maximising coherence is shown in fig . 2d where @xmath46 is plotted as a function of lag @xmath29 . we have chosen @xmath68 ( and is shown in fig . 2b ) from the basic frequency of the oscillators . solid line at @xmath69 indicates @xmath70(see the last part of the methodology for details ) . negative shift indicates that the time series of attractor 1 is in advance and positive shift indicates that the time series of attractor 2 is in advance . since there is a delay of 2 sec , the coherence calculated at @xmath31 is zero ( see above ) . in order to compensate for the reduction in the coherence and thereby to estimate the time delay , time series of the attractors are shifted by a time lag @xmath29 . as the delayed information flows from attractor 2 to attractor 1 , coherence @xmath46 increases from zero ( see above ) when the time series of attractor 1 is shifted * back * in time and reaches a maximum at the @xmath71 which is the time delay between the two time series ( @xmath21 used in the simulation is 2 sec ) . the significance of deviation from the surrogate data @xmath72 is 2.24 and indicates that @xmath73 is not due to spurious correlations . the error - bar for the delay ( as explained in the methodology section ) is calculated as 0.4 sec . thus , the time delay @xmath74 is @xmath75 sec and is well captured by the maximising coherence analysis ( expected delay value is 2 sec for the flow from attractor 2 to attractor 1 ) . sec . horizonal line at the coherence of 0.15 indicates the confidence level ( @xmath67 ) . there is a significant coherence in the frequency range of 0.1 - 0.2 @xmath16 between the two attractors . the value of @xmath28 used for maximising coherence analysis is indicated along with its coherence value @xmath44 . phase estimate @xmath76 of the coupled attractors . error - bars represent 95% confidence interval defined by eq . errors are relatively small in the region of significant coherence which is in accordance to eq . 4 . ( d ) . coherence @xmath46 , as the function of lag @xmath29 . horizontal line at @xmath69 indicates @xmath77 . @xmath39 represents the average over different surrogates . @xmath46 has reached the maximum value at the @xmath29 ( delay ) value of -2.1 sec ( from attractor 2 to 1 ) , and the expected value of @xmath21 is 2 sec for the flow from attractor 2 to attractor 1 . the significance of deviation from the surrogate @xmath72 is 2.24 and indicates that @xmath46 is not due to spurious correlations.,width=480,height=432 ] next , we consider the coupling scheme as shown in fig . 3 , where two rssler attractors are coupled as considered above , but with a bi - directional flow , @xmath78 and @xmath79 . @xmath21 used in the coupling is same as in fig . 1 , which is 2 sec . and @xmath79 . @xmath21 is set to 2 sec.,width=192 ] power spectra of the attractors look almost the same as the two attractors perturb each other almost to the same extent ( see fig . 4a ) . as obtained for uni - directional coupling , in this case also , significant coherence is between 0.1 - 0.2 @xmath16 ( fig . 4b ) . as the number of data points used is the same ( 30000 ) as in the uni - directional coupling , the confidence level @xmath67 is also around the coherence of 0.15 ( see the horizontal line in fig . 4 b which is same as in fig . phase estimate of the bi - directionally coupled systems is shown in fig . 4c with the 95 % confidence interval as the error - bars . error - bars are relatively narrow ( see eq . 4 ) in the frequency band of 0.1 - 0.2 @xmath16 where significant coherence is observed ( as seen in fig . result of the maximising coherence for bi - directionally coupled systems is shown in fig . 4d . the horizontal line at @xmath80 has the meaning as in fig . in this case 0.15 @xmath16 is used as @xmath28 ( and is shown in fig . 4b ) which is the basic frequency of the attractors . @xmath46 shows two maxima , one at @xmath81 sec which corresponds to the flow from attractor 2 to 1 and another at @xmath82 sec which corresponds to the flow from attractor 1 to 2 , while the @xmath21 used in the coupling is 2 sec in both directions ( see fig . 4d ) . also , in this case , the significance of deviation from surrogate calculated by @xmath37 at @xmath29 values -2.5 sec and 1.7 sec are well above 2 ( see fig . 4d ) indicating that the @xmath46 obtained at these two @xmath29 values are not due to spurious correlations . thus , the time delays of the system are @xmath83 sec ( for the flow from attractor 2 to attractor 1 ) and @xmath84 sec ( for the flow from attractor 1 to attractor 2 ) and are well within the expected value of 2 sec in either direction ( see fig . the error - bars are obtained by surrogate analysis . and @xmath79 . result of maximising coherence @xmath46 . @xmath46 has reached maximum value at two places , one at -2.5 sec ( from attractor 2 to 1 ) and another at 1.7 sec ( from attractor 1 to 2 ) , where @xmath21 is set to 2 sec in both directions . the significance of deviation from surrogate at @xmath85 is 6.60 and at @xmath86 is 3.25 and indicates that @xmath87 at @xmath29 values -2.5 sec and 1.7 sec are not due to spurious correlations . the horizontal line at @xmath80 has the same meaning as in fig . 2d.,width=480,height=432 ] results of the theoretically coupled systems are summarised in fig . 5 . for the unidirectional flow , shown in fig . 1 , which is referred to as scenario 1 in fig . 5 , the value of the delay obtained by maximising coherence , shown by @xmath88 with the error - bars ( calculated by surrogate analysis ) shown as vertical line , is close to the expected value of -2 sec ( shown as dot in fig . similarly for the bi - directional flow ( see fig . 3 ) which is referred to as scenario 2 in fig . 5 , the values of the delay obtained by maximising coherence with their error - bars ( vertical lines ) are close to the expected delay of 2 sec in either direction ( see fig . 3 ) . compared to the unidirectional flow , the results of the bi - directionally coupled oscillators show slightly more deviations from the expected value of 2 sec but when we include the error - bar , they are close to the expected value . so when dealing with the biological systems we should consider the delay always with the error - bar before we draw any conclusion from it . for the uni - directionally coupled oscillators there is no significant change in the result when @xmath28 is chosen as 0.19 @xmath16 where there is a significant coherence between the two oscillators . further for the above studied systems similar results can be obtained by cross - correlation analysis and by the time delayed phase synchronisation analysis as well @xcite . . vertical lines are the error - bars of delay ( see text for details).,width=384,height=288 ] tremor is an involuntary , periodic movement of the parts of the body @xcite . it can be classified as normal and pathological tremor depending on amplitude and the conditions under which tremor is activated @xcite . essential tremor is a common movement disorder characterised mostly by postural tremor of the arms . other neurological abnormalities are typically absent in essential tremor @xcite . it has been shown by experimental studies in animals @xcite and by positron emission tomography or functional magnetic resonance imaging studies @xcite on human beings , that different parts of the brain are involved in essential tremor . the correlation between the thalamic activity and forearm electromyogram provides a direct evidence for the involvement of the thalamus in the tremor oscillations @xcite . further , control of essential tremor by stereo - tactic lesions and high frequency stimulation of the ventrolateral thalamus adds support for the involvement of thalamus in the mechanism of essential tremor @xcite . as the cerebellum has its main outputs to thalamus which in turn has strong projections to cortex , it is hypothesised that essential tremor arise from an oscillating cerebello - thalamo - cortical loop the output station of which is the motor cortex @xcite . . for the sake of clarity , the power spectrum of eeg is shifted vertically upwards from its original position by a factor of 10 . coherence spectrum of c4 and extensor muscle . the horizontal line at the coherence of 0.024 indicates the confidence level of @xmath67 . there is a significant coherence at the ( tremor ) frequency of 5 @xmath16 and at 10 - 15 @xmath16 as well . the value of @xmath28 used for the maximising coherence is shown along with its coherence value @xmath44 . phase estimate of the c4 and extensor muscle with 95% confidence interval as the error - bars . error - bars are narrow at the significantly coherent frequency bands of 5 @xmath16 and 10 - 15 @xmath16 as well . coherence as a function of @xmath29 , @xmath46 . it has reached maximum at @xmath89 @xmath0 . the horizontal line at @xmath80 has the same meaning as in fig . the significance of deviation from the surrogate is given as @xmath90.,width=480,height=432 ] cortico - muscular coherence for essential tremor has been studied earlier @xcite but has not been ascertained at the tremor frequency in @xcite . we show below that there is a significant cortico - muscular coherence when the signal to noise ratio of the tremor increases beyond a certain threshold value as seen in @xcite . since the paper is aimed at time delay estimation , the detailed results and discussion of the spectral analysis of the tremor study will be reported elsewhere @xcite . for this purpose we consider five patients ( with mean age 66 years and standard deviation 5.1 years ) , which showed postural essential tremor of the arms . in a dimly lighted room , patients are asked to sit on a comfortable chair in a slightly supine position with both hands held against gravity while the forearms are supported . eeg is recorded with a 64-channel eeg system ( neuroscan ) with standard electrode positions @xcite . surface emg is recorded from emg electrodes attached to wrist flexors and extensors of both arms . for all the subjects eeg and emg are sampled at a rate of 1000 @xmath16 . eeg and emg are bandpass filtered online , respectively , between 0.01 - 200 @xmath16 and 30 - 200 @xmath16 . data are stored in a computer and are analysed offline . in all the cases emg is recorded on both hands . each recording lasted for 1 - 4 minutes . artifacts like base line shift , eye blinks etc . are discarded by visual inspection . before further analysis , emg signals are full wave rectified ( magnitude of the deviations from mean ) and eeg is made reference free by constructing the second ( spatial ) derivative , ( laplacian ) @xcite . we discuss in detail for one of the subjects , subject 1 , which we feel typical and for the remaining subjects we summarise the results in a table . for subject 1 , the power spectra of eeg ( solid line ) and emg ( solid line with dots ) , see fig . 6a , are calculated with @xmath65 and hence the frequency resolution is 1 @xmath16 ( since @xmath91 @xmath16 ) . in fig . 6a , we show the power spectra of one of the eeg channels ( c4 ) in the contralateral ( right ) side to tremor ( left ) hand and the rectified emg ( extensor muscle ) of the same hand . power spectrum of eeg ( shown as solid line in fig . 6a ) is shifted vertically upwards by a factor of 10 from its initial position for the sake of clarity . power spectrum of eeg shows dominant activity in a broad band between 8 - 20 @xmath16 containing the bands of 10 - 13 @xmath16 and 15 - 20 @xmath16 , which correspond , respectively , to the alpha and beta activity of the brain @xcite . muscle spectrum ( solid line with dots in fig . 6a ) shows tremor activity around 5 @xmath16 . correlation between the muscle and c4 ( the central electrode overlaying the primary motor cortex , pmc ) of eeg can not be guessed from the eeg spectrum as there is no significant activity around 5 @xmath16 ( tremor frequency ) . correlation becomes clear when we consider coherence . figure 6b shows coherence between c4 and extensor muscle . the horizontal line around the coherence of 0.024 indicates the confidence level of @xmath67 . there is a significant coherence ( and hence correlation ) between c4 and the tremor activity reflected by the muscle @xcite and also in the frequency band of 10 - 15 @xmath16 . phase estimate of this system is given in fig . 6c with @xmath92 confidence interval as error - bars . it is clear from fig . 6c that phase can be reliably estimated at the ( tremor and the significant coherence ) frequency of 5 @xmath16 ( confidence interval is narrow ) and also in the frequency band of 10 - 15 @xmath16 . in an earlier study _ hellwig et al . _ @xcite investigated time delays based on the value of the phase estimate @xmath22 at the * tremor frequency*. but the results obtained @xcite are not in the interpretable level . the reason postulated in @xcite that the delay might be modulated by frequency dependent mechanism is hardly conceivable in a narrow ( single ) frequency band at which phase estimate is significant . the reason may be due to the weak assumption of the model @xmath93 , where @xmath94 is the indispensable constant @xcite , phase shift between the two processes which is ignored in @xcite . however , we show below that the delay estimated by maximising coherence is reasonably in good agreement with the experimentally observed conduction velocity and is in physiologically interpretable range . application of the maximising coherence method to real life data like eeg is not straight forward . since the maximising coherence method is aimed at to capture the time delay ( by making up for the reduction in the coherence ) by artificially shifting ( local dynamics ) the time series , it is very sensitive to non - stationarities in the time series . however , in order to make coherence analysis robust against non - stationarities within the time series , we use the following way to discard the non - stationary parts of the data : we calculate autocorrelation function @xcite , in each of the segments ( of length 1000 samples ) . then the de - correlation time , time at which autocorrelation function falls to @xmath95 ( as the first value of autocorrelation function is 1 by definition ) @xcite is calculated for all the segments . non - stationarities due to base line shift , eye blinks will introduce artificial trends ( lifting up of the mean value of the signal ) @xcite in the signal , which will have higher de - correlation time than the stationary parts of the signal . in order to discard the non - stationary parts , about 50 % of the median of the de - correlation time calculated in all the segments is taken as threshold and parts of the eeg and emg data with de - correlation time greater than the above defined threshold are discarded in the delay estimation . after discarding the non - stationary parts of the data , the segments left out on either side of the removed parts are stitched as if they are recorded continously . of course , care should be taken in applying the above method . the above method will capture only the non - stationary parts if it is applied to the data of low signal to noise ratio which can be avoided by visual inspection of the data . result of the delay estimation between c4 and muscle , is shown in fig . 6d . the horizontal line at @xmath69 in fig . 6d has the same meaning as in fig . here , we have used @xmath96 @xmath16 ( see fig . 6b ) , where significant coherence is observed . negative shift indicates that the time series of eeg is in advance . in fig . 6d , @xmath46 reaches a maximum value at @xmath89 @xmath0 , indicating the cortex ( time series of emg is in advance for positive values of @xmath29 ) is driving the tremor . the significance of deviation from the surrogate is given by @xmath97 and indicates that @xmath46 obtained is not due to spurious correlations . the error - bar obtained by surrogate analysis is 7 @xmath0 and hence the delay obtained for c4 is 16@xmath987 @xmath0 which is in good agreement with the experimentally observed value of around 15 @xmath0 @xcite . comparison of fig . 6d with fig . 2d , shows that the information flow is in one direction ( i.e. ) from cortex to muscle . but in other subjects , in ( an electrode from ) this region , we have also observed a bi - directional flow , a situation similar to fig . 4d . in this case ) . magnitude of the coherence is indicated by the gray scale - bar juxtaposed to the isocoherence map . parts of the scalp which showed significant coherence are coloured black and the confidence level of @xmath67 is marked with white colour ( which is set to zero , see text for details ) . ( electrodes o5-o6 are just displayed and are not used for coherence analysis).,width=528 ] of the 64-channels of eeg recorded , coherence and delay estimation are explained in detail for c4 , as it is overlaying the primary sensorimotor cortex which is involved in motor control of the contralateral hand affected by essential tremor . for the rest of the electrodes ( see fig . 7a ) we give the results of the coherence analysis as isocoherence ( region of scalp with same coherence ) map . electrodes on the boundary of the scalp are not considered for the analysis as they had low signal to noise ratio . further , we have not used the electrodes o5 - o6 for the study . to construct isocoherence map , we have considered the coherence @xmath99 at the ( tremor ) frequency of 5 @xmath16 for all the thirty five channels ( see above ) shown in fig . 7a , from f5 ( top left of scalp , fig . 7a ) , to p6 ( bottom right of scalp , fig . we have subtracted the confidence level ( @xmath67 ) from the coherence so that the confidence level for coherence is zero . in fig . 7b , surface of the scalp exhibiting significant coherence is marked with black colour and the coherence of the parts which do not cross the confidence level of @xmath67 ( which are set to zero after the above normalisation ) are marked with white colour . in fig . 7b , c4 shows maximum coherence , ( see the gray scale - bar juxtaposed to fig . thus , there is a hot spot ( region of significant coherence ) in the pmc region . similar results have been obtained in the spectral study of essential tremor @xcite . there is another hot spot for subject 1 in the frontal region of the scalp ( fcz ) . the detailed discussion of this will be dealt with in @xcite . in fig . 8 ( a - f ) has the same meaning as in fig . the significance of deviation from the surrogate @xmath37 at @xmath47 is given in all the plots . estimated values of delay by maximising coherence are ( c ) -28 @xmath0 ( d ) -2 @xmath100 0 @xmath0 ( e ) -30 @xmath0 and 31 @xmath0 ( f ) 25 @xmath0 . as the maximum of @xmath101 in fig . 8f is not clearly discernable ( see text for details ) it is magnified in the inset shown in fig . negative shift indicates that the time series of eeg is in advance and positive shift indicates that the time series of emg is in advance . a negative delay indicates the flow from muscle to eeg and a positive delay indicates the flow from eeg to muscle . c2 and c6 do not qualify in the surrogate analysis as their @xmath49 values lie below 2 indicating that the values of @xmath46 ( see fig . 8a and 8b ) obtained for these two electrodes are due to spurious correlations and therefore are not considered for further analysis . , width=384,height=528 ] spectral and delay estimations for c4 and extensor muscle have been analysed in detail showing a uni - directional flow from pmc to muscle , with a delay of 16 @xmath0 . for the rest of the electrodes ( c2 , c6 , cp2 , cp4 , cp6 ) of the central hot spot and fcz of the frontal hot spot , the results of delay estimation are given in fig . the significance of deviation from surrogate @xmath37 at @xmath47 is given in all the plots , see fig . 8(a - f ) . for c2 ( 8a ) and c6 ( fig . 8b ) the value of @xmath49 is below two indicating that the values of @xmath46 obtained for these two electrodes are due to spurious correlations . therefore we do not consider them for further analysis . for cp2 , cp4 , cp6 and fcz , @xmath49 values obtained are above 2 ( see fig . 8(c - f ) ) indicating that @xmath46 values obtained for these electrodes are not due to spurious correlations . delay values exhibited by @xmath30 shown in fig . 8(c - f ) are cp2 -28 @xmath98 3.3 @xmath0 ( see fig . 8c ) , cp4 , -2 @xmath100 0 @xmath0 ( see fig . 8d ) , cp6 -30 @xmath98 7.4 and 31 @xmath98 13.5 @xmath0 ( see fig . 8e ) and fcz 25 @xmath98 10.4 @xmath0 ( see fig . the maximum in @xmath102 for fcz shown in fig . 8f is not clearly discernable and hence this region is magnified in the inset shown in fig . 8f where the maximum at @xmath103 can be seen without any ambiguity . the large fluctuations in @xmath46 for fcz ( fig . 8f ) and c4 ( fig . 6d ) indicate larger uncertainty in the delay estimated by this method and is one of the shortcomings of this method . however , this uncertainty , to some extent is revealed by the error - bars ( see above ) obtained by the surrogate analysis . a positive value of @xmath29 indicates that the time series of muscle is in advance and a negative value of @xmath29 indicates that the time series of eeg is in advance . error - bars are obtained by surrogate analysis as discussed in the methodology section . a zero delay for cp4 ( see fig . 8d ) may be due to the nullification of two strong counter acting forces ( drive from pmc to muscle and opposing drive from muscle to pmc , a situation similar to phase locking in coupled oscillators ) . the positive values of estimated delays including the error - bars are well in agreement with the experimentally observed value of 15 @xmath0 for forearm muscles @xcite . for the negative delay ( flow from muscle to eeg ) , it is at least partly in keeping with normal median somatosensory evoked potential ( sep ) latencies of around 20 @xmath0 . further , in order to make the delay statistically significant we report the delay of the hot spot as a whole as the delay of the system , by weighting the delay obtained for each electrode ( when it is qualified in surrogate analysis ) by a fraction of the coherence @xmath44 value that each electrode contributes to the hot spot . by doing so , electrodes showing higher tremor related correlations will contribute more to the delay of the system . for subject 1 , there is a hot spot ( central hot spot ) in the pmc region of scalp and another in the frontal region of the scalp ( i.e. ) fcz ( see fig . 7b ) . of the electrodes contributing to the central hot spot , we do not consider c2 and c6 ( as they do not qualify in the surrogate analysis ) for further analysis . we weight the delay obtained for rest of the electrodes contributing to the central hot spot by @xmath104 , where @xmath105 is the number of electrodes contributing to hot spot after qualifying the surrogate analysis , and @xmath106 $ ] % is the contribution of the @xmath107th electrode to the hot spot ( it is easy to check that @xmath108 ) . similarly @xmath109 is calculated for each hot spot . further , in the calculation of @xmath109 for positive delay , only those electrodes which display positive delay are considered and similarly for negative delay , only those electrodes which display negative delay are considered . for subject 1 , only one electrode , fcz contributes to the frontal hot spot , for which @xmath110 . we have normalised the error - bars of the delays in a similar way . the final values of delay for both the hot spots are given in table 1 along with the delay values obtained by maximising coherence analysis for four other subjects . .summary of delay estimation by maximising coherence for five essential tremor subjects . positive and negative signs are introduced in maximising coherence analysis just to denote the directions of information flow and have got no real life significance . [ cols="^,^,^,^,^ " , ] of the five subjects considered three show an additional hot spot in the frontal region the delay values of which are also given in table 1 ( see third column ) . as can be seen there is also a cortico - muscular delay between the frontal hot spot and the periphery in all three cases . however , this delay is about double the cortico - muscular delay from the central hot spot . these results indicate that both the central area most likely corresponding to the primary sensorimotor cortex and more frontal area are involved in the generation of essential tremor . while the delays from the central area ( pmc ) between 11 and 16 @xmath0 are in keeping with a direct transmission through fast conducting pyramidal pathways @xcite the cortico - muscular delay from the more frontal area rather indicates another way of interaction with the periphery . thus the method of maximising coherence is capable of distinguishing different types of cortico - muscular interactions which are relevant to the understanding of the pathophysiology of essential tremor , for details see @xcite . the musculo - cortical delays partly agree but are generally slightly lower than the somatosensory conduction delays known from routine sep studies and most likely reflect feed back of the peripheral tremor to the cortex . we have used the method of maximising coherence to obtain the time delay between two series . as a test case , we applied this method to uni- and bi - directionally coupled rssler attractors . in both cases , the delays estimated by maximising coherence match well with the expected values , within the error limits which are obtained by surrogate analysis . for eeg - emg time series of essential tremor subjects , this method yields a delay in the range of 11 to 16 @xmath0 for the primary sensorimotor area and 23 - 27 @xmath0 for the more frontal area involved in the tremor oscillation whereas the experimentally observed value of the conduction time between primary cortex and muscle is 15 @xmath98 2 @xmath0 @xcite . the larger delay observed for the frontal area may indicate a different , possibly indirect interaction with the periphery . the musculo - cortical delays ( 9 to 24 @xmath0 ) are partly in keeping with the delay observed ( around 20 @xmath0 ) in sep studies . one of the reasons for the slight deviation of the delay from experimentally observed value may be as follows : we have employed maximising coherence method to capture the delay between two series , assuming that there will be a continuous delayed flow of information from one time series to another . but , it may not be the case in biological data like eeg and emg . intermittently the flow may be absent . one way to check out this is to divide the time series into small sub - portions and calculate the time delay in each portion . but coherence analysis needs at least a minimum of 30000 data points to decide statistically whether or not the two time series under study are linearly correlated . so we need to analyse the nature of information flows using the nonlinear methods like extended granger causality @xcite , transfer entropy @xcite . this issue will be addressed in our future work . of course other methods like cross - correlation analysis and time delayed phase synchronisation analysis @xcite can be used to estimate the delay between the eeg and emg of the essential tremor subjects . but the problems in using cross - correlation are addressed in @xcite and references therein . there exists an analytical expression for the confidence limit for coherence estimate but this is lacking for phase synchronisation used in @xcite . that is the reason why phase synchronisation is not commonly used for correlation analysis ( subsequently for delay analysis in this work ) though this method is expected to be superior to coherence analysis @xcite . however comparison of the different methods of delay estimation is beyond the scope of the current paper . t. mller , m. lauk , m. reinhard , a. hetzel , c.h . lcking , j. timmer , annals of biomedical engineering , 31 ( 2003 ) 1423 . m. lindemann , j. raethjen , j. timmer , g. deuschl , g. pfister , j. neurosci . meth . , 111 ( 2001 ) 127 . carter , proc . ieee , 75 ( 1987 ) 236 . rothwell , p.d . thompson , b.l . day , s. boyd , c.d . marsden , exp . physiol . , 76 ( 1991 ) 159 . b. hellwig , s. huler , b. schelter , m. lauk , g. guschlbauer , j. timmer , c.h . lcking , lancet , 357 ( 2001 ) 519 . d. rybski , s. havlin , a. bunde , physica a , 320 ( 2003 ) 601 . d.m . halliday , j.r . rosenberg , a.m. amjad , p. breeze , b.a . conway , s.f . farmer , prog . biol . , 64 ( 1995 ) 237 . j. timmer , m. lauk , g. deuschl , electroencephalogr . neurophysiol . , 101 ( 1996 ) 461 . s. salenius , s. avikainen , s. kaakkola , r. hari , p. brown , brain , 125 ( 2002 ) 491 . j. raethjen , r.b . govindan , f. kopper , g. deuschl , brain ( in preparation ) . nonlinear time series analysis , edited by h. kantz and t. schreiber ( cambridge university press , 1997 ) . j. theiler , a. longtin , b. galdrikan , j.d . farmer , physica d , 58 ( 1992 ) 77 . t. schreiber , a. schmitz , phys . 77 ( 1996 ) 635 . j. timmer , phys . lett . , 85 ( 2000 ) 2647 . k. narayanan , r.b . govindan , m.s . gopinathan , phys . e , 57 ( 1998 ) 4594 . a. wolf , j.b . swift , l. swinney , a. vastano , physica d , 16 ( 1985 ) 285 . g. deuschl , p. bain , m. brin , scientific committee , mov . disord . , 13 ( 1998 ) 2 . findley , w.c . koller , neurology , 37 ( 1987 ) 1194 . r. llin@xmath111s , r.a . volkind , exp . brain res . , 18 ( 1973 ) 69 . y. lamarre , animal models of physiological , essential and parkinsonian - like tremors . editors l.j . findley and r. capildeo , movement disorders : tremor . ( macmillan press , london , 1984 ) 183 . m. hallett , r.m . dubinsky , j. neurol . , 114 ( 1993 ) 45 . jenkins , p.g . bain , j.g . colebatch , p.d . thompson , l.j . findley , r.s.j . frackowiak , c.d . marsden , d.j . brooks , ann . neurol . , 34 ( 1993 ) 82 . bucher , k.c . seelos , r. dodel , m. reiser , w.h . oertel , ann . neurol , 41 ( 1998 ) 273 . hua , f.a . lenz , t.a . zirh , p.m. dougherty , j. neurosurg psychiatry , 64 ( 1998 ) 273 . benabid , p. pollak , c. gervason , d. hoffmann , d.m . gao , m. hommel , j.e . perret , j. derougemont , lancet , 337 ( 1991 ) 403 . schuurman , d.a . bosch , p.m.m . bossuyt , g.j . bonsel , e.j.w . van someren , r.m.s . de bie , m.p . merkus , j.d . speelman , n. engl . , 342 ( 2000)461 . d.m . halliday , b.a . conway , s.f . farmer , u. shahani , a.j.c . russell , j.r . rosenberg , lancet , 355 ( 2000 ) 1149 . b. hellwig , s. huler , m. lauk , b. guschlbauer , b. kster , r. kristeva - feige , j. timmer , c.h . lcking , clin . neurophysiol . , 111 ( 2000 ) 806 . klem , h.o . jasper , h.h . the ten - twenty electrode system of the international federation . editors g. deuschl and a. eisen , in . recommendations for the practice of clinical neurophysiology : guidelines of the international federation of clinical neurophysiology , electroencephalogr . suppl . , 52 ( 1999 ) 3 . b. hjorth , electroencephalogr . neurophysiol . , 39 ( 1975 ) 526 . b. hjorth , j. clin . neurophysiol . , 8 ( 1991 ) 391 . guyton , text book of medical physiology , editors m.j . wonsiewicz , ( prism books ( pvt ) ltd . , bangalore , india , 1991 ) , 659 . t. mima , m. hallett , j. clin . neurophysiol . , 16 ( 1999 ) 501 . t. mima , m. hallett , clin . neurophysiol . , 110 ( 1999 ) , 1892 . t. mima , j. steger , a.e . schulman , c. gerloff , m. hallett , clin . neurophysiol . , 111 ( 2000 ) 326 . halliday , b.a . conway , s.f . farmer , j.r . rosenberg , neurosci . lett . , 241 ( 1998 ) 5 . d. kaplan and l. glass , understanding nonlinear dynamics , editors j.e . marsden , l. sirovich , m. golubitsky and w. jger ( springer - verlag , new york , 1995 ) . z. chen , p.ch . ivanov , k. hu , h.e . stanley , phys . e , 65 ( 2002 ) 041107 . y. chen , g. rangarajan , j. feng , m. ding , phys . a , 324 ( 2004 ) 26 . t. schreiber , phys . 85 ( 2000 ) 461 . a. gozolchiani , s. moshel , j.m . hausdorff , e. simon , j. kurths , s. havlin , arxiv.org e - print cond - mat/0410617 .

the phase structure of quantum chromodynamics(qcd ) in the presence of an external magnetic field has been explored extensively in recent years@xcite . this is of great phenomenological relevance for non - central heavy ion collisions and dense neutron stars as well as cosmological evolutions in its early stage . in particular , the peculiar response of the spontaneously breaking of chiral symmetry to a strong magnetic field is among the central issues and posseses a theoretical challenge . the profound effect of the magnetic field on chiral symmetry breaking had been pointed out in quantum electrodynamics(qed ) very early@xcite . it was found that magnetic fields can enhance the chiral symmetry breaking due to an effective dimension reduction in the lowest landau level(lll ) in the charged fermionic sector . this phenomenon is called _ magnetic catalysis_. in qcd , this problem can be addressed reliably by the lattice monte - carlo simulation without sign problem@xcite . it has shown intriguingly and , however , unexpectedly that the chiral critical temperature decreases with an increasing magnetic field , which is in explicit conflict with magnetic catalysis and thus termed as _ inverse magnetic catalysis _ or _ magnetic inhibition_. many attempts already exist in the literatures trying to explain this contradiction@xcite . along with the dimension reduction in the fermionic sector , which accounts for the magnetic catalysis , the significance of magnetic fields manifests in another important aspect in a general way in systems with spontanously symmetry breaking . it enhances the fluctuations through the spatial anisotropy of the spectrum of bosonic modes in the system , although the bosons are still in @xmath0 dimension and do not directly suffer from the mermin - wagner - coleman theorem which forbids the spontaneously breaking of continuous symmetry in @xmath1 and @xmath2 dimensions@xcite . this claim follows by the observation that the ginzburg critical window of the chiral phase transition gets widened in the presence of a magnetic field@xcite . the role played by magnetic fields is thus twofold which competes with each other . a simple and straightforward model to investigate this competition is the bose - einstein condensation(bec ) of neutral bosons composed by two equally and oppositely charged fermions@xcite , which is a physical analog of chiral condensate in a magnetic field , yet analytically tractable . the corresponding non - relativistic case had been studied in @xcite , where we found that the critical temperature for the bec was dramatically affected by the magnetic field . in weak coupling domain ( where the coupling fails to support a two fermion bound state in the absence of a magnetic field ) , dimension reduction dominates and we found the usual magnetic catalysis effect . in strong coupling domain ( where a two fermion bound state exists in the absence of a magnetic field ) , however , the fluctuations dominate and the situation gets reversed , we found the inverse magnetic catalysis , i.e. , the critical temperature of bec decreases as an increasing magnetic field . in the present paper , we shall extend the above analysis to the relativistic fermions . technically , this amounts to replace the non - relativistic fermionic propagator by the dirac one in a magnetic field@xcite . the corresponding one loop diagram underlying the theory will be quadratically divergent . the leading divergence can be removed by the coupling renormalization , similar to the non - relativistic case , where the renormalized coupling is in terms of the scattering length , but the logarithmic subleading divergence remains and a uv cutoff has to be introduced by hand . consequently , the results will be explicitly cutoff dependent , like all relativistic field theories with 4-fermion interactions . nevertheless , the cutoff does not change the results qualitatively . we found that the low lying spectrum of the bound pairs is given by @xmath3 where @xmath4(@xmath5 ) is the momentum component parallel ( perpendicular ) to the magnetic field , @xmath6 is an anisotropicity factor and @xmath7 is the rest energy of a bound pair . this spectrum implies a nonrelativistic like condensation temperature @xmath8^{2/3}\frac{\pi}{\bar\mu}\label{transitiontemperature}.\ ] ] at sufficiently low density . unlike the nonrelativistic case where only the anisotropicity factor @xmath9 is magnetic field dependent , the `` mass '' @xmath10 here depends on the magnetic field as well . in the weak coupling domain , the magnetic field dependence of both @xmath9 and @xmath10 support magnetic catalysis and @xmath11 increases with an increasing magnetic field as in the nonrelativistic case . the situation in strong coupling domain , however , is very different from that in the nonrelativistic case , in which the inverse magnetic catalysis prevails regardless of the strength of the magnetic field . in the present case , the inverse magnetic catalysis only appears when magnetic field is low and the magnetic catalysis effect takes place again when magnetic field is sufficiently high because of opposite roles played by @xmath9 and @xmath10 there . consequently , @xmath11 decreases first and then increases with the magnetic field . the rest of the paper is organized as follows : in section ii we lay out the general formulation and present the mean field approximation . the fluctuations beyond the mean field theory , which is necessary for bec , is calculated under the gaussain approximation in section iii . the numerical solutions to the gap equation and the condensation temperature are presented in section iv . section v is devoted to the conclusions and outlooks . some calculation details and useful formulas are presented in the appendices a , b , c and d. throughout the paper , we will work in euclidean space with the four vector represented by @xmath12 with @xmath13 the matsubara frequency for bosons @xmath14 or for fermions @xmath15 with @xmath16 the temperature . we consider a system consisting of relativistic fermions of mass @xmath17 and chemical potential @xmath18 with opposite electric charge interacting through a short ranged instantaneous attractive interaction . in a quark matter , the attractive interaction stems from the nonperturbative qcd effects and the long range coulomb interaction is perturbative and can be ignored in the leading order . the lagrangian density reads @xmath19 where @xmath20 and @xmath21 with @xmath22 the covariant derivative and @xmath23 the charge magnitude . we also defined the charge conjugate operator @xmath24 and @xmath25 . in the following , we shall consider the situation of a constant magnetic field and work in landau gauge , in which the vector potential @xmath26(the field @xmath27 being along the @xmath28 direction ) . introducing the usual hubbard - stratonovich field @xmath29 coupled to @xmath30 , we obtain the partition function @xmath31.\ ] ] where @xmath32 is the normalization constant and the nambu - gorkov(ng ) spinors are defined as @xmath33 the effective action is given by @xmath34,\ ] ] with @xmath35 the thermodynamic potential reads @xmath36 and the number density of fermions is given by @xmath37 with @xmath38 the pressure and @xmath39 the spatial volume of the system . integrating out the ng degrees of freedom , the partition function becomes @xmath40\right),\ ] ] with @xmath41-\int d^4 x\frac{|\delta(x)|^2}{g}.\label{effectiveaction}\ ] ] where the trace runs over space , imaginary time , dirac and ng indices . the mean field approximation ignores the fluctuations in @xmath29 field , which amounts to set @xmath42 . in this approximation , we obtain @xmath43 ^ 2\right ) \left((i\omega_n)^2-\left[e_l^-(\delta_0)\right]^2\right)\right],\end{aligned}\ ] ] where @xmath44 and @xmath45 labels the landau levels ( ll ) . notice that , the factor @xmath46 reflects the non - degeneracy in the lowest landau level ( lll ) . we defined @xmath47^{1/2}.\ ] ] which are the energy spectrum of the fermionic exitations in the system . the thermodynamic potential under mean - field approximation is @xmath48 . \label{pressure0}\ ] ] from the saddle point condition @xmath49}{\delta\delta_0}=0,\ ] ] one obtains the gap equation @xmath50.\label{gapequation}\end{aligned}\ ] ] and @xmath51 at the transition temperatute @xmath52 . the number density under mean field approximation follows from eq.([fermionnumberdensity ] ) and reads @xmath53}{\partial\mu}\right)_t . \label{fermionnumberdensity0}\ ] ] in bcs limit , the critical temperature coincides with the threshold pairing temperature which , at a given chemical potential , is determined by ( [ gapequation ] ) at @xmath51 first and the chemical potential is in turn solved by the density equation ( [ fermionnumberdensity ] ) . in bec limit , however , the roles of ( [ gapequation ] ) and ( [ fermionnumberdensity ] ) reversed because the critical tempearture becomes separate from the pairing . ( [ gapequation ] ) determines the chemical potential@xcite . the bec critical temperature is determined by eq.([fermionnumberdensity ] ) equals ( [ fermionnumberdensity0 ] ) plus the contribution from the bosonic excitation spectrum , whose low momentum behavior will be given in the next section . there we shall also see that the quantity @xmath54 with @xmath18 the solution of the gap equation measures the binding energy of a relativistic pair and locates the bosonic pole below the two - fermion cut at zero total momentum . to make the formulation analytically tractable , we assume that @xmath55 and @xmath56 the gap equation ( [ gapequation ] ) can be approximated by @xmath57.\label{zerotemgapeq}\end{aligned}\ ] ] where we have set @xmath51 , which amounts to approach the critical point from normal to superconducting phase . obviously , this equation is quadratically divergent and needs to be regularized . to maintain the relativistic invariance , we shall use the pauli - villars regularization scheme . introducing the pv regulators , we obtain @xmath58.\label{pvgapeq}\end{aligned}\ ] ] with @xmath59 and @xmath60 . throughout the paper , we shall assume that the regulator mass @xmath61 but finite . carrying out the integral over the momentum under the constrains imposed by the regularization conditions , we obtain @xmath62.\label{gapeqinappendix}\end{aligned}\ ] ] dropping all terms with negative powers of @xmath63 since @xmath64 by definition , we end up with @xmath65\right.\\ \nonumber&+\left.2\sum_{l=0}^{\infty}\left[\frac{\mu}{\sqrt{2leb+m^2-\mu^2}}\tan^{-1}\frac{\mu}{\sqrt{2leb+m^2-\mu^2}}-\frac{\mu^2}{2leb+m^2-\mu^2}\right]\right\}\\ & + \frac{1}{4\pi^2}\left[m^2\sum_{s>0}c_s\lambda_s^2\ln\lambda_s^2 - 2\mu^2\sum_{s>0}c_s\ln\lambda_s^2+m^2\left(1+\ln\frac{2eb}{m^2}\right)-2\mu^2\ln\frac{2eb}{m^2}-eb\ln\frac{4\pi eb}{m^2}\right].\label{regularizedgapeq}\end{aligned}\ ] ] with @xmath66 and @xmath67 the gamma function , and @xmath68 . the detailed steps leading from ( [ gapeqinappendix ] ) to ( [ regularizedgapeq ] ) will be shown in appendix a. the quadratic term in the regulator masses in the last line can be absorbed into the coupling constant as usual and the logarithmic term in the regulator ones remains , leaving the gap equation explicitly depends on an uv cutoff . in the appendix b , we have proved that solutions to this gap equation is a decreasing function of an increasing magnetic field . the gap equation at zero magnetic field can be obtained straightforwardly from ( [ regularizedgapeq ] ) and reads @xmath69 where the critical coupling @xmath70 had absorbed the quadratically divergent term . the strong and weak coupling domain is characterized by a critical coupling , which is the threshold coupling for a bound state to show up in the absence of a magnetic field . it follows then the critical coupling constant is defined as the one for @xmath71 . moreover , we shall denote @xmath72 with @xmath73 an uv cutoff . we thus have @xmath74 this definition establishs an explicit dependence of the dimensionless critical coupling constant @xmath75 on the uv cutoff . in this section , we shall include all nonzero momentum component of the hubbard - stratonovich field to the quadratic order ( gaussian fluctuations ) in the partition function . approching the condensate from the normal phase , we shall reproduce the gap quation ( [ zerotemgapeq ] ) by the thouless critirion and obtain the contribition from the bosonic excitations to the number density ( [ fermionnumberdensity ] ) , which is necessary to determine the critical temperature in bec limit . for the simplicity of notation , we shall not explicitly write out the pauli - villars regulators in the following calculations . but they will be restored whenever the integral is divergent . expanding the bosonic field @xmath29 around the trivial saddle point @xmath51 in the effective action ( [ effectiveaction ] ) , we obtain that @xmath76&\simeq { \cal s}_{\rm eff}[0]-\frac{1}{g}\int d^4 x|\delta(x)|^2\\ & + { \rm tr}\left[g_0 ^ -(x , y)\gamma^5\delta^*(y ) g_0^+(y , x)\gamma^5 \delta(x)\right],\label{guanssianaction}\end{aligned}\ ] ] with @xmath77^{-1}$ ] the diagonal elements in the ng propagator ( [ ngpropagator ] ) , which are nothing but the inverse dirac operators in a magnetic field in coordinate space . the dirac propagator had been firstly obtained by schwinger using the proper time method . here we shall employ the form with explicit landau levels dependence , given in@xcite @xmath78^{-1}\delta(x - y)\\ = & e^{i\phi(x , y)}t\sum_n\int\frac{d^3{\bf k}}{(2\pi)^3}e^{-ik\cdot z}\sum_{l=0}^{\infty}g_l^\pm(\omega_n,{\bf k}),\label{diracpropagator}\end{aligned}\ ] ] with @xmath79\left[2\pi_+l_l\left(\frac{2{\bf k}_\perp^2}{eb}\right)-2\pi_-l_{l-1}\left(\frac{2{\bf k}_\perp^2}{eb}\right)\right]\right.\\ & + \left.4({\bf k}\cdot\gamma)_\perp l_{l-1}^1\left(\frac{2{\bf k}_\perp^2}{eb}\right ) \right\}.\end{aligned}\ ] ] where @xmath80 and we also defined the projectors @xmath81 , @xmath82 is the generalized laguerre polynomials(@xmath83 and @xmath84 by definition ) . the gauge dependent phase factor is given by @xmath85 which cancels out in the combination of ( [ guanssianaction ] ) . consequently , the kernel mediating @xmath86 and @xmath29 there is explictly translational invariant , i.e. , depends only on @xmath87 , as expected . in momentum space , ( [ guanssianaction ] ) becomes @xmath88={\cal s}_{\rm eff}[0]-t\sum_n\int \frac{{d^3\bf p}}{(2\pi)^3}\gamma^{-1}(\omega_n,{\bf p})|\delta(\omega_n,{\bf p})|^2,\label{gaussianeffectiveaction}\ ] ] where the dependence of the coefficient @xmath89 on @xmath90 and @xmath27 is implicit . the instability towards the superconducting phase is then signaled by the condition @xmath91 which is just the thouless criterion for the onset of a long range order . the number density equation is determined by @xmath92 where @xmath93 $ ] with @xmath94 the fermi - dirac distribution function , is the fermionic contribution defined in ( [ fermionnumberdensity ] ) . ( [ thoulesseq ] ) and ( [ gaussiannumbereq ] ) are two basic equations to solve for bcs / bec crossover@xcite . following the convention in@xcite , one can write the number density ( [ gaussiannumbereq ] ) in terms of a phase shift defined by @xmath95 $ ] . we have @xmath96 with @xmath97 $ ] the bose - einstein distrubution function . to calculate @xmath89 , we make use of the eq.([diracpropagator ] ) and the fourier transformation of the fluctuation field @xmath98 and obtain @xmath99\left[l_{l^\prime}\left(\frac{2{\bf p}_\perp^2}{eb}\right)l_l\left(\frac{2{\bf q}_\perp^2}{eb}\right)+l_{l^\prime-1}\left(\frac{2{\bf p}_\perp^2}{eb}\right)l_{l-1}\left(\frac{2{\bf q}_\perp^2}{eb}\right)\right]\right.\\ & \left.-8\left({\bf k}\cdot{\bf q}\right)_\perp l^1_{l^\prime-1}\left(\frac{2{\bf p}_\perp^2}{eb}\right)l^1_{l-1}\left(\frac{2{\bf q}_\perp^2}{eb}\right)\right\}-\frac{1}{g}.\end{aligned}\ ] ] with @xmath100 . working out the sum over matsubara frequecies , one can verify that the condition ( [ thoulesseq ] ) will yield the same gap equation as ( [ zerotemgapeq ] ) . let s proceed by employing the proper - time representation of the propagators introduced by schwinger @xmath101 where the feynmann parametrization scheme had been used . we obtain @xmath102+(1-x)\left[(\omega'+\omega+2\mu)^2-q_3 ^ 2 - 2leb - m^2 \right ] ) } \\ \nonumber&\times\left\ { \left[\omega^{\prime}(\omega^{\prime}+\omega+2\mu)-k_3q_3-m^2\right]\left[l_{l^\prime}\left(\frac{2{\bf p}_\perp^2}{eb}\right)l_l\left(\frac{2{\bf q}_\perp^2}{eb}\right)+l_{l^\prime-1}\left(\frac{2{\bf p}_\perp^2}{eb}\right)l_{l-1}\left(\frac{2{\bf q}_\perp^2}{eb}\right)\right]\right.\\ & \left.-8\left({\bf k}\cdot{\bf q}\right)_\perp l^1_{l^\prime-1}\left(\frac{2{\bf p}_\perp^2}{eb}\right)l^1_{l-1}\left(\frac{2{\bf q}_\perp^2}{eb}\right)\right\}-\frac{1}{g}.\label{inversegamma}\end{aligned}\ ] ] where we analytically continuated the matsubara frequency to the continuous one as @xmath103 and shift the frequency from @xmath104 to @xmath105 . the same approximation as in the preceding section , @xmath106 has been made in ( [ inversegamma ] ) that turns the matsubara sum over @xmath107 to the integral over @xmath108 using the generating function of laguerre polynomials @xmath109 one can carry out the sum over landau levels and has @xmath110}\\ & \times \left\ { \left[k\cdot q - m^2\right]\frac{1+tt'}{(1-t)(1-t')}-8\left({\bf k}\cdot{\bf q}\right)_\perp\frac{tt'}{(1-t)^2(1-t')^2 } \right\}-\frac{1}{g}.\label{covariantgamma}\end{aligned}\ ] ] with @xmath111 and @xmath112 . we defined @xmath113 , @xmath114 and @xmath115 , by which eq.([covariantgamma ] ) can be written in a covariant form . one may observe that all momentum integrals become gaussian ones , which is the advantage of using the proper - time representations . the integral over @xmath116 can be carried out by shifting the variable to @xmath117 and wick rotating @xmath118 with @xmath119 from @xmath120 to @xmath121 , and we obtain @xmath122}\\ & \times \left\ { \left[\frac{1}{\lambda}+x(1-x)p^2+m^2\right]\frac{1+tt'}{(1-t)(1-t')}+8\left({\bf k}\cdot{\bf q}\right)_\perp\frac{tt'}{(1-t)^2(1-t')^2 } \right\}-\frac{1}{g}.\end{aligned}\ ] ] the other integral over momentum @xmath123 can be performed by noticing that @xmath124 and @xmath125 we end up with @xmath126 -\frac{\tanh(\lambda xeb)\tanh(\lambda(1-x)eb)}{\tanh(\lambda xeb)+\tanh(\lambda(1-x)eb)}\frac{{\bf p}_\perp^2}{eb}}\left\ { \left[\frac{1}{\lambda}+x(1-x)p^2+m^2\right]\coth(\lambda eb)\right.\\ & + \left.\frac{1}{\sinh^2(\lambda eb ) } \left[eb -\frac{\tanh(\lambda xeb)\tanh(\lambda(1-x)eb)}{\tanh(\lambda xeb)+\tanh(\lambda(1-x)eb)}{\bf p}_\perp^2\right ] \right\}-\frac{1}{g}.\label{finalgamma}\end{aligned}\ ] ] the singularity structure of @xmath127 in the entire complext @xmath128-plane reflects the two - particle spectrum . there will be an isolated real pole representing the two - body bound pair and a branch cut along the real axis representing the continuum of two - particle excitations . for sufficiently strong coupling and low temperature , the contribution to the number density is dominated by the bound pair pole , which is determined by @xmath129 with @xmath130 the solution to the mean field equation ( [ regularizedgapeq ] ) . we henceforth consider the expansion of ( [ finalgamma ] ) around this pole and obtain @xmath131 with @xmath132}\\ & \times\left\{\left[\frac{2}{\lambda}+4x(1-x)\bar{\mu}^2+m^2\right]\coth(\lambda eb)+\frac{eb}{\sinh^2(\lambda eb ) } \right\ } , \label{coeffa}\end{aligned}\ ] ] and @xmath133}\\ \nonumber&\times\frac{\tanh(\lambda xeb)\tanh(\lambda(1-x)eb)}{\tanh(\lambda xeb)+\tanh(\lambda(1-x)eb)}\\ & \times \left\{\left[\frac{1}{\lambda}+4x(1-x)\bar{\mu}^2+m^2\right]\coth(\lambda eb)+\frac{2eb}{\sinh^2(\lambda eb)}\right\}. \label{coeffb}\end{aligned}\ ] ] notice that , the kinetic term in ( [ gammaundergaussian ] ) becomes anisotropic with respect to the directions along and perpendicular to the magnetic field due to the rotational symmetry breaking by the magnetic field . one can readily show that @xmath134 at zero magnetic field as expected . moreover , the coefficients satisfy the inequality @xmath135 regardless of solutions to the gap equation(see appendix d for details ) . the bosonic spectrum at low momentum is determined by the pole of @xmath127 and reads @xmath136 with the anisotropicity factor @xmath137 in contrast to the non - relativistic case , in which the mass of the pair is @xmath138 , the mass is replaced by the @xmath139 , which is magnetic field dependent . once @xmath140 becomes smaller than @xmath17 , stable bosonic pairs appear and they dominate the total number density at sufficienly low temperature . the condensation temperature is determined by setting the chemical potential at the solution of the mean field equation ( [ regularizedgapeq ] ) , i.e. @xmath141 , and the phase factor follows @xmath142 where @xmath143 is the heaviside step function . substituting ( [ phasefactor ] ) into ( [ numberdensity ] ) , we obtain @xmath144^{-1 } , \label{beccondition}\ ] ] where @xmath145 the fermionic contribution had been ignored under the condition @xmath146 of our approximation . solving ( [ beccondition ] ) for @xmath11 , we end up with the formula ( [ transitiontemperature ] ) reported in section i. in this section , we shall explore the bec critical temperature in view of ( [ regularizedgapeq ] ) and ( [ transitiontemperature ] ) . the results display a nontrivial dependence on the strength of the coupling constant and magnetic field . the strong coupling and weak coupling domains correspond to @xmath147 and @xmath148 respectively . for the discussion below , it is convenient to introduce a dimensionless critical temperature @xmath149 where @xmath150 is given by ( [ transitiontemperature ] ) with @xmath151 and @xmath152 . in the strong coupling domain , @xmath153 is the solution of the gap equation at zero magnetic field and @xmath150 is the critical temperature of a bound pair at zero magnetic field . in the weak coupling domain , we set @xmath154 and @xmath150 the critical temperature of a nonrelativistic boson with mass @xmath138 at zero magnetic field . the ( inverse)magnetic catalysis implies increasing(decreasing ) @xmath155 with magnetic field . let us consider first the weak magnetic field limit @xmath156 , where analytic method can be applied . in the strong coupling domain , we expand the rhs of the gap equation to the quadratic order in @xmath157 with the aid of stirling formula for the gamma and di - gamma functions in ( [ regularizedgapeq ] ) and the euler - maclaurin formula for the infinite series there and obtain @xmath158 with the solution @xmath159,\ ] ] where @xmath160 and @xmath161 is the derivative of rhs of ( [ regularizedgapeq ] ) with respect to @xmath18 at @xmath162 and @xmath163 , given by @xmath164 the anisotropicity in this approximation reads @xmath165 with @xmath166}\left\{\frac{1}{\lambda}(2+x-6x^2 + 3x^3)+x(1-x)^2\left[4x(1-x)\bar{\mu}_0 ^ 2+m^2\right]\right \}\\ & = \frac{3m^2 - 2\bar{\mu}_0 ^ 2}{32\bar{\mu}_0 ^ 2}-\frac{9m^4 - 28m^2\bar{\mu}_0 ^ 2 + 16\bar{\mu}_0 ^ 4}{96\bar{\mu}_0 ^ 3\sqrt{m^2-\bar{\mu}_0 ^ 2}}\tan^{-1}\frac{\bar{\mu}_0}{\sqrt{m^2-\bar{\mu}_0 ^ 2}},\end{aligned}\ ] ] and @xmath167}\\ \nonumber&\times\left[\frac{3}{\lambda}+4x(1-x)\bar{\mu}_0 ^ 2+m_s^2\right ] \\ = & \frac{1}{2}\ln\lambda^2-\frac{m^2 - 2\bar{\mu}_0 ^ 2}{2\bar{\mu}_0\sqrt{m^2-\bar{\mu}_0 ^ 2}}\tan^{-1}\frac{\bar{\mu}_0}{\sqrt{m^2-\bar{\mu}_0 ^ 2}}+\frac{2}{3}.\end{aligned}\ ] ] it follows that the dimensionless critical temperature @xmath168 with the coefficient @xmath169 the numerical values of the coefficient @xmath170 are tabulated in table i for different couplings and cutoffs in the strong coupling domain and indicate an inverse magnetic catalysis at a weak magnetic field . the scaled bound state mass versus the dimensionless magnetic field in strong coupling domain . the solid and dashed line corresponds to an uv cutoff @xmath171 and @xmath172 , respectively.,height=264 ] the scaled bound state mass versus the dimensionless magnetic field in weak coupling domain . the solid and dashed line corresponds to an uv cutoff @xmath171 and @xmath172 , respectively.,height=264 ] coming to the weak coupling domain , the bound magnetic field is the only catalyst of the bound state and a weak magnetic field triggers a weak bound state , nonrelativistic formulas in @xcite can be carried over in the weak field limit . @xmath173 and @xmath174 with the scattering length @xmath175 extracted from @xmath176 the lll approximation works here and the usual magnetic catalysis emerges ( increasing with @xmath27 ) . the details of the non - relativistic approximation of the gap equation can be found in appendix d. .[tab : table]numerical values of @xmath170 for different couplings and cutoffs in the strong coupling domain . [ cols="^,^,^",options="header " , ] the anisotropicity @xmath9 versus the dimensionless magnetic field in strong coupling domain . the solid and dashed line corresponds to solutions to the gap equation with an uv cutoff @xmath171 and @xmath172 , respectively.,height=264 ] the anisotropicity @xmath9 versus the dimensionless magnetic field in weak coupling domain . the solid and dashed line corresponds to solutions to the gap equation with an uv cutoff @xmath171 and @xmath172 , respectively.,height=264 ] the ratio @xmath177 versus the dimensionless magnetic field in strong coupling domain . the solid and dashed line corresponds to solutions to the gap equation with an uv cutoff @xmath171 and @xmath172 , respectively.,height=264 ] the ratio @xmath177 versus the dimensionless magnetic field in weak coupling domain . the solid and dashed line corresponds to solutions to the gap equation with an uv cutoff @xmath171 and @xmath172 , respectively.,height=264 ] the condensation temperature versus the dimensionless magnetic field in strong coupling domain . the solid and dashed line corresponds to solutions to the gap equation with an uv cutoff @xmath171 and @xmath172 , respectively.,height=264 ] the condensation temperature versus the dimensionless magnetic field in weak coupling domain . the solid and dashed line corresponds to solutions to the gap equation with an uv cutoff @xmath171 and @xmath172 , respectively.,height=264 ] beyond the weak field limit , we solve the gap equation ( [ regularizedgapeq ] ) numerically at two different coupling domains that are mildly above and blow the critical coupling with two different sets of uv cutoff and the results are presented in fig.1 and fig.2 . in both cases , the chemical potential at the critical temperature decreases with an increasing magnetic field indicating an increasing binding energy of the bound pairs , consistent with the magnetic catalysis . in strong coupling domain , however , bound pairs may appear at zero magnetic field in contrast to the case in weak coupling domain , in which the pairs can only support by magnetic catalysis . therefore , the curve in fig.1 starts from @xmath178 at @xmath162 while the curve in fig.2 from @xmath179 at @xmath162 . in both cases , the results display an explicit dependence on the uv cutoff , which is intrinsic in the ( 3 + 1 ) dimensional model with contact interaction . but the dependence does not change the results qualitatively . the reason for the cross between the solid and dashed curves at lower magnetic field can be explained as follows . one can show that @xmath180^{-1}.\label{muoflambda}\ ] ] with @xmath181 and @xmath182 the rhs of the gap equation ( [ regularizedgapeq ] ) . it had been proved in the appendix that @xmath183 and ( [ muoflambda ] ) thus changes sign at a critical value @xmath184 . one therefore expects a crossing of the two sets of solutions corresponding to different uv cutoffs at strong coupling domain with @xmath185 as shown in fig.1 . such crossing should be absent in weak coupling since @xmath186 . the magnetic field dependence of the dimensionless critical temperature ( [ dimltc ] ) is determined by that of the anosotropicity factor @xmath9 and the ratio @xmath187 , which are plotted in fig.3 to fig.8 . as we can see , the anisotropicity decreases ( increases ) with magnectic field in strong ( weak ) coupling domains while the ratio @xmath188 increases with magnetic field in both domains . consequently , the competetion between the two opposite field dependence of @xmath189 and @xmath188 in the strong coupling domain yields an inverse magnetic catalysis at lower magnetic field but magnetic catalysis at higher magnetic field . in the weak coupling domain , however , both factors increases with the magetic field and the usual magnetic catalysis emerges . in this paper , we have investigated the bec of bound pairs composed by two equally and oppositely charged relativistic fermions in the presence of a magnetic field . this is inspired by the inverse magnetic catalysis effect found in lattice qcd and is a generalization of our non - relativistic work @xcite . the purpose is to explore the interplay between the pairing dynamics and the fluctuations of bound pairs in a magnetic field , the latter of which underlies the mermin - wigner - coleman theorem on the absence of a long range order in lower dimensions . a nambu - jona - lasinio(njl)-like pairing interaction is considered and the fluctuation is examined within the gaussian approximation , which is valid at a sufficiently low density . unlike the case without a magnetic field where a bound state emerges for the coupling strength above a certain threshold , the bound states occurs also in the weak coupling domain because of the @xmath190 nature of the fermionic spectrum within lll . the binding energy of a bound pair is found to increase with the magnetic field as a consequence of the magnetic catalysis . the dependence of the bec critical temperature on the magnetic field depends on the coupling strength . in weak coupling domain , we found the usual magnetic catalysis . in strong coupling domain , unlike the nonrelativitic case , the inverse magnetic catalysis appears only in small magnetic field region and magnetic catalysis takes place again when magnetic field becomes sufficiently large . the reason for this difference is the dependence of the mass of bound pairs on the magnetic field in relativistic case . the condensation temperature is thus determined by two factors , the anisotropy and the ratio between the mass of bound pairs at zero magnetic field and the one with magnetic field dependence , both depends on the magnetic field . in strong coupling domain , the two factors have opposite field dependence and yield the inverse magnetic catalysis in small magnetic filed region but magnetic catalysis at large field . in weak coupling domain , both factors increase with the magnetic field and magnetic catalysis prevails . this variation indicates that the enhanced fluctuations by the anisotropy in the kinetic terms of the bound pairs is less dominant as compared with the nonrelativistic case . although the system we have studied shares the same physics as the chiral condensate in qcd regarding general prperties of a long range order and its fluctuations and is analytically tractable . the qcd dynamics , however , are expected to be far more complicated and other mechanisms may also contribute the simulated magnetic field dependence of the condensation temperature . for example , the quark loop that is influenced by the magnetic field can impact on the running of qcd coupling . in recently works based on functional renormalization group studies@xcite , inverse magnetic catalysis at a low magnetic field and a delayed magnetic catalysis at a high field had been found . the magnetic field at the turning point ( minimum @xmath52 ) between decreasing and increasing @xmath52 is expected to be of order of several @xmath191 . ads / qcd model suggests the same behavior and similar value of magnetic field at the turning point@xcite . it is shown by one lattice calculation@xcite , however , that even at an extremely strong field @xmath192 , inverse magnetic catalysis still prevails . the reconciliation is still lacking at the present stage . in addition to the theoretical values of our present study , the physics involved may also be applicable in cold dense quark matter , such as the color - flavor - locking superconducting phase and the planar phase in single flavor superconductivity@xcite . they are of interests in dense neutron stars , where magnetic field up to @xmath193 to @xmath194 appears on the surface and even higher magnetic field exists in the interior . this research is partly supported by the ministry of science and technology of china ( mstc ) under the `` 973 '' project no . 2015cb856904(4 ) . b. f. is supported by nsfc under grant no . 11305067 , no . 11535005 and the fundamental research funds for the central universities , hust : no . hou and h - c ren are partly supported by nsfc under grant nos . 11375070 , 11135011 and 11221504 . p - p wu is supported by the exploratory fund of henan polytechnic university , no . 13156041 . in this appendix , the reduction of the gap equation ( [ gapeqinappendix ] ) to the form ( [ regularizedgapeq ] ) shall be presented . the first term inside the brackets of ( [ gapeqinappendix ] ) can be evaluated as @xmath195.\end{aligned}\ ] ] using the infinite product representation of gamma function @xmath196,\ ] ] one has @xmath197 therefore the gap equation ( [ gapeqinappendix ] ) becomes @xmath198\label{intermediate}.\end{aligned}\ ] ] with @xmath199 it can be further simplified for @xmath200 and @xmath201 by eliminating the terms that vanishes in the limit @xmath202 with @xmath203 since @xmath204 . ] . for this purpose , we rewrite ( [ intermediate ] ) as @xmath205\right\}.\end{aligned}\ ] ] we have @xmath206 and the other infinite series converges for each @xmath207 , which vanishes in the limit @xmath208 for @xmath203 and can be ignored . applying the stirling formulas @xmath209 and @xmath210 with @xmath211 the bernoulli numbers , @xmath212 . we thus have , for @xmath203 @xmath213 and @xmath214 we end up with the form of ( [ regularizedgapeq ] ) of the gap equation . in this appendix , we shall provide analytical proofs of the monotinic dependence of the bound state mass and the anisotropy factors on the magnetic field . because of the complication introduced by pauli - villars regulators , we limit our scope within the approximaton that maintains only the logarithmic terms in the regulator masses . denoting the rhs of the gap equation ( [ gapeqinappendix ] ) by @xmath215 , we have _ lemma _ : ( 1 ) for a given @xmath27 , @xmath216 is a monotonically increasing function of @xmath18 for @xmath217 ; ( 2 ) for a given @xmath18 within @xmath218 , @xmath216 is a monotonically increasing function of @xmath27 . proof : ( 1 ) the partial derivative of @xmath216 with respect to @xmath18 reads : @xmath219 with @xmath220 defined in ( [ thetas ] ) and @xmath221 and @xmath222 . let us divide the above derivative into two terms , @xmath223 and @xmath224 , with @xmath224 carrying the logarithmic uv divergence and @xmath223 not . we have @xmath225\\ \simeq&\frac{eb\mu}{\pi^2}\left[\sum_{l=0}^\infty \alpha_l\left(\frac{1}{2leb+m^2-\mu^2}-\frac{1}{2leb+m^2}\right)\right]\ge 0,\end{aligned}\ ] ] and @xmath226.\ ] ] where we have assumed that the terms with negative power of the regulator masses dropped from @xmath223 are too small to offset the direction of the inequality . it follows from the binet formula @xmath227 and its derivative @xmath228 that @xmath229 . consequently @xmath230 \(2 ) taking the derivative of ( [ gapeqinappendix ] ) with respect to @xmath27 , we find that @xmath231 where @xmath232\\ \nonumber\ge&\frac{e\mu}{4\pi^2}\sum_{l=0}^\infty\alpha_l\left[\frac{leb+m^2-\mu^2}{(2leb+m^2-\mu^2)^{\frac{3}{2 } } } \sin2\theta_0\right.\\ \nonumber&-\left.\frac{leb\mu}{(2leb+m^2)(2leb+m^2-\mu^2)}\right]\\ = & \frac{e\mu^2}{2\pi^2}\sum_{l=0}^\infty\alpha_l\frac{m^2-\mu^2}{(2leb+m^2)}\ge 0,\end{aligned}\ ] ] and @xmath233\nonumber\\ = & \frac{1}{2\pi^2}\frac{\partial}{\partial b}\left\{eb\left[\ln\gamma\left(\frac{m^2}{2eb}\right)-\frac{m^2}{2eb}\left(\ln \frac{m^2}{2eb}-1\right)\right.\right.\nonumber\\ & -\left.\left.\frac{1}{2}\ln\frac{4\pi eb}{m^2}\right]\right\}\ge 0.\end{aligned}\ ] ] with the inequality following from ( [ b1 ] ) and ( [ b2 ] ) . the lemma is proved . for the solution of the gap equation , @xmath234 and we arrive at _ theorem _ : the bound state mass is a decreasing function of the external magnetic field . @xmath239 \frac{ebs's\cosh((s'+s)eb)}{s'+s}+\frac{(eb)^2s's}{(s'+s)\sinh((s'+s)eb ) } -\left[\frac{1}{s'+s}+\frac{4s's\bar{\mu}^2}{s'+s}+m^2\right]\\ \nonumber & \times\frac{\tanh ( s'eb)\tanh ( s eb)}{\tanh ( s'eb)+\tanh ( s eb ) } -\frac{2eb\tanh ( s'eb)\tanh ( s eb)}{\sinh((s'+s)eb)(\tanh ( s'eb)+\tanh ( s eb))}\nonumber\\ \nonumber = & \left[\frac{1}{s'+s}+\frac{4s's\bar{\mu}^2}{s'+s}+m^2\right ] \left[\frac{ebs's}{s'+s}\cosh((s'+s)eb)-\frac{\tanh ( s'eb)\tanh ( s eb)}{\tanh ( s'eb)+\tanh ( s eb)}\right ] + \frac{ebs's}{(s'+s)^2}\cosh((s'+s)eb)\\ \nonumber & + \frac{(eb)^2s's}{(s'+s)\sinh((s'+s)eb ) } -\frac{2eb\tanh ( s'eb)\tanh ( s eb)}{\sinh((s'+s)eb)(\tanh ( s'eb)+\tanh ( s eb))}\nonumber\\ \geq & \frac{eb s's}{s'+s}\left[\frac{1}{s'+s}+\frac{4s's\bar{\mu}^2}{s'+s}+m^2\right ] \left[\cosh((s'+s)eb)-1\right ] + \frac{ebs's}{(s'+s)^2}\left[\cosh((s'+s)eb)-\frac{(s'+s)eb}{\sinh((s'+s)eb)}\right]\nonumber \\ \geq & \frac{ebs's}{(s'+s)^2\sinh((s'+s)eb)}\left[\sinh ( 2(s'+s)eb)-2(s'+s)eb\right ] \geq 0.\end{aligned}\ ] ] a crucial inequality employed above is that @xmath240 for positive @xmath241 and @xmath242 . to prove it , we notice that the function @xmath243 is monotonically increasing for @xmath244 and @xmath245 . setting @xmath246 , it follows from the inequality @xmath247 that @xmath248 repeating the same argument with @xmath249 , we end up with @xmath250 and the inequality is proved . in this section , we shall take the nonrelativistic ( nr ) approximation of the gap equation ( [ regularizedgapeq ] ) . the conditions for the nr approximation are @xmath251 , @xmath252 with the nr binding energy @xmath253 and cyclotron frequency @xmath254 . let us break the infinite series on rhs of ( [ regularizedgapeq ] ) into two pieces , @xmath255,\ ] ] and @xmath256.\ ] ] with @xmath220 defined in ( [ thetas ] ) and @xmath257 . consequently @xmath258,\end{aligned}\ ] ] and the summation in @xmath259 can be approximated by an integral , i.e. @xmath260 using the formula @xmath261 with @xmath262 a contour on @xmath263-plane going around the positive real axis counterclockwisely , we find @xmath264 it follows from the hermite formula and stirling formula for @xmath265 that @xmath266 and @xmath267 consequently @xmath268 . \label{series}\end{aligned}\ ] ] applying the stirling formula to the gamma function on rhs of ( [ regularizedgapeq ] ) and combining with ( [ series ] ) , we end up with @xmath269.\ ] ] which , upon identifying @xmath270 with @xmath271 , is the nonrelativistic gap equation obtained in @xcite . @xmath274}\left\{\left[\frac{2}{\lambda}+4x(1-x)\bar{\mu}^2+m^2\right]\coth(\lambda eb)+\frac{eb}{\sinh^2(\lambda eb ) } \right\}\\ = & \frac{{\bar{\mu}}}{\pi^2eb}\int_0 ^ 1 dx\int_0^\infty d\xi \xi^{s+\frac{1}{2 } } x(1-x)e^{-\frac{m}{\omega_b}\xi(2x-1)^2 + 8\frac{b}{\omega_b}x(1-x)}\left\{\left[\frac{2eb}{\xi}+4x(1-x)\bar{\mu}^2+m^2\right]\coth(\xi)+\frac{eb}{\sinh^2(\xi ) } \right\},\end{aligned}\ ] ] @xmath275}\frac{\tanh(\lambda xeb)\tanh(\lambda(1-x)eb)}{\tanh(\lambda xeb)+\tanh(\lambda(1-x)eb)}\\ \nonumber & \times\left\{\left[\frac{1}{\lambda}+4x(1-x)\bar{\mu}^2+m^2\right]\coth(\lambda eb)+\frac{2eb}{\sinh^2(\lambda eb)}\right\}\\ \nonumber = & \frac{{\bar{\mu}}}{\pi^2}\int_0 ^ 1 dx\int_0^\infty d\xi \xi^{s-\frac{1}{2 } } e^{-\frac{m}{\omega_b}\xi(2x-1)^2 + 8\frac{b}{\omega_b}x(1-x ) } \frac{\tanh(\xi x)\tanh(\xi(1-x))}{\tanh(\xi x)+\tanh(\xi(1-x))}\\ & \times\left\{\left[\frac{eb}{\xi}+4x(1-x)\bar{\mu}^2+m^2\right]\coth(\xi)+\frac{2eb}{\sinh^2(\xi)}\right\}.\end{aligned}\ ] ] and define @xmath272 and @xmath273 as their analytic continuations at @xmath276 , i.e. @xmath277 and @xmath278 . both integrals are convergent for @xmath279 and the main contribution to the integral comes from @xmath280 . in the nonrelativistic limit , @xmath252 and @xmath251 , the integrand is peaked at @xmath281 and only the 2nd and 3rd terms inside the brackets dominate . it follows that @xmath282 upon writing @xmath283 and using the integral representation of the hurwitz zeta function , we find that @xmath284.\ ] ] under the same approximation , we have @xmath285 using the identity @xmath286,\ ] ] and integration by part , we find d. e. kharzeev , k. landsteiner , a. schmitt and h - u yee , lect . notes phys . * 871 * , 1 ( 2013 ) and _ references therein _ ; v. miransky and i. shovkovy , phys . rep * 576 * ( 2015 ) 1 . k. klimenko , theor . phys . * 89 * , 1161 ( 1991 ) ; z. phys . c * 54 * , 323 ( 1992 ) . v. gusynin , v. miransky and i. shovkovy , phys . lett . * 73 * , 3499 ( 1994 ) ; phys . d * 52 * , 4718 ( 1995 ) ; phys . b * 349 * , 477 ( 1995 ) ; nucl . * b 462 * , 249 ( 1996 ) . g. bali , f. bruchmann , g. endrdi , z. fodor , s.d . katz , s. krieg , a. sch@xmath288fer and k.k . szab@xmath289 , j. high energy phys . 02 , ( 2012 ) 044 ; g. bali , f. bruchmann , g. endrdi , z. fodor , s.d . katz , s. krieg and a. sch@xmath288fer , phys . d * 86 * , 071502 ( 2012 ) . f.preis , a. rebhan and a. schmitt , j. high energy phys . 03 ( 2011 ) 033 . k. fukushima and j. m. pawlowski , phys . d * 86 * , 076013 ( 2012 ) . k. fukushima and y. hidaka , phys . lett . * 110 * , 031601 ( 2013 ) . t. kojo and n. su , phys . lett . * b 720 * , ( 2013 ) . f. bruckmann , g. endrodi , and t. g. kovacs , j. high energy phys . 1304 , 112 ( 2013 ) . j. chao , p. chu and m. huang , phys . rev d * 88 * , 054009 ( 2013 ) ; l. yu , h. liu and m. huang , phys . d * 90 * , 074009 ( 2014 ) ; l. yu , j. van doorsselaere and m. huang , phys . d * 91 * , 074011 ( 2015 ) . r.farias , k.gomes , g. krein and m. pinto , phys . c * 90 * , 025023 ( 2014 ) . b. feng , d.f . hou and h.c . ren , phys . d * 92 * , 065011 ( 2015 ) . e.j.ferrer , v.de la incera and x. wen , phys . d * 91 * , 054006 ( 2015 ) . n. mueller and j. m. pawlowski , phys . d * 91 * , 116010 ( 2015 ) . j. braun , w. a. mian and s. rechenberger , arxiv:1412.6025[hep - ph ] . k. a. mamo , j. high energy phys . 05 , 121 ( 2015 ) . a. flachi , k. fukushima and v. vitagliano , phys . * 114 * , 181601 ( 2015 ) . chen , k. fukushima , x.g . huang and k. mameda , arxiv:1512.08974[hep - ph ] . k. hattori , t. kojo and n. su , arxiv:1512.07361[hep - ph ] . n. mermin and h. wagner , phys . rev . lett . * 17 * , 1133 ( 1966);p . hohenberg , phys . rev . * 158 * , 383 ( 1967 ) ; s.coleman , commun . math . phys . * 31 * , 259 ( 1973 ) . b. feng , d - f hou and h - c ren , _ in preparation_. j. schwinger , phys . rev . * 82 * , 664 ( 1951 ) ; a. chodos , i. everding and d. a. owen , phys . rev . d * 42 * , 2881 ( 1990 ) . c. s de melo , m. randeria and j.r.engelbrecht , phys . lett . * 71 * , 19 ( 1985 ) . p. nozires and s schmitt - rink , j. low temp . phys . * 59 * , 195 ( 1985 ) . y. nishida and h. abuki , phys . d * 72 * , 096004 ( 2005 ) ; h. abuki , nucl . phys . * a 791 * , 117 ( 2007);h . abuki , g. baym , t. hatsuda and y. yamamoto , phys . d * 81 * , 125010 ( 2010 ) . j. deng , a. schmitt and q. wang , phys . d * 76 * , 034013 ( 2007 ) . j. wang , v. de la incera , e.j . efrain and q. wang , phys . d * 84 * , 065014 ( 2011 ) ; e.j . efrain , v. de la incera , j. keith and i. portillo , arxiv:1405.7422[hep - ph ] . g. sun , l. he and p. zhuang , phys . d * 75 * , 096004 ( 2007);l . he and p. zhuang , phys . d * 76 * , 056003 ( 2007 ) . g. endrdi , arxiv:1504.08280[hep - lat ] . ferrer , v. de la incera , phys . * 102 * , 050402 ( 2009 ) ; e.j . ferrer , v. de la incera , nucl . * b 824 * , 217 ( 2010);b . feng , e. j. ferrer and v. de la incera , nucl . * b 853* 213 ( 2011 ) ; phys . * b 706 * , 232 ( 2011 ) ; phys . d * 85 * , 103529 ( 2012 ) . a. schmitt , q. wang and d. h. rischke , phys . lett . * 91 * , 242301 ( 2003 ) ; b. feng , d. f. hou , h. c. ren and p. p. wu , phys . lett . * 105 * , 042001 ( 2010 ) .

over the last two decades , the application of ion traps has expanded from mass spectrometry @xcite and frequency standards @xcite towards engineering of quantum systems which can be used for quantum computation @xcite and quantum simulation @xcite . it is commonly accepted that large - scale trapped ion quantum information processors require micrometer - scale ion traps @xcite . such traps usually suffer from excessive electric field noise close to metallic surfaces at room temperature , but at cryogenic temperatures this noise is strongly reduced @xcite . thus , it seems natural to move towards cryogenic experimental setups which have the additional advantage that the ambient pressure is usually a few orders of magnitude lower than at room temperature . this enables trapping of longer ion strings due to fewer collisions with background gas . furthermore , no bake - out procedure of the vacuum vessel is required , allowing for rapid trap cycle times . + the operation of paul traps requires high radio - frequency ( rf ) voltages , which are usually generated with the aid of the voltage gain present in rf resonators . for this purpose , helical resonators are typically used in the frequency regime up to 50mhz , whereas for experiments requiring higher drive frequencies coaxial resonators have been used as well @xcite . in cryogenic experiments , the resonator needs to fulfill different criteria than in room temperature experiments where the resonator can be placed outside the vacuum vessel . in particular , the connections in a cryostat need to have low thermal conductivity to limit the thermal load . following the wiedemann - franz law , this results in a low electrical conductivity between room - temperature and the cryogenic parts of the experiment . thus , the resonators have to be operated at the cold - stage . moreover , space constraints are stricter in cryogenic systems which makes helical resonators undesirable as they are generally bulky . to minimize the volume of the resonator , an rlc - series - resonator can be used @xcite . + in section [ sec : general considerations ] , we focus on the choice of the trap drive frequency , the required voltage gain , and voltage monitoring . section [ sec : coil design ] covers coil design for three types of coils , and in the following section [ sec : matching ] , we discuss impedance matching of the resonator and present an efficient way to match cryogenic resonators . section [ sec : rf shield ] focuses on the design of rf shielding , and finally , we present the results in section [ sec : results ] . in trapped ion experiments , rf - voltages of several hundreds of volts are commonly applied to the trap , which has a simple capacitor as its electrical circuit equivalent . thus , we need to design a resonator that maximizes the rf - voltage at this capacitor for a given input power . in this section , we discuss general considerations in resonator design which are not limited to a specific resonator type . at first , we consider the losses in an rlc - resonator and its scaling with frequency . from the solution of the equation of motion of a trapped ion in a paul trap ( mathieu equation ) , one obtains the trap voltages with the stability parameter @xmath0 @xcite , which scales as @xmath1 where @xmath2 is the amplitude of the rf - voltage at the trap and @xmath3 is the trap drive ( angular ) frequency . hence , for constant @xmath0 the loss power scales as @xmath4 a low trap drive frequency will reduce the losses . on the other hand , high secular motion frequencies of the ion in the trap are desirable for operation . higher secular frequencies require a higher @xmath3 , and thus the desired secular frequencies set a lower limit for @xmath3 . in this study , we aim for a @xmath0 around 0.25 , an axial secular frequency of 1mhz , and both radial frequencies to be 3.7mhz . this will require a trap drive frequency of @xmath5 42.6mhz . simulations of our trap show that we need a drive voltage of about @xmath6 to reach the desired trap frequencies . in order to limit the thermal load onto the cryostat from dissipated power in the resonator , this voltage should be reached with less than 100mw of rf input power . this power corresponds to a consumption of about 1/7l of liquid helium per hour when operating a wet cryostat at 4.2k @xcite . ( 0,1.75 ) to [ short , color = black , o- ] ( 5.2,1.75 ) ; ( 5,1.75 ) to [ r=@xmath7 ( 6.5,1.75 ) ; ( 6.3,1.75 ) to [ short , color = black ] ( 6.9,1.75 ) ; ( 6.9,1.75 ) to [ l=@xmath8 ( 8,1.75 ) ; ( 8,1.75 ) to [ short , color = black , i=@xmath9 ( 9,1.75 ) to [ short , color = black ] ( 9,1.15 ) ; ( 9,1.2 ) to [ c=@xmath10 ( 9,0.8 ) ; ( 9,0.85 ) to [ short , color = black ] ( 9,0.25 ) to [ short , color = black , -o ] ( 0,0.25 ) ; ( 0,0.35 ) to [ open , v^=@xmath11 ( 0,1.75 ) ; ( 9,0.25 ) to [ short , color = black , * -o ] ( 11,0.25 ) ; ( 11,0.35 ) to [ open , v=@xmath12 ( 11,1.75 ) ; ( 11,1.75 ) to [ short , color = black , o- * ] ( 9,1.75 ) ; at ( 2.5,1 ) * matching * ; fig . [ fig : general matching ] shows an rlc - series - resonator driven through a matching network . in the following , we will assume perfect impedance matching with a loss - less matching network . thus , we can set the input power equal to the loss power in the resonator @xmath13 , where @xmath14 with @xmath15 being the input voltage supplied to the circuit and @xmath16 the wave impedance of the connecting cable , commonly @xmath17 . we can further write @xmath18 where @xmath19 is the current in the resonator , @xmath20 its effective loss resistance , @xmath3 its frequency , @xmath21 the voltage at the capacitor , and @xmath22 the capacitor representing the trap . thus , we obtain @xmath23 where @xmath24 is the voltage gain of the circuit , defined as @xmath25 . + the quality factor @xmath26 of a resonator is defined as the resonance frequency @xmath27 divided by the bandwidth @xmath28 @xmath29 in trapped ion experiments , we are usually not interested in a small bandwidth around @xmath3 but rather in a large voltage gain at the trap drive frequency . it should be noted that the voltage gain and bandwidth are qualitatively but not necessary quantitatively the same for different types of resonators . the quantity that is of direct interest for our application is the voltage gain , which can be derived from ( [ eq : matching 1 ] ) as @xmath30 with the quality factor of an ideal rlc resonator @xmath31 . this indicates that both , the capacitive load , and the effective resistance of the resonator should be minimized . ( 0,1.75 ) to [ short , color = black , o- ] ( 3.5,1.75 ) to [ short , color = black , i=@xmath9 ( 4.5,1.75 ) ; ( 4.4,1.75 ) to [ /tikz / circuitikz / bipoles / length=1cm , r=@xmath7 ( 5.4,1.75 ) ; ( 5.3,1.75 ) to [ short , color = black ] ( 5.9,1.75 ) ; ( 5.9,1.75 ) to [ l , l_=@xmath8 ( 7,1.75 ) ; ( 7,1.75 ) to [ short , color = black ] ( 7.5,1.75 ) to [ short , color = black ] ( 7.5,1.1 ) ; ( 7.5,1.75 ) to [ /tikz / circuitikz / bipoles / length=1cm , c , l_=@xmath10 ( 7.5,0.25 ) ; ( 7.5,0.9 ) to [ short , color = black ] ( 7.5,0.25 ) to [ short , color = black , -o ] ( 0,0.25 ) ; ( 0,0.35 ) to [ open , v^=@xmath11 ( 0,1.75 ) ; ( 7.5,1.75 ) to [ /tikz / circuitikz / bipoles / length=1cm , c=@xmath32 , color = black , * - ] ( 9.5,1.75 ) to [ /tikz / circuitikz / bipoles / length=1cm , c=@xmath33 ( 9.5,0.25 ) to [ short , color = black , - * ] ( 7.5,0.25 ) ; ( 7.5,1.75 ) to [ short , color = black , * - ] ( 8.4,1.75 ) ; ( 9.5,0.25 ) to [ short , color = black , * -o ] ( 11.5,0.25 ) to [ open , v=@xmath34 ( 11.5,1.75 ) to [ short , color = black , o- * ] ( 9.5,1.75 ) ; ( 8.6,1.75 ) to [ short , color = black , -o ] ( 11.5,1.75 ) ; ( 11.5,0.25 ) to [ short , color = black , o- * ] ( 7.5,0.25 ) ; ( 9.5,0.25 ) to [ short , color = black , * - ] ( 9.5,0.9 ) ; ( 9.5,1.75 ) to [ short , color = black , * - ] ( 9.5,1.1 ) ; ( 7.4,3 ) to [ short , color = black ] ( 7.4,2.5 ) to [ short , color = black ] ( 7,2.5 ) ; ( 7,2.5 ) to [ l , l_=@xmath35 ( 5.9,2.5 ) ; ( 5.9,2.5 ) to [ short , color = black ] ( 5.4,2.5 ) to [ short , color = black ] ( 5.4,3 ) ; ( 7.4,3 ) to [ short , color = black , -o ] ( 7.4,3.5 ) ; ( 7.4,3.5 ) to [ open , v=@xmath36 ( 5.4,3.5 ) ; ( 5.4,3.5 ) to [ short , color = black , o- ] ( 5.4,3 ) ; ( 7.7 , 0.3 ) to [ open , v=@xmath12 ( 7.7,1.75 ) ; at ( 2.25,1 ) * matching * ; voltage pick - ups are not required for the operation of the trap , but are useful to measure the voltage on the trap and are required to actively stabilize the rf voltage on the trap . in this section , we discuss inductive and capacitive pick - up , as shown in fig . [ fig : general pickup ] . + a capacitive pick - up consists of a voltage divider parallel to the trap , where @xmath37 , which leads to the pick - up voltage @xmath38 @xmath39 has to be small compared to the trap capacitance @xmath22 , or the capacitive load of the rlc - resonator will increase significantly , reducing @xmath24 as shown in ( [ eq : matching 2 ] ) . typical values for @xmath40 are several hundred times the value of @xmath39 . + in an ideal rlc - series - resonator , the voltage on the coil is the same as the voltage at the capacitor but with a phase shift . we can monitor the voltage at the coil as a signal proportional to the voltage at the trap since we are only interested in the amplitude of the trap voltage . here , @xmath41 and @xmath42 are coupled , and the pick - up voltage can be estimated with the derivations from ref . + in order to maintain low losses , the pick - up should not add significant losses in the resonator . if one uses a coaxial cable with a wave impedance @xmath16 to monitor the pick - up signal , a resistor @xmath16 is used as a termination to avoid reflections . the losses are then @xmath43 which need to be small compared to the losses in the resonator @xmath44 . + in general , we recommend using an inductive pick - up , because it does not increase the capacitive load , which would reduce the voltage gain . however for experiments , which frequently test different traps , a capacitive pick - up may be preferable since the more accurate ratio between pick - up and applied voltage facilitates estimating trapping parameters . surface traps for cryogenic environments typically use a high - quality dielectric carrier material and thus the losses in the resonator are usually dominated by the coil . in this study , we demonstrate the design and production of compact and low - loss coils that should allow us to drive our resonator under the conditions stated in section [ sec : gen considerations - choose trap drive ] . coils produced with machines are desirable because the production process is reproducible . furthermore , we prefer toroid coils as they guide the magnetic field in their center , making them less sensitive to their environment . in our application , ferromagnetic core materials are undesirable , because they would induce spatial inhomogeneities of the magnetic field near the trap . we estimate the load capacitance of our resonator to be 10pf , which at a given resonance frequency of 42.6mhz yields a required inductance of 1.4@xmath45h . kamby and coworkers @xcite demonstrated the integration of toroidal rf - inductors into printed circuit boards ( pcb ) . these coils can be fully produced by machines and in this study we will refer to these inductors as pcb coils . @xcite provides specific formulas for the inductance @xmath41 , the resistance of one segment @xmath46 , and the resistance of one via @xmath47 of pcb coils . in order to minimize the losses of the coil we want to minimize the resistance of one winding while maintaining a constant cross - section to keep the inductance constant the resistance of one winding is defined by @xmath48 where @xmath49 and @xmath50 are the outer and the inner radius of the pcb coil , @xmath51 is the thickness of the pcb , and @xmath52 and @xmath53 are the average length - dependent resistances of a segment and a via . @xmath52 and @xmath53 depend on the exact geometry , but usually vias have similar cross - sections as the segments making their length - dependent resistances approximately equal . that means , in order to minimize ( [ eq : pcb coils minimize resistance ] ) and to keep the cross section of the coil constant , @xmath54 and @xmath51 should be similar in size . + we chose rogers 4350b as the substrate material , because it has a small loss tangent , a similar thermal expansion coefficient as copper , and is ultra - high vacuum compatible . the thickest available rogers 4350b had 3.1 mm thickness , which is unfortunately too thin to minimize the resistance for a given cross - section . + we produced this type of coil in - house and during multiple iterations , we varied the size and the number of vias . the lowest losses for our production process could be achieved with the design depicted in fig . [ fig : pcb coil ] . the pcb coil has an outer diameter of 35 mm and an inner diameter 20.4 mm with a substrate thickness of 3.1 mm . its 64 windings result in an inductance of 1.4@xmath45h . we added one winding which is galvanically isolated from the rest of the coil . this winding acts as an inductive pick - up and should thus be as reproducible as the entire coil . for the second type of coils , wire coils , we wound a single wire around a rigid structure leading to a similar geometry as the pcb coils . such a coil is less reproducible than a pcb coil since the wire length will vary for each coil . hence , its inductance and its resistance will vary as well . a coil with the same geometry as for the pcb coils above , made out of silver - plated copper on a rogers 4350b substrate , is shown in fig . [ fig : wire coil ] . + in wire coils , the cross - section of the core of a wire coil is bigger than that of a pcb coil , because the wire bends are less sharp . hence , we expect the inductance of a wire coil to be higher than the one of a pcb coil with the same nominal geometry . the 0.4 mm thick silver - plated copper wire , which we used for the wire coil , had a cross - section similar to the average cross - section of a pcb coil . therefore , we expect similar losses in both types of coils with the same geometries . these considerations allow us to characterize the production processes of the pcb coils by comparing the two types of coils . higher losses in the pcb coils are an indication that the production of the pcb coils can be improved . + since we do not need traces on the substrate , it is possible use any machinable material with a low loss tangent . we also produced wire coils with a teflon core since it has a lower loss tangent and a lower dielectric constant than rogers 4350b . thus , the losses will be less affected by the core material and the self - resonance frequency of the coil will be higher . another option to minimize the losses in the coil can be to use a superconducting wire to reduce the ohmic losses of the resonator . the quality factor of the presented coils are limited by the ohmic resistance of the conductor . thus , a coil based on a superconductor is expected to result in lower losses in the resonator . furthermore , it could be necessary to operate the experiment at temperatures above 15k to reduce the liquid helium consumption in wet cryostats . at these temperatures , a superconductor with a high critical temperature such as a high temperature superconductor ( hts ) is required . however , a coil made from a readily available hts material needs to be a two - dimensional ( 2d ) structure due to manufacturing constraints , and we choose to use a spiral coil as shown in fig . [ fig : hts coil ] . + since such coils require only 2d structuring , they can be manufactured in a very accurate and reproducible way . for the demonstrated spiral coil , a high - temperature superconductor ( hts ) was produced by _ _ ceraco ceramic coating__. our coils were made out of an yttrium barium copper oxide ( ybco ) film of 330 nm thickness on a sapphire wafer . to facilitate soldered connections , a 200 nm film of gold was placed on top of the superconductor . the critical temperature of this ybco film is above 87k , and its critical current density is higher than @xmath55a/@xmath56 at 77k . we can derive the peak current in the coil from section [ sec : general considerations ] to be 0.64a . hence , traces with a width of more than 100@xmath45 m are required to stay below the critical current density . + for our square spiral coils , we chose a trace width of 300@xmath45 m , a trace gap of 150@xmath45 m , and 10 windings . we designed 9 different coils varying the outer diameter between 14 and 20 mm . equation 4.1 of reference @xcite gives an estimate for the inductance , which results in target inductances between 1.2@xmath45h and 1.6@xmath45h for our coils . this design should allow for coils with very high quality factors already at temperatures accessible with liquid nitrogen instead of liquid helium . + it should be noted here that superconductors are perfect diamagnets , which are spatial inhomogeneities for magnetic fields . one can align the 2d plane of the spiral coil with the spatially homogeneous magnetic field of the experiment to minimize this effect , but one should always keep this in mind . the resonator is connected to the source through a cable with a given wave resistance @xmath16 , typically 50@xmath3 . the complex voltage reflection coefficient @xmath57 at the transition from the cable to the resonator @xcite is then @xmath58 where @xmath59 is the complex reflected voltage , @xmath60 is the complex incoming voltage , and @xmath61 is the complex impedance of the resonator . the reflected power coefficient @xmath62 is then just @xmath63 . since rlc - series resonators have a very low impedance on resonance , @xmath64 , the transition from a 50@xmath3-cable to the resonator on resonance will reflect almost all the inserted power . in order to minimize the reflection , a properly designed impedance matching network is required . the most commonly used matching network is the @xmath41-network as it is the simplest , consisting only of two circuit elements . + ( 0,5.75 ) to [ short , color = black , o- ] ( 1.15,5.75 ) ; ( 1.05,5.75 ) to [ /tikz / circuitikz / bipoles / length=1cm , c=@xmath65 ( 1.45,5.75 ) ; ( 1.35,5.75 ) to [ short , color = black ] ( 3.1,5.75 ) ; ( 3,5.75 ) to [ /tikz / circuitikz / bipoles / length=1cm , r=@xmath7 ( 4,5.75 ) ; ( 3.9,5.75 ) to [ short , color = black ] ( 5.05,5.75 ) ; ( 4.9,5.75 ) to [ /tikz / circuitikz / bipoles / length=1 cm , l , l=@xmath8 ( 6,5.75 ) ; ( 5.8,5.75 ) to [ short , color = black ] ( 6.5,5.75 ) to [ short , color = black ] ( 6.5,4.85 ) ; ( 6.5,4.95 ) to [ /tikz / circuitikz / bipoles / length=1cm , c , l_=@xmath10 ( 6.5,4.55 ) ; ( 6.5,4.65 ) to [ short , color = black ] ( 6.5,3.75 ) to [ short , color = black , -o ] ( 0,3.75 ) ; ( 0,3.85 ) to [ open , v^=@xmath11 ( 0,5.75 ) ; ( 2.25,5.75 ) to [ short , color = black , * - ] ( 2.25,5.05 ) ; ( 2.25,5.25 ) to [ /tikz / circuitikz / bipoles / length=1 cm , l=@xmath66 , color = black ] ( 2.25,4.15 ) ; ( 2.25,4.35 ) to [ short , color = black , - * ] ( 2.25,3.75 ) ; ( 6.7 , 3.8 ) to [ open , v=@xmath12 ( 6.7,5.75 ) ; ( -1.5,6 ) node [ ] ; ( 0,2.25 ) to [ short , color = black , o- ] ( 0.85,2.25 ) ; ( 0.7,2.25 ) to [ /tikz / circuitikz / bipoles / length=1cm , l=@xmath66 ] ( 1.8,2.25 ) ; ( 1.6,2.25 ) to [ short , color = black ] ( 3.1,2.25 ) ; ( 3,2.25 ) to [ /tikz / circuitikz / bipoles / length=1cm , r=@xmath7 ( 4,2.25 ) ; ( 3.9,2.25 ) to [ short , color = black ] ( 5.05,2.25 ) ; ( 4.9,2.25 ) to [ /tikz / circuitikz / bipoles / length=1cm , l , l=@xmath8 ( 6,2.25 ) ; ( 5.8,2.25 ) to [ short , color = black ] ( 6.5,2.25 ) to [ short , color = black ] ( 6.5,1.35 ) ; ( 6.5,1.45 ) to [ /tikz / circuitikz / bipoles / length=1cm , c , l_=@xmath10 ( 6.5,1.05 ) ; ( 6.5,1.15 ) to [ short , color = black ] ( 6.5,0.25 ) to [ short , color = black , -o ] ( 0,0.25 ) ; ( 0,0.35 ) to [ open , v^=@xmath11 ( 0,2.25 ) ; ( 2.25,2.25 ) to [ short , color = black , * - ] ( 2.25,1.35 ) ; ( 2.25,1.45 ) to [ /tikz / circuitikz / bipoles / length=1cm , c=@xmath68 , color = black ] ( 2.25,1.05 ) ; ( 2.25,1.15 ) to [ short , color = black , - * ] ( 2.25,0.25 ) ; ( 6.7 , 0.3 ) to [ open , v=@xmath12 ( 6.7,2.25 ) ; ( -1.5,2.5 ) node [ ] ; fig . [ fig : matching networks ] shows the two possible @xmath41-networks where the reactance parallel to the resonator transforms the real part of the impedance to be @xmath16 . the resulting imaginary term is small compared to @xmath16 ( for low loss resonators ) and can be compensated with the series reactance . + and matching coil @xmath69 for the matching networks of fig . [ fig : matching networks ] over a varying quality factor of the coil . the matching is calculated for a 50@xmath3-cable , an inductor of 1.4@xmath45h , and a capacitor of 10pf . ] the resonator s quality factor from ( [ eq : quality factor - bandwidth ] ) can be written as @xmath70 for an rlc - series resonator . [ fig : matching network simulation ] shows the required values for the matching capacitor @xmath71 and inductance @xmath69 as a function of the quality factor of the coil with inductance of 1.4@xmath45h . for quality factors around 1000 , the required inductance @xmath69 is in the range of 10nh . thin traces on a pcb have approximately a length - dependent inductance of 1nh / mm . therefore , small deviations in the trace length will lead to inefficient matching . the strong dependence of the matching on the reactance parallel to the resonator makes the matching network from fig . [ fig : matching networks]b ) more favorable , because capacitances of several hundred pf can be adjusted with much more precision than inductances of about 10nh . one can incorporate the inductance @xmath69 into the trace on the pcb from the coaxial connector to the matching capacitance , by setting this distance between 1 and 2 cm . hence , we expect a power reflection of 1% or less even without adding an additional inductor . + ( -0.5,2.25 ) to [ short , color = black , o- ] ( 3.6,2.25 ) ; ( 3.5,2.25 ) to [ /tikz / circuitikz / bipoles / length=1cm , r=@xmath7 ( 4.5,2.25 ) ; ( 4.4,2.25 ) to [ short , color = black ] ( 5.45,2.25 ) ; ( 5.4,2.25 ) to [ l , l=@xmath8 ( 6.6,2.25 ) ; ( 6.5,2.25 ) to [ short , color = black ] ( 7,2.25 ) to [ short , color = black ] ( 7,1.35 ) ; ( 7,1.45 ) to [ /tikz / circuitikz / bipoles / length=1cm , c , l_=@xmath10 ( 7,1.05 ) ; ( 7,1.15 ) to [ short , color = black ] ( 7,0.25 ) to [ short , color = black , -o ] ( -0.5,0.25 ) ; ( -0.5,0.25 ) to [ open , v^=@xmath11 ( -0.5,2.25 ) ; ( 1.75,2.25 ) to [ short , color = black , * - ] ( 1.75 , 1.35 ) ; ( 1.75,1.45 ) to [ /tikz / circuitikz / bipoles / length=1cm , c , l_=@xmath68 , color = black ] ( 1.75,1.05 ) ; ( 1.75,1.15 ) to [ short , color = black , - * ] ( 1.75 , 0.25 ) ; ( 2.75,2.25 ) to [ short , color = black , * - ] ( 2.75 , 1.65 ) ; ( 2.75,1.85 ) to [ /tikz / circuitikz / bipoles / length=1cm , l=@xmath72 , color = black ] ( 2.75,0.65 ) ; ( 2.75,0.85 ) to [ short , color = black , - * ] ( 2.75 , 0.25 ) ; ( 7.2 , 0.3 ) to [ open , v=@xmath12 ( 7.2,2.25 ) ; rf transformers or baluns on the rf input side are commonly used to avoid ground loops between the rf source and the experiment . when using the circuit from fig . [ fig : matching networks]b ) in combination with a transformer , the dc potential of the trap electrodes is not defined . we mitigate this by adding an rf - choke parallel to the matching capacitor , where @xmath73 , as depicted in fig . [ fig : matching network choice ] . + -cable , an inductor of 1.4@xmath45h , and a capacitor of 10pf , thus a resonance frequency of 42.6mhz . the red lines show the values for a fixed capacitance @xmath71 chosen for an ideal matching of @xmath74 . in a ) , one can see the required matching capacitor as a function of the quality factor of the coil @xmath75 . b ) shows the change in resonance frequency of an ideal match . in c ) , the voltage gain is plotted over @xmath75 . d ) shows the reflection coefficient of the matching network . ] in a cryostat , it is impractical to adjust the matching capacitance @xmath71 during or after a cool down . hence , we have to find an efficient way of choosing the best value for @xmath71 with a minimum number of temperature cycles . for this , we simulated our matching circuit for varying and also for a constant value of @xmath71 . [ fig : matching reflection ] shows results of simulations where the fixed capacitor value was chosen for a match at @xmath74 . fig . [ fig : matching reflection]a ) depicts an increase of the optimum @xmath71 with increasing @xmath75 . [ fig : matching reflection]b ) illustrates the dependence of the matched resonance frequency on @xmath75 . in fig . [ fig : matching reflection]c ) , the simulations confirm the dependence of the voltage gain on @xmath75 , following from ( [ eq : matching 2 ] ) . if the physical value for @xmath75 is smaller than the expected value , used to chose the fixed matching capacitor , the voltage gain is about the same than for an ideal match . but if the actual quality factor @xmath75 is higher than the expected value , the voltage gain will be lower than with an ideal match . hence underestimating the quality factor of the coil should be avoided since it would result in a significantly lower voltage gain . [ fig : matching reflection]d ) shows the reflection coefficient as a function of @xmath75 . the effect of @xmath69 is small and thus it was omitted for these simulations . we can see that even without @xmath69 , but with an ideally matched @xmath71 , a low reflection coefficient can be achieved . if one chooses a constant value for @xmath71 at a match for @xmath74 , the reflected power will stay below 5% even if @xmath75 varies between 800 and 2000 . + the parameters for exact impedance matching can be estimated with one additional cooling cycle . for the first cycle , one has to guess the value for @xmath75 and adjust @xmath71 according to simulations following the approach of fig . [ fig : matching reflection]a ) . when the resonator is cold , the reflection coefficient is measured and compared with the simulations in fig . [ fig : matching reflection]d ) . the measured reflection coefficient can then only correspond to two values for @xmath75 . if the reflection coefficient passed a minimum during cool - down , one should choose the higher value , if not , the lower value of @xmath75 . with this new value for @xmath75 , one can calculate the correct matching capacitor value @xmath71 . our experience shows that this method will yield a power reflection coefficient of less than 2% for the second cool down . rf resonators emit electromagnetic radiation , which needs to be shielded to minimize unwanted rf interference in other parts of the experiment . the rf shield needs to be grounded , and thereby generates a capacitance between the resonator and ground . this capacitance has to be minimized as suggested in ( [ eq : matching 2 ] ) . in order to keep the resonator compact , we have to find a trade - off between the size of the shield and the capacitance added by it . + ( 1,2.5 ) to [ short , color = black , o- ] ( 2.1,2.5 ) ; ( 2,2.5 ) to [ /tikz / circuitikz / bipoles / length=1cm , r=@xmath7 ( 3,2.5 ) ; ( 2.9,2.5 ) to [ short , color = black ] ( 3.55,2.5 ) ; ( 3.4,2.5 ) to [ /tikz / circuitikz / bipoles / length=1 cm , l=@xmath8 ( 4.5,2.5 ) ; ( 4.3,2.5 ) to [ short , color = black ] ( 5.5,2.5 ) to [ short , color = black ] ( 5.5,2.25 ) to [ short , color = black ] ( 6.2,2.25 ) to [ short , color = black ] ( 6.2,1.735 ) ; ( 6.2,1.825 ) to [ /tikz / circuitikz / bipoles / length=1cm , c=@xmath10 ( 6.2,1.425 ) ; ( 6.2,1.525 ) to [ short , color = black ] ( 6.2,1 ) to [ short , color = black ] ( 5.5,1 ) to [ short , color = black ] ( 5.5,0.75 ) to [ short , color = black , -o ] ( 1,0.75 ) ; ( 5.5,2.25 ) to [ short , color = black , * - ] ( 4.8,2.25 ) to [ short , color = black ] ( 4.8,1.725 ) ; ( 4.8,1.825 ) to [ /tikz / circuitikz / bipoles / length=1cm , c , l_=@xmath76 ( 4.8,1.425 ) ; ( 4.8,1.525 ) to [ short , color = black ] ( 4.8,1 ) to [ short , color = black , - * ] ( 5.5,1 ) ; ( 5.1,3 ) to [ short , color = black ] ( 5.6,3.5 ) to [ short , color = black ] ( 5.362,3.5 ) to [ short , color = black ] ( 5.4,3.5 ) to [ short , color = black ] ( 5.4,4.538 ) to [ short , color = black ] ( 5.4,4.5 ) to [ short , color = black ] ( 4.762,4.5 ) to [ short , color = black ] ( 4.8,4.5 ) to [ short , color = black ] ( 4.8,3.462 ) to [ short , color = black ] ( 4.8,3.5 ) to [ short , color = black ] ( 4.6,3.5 ) to [ short , color = black ] ( 5.127,2.973 ) ; ( -1.5,7 ) to [ short , color = black , o- ] ( -0.9,7 ) ; ( -1,7 ) to [ /tikz / circuitikz / bipoles / length=1 cm , r=@xmath77 ( 0,7 ) ; ( -0.1,7 ) to [ short , color = black ] ( 0.35,7 ) ; ( 0.25,7 ) to [ /tikz / circuitikz / bipoles / length=1 cm , l=@xmath78 ( 1.25,7 ) ; ( 1.15,7 ) to [ short , color = black ] ( 1.75,7 ) ; ( 1.5,7 ) to [ short , color = black , * - ] ( 1.5,6.35 ) ; ( 1.5,6.45 ) to [ /tikz / circuitikz / bipoles / length=1 cm , c , l_=@xmath79 ( 1.5,6.05 ) ; ( 1.5,6.15 ) to [ short , color = black , - * ] ( 1.5,5.5 ) ; ( 1.75,5.5 ) to [ short , color = black , -o ] ( -1.5,5.5 ) ; ( 2.24,6.25 ) to [ short , color = black , - * ] ( 2.25,6.25 ) ; ( 2.51,6.25 ) to [ short , color = black , - * ] ( 2.5,6.25 ) ; ( 2.74,6.25 ) to [ short , color = black , - * ] ( 2.75,6.25 ) ; ( 3.25,7 ) to [ short , color = black ] ( 3.6,7 ) ; ( 3.5,7 ) to [ /tikz / circuitikz / bipoles / length=1 cm , r=@xmath77 ( 4.5,7 ) ; ( 4.4,7 ) to [ short , color = black ] ( 4.85,7 ) ; ( 4.75,7 ) to [ /tikz / circuitikz / bipoles / length=1 cm , l=@xmath78 ( 5.75,7 ) ; ( 5.65,7 ) to [ short , color = black ] ( 8,7 ) to [ short , color = black ] ( 8,6.35 ) ; ( 6,7 ) to [ short , color = black , * - ] ( 6,6.35 ) ; ( 6,6.45 ) to [ /tikz / circuitikz / bipoles / length=1 cm , c , l_=@xmath79 ( 6,6.05 ) ; ( 6,6.15 ) to [ short , color = black , - * ] ( 6,5.5 ) ; ( 8,6.45 ) to [ /tikz / circuitikz / bipoles / length=1 cm , c=@xmath10 ( 8,6.05 ) ; ( 8,6.15 ) to [ short , color = black ] ( 8,5.5 ) to [ short , color = black ] ( 3.25,5.5 ) ; ( -1.5,7.75 ) to[out=-90,in=-180 ] ( -1.4,8 ) to[out=0,in=-90 ] ( 0.125,8.25 ) to[out=-90,in=-180 ] ( 1.65,8 ) to[out=-0,in=-90 ] ( 1.75,7.75 ) ; ( 0.125,8.75 ) circle(0pt)node[anchor = north]half a winding ; ( -1.5,5.25 ) to[out=-90,in=180 ] ( -1.25,5 ) to[out=0,in=180 ] ( 1.5,5.25 ) to[out=0,in=+90 ] ( 2.375,4.7 ) to[out=+90,in=+180 ] ( 3.25,5.25 ) to[out=0,in=180 ] ( 6,5 ) to[out=0,in=-90 ] ( 6.25,5.25 ) ; ( 2.375,4.05 ) circle(0pt)node[anchor = south]x 2n ; first , we need to find a model for the shield capacitance @xmath80 . [ fig : shield cap model ] depicts a model where we divide the coil in 2@xmath81 half windings . each has an ohmic resistance of @xmath82 , an inductance @xmath83 , and a capacitance to the shield @xmath84 . we can estimate @xmath84 with the standard formula for a parallel - plate capacitor if we assume that the grounded shield is parallel to the coil . from the calculations of the impedance of the entire circuit , we can extract the effective capacitance @xmath80 that occurs due to the presence of the shield . + the results of these simulations for our pcb coils with an outer radius of 17.5 mm and an inner radius of 10.2 mm are shown in fig . [ fig : shield cap simulation ] . it can be seen that a distance of about 2 mm is sufficient to keep the additional capacitive load below 1pf . adapted simulations for the wire coils yield results similar to the ones for the pcb coils . hence , we can use the same shield for both types of coils . + we expect that the capacitance caused by the shield for the spiral coils is even smaller , because its surface is smaller . additionally , if we place the spiral coil in the center of the same shield , the distance to the shield will be 1.55 mm bigger than for the other coils . hence , the capacitance to ground will decrease further . moreover , the shield has an influence on the magnetic field lines of the spiral coil in this configuration and thereby also on the inductance and the losses of the coil . + at room temperature , the penetration depth due to skin effect in copper at frequencies around 40mhz is about 10@xmath45 m . thus , any mechanically stable shield , of a couple of 1/10 mm thickness , should yield a suitable attenuation . [ fig : resonator with shield ] shows a resonator encased in a silver - plated copper shield . the dimensions of this resonator are 57 mm x 40 mm x 10.2 mm . on the left side of the resonator , one can see a pcb with the matching network at the top and the smt - connector for the inductive pick - up at the bottom . on the right side is the pcb coil , and the two pcbs are connected together by soldered joints . in our measurement setup , we used a capacitive pick - up to measure the voltage at the trap . with the known ratio of this capacitive voltage divider , we could directly measure the voltage at the trap and calibrate the inductive pick - up of the coil . + lcccc * name * & * t * & @xmath85 & @xmath86 & @xmath87 + pcb coil & 295k & 42.4mhz & 19.4 & 60 + pcb coil & @xmath8880k & 42.8mhz & 27.1 & 115 + pcb coil & @xmath8810k & 42.9mhz & 41.5 & 208 + wire coil & 295k & 36.8mhz & 25.2 & 89 + wire coil & @xmath8880k & 37.2mhz & 36.5 & 184 + wire coil & @xmath8810k & 37.9mhz & 58.9 & 624 + hts coil & @xmath8988k & 44.3mhz & 92.1 & 1172 + table [ tab : measurement results ] shows our measurement results . the pcb coils had an inductance very close to the desired value . unfortunately , we could not achieve the expected quality factor . we had designed the wire coils to have losses similar to the pcb coils , but the pcb coils showed much higher losses than the wire coils . we think this is due to an imperfect production process that we could not improve further with our in - house techniques . + additionally to the pcb coils in copper , we produced them in silver as well . the inductances were similar , but the losses in the silver coils were even higher . this was unexpected as silver is a better conductor than copper . but the silver coils needed to be reworked after fabrication to remove shorts between the segments . during this process , the silver coating might have been scratched , resulting in higher losses . we expect that silver plating after structuring should reduce the losses compared to a copper coil , but we were unable to do so with our in - house fabrication process . + the voltage gain of about 60 of the wire coil would require 160mw rf input power to get to the desired 170@xmath90 at the trap . if we adjust the voltage to keep the stability parameter @xmath91 , only 100mw rf input power are required at the modified resonance frequency . + we expect that one can improve the quality factor of the wire coil by using a thicker wire . we also tested a resonator with a lead - plated wire , which did not show an increased voltage gain . we either could not reach the critical temperature of lead in our test setup , or the lead was not pure enough to become superconducting . + the spiral coil showed the highest voltage gain , although the shield influences the magnetic field generated by the coil . we observed a change of the resonance frequency of a couple of percent when cooling down from liquid nitrogen temperatures to liquid helium temperatures , which we attribute to the shield surrounding the coil . with a voltage gain of 92 , one would need about 70mw of rf input power to get 170@xmath90 . with a stability parameter @xmath92 at the higher resonance frequency , it would be 82mw . we have demonstrated three designs of rf resonators for trapped ion experiments with paul traps . the pcb coils have favorable mechanical properties but still need an improved production process , such as e.g. thermal annealing . the wire coils fulfill our criteria and can be further improved by using a thicker or superconducting wire . the hts spiral coil showed the highest voltage gain , which was already accessible at temperatures reached with liquid nitrogen cooling . + efficient matching is usually difficult in a cryogenic environment because normally one can not tune the matching parameters when the cryostat is cold . we have shown a recipe to efficiently match rlc - resonators without precise knowledge of their quality factors . during this empirical study we regularly achieved matching with a power reflection of less than 2% on the second cool - down . additionally , we designed and implemented a shield for resonators to minimize rf pick - up in other parts of the experiment . we thank gerhard hendl from the institute of quantum optics and quantum information of the austrian academy of science for the production of pcbs , and pcb coils for this study . furthermore , we want to thank florian ong for feedback on the manuscript . this research was funded by the office of the director of national intelligence ( odni ) , intelligence advanced research projects activity ( iarpa ) , through the army research office grant w911nf-10 - 1 - 0284 . all statements of fact , opinion or conclusions contained herein are those of the authors and should not be construed as representing the official views or policies of iarpa , the odni , or the u.s . government . w. paul , rev . * 62 * ( 1990 ) 531 - 540 m. chwalla , j. benhelm , k. kim , g. kirchmair , t. monz , m. riebe , p. schindler , a.s . villar , w. haensel , c. f. roos , r. blatt , m. abgrall , g. santarelli , g. d. rovera , ph . laurent , phys . * 102 * ( 2009 ) 023002 h. s. margolis , g. huang , g. p. barwood , s. n. lea , h. a. klein , w. r. c. rowley , p. gill , r. s. windeler , phys . rev . a * 67 * ( 2003 ) 032501 j.i . cirac , p. zoller , phys . * 74 * ( 1995 ) 4091 - 4094 d. kielpinski , c. monroe , d.j . wineland , nature * 417 * ( 2002 ) 709 - 711 t. monz , p. schindler , j.t . barreiro , m. chwalla , d. nigg , w.a . coish , m. harlander , w. hnsel , m. hennrich , r. blatt phys . * 106 * ( 2011 ) 130506 d. porras , j.i . cirac , phys . * 92 * ( 2004 ) 207901 p. jurcevic , b.p . lanyon , p. hauke , c. hempel , p. zoller , r. blatt , c.f . roos , nature * 511 * ( 2014 ) 202 - 205 q.a . turchette , b.e . king , d. leibfried , d.m . meekhof , c.j . myatt , m.a . rowe , c.a . sackett , c.s . wood , w. itano , c. monroe , d. j. wineland , physical review a * 61 * ( 2000 ) 063418 j. labaziewicz , y. ge , p. antohi , d. leibrandt , k. r. brown , i. l. chuang , physical review letters * 100 * ( 2008 ) 013001 w.w . macalpine , r.o . schildknecht , proc . ire * 47 * ( 1959 ) 2099 - 2105 j.d . siverns , l.r . simkins , s.weidt , w.k . hensinger , appl . b * 107 * ( 2012 ) 921 - 934 s.r . jefferts , c. monroe , e.w . bell , d.j . wineland , phys . rev . a * 51 * ( 1995 ) 3112 - 3116 m. e. poitzsch , j. c. bergquist , w. m. itano , d. j. wineland , rev . sci . instrum . * 67 * ( 1996 ) 129 - 134 d. gandolfi , m. niedermayr , m. kumph , m. brownnutt , r. blatt , rev . * 83 * ( 2012 ) 084705 d. j. wineland , c. monroe , w. m. itano , d. leibfried , b. e. king , d. m. meekhof , journal of research of the national institute of standards and technology * 103 * ( 1998 ) 259 - 328 j. w. ekin , _ experimental techniques for low - temperature measurements _ ( oxford university press , 2006 ) 504 - 505 p. kamby , a. knott , m.a.e . andersen , iecon 2012 - 38th annual conference on ieee industrial electronics society ( 2012 ) 680 - 684 s. s. mohan , _ the design , modeling and optimization of on - chip inductor and transformer circuits _ thesis , department of electrical engineering , stanford university , 1999 ) d.m . pozar , _ microwave engineering _ ( john wiley & sons , 2011 )

of the various types of supernovae ( sne ) we observe , only certain subtypes of type ii sne have been conclusively connected to their progenitors . pre - explosion imaging has shown that sne ii - plateau ( sne iip ) , for example , come from red supergiants ( see review by @xcite ) . owing to their cosmological significance , the progenitor systems of sne ia have been pursued relentlessly over the last two decades , but their nature is still debated ( see for a recent review ) . sne ii and stripped - envelope sne ( se sne ; i.e. , sne iib , ib , ic , and broad - lined ic ; e.g. , ) are attributed to the core collapse ( cc ) of stars more massive than @xmath0 . spectroscopically , se sne are distinguished from sne ii by the lack of hydrogen features ( either partial , as in sne iib , or nearly complete , as in sne ib ; e.g. , ) . sne ic also lack helium features . in order to explain this lack of hydrogen and helium , se sne are thought to be the explosions of stars that have had their outer envelopes stripped away before the explosion ( hence their name ) . of all the processes suggested to explain this stripping , the leading models make use of either stellar winds ( e.g. , @xcite ) , interaction with a binary companion ( e.g. , ) , or a combination of both ( e.g. , @xcite ) . in the case of broad - lined sne ic connected to gamma - ray bursts , chemically homogeneous evolution ( e.g. , ) and explosive common - envelope ejection @xcite have also been suggested . pre - explosion imaging of the sites of these sne has so far failed to reveal the nature of their progenitors conclusively ( e.g. , @xcite ) , though the case for yellow super - giants in binary systems as the progenitors of sne iib is gaining traction ( e.g. , ) . new observational methods are required to address this question . sne ia are thought to be thermonuclear explosions of carbon - oxygen white - dwarf remnants of @xmath1 @xmath2 stars . in order to disrupt the otherwise stable white dwarf , most models place it in a binary system where it can grow in mass , raising the temperature in the core until the carbon is ignited in a thermonuclear runaway . to grow in mass , the white dwarf can either siphon matter off of a main - sequence or evolved companion star ( the so - called `` single - degenerate '' scenario ; @xcite ) or merge with a second carbon - oxygen white dwarf after the two spiral in owing to loss of energy and angular momentum to gravitational waves ( the `` double degenerate '' scenario ; @xcite ) . recently , direct collisions of white dwarfs have also been suggested as a possible progenitor channel ( e.g. , @xcite ; but see also @xcite ) . many methods have been used to constrain these various sn progenitor models , including ( but far from limited to ) direct imaging of the explosion sites either before or long after the sn explosion ( e.g. , @xcite ) , multiwavelength follow - up observations ( e.g. , @xcite ) , and analyses of sn remnants ( e.g. , @xcite ) . in this work , we concentrate on correlations between sn rates and various galaxy properties and what those correlations can tell us about sn progenitors . @xcite showed that sn ia rates per unit mass decreased with increasing galaxy stellar mass . * hereafter l11 ) showed the same effect for all sn types , in all types of galaxies ( but see section [ sec : snia ] ) , and dubbed this the `` rate - size , '' or rate - mass , relation . we confirmed this trend for sne ia in star - forming galaxies in @xcite and for sne ii in ( * ? ? ? * hereafter g15 ) . following @xcite , we showed that the dependence of the sn ia rates on stellar mass results from a combination of galaxy scaling relations ( older galaxies are more massive , on average , than younger ones ; @xcite ) and the sn ia delay - time distribution ( dtd ) , which behaves as a power law with an index of @xmath3 ( e.g. , observations by @xcite ; and reviews by ) . , @xcite , @xcite , and g15 also measured sn ia rates as a function of the galaxies specific star - formation rate ( ssfr ) . in g15 , we showed that our explanation for the rate - mass correlation also explained the observed trend between sn ia rates and ssfr , where the rates are constant in passive galaxies but rise with increasing ssfr in star - forming galaxies . in g15 , we also measured sn rates as a function of stellar mass and ssfr for sne ii and claimed that their rate - mass relation was simply the result of their progenitors short lifetimes : because sne ii come from massive ( @xmath4 ) stars , their rates track the star - formation rates of their galaxies . similar measurements of cc sn rates ( i.e. , combining sne ii and se sne ) were made by @xcite . l11 was part of a series of papers that explored the sn sample collected by the lick observatory supernova search ( loss ; l11 ; @xcite ) . loss is an ongoing survey for sne in local galaxies using the 0.76-m katzman automatic imaging telescope . for detailed descriptions of the survey , see @xcite , @xcite , and @xcite . here , we use the loss sample to remeasure and re - analyze the sn rates originally published by l11 . in section [ sec : galaxies ] , we match between the loss sample and spectroscopy from the sloan digital sky survey ( sdss ; @xcite ) in order to go beyond l11 and measure the sn rates not only as a function of galaxy stellar mass but also of ssfr and metallicity , as expressed by the abundance of gas - phase oxygen in the centers of the loss galaxies . in section [ sec : addendum ] , we make several addendums to the loss sample . we publish the control times necessary to measure sn rates with this sample , as well as updated galaxy - property tables and sn rates . we deal with sne ia in section [ sec : snia ] , and with cc sne in section [ sec : ccsn ] . we find significant correlations between the sn rates and the various galaxy properties . most importantly , we find that the cc sn rates behave differently in different types of galaxies : the se sn rates are shown to be deficient , relative to the sn ii rates , in galaxies with low stellar mass , high ssfrs , and low metallicity values . other studies have reported similar trends through measurements of correlations between fractions of sne within a given sample and metallicity ( e.g. , ) , or by splitting sn fractions between different types of galaxies , which encompass different metallicity regimes @xcite . we conduct an in - depth comparison of our results with these works in section [ subsec : ccsn_context ] . in section [ sec : discuss ] , we discuss how the various rate correlations are not independent . for all sn types , a correlation with one galaxy property can be transformed into the measured correlations with any other galaxy property by using well - known galaxy scaling relations . this makes it impossible to distinguish causation from correlation , especially for the deficiency of se sne in lower - mass galaxies . however , we argue that the structure seen in the correlations ( e.g. , the way the ratio between the se sn and sn ii rates depends on galaxy stellar mass ) can be used to constrain progenitor models . paper ii in this series will use population fractions , as measured from the loss volume - limited subsample of sne , to strengthen the results presented here and add further constraints on sn progenitor scenarios . for paper ii , we rely on a reclassification of the sne in this subsample , as reported by shivvers et al . ( in prep . ) . between march 1998 and december 2008 , loss discovered a total of 1,036 sne . most of these were discovered among the 14,882 galaxies directly targeted by the search ( at a median distance of @xmath5 mpc , where the upper and lower bounds contain 68% of the galaxies in the sample ) , but a few dozen were also discovered in background galaxies in the loss fields . this sample , along with the subsamples used to measure the loss sn luminosity functions @xcite and rates ( l11 ) , are described in detail by @xcite . loss classified their sne into three broad categories ( e.g. , ): sne ia , ib / c , and ii . the first category included all sn ia subtypes , including the subluminous sn 1991bg - like sne , overluminous sn 1991t - like sne , and sn 2002cx - like sne ia ( now referred to as sne iax ; @xcite ) . sne ib / c included sne ib , ic , and `` peculiar '' se sne ( such as broad - lined sne ic ; see , e.g. , for a review ) . the final category comprised sne iip , iil , iin , and iib . however , as mentioned earlier , sne iib are characterized by hydrogen deficiency , which is an indication of envelope stripping . thus , they should be grouped with the sne ib / c ( e.g. , @xcite ) . in this work , we use `` se sne '' instead of `` sne ib / c . '' however , because we use the control times calculated by l11 for the loss sample ( see sections [ sec : addendum ] and [ subsec : rates_mass ] , below ) , we must keep the sne iib grouped with the sn ii class when calculating rates . fourteen sne had no spectroscopic classification . @xcite divided these sne into the three sn categories `` according to the statistics of the sne having spectroscopic classification '' and used them in the l11 rate calculations . we choose to exclude these sne from the rate measurements performed in this work . @xcite divided the loss galaxies into eight hubble types e , s0 , sa , sb , sbc , sc , scd , and irr ( irregular ) according to their designations in the nasa extragalactic database . we keep these designations in the current work . we divide the complete loss galaxy sample into several subsamples , which we label the `` loss , '' `` ssfr , '' and `` metallicity '' samples . the loss sample is similar to the sample used by l11 to measure the original loss sn rates ( i.e. , the `` full - optimal '' subsample ; 10,121 galaxies from the targeted list only , i.e. , excluding sne discovered in background galaxies ) and is used here to measure sn rates per unit mass ( `` specific '' rates ) as a function of galaxy stellar mass . the ssfr sample comprises 2,415 galaxies targeted by loss that had spectra acquired by the sdss and were analyzed by both the mpa - jhu galspec pipeline @xcite and the nasa - sloan atlas ( nsa ) , which is based on the sdss dr8 spectroscopic catalog @xcite . this sample is used to correlate the sn rates with ssfr . the `` metallicity '' sample is a subset of 1,000 galaxies from the mpa - jhu galspec catalog , which we use to measure specific sn rates as a function of the central oxygen abundances of the galaxies . we detail the selection criteria for each of the subsamples in appendix [ appendix : samples ] . in appendix [ appendix : consistency ] , we show that the rates measured with each sample are mutually consistent . the numbers of galaxies and sne in each subsample are summarized in table [ table : sample_general ] and the properties of the galaxies in the entire loss sample are given in table [ table : gal_prop ] . cc & + before delving into correlations between sn rates and galaxy properties , we use this section to present several updates and extensions to the original loss papers . figures 3 and 4 of l11 show measurements of specific sn rates as a function of galaxy stellar mass in galaxies of different hubble types . however , in these figures , the rates for each type are scaled to match the rates as measured in sbc galaxies . this means that the scatter of the rates in different galaxy types is not visible in the figures ( though it can be reconstructed to some degree from the parameters of the fits to the original , unscaled measurements , which appear in table 4 of l11 ) , and the rate - mass correlation appears tighter than it really is . in figure [ fig : rates_all ] , we show similar measurements , but do not scale them , so that their spread is apparent . in the sn ii rates , where the scatter is largest , there is a clear progression from late - type ( i.e. , younger and less massive ) galaxies , where the rates are highest , to early - type ( older , more massive ) galaxies . in the se sn and sn ia rates , where the scatter is smaller , this trend is not as clear . we publish our newly calculated loss rates in table [ table : rates ] . l11 fit power laws to their rates and published the fits in their table 4 . they did not , however , publish the measurements themselves . our rates are not identical to those measured by l11 , as we use the entire full - optimal subset of galaxies , after filling in missing masses ( see appendix [ appendix : samples ] ) , but leave out the 14 sne without spectroscopic classifications ( see section [ sec : galaxies ] , above ) . however , the differences between the loss sample used here and the one used by l11 are minor and have little effect on the resultant rates . l11 divided their sne into bins so that each rate bin contained roughly the same number of sne ( `` sn - fixed '' bins ) . this is a common practice in studies where the sn sample sizes are either small to begin with , or reduced owing to binning ( in this case , by binning the host galaxies according to hubble type ) ; we will follow this binning scheme in the following sections . once the sample size is increased , though , other binning schemes become available , such as using bins of constant stellar mass ( `` mass - fixed '' bins ) . it then becomes necessary to formulate an objective , data - driven method to determine the binning scheme that will extract the most information out of the sample . to better trace the correlation between the sn rates and stellar mass , we use a sliding mass bin of constant width . the size of this bin depends on the size of the sn sample and is determined by knuth s rule @xcite , with the added constraint that the bin width not exceed 1 dex of @xmath2 . in each iteration , the bin either gains or loses the nearest sn to its current borders , such that in some iterations , losing one sn on one end may lead to a gain of several sne on the other . in each bin @xmath6 , the specific sn rate , @xmath7 , is measured according to @xmath8 where @xmath9 is the number of sne in the bin , @xmath10 is the stellar mass of the @xmath11-th galaxy in the bin , and @xmath12 is the control ( or visibility ) time of the @xmath11-th galaxy i.e . , the time during which a given sn type could have been detected by loss during the survey . these control times take into account the detection efficiency of the survey , as well as our broad knowledge of sn characteristics ( i.e. , light curve shapes and luminosity functions ) . we use the original loss control times as computed by l11 ( see also @xcite and @xcite ; see section [ subsec : rates_mass ] for a discussion of how the inclusion of sne iib in the sn ii control times may affect the rates measured here).these control times were not included in tables 2 and 4 of @xcite , which laid out most of the properties of the loss galaxy and sn samples . in order for our work to be reproducible , and for others to continue to explore the loss sample , we publish the control times in table [ table : sample_general ] . as can be seen in figure [ fig : rates_all ] , the sliding - bin rates diverge wildly at the edges of the mass range in each panel . as the number of galaxies that go into the denominator in equation [ eq : rate ] grows smaller ( as shown in the stellar mass histograms in the bottom panels ) , and the number of sne decreases , the poisson noise dominates , and the rate eventually diverges . this problem also affects rates measured with sn - fixed or mass - fixed bins ( as shown in figure [ fig : all_samples ] ) , but is usually not as apparent . the sliding - bin rates are useful for visual recognition of possibly interesting features , such as the break and rate ratio discussed in section [ subsec : rates_mass ] , below . however , they are not independent measurements and so can not be used for curve fitting . thus , throughout this work , we use sn - fixed bins for fitting purposes . the number of bins for each fit is chosen to maximize the number of sne in each bin while also revealing as much structure in the rates . we repeat each fit with a different number of bins to make sure that the results do not change appreciably . all rates in this paper are presented in units of @xmath13 , which are abbreviated to `` snum '' ( i.e. , sn rate per unit mass ) . in all the sn rate figures presented here , vertical error bars are 68% poisson uncertainties , while horizontal error bars denote the 16th and 84th percentiles of the distribution of galaxies within each bin . [ cols="^,^ " , ] in this work , we measure specific sn rates using the loss `` full - optimal '' sample and two subsamples of this sample , labeled `` ssfr '' and `` metallicity , '' which are described in section [ sec : galaxies ] . as one chooses progressively smaller subsamples of a given sample , there arises the possibility that any resulting measurements from those subsamples would be biased , relative to measurements performed with the main sample . to test for any such biases , and whether the sliding - bin rates are a good representation of the rates , figure [ fig : all_samples ] shows the rates for each sn type as measured from the different subsamples and with various binning schemes . `` sn - fixed '' bins are chosen so that they contain roughly the same number of sne in each bin while `` mass - fixed '' bins contain roughly the same amount of stellar mass , in log scale . finally , we also show the rates in fixed bins of varying width , as calculated with the astroml @xcite realization of the @xcite bayesian blocks algorithm . the various binning schemes produce measurements that generally agree with each other , except at the lower end of the mass range , where the small number of galaxies and sne can cause relatively large fluctuations in the rates . it is also clear that the rates as measured with a sliding bin are a good representation of the data . moreover , the rates measured with the ssfr subsample are consistent with those measured with the main loss sample . the rates measured from the metallicity subsample , however , are markedly higher in the largest - mass bin for both sne ii and se sne ; no such bias is noticeable for the sn ia rates . we ascribe this bias to the small size of the metallicity subsample ( @xmath14% of the main loss sample ) . this bias means that a larger sample is required to test the validity of the correlations between the sn rates and metallicity shown in figure [ fig : rates_oh ] . the connection between these correlations and the galaxy scaling relations between metallicity and stellar mass , however , should not be affected by this bias , as in this work we measure the latter scaling relation directly from the galaxies and sn hosts in the metallicity subsample . the likelihood ratio test is used to compare the goodness - of - fit of nested models , such as polynomials of increasing order . in this work , we use the likelihood ratio test to compare between zeroth- , first- , and second - order polynomials as fits to various datasets . the likelihood ratio ( @xmath15 ) is simply the ratio of the likelihoods ( @xmath16 ) of the data ( @xmath17 ) given the best - fitting parameters of each type of fit , @xmath18 for the null hypothesis ( the lower - order polynomial ) and @xmath19 for the model being tested : @xmath20 the likelihood ratio is distributed as a @xmath21 distribution , with the number of degrees of freedom equal to the difference in the degrees of freedom of the models tested ( one or two in our case ) . @xmath22 , so that through the likelihood ratio we can obtain a @xmath23-value for the significance of the rejection of the null hypothesis . these @xmath23-values can then be translated into gaussian standard deviations , so that here we report the significance of the tests in multiples of @xmath24 , with @xmath25 as the minimal significance for a `` discovery . '' when fitting the sn rates , the likelihood function is simply that of the poisson probability density function ( pdf ) , as the uncertainties of the rates are dominated by the sizes of the sn samples . the likelihood function is then @xmath26 where @xmath27 is the number of bins in which the rates are measured , @xmath28 is the observed number of sne , and @xmath29 is the number of sne resulting from the best fit to the rates . the likelihood function of the rate - ratio measurements shown in figure [ fig : rate_ratio ] is more complicated . formally , it should be the ratio ( @xmath30 ) of the poisson pdfs of the se sne , @xmath31 , and sne ii , @xmath32 , in each bin . however , as @xcite notes , if @xmath33 , the denominator vanishes , and @xmath15 is undefined . @xcite uses a truncated version of @xmath32 , so that @xmath34=1 $ ] , to solve this problem , but the resultant pdf of @xmath30 is hard to compute . however , for sufficiently large values of @xmath35 and @xmath36 , the poisson pdfs will approach the gaussian pdfs @xmath37 and @xmath38 , with means @xmath39 and @xmath40 , and standard deviations @xmath41 and @xmath42 . @xcite calculated the ratio of normal pdfs . for gaussian functions , @xcite have shown that given that @xmath17 and @xmath43 are uncorrelated , and that the coefficients of variation satisfy @xmath44 and @xmath45 , the ratio @xmath30 can be transformed via the geary - hinkley transformation into @xmath46 and the pdf of @xmath47 will then be a normal distribution with @xmath48 and @xmath49 . for @xmath37 and @xmath38 , the conditions on the coefficients are satisfied for @xmath50 and @xmath51 . the latter conditions are satisifed for all binning methods of the rate ratio . however , owing to the small sn samples , the basic condition for the approximations described above that the numbers of sne are large enough that their poisson pdfs approach gaussian ones is clearly not satisifed . to alleviate this problem , when fitting the polynomials to the measurements , we take the upper uncertainty ( which , for a poisson distribution , is always larger than the lower uncertainty ) of each measurement as the overall uncertainty . , a. v. , li , w. d. , treffers , r. r. , & modjaz , m. 2001 , in astronomical society of the pacific conference series , vol . 246 , iau colloq . 183 : small telescope astronomy on global scales , ed . b. paczynski , w .- p . chen , & c. lemme , 121 , w. d. , filippenko , a. v. , treffers , r. r. , et al . 2000 , in american institute of physics conference series , vol . 522 , american institute of physics conference series , ed . s. s. holt & w. w. zhang , 103106 lccccccccccccbcccccccccccc ugca_017 & 0 & 1 & 0 & 0 & 8.222 & 7.817 & 7.799 & 0.2599 & 1.4566 & 1.0383 & 25.78 & sc & 21.5601 & -6.0942 & & & & & & & & & & + ugca_024 & 1 & 0 & 0 & 0 & 9.772 & 9.534 & 9.599 & 0 & 0.1499 & 0 & 17.44 & scd & 31.1308 & -6.1989 & & & & & & & & & & + ugc_03825 & 1 & 0 & 0 & 1 & 6.776 & 2.215 & 4.993 & 10.0454 & 3.3206 & 15.1121 & 115.29 & sbc & 110.8882 & 41.4350 & 1864 & 53313 & 171 & 0.0276 & 6.7895 & 0.1012 & 0.0137 & & 4.3351 & 0.2624 + ugc_03944 & 0 & 0 & 0 & 0 & 8.730 & 7.540 & 6.696 & 0.9475 & 1.2143 & 2.2752 & 55.03 & scd & 114.6521 & 37.6335 & 431 & 51877 & 34 & 0.0130 & 0.8508 & 0.4333 & 0.4667 & 9.0732 & 0.6577 & 0.7980 + ugc_04226 & 1 & 0 & 1 & 0 & 6.624 & 4.436 & 3.355 & 5.3292 & 2.4379 & 8.9522 & 110.23 & scd & 121.8392 & 40.3983 & 545 & 52202 & 111 & 0.0264 & 4.4377 & 0.0777 & 0.0161 & & 4.4055 & 0.0253 + columns : + @xmath52galaxy name . + @xmath53whether galaxy belongs to the loss `` full - optimal '' subsample used to measure sn rates here and by l11 ( 1 = yes , 0 = no ) . + @xmath54number of sne ia , se sne , and sne ii ( respectively ) discovered in each galaxy . + @xmath55sn control time for sne ia , se sne , and sne ii ( respectively ) . + @xmath56galaxy stellar mass measured by l11 from the @xmath57- and @xmath58-band luminosities . + @xmath59@xmath57- and @xmath58-band luminosities ( respectively ) . + @xmath60distance to galaxy . + @xmath61hubble type of galaxy , according to the system adopted by @xcite . + @xmath62right ascension and declination ( respectively ) of galaxy , in decimal units . + @xmath63sdss plate , modified julian date ( mjd ) , and fiber identifier ( respectively ) . + @xmath64galaxy redshift , measured from the sdss spectrum . + @xmath65galaxy stellar mass , sfr , ssfr , and oxgen abundance ( respectively ) measured by the sdss mpa - jhu galspec pipeline . + @xmath66galaxy stellar mass and ssfr ( respectively ) measured from nsa petrosian and sersic photometry . + lccccccccccccc + e & 6.1^+4.4_-5.0 & 0.092^+0.042_-0.030 & 9 & 14.1^+2.3_-1.6 & 0.095^+0.047_-0.033 & 8 & 21.1^+2.6_-2.3 & 0.086^+0.040_-0.028 & 9 & 33.4^+15.4_-6.3 & 0.034^+0.016_-0.011 & 9 & 35 + s0 & 2.2^+1.8_-1.7 & 0.177^+0.061_-0.047 & 14 & 7.0^+2.2_-1.7 & 0.073^+0.025_-0.019 & 14 & 13.6^+3.7_-2.1 & 0.052^+0.018_-0.014 & 14 & 26.5^+10.9_-5.1 & 0.051^+0.018_-0.014 & 14 & 56 + sab & 1.6^+1.2_-1.2 & 0.253^+0.102_-0.075 & 11 & 4.4^+0.5_-0.7 & 0.188^+0.080_-0.058 & 10 & 7.5^+2.4_-1.7 & 0.074^+0.030_-0.022 & 11 & 15.8^+6.4_-3.5 & 0.067^+0.027_-0.020 & 11 & 43 + sb & 1.2^+1.0_0.9 & 0.284^+0.108_-0.081 & 12 & 3.7^+0.8_-0.7 & 0.233^+0.084_-0.064 & 13 & 7.4^+2.6_-1.9 & 0.067^+0.025_-0.019 & 12 & 15.1^+5.9_-2.7 & 0.082^+0.030_-0.022 & 13 & 50 + sbc & 1.0^+1.1_-0.8 & 0.228^+0.104_-0.075 & 9 & 3.9^+0.8_-0.8 & 0.167^+0.076_-0.055 & 9 & 6.5^+1.2_-1.0 & 0.147^+0.067_-0.048 & 9 & 11.3^+6.4_-2.2 & 0.078^+0.035_-0.025 & 9 & 36 + sc & 0.7^+0.9_-0.5 & 0.237^+0.117_-0.082 & 8 & 3.1^+0.9_-0.7 & 0.194^+0.096_-0.067 & 8 & 6.0^+1.2_-1.0 & 0.176^+0.087_-0.061 & 8 & 11.3^+6.5_-2.7 & 0.103^+0.051_-0.036 & 8 & 32 + scd & 0.1^+0.3_-0.1 & 0.217^+0.211_-0.118 & 3 & 1.2^+0.4_-0.3 & 0.406^+0.200_-0.141 & 8 & 2.9^+1.3_-0.8 & 0.111^+0.075_-0.048 & 5 & 9.1^+6.1_-2.5 & 0.128^+0.076_-0.051 & 6 & 22 + e scd & 0.9^+1.4_-0.8 & 0.236^+0.032_-0.029 & 68 & 4.7^+1.3_-1.0 & 0.140^+0.019_-0.017 & 69 & 9.3^+2.5_-1.9 & 0.086^+0.012_-0.011 & 67 & 19.5^+11.4_-5.0 & 0.056^+0.008_-0.007 & 70 + all galaxies & 0.8^+1.4_-0.7 & 0.231^+0.032_-0.028 & 68 & 4.6^+1.3_-1.0 & 0.139^+0.019_-0.017 & 69 & 9.3^+2.5_-1.9 & 0.085^+0.012_-0.010 & 67 & 19.5^+11.4_-5.0 & 0.056^+0.008_-0.007 & 70 + passive & 5.5^+2.9_-3.4 & 0.069^+0.037_-0.025 & 7 & 11.6^+1.4_-1.0 & 0.119^+0.058_-0.041 & 8 & 19.6^+6.7_-4.2 & 0.043^+0.023_-0.016 & 7 & & & 22 + star forming & 0.3^+0.7_-0.2 & 0.704^+0.224_-0.174 & 16 & 2.7^+1.1_-0.9 & 0.215^+0.071_-0.055 & 15 & 7.4^+5.3_-2.3 & 0.085^+0.027_-0.021 & 16 & & & 47 + + e & 1.2^+1.6_-1.1 & 0.132^+0.304_-0.109 & 1 & 8.8^+2.8_-3.4 & 0.000^+0.031_-0.000 & 0 & 21.7^+11.0_-6.9 & 0.007^+0.015_-0.006 & 1 & 55.6^+21.0_-7.3 & 0.040^+0.093_-0.033 & 1 & 3 + s0 & 2.9^+2.1_-2.2 & 0.010^+0.024_-0.009 & 1 & 7.4^+0.9_-0.9 & 0.018^+0.042_-0.015 & 1 & 9.2^+0.3_-0.3 & 0.075^+0.172_-0.062 & 1 & 15.8^+10.8_-4.5 & 0.004^+0.010_-0.004 & 1 & 4 + sab & 0.8^+0.8_-0.5 & 0.403^+0.319_-0.193 & 4 & 2.7^+0.5_-0.5 & 0.292^+0.197_-0.126 & 5 & 4.5^+0.6_-0.7 & 0.122^+0.097_-0.059 & 4 & 9.3^+7.3_-2.8 & 0.041^+0.028_-0.018 & 5 & 18 + sb & 0.6^+0.5_-0.5 & 0.625^+0.423_-0.270 & 5 & 1.9^+0.2_-0.2 & 0.558^+0.378_-0.241 & 5 & 4.1^+1.5_-1.3 & 0.106^+0.072_-0.046 & 5 & 10.4^+5.5_-2.8 & 0.056^+0.038_-0.024 & 5 & 20 + sbc & 0.7^+0.7_-0.5 & 0.487^+0.329_-0.210 & 5 & 3.3^+1.4_-1.2 & 0.128^+0.087_-0.055 & 5 & 6.6^+0.6_-0.7 & 0.358^+0.214_-0.142 & 6 & 10.7^+6.2_-2.5 & 0.120^+0.081_-0.052 & 5 & 21 + sc & 0.5^+0.6_-0.3 & 0.511^+0.275_-0.188 & 7 & 2.3^+0.9_-0.6 & 0.255^+0.126_-0.088 & 8 & 4.8^+1.2_-0.9 & 0.241^+0.130_-0.089 & 7 & 9.4^+5.4_-2.1 & 0.159^+0.078_-0.055 & 8 & 30 + scd & 0.2^+0.5_-0.2 & 0.225^+0.178_-0.108 & 4 & 1.4^+0.2_-0.2 & 0.538^+0.426_-0.258 & 4 & 2.3^+0.6_-0.5 & 0.295^+0.199_-0.127 & 5 & 5.2^+5.7_-1.7 & 0.091^+0.072_-0.043 & 4 & 17 + irr & 0.1^+0.1_-0.0 & 1.494^+3.435_-1.236 & 1 & 0.2^+0.0_-0.0 & 0.000^+25.3_-0.000 & 0 & 0.4^+1.1_-0.2 & 0.357^+0.820_-0.295 & 1 & 5.3^+8.8_-2.4 & 0.066^+0.152_-0.055 & 1 & 3 + sab scd & 0.4^+0.7_-0.3 & 0.402^+0.098_-0.080 & 25 & 2.3^+0.7_-0.6 & 0.261^+0.059_-0.049 & 28 & 4.7^+1.1_-0.9 & 0.151^+0.037_-0.030 & 25 & 10.2^+6.3_-2.7 & 0.088^+0.020_-0.017 & 28 + all galaxies & 0.4^+0.7_-0.3 & 0.377^+0.086_-0.071 & 28 & 2.4^+0.7_-0.6 & 0.203^+0.044_-0.037 & 30 & 4.9^+1.2_-0.9 & 0.112^+0.025_-0.021 & 29 & 12.1^+10.2_-4.0 & 0.035^+0.008_-0.007 & 29 + + s0 & 1.8^+1.4_-1.4 & 0.016^+0.037_-0.013 & 1 & 4.0^+0.0_0.0 & 0.985^+1.299_-0.636 & 2 & 5.2^+1.0_-0.8 & 0.020^+0.026_-0.013 & 2 & 12.9^+10.8_-4.5 & 0.004^+0.005_-0.002 & 2 & 7 + sab & 0.9^+0.8_-0.6 & 0.645^+0.259_-0.192 & 11 & 3.6^+1.1_-1.0 & 0.126^+0.054_-0.039 & 10 & 7.3^+1.9_-1.4 & 0.098^+0.039_-0.029 & 11 & 14.4^+6.9_-3.0 & 0.075^+0.030_-0.022 & 11 & 43 + sb & 1.4^+1.2_-1.1 & 0.303^+0.105_-0.080 & 14 & 4.4^+0.9_-0.8 & 0.223^+0.081_-0.061 & 13 & 8.0^+1.7_-1.7 & 0.146^+0.050_-0.039 & 14 & 14.0^+6.2_-2.1 & 0.123^+0.042_-0.032 & 14 & 55 + sbc & 0.7^+0.6_-0.5 & 1.317^+0.365_-0.292 & 20 & 2.7^+0.9_-0.7 & 0.451^+0.129_-0.103 & 19 & 5.1^+1.2_-0.8 & 0.382^+0.106_-0.085 & 20 & 10.2^+5.7_-2.5 & 0.204^+0.057_-0.045 & 20 & 79 + sc & 0.4^+0.5_-0.3 & 2.096^+0.642_-0.503 & 17 & 1.5^+0.3_-0.3 & 1.241^+0.380_-0.298 & 17 & 3.5^+1.8_-1.2 & 0.419^+0.124_-0.098 & 18 & 9.3^+5.5_-2.1 & 0.403^+0.123_-0.097 & 17 & 69 + scd & 0.1^+0.2_-0.1 & 2.256^+0.717_-0.558 & 16 & 0.7^+0.3_-0.2 & 1.442^+0.458_-0.357 & 16 & 1.7^+0.5_-0.3 & 1.194^+0.365_-0.287 & 17 & 4.1^+5.0_-1.2 & 0.355^+0.113_-0.088 & 16 & 65 + irr & 0.0^+0.0_-0.0 & 1.416^+3.255_-1.171 & 1 & 0.4^+1.1_-0.2 & 0.725^+0.956_-0.468 & 2 & 3.3^+0.8_-1.0 & 0.316^+0.727_-0.262 & 1 & 8.8^+7.3_-2.0 & 0.194^+0.256_-0.126 & 2 & 6 + sab scd & 0.3^+0.5_-0.2 & 1.52^+0.19_-0.17 & 78 & 2.0^+0.7_-0.6 & 0.547^+0.070_-0.062 & 77 & 4.6^+1.4_-1.1 & 0.253^+0.032_-0.029 & 78 & 10.7^+6.3_-2.8 & 0.139^+0.018_-0.016 & 78 + all galaxies & 0.3^+0.5_-0.2 & 1.33^+0.16_-0.15 & 81 & 2.0^+0.7_-0.6 & 0.426^+0.053_-0.047 & 81 & 4.7^+1.4_-1.1 & 0.175^+0.022_-0.019 & 81 & 12.2^+10.2_-4.0 & 0.052^+0.006_-0.006 & 81 + cccc|cccc|cccc & & + + 0.035^+0.016_-0.013 & 2.2^+5.1_-1.8 & 1 & 171 & 0.011^+0.007_-0.007 & 8.0^+18.5_-6.7 & 1 & 166 & 0.015^+0.011_-0.007 & 4.2^+9.8_-3.5 & 1 & 189 + 0.084^+0.036_-0.030 & 0.4^+0.3_-0.9 & 1 & 454 & 0.075^+0.045_-0.049 & 0.4^+0.9_-0.3 & 1 & 592 & 0.09^+0.07_-0.04 & 0.3^+0.6_-0.2 & 1 & 623 + 0.09^+0.04_-0.03 & 1.1^+1.1_-0.6 & 3 & 456 & 0.08^+0.07_-0.05 & 0.9^+0.8_-0.5 & 3 & 670 & 0.14^+0.11_-0.06 & 1.2^+0.6_-0.4 & 8 & 728 + 0.10^+0.06_-0.03 & 0.9^+0.8_-0.5 & 3 & 483 & 0.10^+0.10_-0.05 & 0.6^+0.6_-0.3 & 3 & 784 & 0.15^+0.11_-0.07 & 2.5^+0.8_-0.6 & 17 & 727 + 0.14^+0.07_-0.05 & 0.9^+0.7_-0.4 & 4 & 478 & 0.11^+0.11_-0.06 & 0.7^+0.5_-0.3 & 4 & 865 & 0.15^+0.11_-0.07 & 2.4^+0.8_-0.6 & 16 & 713 + & & & & & & & & & & & + & & & & & & & & & & & + + 0.035^+0.051_-0.022 & 2.7^+6.2_-2.2 & 1 & 143 & 0.012^+0.018_-0.006 & 13.7^+31.5_-11.3 & 1 & 90 & 0.02^+0.03_-0.01 & 6.9^+15.9_-5.7 & 1 & 108 + 0.09^+0.12_-0.05 & 0.5^+1.1_-0.4 & 1 & 374 & 0.050^+0.071_-0.026 & 1.6^+3.8_-1.4 & 1 & 208 & 0.09^+0.13_-0.05 & 0.5^+1.2_-0.4 & 1 & 372 + 0.09^+0.13_-0.06 & 1.3^+1.3_-0.7 & 3 & 379 & 0.08^+0.12_-0.04 & 1.2^+1.5_-0.7 & 2 & 373 & 0.10^+0.15_-0.06 & 2.3^+1.4_-0.9 & 6 & 402 + 0.10^+0.16_-0.07 & 1.0^+1.0_-0.5 & 3 & 416 & 0.2^+0.3_-0.1 & 0.3^+0.4_-0.2 & 2 & 601 & 0.13^+0.20_-0.08 & 1.7^+1.0_-0.7 & 6 & 435 + 0.14^+0.21_-0.09 & 1.0^+0.8_-0.5 & 4 & 430 & 0.3^+0.5_0.2 & 0.2^+0.2_-0.1 & 3 & 744 & 0.13^+0.21_-0.08 & 2.3^+1.1_-0.8 & 8 & 433 + & & & & & & & & & & & + & & & & & & & & & & & + columns : + @xmath67stellar mass of the galaxies in the bin . error bars represent the upper and lower limits of the bin . + @xmath68specific sn rate . error bars are 68% poisson uncertainties stemming from the number of sne in each bin . + @xmath69number of sne in the given bin . + @xmath70number of galaxies in the given bin . + cccc 0.32^+0.12_-0.09 & 0.13^+0.09_-0.08 & 3 & 26 + 0.84^+0.28_-0.23 & 0.35^+0.11_-0.11 & 14 & 47 + 2.1^+0.7_-0.6 & 0.59^+0.13_-0.11 & 34 & 77 + 5.1^+1.8_-1.3 & 0.64^+0.14_-0.11 & 38 & 93 + 12^+4_-3 & 0.64^+0.15_-0.15 & 20 & 58 + cccccccccc & 8.83^+0.17_-0.24 & 0.55^+0.29_-0.20 & 7 & -12.16^+0.19_-0.25 & 0.087^+0.028_-0.022 & 16 & -1.16^+0.29_-0.42 & 0.118^+0.032_-0.026 & 21 + & 9.113^+0.036_-0.042 & 0.20^+0.10_-0.07 & 8 & -11.20^+0.49_-0.46 & 0.073^+0.023_-0.018 & 16 & -0.23^+0.27_-0.34 & 0.100^+0.026_-0.021 & 22 + & 9.224^+0.050_-0.045 & 0.30^+0.16_-0.11 & 7 & -10.33^+0.12_-0.14 & 0.16^+0.05_-0.04 & 17 & 0.34^+0.25_-0.14 & 0.135^+0.035_-0.029 & 22 + & & & & -9.80^+0.38_-0.25 & 0.31^+0.10_-0.08 & 16 & + & 8.80^+0.16_-0.22 & 0.76^+0.60_-0.37 & 4 & -11.84^+0.57_-0.45 & 0.008^+0.019_-0.007 & 1 & -0.68^+0.45_-0.64 & 0.045^+0.027_-0.018 & 6 + & 9.088^+0.038_-0.042 & 0.26^+0.21_-0.12 & 4 & -10.43^+0.17_-0.30 & 0.09^+0.03_-0.05 & 6 & 0.11^+0.13_-0.13 & 0.13^+0.08_-0.05 & 6 + & 9.174^+0.075_-0.025 & 0.24^+0.19_-0.12 & 4 & -10.08^+0.09_-0.07 & 0.28^+0.19_-0.12 & 5 & 0.45^+0.21_-0.12 & 0.13^+0.08_-0.05 & 6 + & & & & -9.67^+0.32_-0.20 & 0.35^+0.21_-0.14 & 6 & + & 8.72^+0.15_-0.18 & 3.4^+1.3_-1.0 & 12 & -11.84^+0.57_-0.45 & 0.033^+0.015_-0.011 & 9 & -0.78^+0.40_-0.59 & 0.08^+0.02_-0.02 & 20 + & 9.010^+0.041_-0.048 & 1.39^+0.53_-0.39 & 12 & -11.51^+0.15_-0.28 & 0.17^+0.05_-0.04 & 17 & -0.05^+0.11_-0.11 & 0.30^+0.09_-0.07 & 19 + & 9.110^+0.020_-0.031 & 0.70^+0.25_-0.19 & 13 & -10.20^+0.06_-0.08 & 0.56^+0.17_-0.13 & 18 & 0.22^+0.09_-0.08 & 0.35^+0.10_-0.08 & 20 + & 9.181^+0.069_-0.031 & 0.43^+0.17_-0.12 & 12 & -9.95^+0.10_-0.11 & 0.92^+0.28_-0.22 & 17 & 0.51^+0.22_-0.11 & 0.32^+0.09_-0.07 & 20 + & & & & -9.55^+0.34_-0.17 & 1.35^+0.40_-0.31 & 18 & + cccccc ( r_ia / snum ) & log(m_/m _ ) & -0.44 0.06 & 3.8 0.6 & 2.7/2 & > 5 + log(r_ia / snum ) & log(m_/m _ ) & -0.43 0.08 & 3.7 0.8 & 3.9/2 & > 5 + log(r_ia / snum ) & log(ssfr / yr^-1 ) & 0.23 0.08 & 1.7 0.8 & 3.2/2 & > 3 + log(r_ia / snum ) & 12+log(o / h ) & -1.3 0.7 & 11 7 & 1.0/1 & > 1 + ( r_se / snum ) & log(m_/m _ ) & -0.64 0.09 & 5.8 0.9 & 7.3/2 & > 4 + log(r_se / snum ) & log(m_/m _ ) & -0.46 0.10 & 4.1 1.0 & 2.1/2 & > 4 + log(r_se / snum ) & log(ssfr / yr^-1 ) & 0.8 0.5 & 7.0 4.8 & 0.5/1 & > 5 + log(r_se / snum ) & 12+log(o / h ) & -1.5 1.0 & 13 9 & 10 ^ -5/1 & > 1 + ( r_ii / snum ) & log(m_/m _ ) & -0.84 0.05 & 8.2 0.5 & 21/4 & > 4 + log(r_ii / snum ) & log(m_/m _ ) & -0.68 0.05 & 6.6 0.5 & 5.6/4 & > 4 + log(r_ii / snum ) & log(ssfr / yr^-1 ) & 0.9 0.2 & 8.8 1.9 & 2.7/2 & > 5 + log(r_ii / snum ) & 12+log(o / h ) & -1.73 0.45 & 16 4 & 0.6/2 & > 5 + ( m _ ) & 12+log(o / h ) & 2.5 & -12.6 & & + log(m_,ia ) & 12+log(o / h ) & 2.31 0.13 & -10.8 1.2 & 481/24 & + log(m_,se ) & 12+log(o / h ) & 3.39 0.11 & -20.4 1.2 & 16/10 & + log(m_,ii ) & 12+log(o / h ) & 3.33 0.08 & -20.1 0.7 & 321/45 & +

gamma - ray bursts ( grbs ) are the brightest electromagnetic explosions in the universe ( e.g. , @xcite ) . for the short time of their activity , they outshine all other sources in the sky in gamma rays . these energetic explosions originate from cataclysmic cosmic events , whose inner engine " that drives the observed emission is confined to volumes mere tens of kilometers across , as indicated by the duration and variability of gamma - ray emission ( e.g. , @xcite ) . this inner engine is hidden from direct electromagnetic observations , even though the variability of the emission is directly related to the engine s activity . gravitational waves ( gws ) , on the other hand , carry information directly from the inner engine to the observer . with the near completion of advanced gw detectors @xcite , we will soon be able to directly probe the dynamical processes leading to the creation of grbs , which are less ( or not ) accessible via other messengers . the advanced ligo @xcite and virgo @xcite detectors are planned to begin observation in 2015 , albeit initially below design sensitivity , gradually reaching their design sensitivity around the end of the decade @xcite . upon reaching their design sensitivity , the advanced ligo and virgo detectors are expected to be about 10 times more sensitive ( at their most sensitive frequency band around @xmath0hz ) than the initial ligo / virgo detectors . further , advanced detectors will have a wider sensitive frequency band , reaching down to @xmath1hz . this will improve their sensitivity to wide - band sources such as compact binaries . the global gw detector network will further increase its reach in both distance and source direction reconstruction by the addition of further detectors @xcite , such as the japanese kagra ( formerly lcgt ; @xcite ) detector that is under construction , as well as the planned third ligo observatory in india @xcite . there are a number of mechanisms conjectured to result in a grb . potential mechanisms include the formation of a central object surrounded by an accretion disk from the merger of binary neutron stars @xcite or a neutron star and a low - mass black hole @xcite , the collapse of the core of a massive star @xcite , the global reconfiguration of the magnetic fields in magnetized neutron stars @xcite , the accretion - induced collapse of white dwarfs following accretion from a non - degenerate companion @xcite or following the merger of a white dwarf binary @xcite , or even cosmic strings @xcite . identifying the mechanisms behind the creation of individual grbs will be greatly aided by gw detections . this review intends to survey the prospects of gw measurements in understanding the central engines of grbs . while our theoretical view of gw emission mechanisms connected to grbs is rapidly evolving , the main directions , the opportunities and limitations of observational gw astrophysics are becoming clearer . with this review we aim to provide guidance to ( i ) astronomers interested in utilizing the capabilities of gw measurements to complement electromagnetic or neutrino observations in understanding grbs , and ( ii ) gw scientists interested in the open questions in grb astrophysics that can be addressed through gw measurements . compact binary mergers , the likely progenitors of most short grbs ( e.g. , @xcite , although see @xcite for alternative models ) , are one of the primary targets of gw searches ( e.g. , @xcite ) . advanced gw detectors will be able to , among others , measure the properties of the binary @xcite , directly probe the dynamics of the central engine , and correlate the binary s masses and orientation with prompt and delayed electromagnetic counterparts . these observations will constrain the central engine mechanism , provide standard sirens for measuring cosmological parameters @xcite , and constrain the nature of nuclear matter ( see section [ section : cbcphysics ] ) . gw emission from isolated central engines , such as stellar core collapses , accretion induced collapses or magnetars , is typically at higher frequencies and at lower amplitudes than for compact binaries . gw emission depends on the development of varying quadrupole moment , which is highly source and model dependent . the detection of gws from isolated objects would help us understand the internal processes in massive stellar collapses , including the development of differentially rotating protoneutron stars , the nature of nuclear matter , or the creation of massive accretion disks ( sections [ section : collapsarphysics ] , [ section : magnetarphysics ] & [ section : millisecondmagnetars ] ) . for many topics touched upon here , we can refer the interested reader to existing , excellent reviews ; on grb blastwaves @xcite ; on short and long grb observations ; on models of central engines involving core - collapse events @xcite , neutron star mergers @xcite , and isolated neutron stars @xcite ; on gw emission from core - collapse events @xcite and binary mergers @xcite ; on short grb rates @xcite , long grb rates @xcite , and compact object merger rates @xcite . some investigations of isotropic prompt and delayed electromagnetic counterparts have recently been summarized in @xcite ( for short grbs ) . this review is organized as follows . section [ section : emissionprocesses ] outlines gw emission processes from grb progenitors , focusing on the emission scenarios that may produce signals which could be detectable with advanced gw detectors . section [ section : astrophysics ] presents some of the intriguing astrophysical questions that could be answered through the detection of gws from grb progenitors , in some cases in coincidence with other messengers . in section [ section : observationalstrategies ] we review gw observation strategies from single gw to multimessenger searches , also outlining prospects for the advanced detector era . finally , section [ section : conclusion ] we briefly summarizes the presented results and the near future of grb astrophysics with gws . in this section , we outline the main grb progenitor models with an emphasis on their expected gw signature . it should be clear that , depending on the progenitor model , the gw emission from grb progenitors varies widely with respect to predicted signal strength and characteristic frequency . the majority of short - hard grbs are thought to be powered by the merger of ns - ns or ns - bh binaries @xcite . compact binaries are also strong gw sources : the gws produced in the minutes before the merger should be detectable , by second generation instruments ( i.e. , advanced ligo and virgo ) , reliably out to @xmath2mpc for ns - ns binaries , and further for ns - bh systems @xcite . theoretical studies provide detailed information about the binary evolution and the gravitational waveform expected from these systems . moreover , numerical simulations , which include the physics of the ns matter , continue to improve ( see refs . @xcite for reviews ) . here we present an overview of compact binary evolution , which is summarised in fig . [ fig : cbcflowchart ] . at the highest level , there can be four phases : the inspiral phase during which the orbit shrinks due to energy and angular momentum loss via gws ; the merger phase during which dynamical effects and microphysics of the ns become important ; the accretion phase when material stripped from the ns(s ) accretes onto the merged objects ; and the ringdown phase when the merged objects settle down to an unperturbed ( kerr ) bh or a ns . the details vary by ns - ns and ns - bh merger and by mass ( and other parameters ) of the merging objects . the accretion stage is likely essential to the formation of short grbs from binary coalescences , although the formation and evolution of protomagnetars following ns - ns merger has also been proposed as a plausible central engine @xcite . figure [ fig : cbcspectrum ] presents the main features of gw emission in a schematic spectrum of the gw effective amplitude , separately for ns - ns and bh - ns mergers . below we discuss the different stages of the evolution in more detail . early in the evolution of a compact binary system , the two objects are separated by a relatively large distance compared to their radii . the binary elements spiral towards each other by losing angular momentum via the emission of gws . during the early stages of the inspiral , the two compact objects can be approximated as point masses for the purposes of gw emission @xcite . at the late stages of the inspiral , the internal structure of the objects become increasingly important . for example , the tidal deformation of nss in a binary system can ( slightly ) affect the orbital period ( and therefore the gravitational waveform ) in the late inspiral phase @xcite . further , general relativistic spin - spin or spin - orbit coupling can cause the binary s orbital plane to precess , affecting the binary s evolution and gw emission @xcite . nevertheless , the dominant features of the gw signal from the inspiral phase are captured by neglecting the spins and internal structure of the binary elements . as the objects spiral together , their orbital frequency increases producing a gw signal that sweeps upward in frequency . about @xmath3 minutes before merger , the gw from the inspiral of a ns - ns binary begins to sweep upward from @xmath1hz through the band of earth - based gw interferometers . the effective amplitude @xmath4 of the gw signal from a binary system decreases as @xmath5 @xcite , up to a mass - dependent cutoff frequency @xmath6khz @xcite . the frequency ranges @xmath7khz and @xmath8khz are traditionally considered the inspiral and early merger phases , respectively . for @xmath9 the merger retains a binary - like structure and consequently emits relatively strong gws @xcite . advanced detectors will be able to detect a ns - ns inspiral up to @xmath10mpc , while ns - bh inspirals will be detectable up to @xmath11mpc @xcite ( the distances are given for untriggered searches , with optimal source orientation and direction ; for further details see section [ section : allsky ] ) . the effective survey volume determined by averaging over sky location and inclination of the sources is @xmath12 @xcite . using the current best - guess rates of mergers , this gives tens of ns - ns and a few ns - bh binaries detected with advanced detectors each year @xcite . additional advanced detectors , such as kagra @xcite or ligo india @xcite , can significantly increase this range @xcite . third generation detectors are expected to reach an order of magnitude farther than advanced detectors , i.e. to several gpc , and hence will be able to observe tens of thousands of events a year ( e.g. , @xcite ) . depending on the binary system , the merger can progress in multiple distinct directions with qualitatively different gw and gamma - ray emission . the formation of a massive accretion disk is probably crucial to the generation of a grb from binary mergers @xcite . accretion disks can be created via the tidal disruption of a ns at some point during the merger @xcite . alternatively , for ns - ns mergers , material with centrifugal support can be left outside the newly formed bh . whether an accretion disk is formed or not depends on the binary properties ( mass , spin , etc . ) , as well as the ns equation of state ( eos ) . a disk mass of @xmath13m@xmath14 , where m@xmath14 is the mass of the sun , is probably sufficient to supply the energy for the creation of a short grb @xcite . in a recent comparison between numerical simulations and observation , giacomazzo et al . @xcite found that the observed emission of short grbs implies torus masses @xmath15m@xmath14 , which favors high - mass " ns - ns mergers ( @xmath16m@xmath14 ) . further , they find that bh - ns mergers , while can not be excluded , would require a very rapidly spinning bh ( with spin @xmath17 ) . here we list the possible outcomes of the merger and outline the scenarios that can lead to them . 1 . * bh with no accretion disk * for bh - ns binaries , if the radius of the innermost stable circular orbit ( isco ) is greater than the tidal disruption radius , the ns plunges into the bh before it could be tidally disrupted , resulting in a bh with no accretion disk . this will be the case for binaries with relatively high bh : ns mass ratio ( @xmath18 ) @xcite . this ratio strongly depends on the bh spin ( due to the change in the location of the isco ) and the ns eos @xcite . + for ns - ns binaries with equal masses , if the binary mass exceeds a threshold @xmath19 , the nss will promptly collapse to a bh upon merger @xcite , leaving essentially no accretion disk behind @xcite . @xmath19 depends on the ns eos . kiuchi et al . @xcite found that a bh is promptly formed if total mass of the binary system is @xmath20m@xmath14 . + in this scenario no accretion - powered grb will be produced , although there are possible channels through which even such a system can emit electromagnetic radiation @xcite . * bh with accretion disk * for bh - ns binaries with relatively low mass ratio ( @xmath21 ) , the ns will be tidally disrupted before falling into the bh , which leads to the formation of a massive accretion disk @xcite . for spinning bhs , the mass ratio below which disruption occurs is even higher @xcite , and the mass of the formed disk can greatly depend on bh spin and spin - alignment @xcite . prior to tidal disruption , the emitted gw frequency reaches @xmath22khz @xcite . + for ns - ns binaries with unequal ns masses and sufficiently large total mass ( @xmath20m@xmath14 @xcite ) , the less massive ns will be tidally disrupted , followed by the more massive ns s prompt collapse into a bh , leaving a potentially massive accretion disk behind @xcite . + in this scenario , after the merger of a ns - ns binary , if the forming disk features azimuthal variations , gws may be emitted by the material orbiting the central object around a peak amplitude @xmath23khz @xcite . for bh - ns binaries , @xmath24 can be significantly lower , @xmath25inversely proportional to the total mass of the system @xcite . for instance a bh - ns binary with @xmath26m@xmath14 bh mass has @xmath27khz , depending on , e.g. , the bh spin @xcite . + this scenario is a good candidate for the creation of accretion - powered grbs . * hypermassive ns formation * ns - ns binaries with total mass @xmath28m@xmath14 ( this mass threshold depends on the ns eos @xcite ) will not promptly collapse into a bh , but will first form a so - called _ hypermassive _ nsm@xmath14 ) nss , it is possible that the resulting post - merger ns mass can be supported even without differential rotation ( i.e. it is not hypermassive ) . in this case the ns can be long lived ( see , e.g. , @xcite ) . ] @xcite , supported by differential rotation and thermal pressure @xcite . the hypermassive ns eventually collapses into a bh with a delay of @xmath29ms-1s due to ( i ) losing angular momentum via gws or magnetic processes @xcite ( magnetorotational instability @xcite or magnetic winding @xcite ) , ( ii ) its rotation becoming more uniform due to magnetic braking and viscosity @xcite , in which case differential rotational no longer supports the ns , and/or ( iii ) cooling due to , e.g. , neutrino emission , so thermal pressure provides less support against the gravitational pull @xcite . + a rapidly rotating hypermassive ns will assume a ( non - axisymmetric ) ellipsoidal shape @xcite , which is energetically favorable over a spheroid @xcite . such an ellipsoidal hypermassive ns will emit a strong gw signal at twice its ( quasiperiodic ) rotational frequency @xmath30khz @xcite . + hypermassive nss may leave a massive accretion disk behind after collapsing into a bh @xcite . it seems that the outcome depends on the binary mass . as we saw above , for high - mass nss , the merger promptly forms a bh . at the highest masses for which a hypermassive ns is formed , a very short - lived ns is formed , with likely suppressed accretion disk formation @xcite . for lower - mass binaries , the formed hypermassive ns is longer - lived , resulting in stronger gw emission . these lower - mass binaries also result in the formation of a more massive accretion disk . + in this scenario , in which a hypermassive ns is formed , a significant amount of gw energy can be emitted from the hypermassive ns at around its quasiperiodic rotation frequency . quasiperiodic gw emission from a hypermassive ns would be detectable with advanced detectors from @xmath31mpc @xcite , especially because it would be accompanied by an inspiral phase with significantly higher signal - to - noise ratio ( snr ) . this scenario is a good candidate for the creation of grbs , with either bh - torus or protomagnetar central engines . formation of stable ns * if the total mass @xmath32 of a ns - ns binary is below the mass limit @xmath33 of non - rotating nss , or if the merged ns s mass is @xmath34 due to , e.g. , tidal disruption , a stable , long - lived ns can form from the merger . while typical observed ns - ns binary masses @xcite are likely above @xmath35m@xmath14 @xcite , it is plausible that some mergers end up forming a stable ns . accretion disks themselves may contribute to the gw emission of binaries @xcite . bh - torus systems can be unstable to non - axisymmetric perturbations that may give rise to non - axisymmetric torus structure , resulting in gw emission at a few - hundred hz . see also section [ section : accretiondiskinstabilities ] for gw emission from accretion disks . when a bh is formed , or when substantial matter with non - axisymmetric structure ( i.e. , ns or fragmented accretion disk ) falls into the bh , the event horizon of the bh is initially perturbed . subsequently , it quickly approaches the non - perturbed ( kerr ) solution via the emission of gws . this process is called ringdown , with characteristic frequencies of @xmath36khz ( e.g. , @xcite),khz of a particle around a schwarzschild bh of mass @xmath37 at the innermost stable circular orbit ( isco ) . the orbital frequency greatly depends on the bh spin ( e.g. , @xcite for the specific dependence ) . ] given the typical mass of the binary ( @xmath38m@xmath14 ) . at and above @xmath24 , the gw spectrum is qualitatively independent of the properties of the binary ( although quantitatively the characteristic frequency decreases with increasing binary mass ) . it decays exponentially due to the @xmath39 quasinormal mode oscillation of the final black hole @xcite . for solar - mass binaries , gw radiation from the bh ringdown is undetectable with advanced gw interferometers due to its high frequency . we note that bh ringdown can be suppressed or even choked in the presence of intense accretion onto the bh , as is likely the case for unequal - mass ns - ns binaries @xcite . massive stars develop an inert iron core supported by non - thermal degeneracy pressure in their center as the final stage of nuclear fusion . the growing iron core becomes gravitationally unstable upon reaching a mass around the chandrasekhar mass ( @xmath40m@xmath14 ) due to a softening of the eos as the degenerate fermions become relativistic @xcite , collapsing into a ns or a bh . such a collapsing stellar system may result in a supernova explosion ( a core - collapse supernova ) , and/or a grb . the central engine driving grb emission may be an accreting bh ( e.g. , @xcite ) or a rapidly spinning , strongly magnetized protoneutron star ( a _ millisecond proto - magnetar _ ; e.g. , @xcite and references therein ) . the core collapse of a massive star may emit gws through various mechanisms @xcite . below we focus on the mechanisms producing gws that may be sufficiently strong to be detected at distances relevant to grbs , i.e. that may be detectable from @xmath41mpc with 2@xmath42 or @xmath43 generation gw interferometers . a schematic diagram of these emission processes are shown in fig . [ fig : collapsargw ] . for massive stars with initial stellar masses @xmath44m@xmath45m@xmath14 , the collapsing core is expected to form a so - called protoneutron star @xcite . the resulting rotating protoneutron star can be unstable to non - axisymmetric deformations , potentially giving rise to copious gw emission @xcite . the onset of rotational instabilities depends on the rotational rate of the star , which can be conveniently parameterized by @xmath46 , i.e. the ratio of the star s rotational kinetic ( @xmath47 ) and gravitational potential ( @xmath48 ) energy @xcite . the resulting non - axisymmetric structure may be a bar - like @xmath39 mode , giving rise to a characteristic gw emission . higher @xmath49 modes may also arise , albeit they have longer growth time @xcite . while there are still various uncertainties in the evolution and role of rotational instabilities ( e.g. , non - linear mode - coupling effects can severely limit the deformation @xcite , or the role of viscosity and neutrino cooling ) , the emerging picture is that rotational instabilities are likely viable emitters of gws and could play an important role in the future detection and understanding of grb progenitors through their gw signature . the energy potentially available for gw emission is abundant . the rotational energy of a typical ns with 1@xmath50khz rotational frequency is @xmath51m@xmath52c@xmath53 ( e.g. , @xcite ) . even a fraction of this energy , if radiated away in gws , could be detectable at large distances ( @xmath41mpc ) with advanced detectors . the protoneutron star may also accrete supernova fallback material . such an accretion further increases the angular momentum and energy available that may be radiated away via gws @xcite . the amplitude of a gw signal emitted by a rotating bar scales as @xmath54 , where @xmath37 , @xmath55 , @xmath56 and @xmath57 are the mass , radius , gw frequency ( i.e. twice the rotational frequency ) , and distance of the ns , respectively @xcite . the energy radiated away in gws , in the newtonian quadrupole approximation , can be estimated as ( see , e.g. , @xcite ) @xmath58 where @xmath59 is the ellipticity of the bar and @xmath60 is the duration of the presence of the instability . protoneutron stars are subject to different rotational instabilities , the two main categories being dynamical and secular instabilities . a star is dynamically unstable if it is unstable to non - axisymmetric perturbations even in the absence of dissipation ( i.e. if a slightly non - axisymmetric shape that conserves angular momentum is energetically favorable ) . a star is secularly unstable if it is unstable to non - axisymmetric perturbations only if dissipative effects are relevant , i.e. if the change towards a non - axisymmetric shape requires the radiation or redistribution of angular momentum ( see @xcite for interesting historical overviews ) . below we discuss these instabilities further in detail . 1 . * dynamical instabilities * rapidly rotating stars will be subject to dynamical instabilities driven by hydrodynamical and gravitational effects @xcite . dynamical instability is the simplest " form of ns instabilities , since its development is quick ( on the time scale of the rotational period of the ns ) , and does not require dissipation . a uniformly rotating , classical fluid body becomes dynamically unstable at a rotation rate @xmath61 . the maximum rotation of a compact star is limited by mass - shedding to @xmath62 @xcite . ] stability conditions are essentially the same for relativistic stars , for which @xmath63 @xcite . furthermore , differentially rotating stars are subject to non - axisymmetric instabilities even at much slower rotation with @xmath64 . such low-@xmath65 instabilities in differentially rotating protoneutron stars are probably analogous to the papaloizou - pringle instability ( @xcite ; also see section [ section : accretiondiskinstabilities ] ) : the protoneutron star core is surrounded by a fluid rotating at the frequency of a non - axisymmetric mode of the core , hence exciting this mode @xcite . as numerical simulations so far have been too short to capture the long - term behavior of some dynamical instabilities , they are not conclusive in terms of the total energy emitted via gws from dynamical instabilities . nevertheless , gw emission seems to be fast relative to the cooling time of the protoneutron star , or compared to energy loss due to viscosity . consequently , if competing mechanisms that radiate away angular momentum ( e.g. , magnetic fields ) are weak , gws can carry away a significant fraction of the protoneutron star s rotational energy , producing a signal that may be detectable from @xmath66 for narrow - band , circularly polarized gw signals from optimal source direction sensitivity improvement for advanced detectors ; based on the results of @xcite that used only one gw detector multiple detectors could see even farther . ] . the frequency scaling of the distance only applies to @xmath67hz . * secular instabilities * stars with lower rotation rates ( i.e. for which the faster - developing dynamical instabilities are not prevalent ) can be subject to secular ( i.e. dissipation - driven ) non - axisymmetric instabilities @xcite . dissipation can occur via gravitational radiation @xcite or fluid viscosity @xcite . gw emission drives frame - dragging ( so - called chandrasekhar - friedman - schutz ) instabilities of modes that are retrograde with respect to the star but prograde with respect to the observer @xcite . among gw - driven instabilities , fundamental @xmath56-mode bar instabilities have the shortest growth time : @xmath68s for @xmath69 @xcite . in the uniformly rotating approximation of a relativistic star , the protoneutron star becomes unstable to gw - driven @xmath39 @xmath56-mode instabilities for @xmath70 depending on the eos and stellar mass @xcite ( compare with the newtonian limit of @xmath71 @xcite ) . at lower rotation rates , stars become unstable to higher - multipole @xmath56 modes , albeit higher modes have longer growth times @xcite . a recent work of passamonti et al . @xcite indicates that the most unstable @xmath56 modes developing in the aftermath of a supernova explosion may in fact be @xmath72 and 4 modes . + as angular momentum is radiated away through gws ( and potentially other channels ) , the protoneutron star s rotation frequency and therefore the emitted gw frequency decreases , sweeping towards the most sensitive band of ligo - like detectors . some analytical and numerical results indicate that such a gw signal may be detectable at large distances , up to @xmath73mpc with advanced interferometers @xcite . some other recent simulations are also promising , even though they only cover the first few milliseconds after core collapse and therefore can not capture the long - term evolution of the bar mode @xcite . nevertheless , some recent , realistic simulations ( e.g. , @xcite ) predict much smaller detectable range . + rotating protoneutron stars are also unstable to gw - driven _ r - mode _ oscillations ( the restoring force being the coriolis force ) at any ( i.e. arbitrarily low ) rotation rate @xcite . @xmath55 modes are important only if their growth time is less than the damping time of viscous forces . further , @xmath74-mode instability is expected to be saturated at low amplitude due to dissipative effects @xcite , and is suppressed in the presence of magnetic fields @xcite . nevertheless , under favorable conditions , gw signals from protoneutron star @xmath74 modes may be detectable for several years after core collapse @xcite . given such a long duration , the emitted gw signal may be integrated for a measurement of @xmath75yr that would give a detectable signal to distances of @xmath76mpc @xcite ( or to even farther if the protoneutron star is a _ strange quark star _ further calculations ( see @xcite and references therein ) suggest that the saturation ( i.e. maximum ) amplitude of @xmath74-mode instabilities is limited by its non - linear coupling to other inertial modes . 3 . * magnetic distortion * toroidal magnetic fields ( @xmath77 g ) can distort a rotating ns into a prolate shape @xcite . such configuration is unstable to the growth of the angle between the ns s angular momentum and the magnetic axis . this angle increases until the angular momentum and magnetic axis are orthogonal . the resulting configuration is a rotating , non - axisymmetric body that is an efficient emitter of gws @xcite . fast rotating magnetars can also lose spin energy via magnetic dipole radiation and/or magnetized , relativistic winds . nevertheless , observations indicate that such energy losses are not typical ( @xcite and references therein ) , suggesting that gw emission may be relevant in the early phase of newly born magnetars . given that magnetic distortions indeed result in efficient gw emission , dallosso et al . @xcite find that such a gw signal can be detected out to the virgo cluster with advanced detectors . besides their initial rotation , protoneutron stars can accrete material from the infalling matter after core collapse @xcite . it is possible that nss radiate away this angular momentum that they gain from accretion via gws , e.g. , via dynamical and/or secular instabilities discussed above . such gw signal may be detectable to @xmath78mpc with advanced detectors @xcite . upon the core collapse of a massive star ( @xmath79m@xmath80 @xcite ) , a plausible scenario is the formation of a central bh surrounded by an accretion disk @xcite . such a bh - torus system can be the source of copious gw emission if the disk assumes a finite quadrupole structure due to non - axisymmetric instabilities @xcite . the emergence of such a non - axisymmetric structure on the time scales comparable to the lifetime of the accretion disk requires a stellar progenitor with sufficiently high angular momentum . a high rotation rate is also required for the creation of a grb ( e.g. , @xcite ) . below we outline some of the possible scenarios through which non - axisymmetric instabilities in accretion disks may result in strong gw emission , and address whether existing or future gw detectors can probe that signal . as an example to the sensitivity of a gw search to accretion - disk non - axisymmetric instabilities , we consider a 3m@xmath14 bh with a @xmath81m@xmath14 clump in its accretion disk . such a system emits gws similar to an isolated low - mass binary coalescence , and may be seen to a horizon distance ( for definition see section [ section : allsky ] ) @xcite @xmath82 for an advanced detector with optimal source position and orientation ( this distance is reduced by a factor 2.26 for average sky location and orientation ) , where @xmath83 is the so - called _ chirp mass _ ( @xmath84m@xmath14 for the example above ) . 1 . * disk fragmentation via gravitational instability * accretion disks with sufficiently large angular momenta are gravitationally unstable @xcite . a gravitationally unstable disk will fragment if the disk cooling time is sufficiently short ( @xmath85 orbital period ) @xcite . relevant accretion disks cool rapidly , e.g. , via neutrinos or strong winds @xcite , favoring fragmentation , although the emergence of magnetorotational instability ( mri ) in the disk can heat the disk through ohmic dissipation @xcite . the resulting fragmented disk will emit a strong , chirp - like gw signal @xcite . viscosity and gw emission drive the angular momentum loss of the disk . consequently , low disk viscosity favors stronger gw emission , which may be detectable from @xmath73mpc with advanced ligo / virgo @xcite . the gw frequency at its highest snr is probably in the @xmath86hz range , depending on the strength of viscous forces . the duration of gw emission may be similar to that of gamma - ray emission . furthermore , the tidal disruption of fragments in the accretion disk may be behind some of the x - ray flares observed in grb afterglows @xcite . accretion disk fragmentation has been observed in approximate numerical simulations @xcite . it is not clear , however , whether such a fragmentation would occur in more realistic cases as well , unless the pressure support of the star is drastically reduced , e.g. , via neutrino losses @xcite . * non - axisymmetric structure via papaloizou - pringle instability * differentially rotating accretion disks can be subject to global non - axisymmetric instabilities . one such instability , discovered by papaloizou & pringle @xcite , develops in accretion disks in which azimuthal pressure gradients ( due to high internal temperature ) give rise to differential rotation @xcite , i.e. the angular velocity @xmath87 of the disk depends on the radius @xmath55 . in such disks , large - scale spiral pressure waves can emerge with fixed pattern speed @xmath88 . if there is a radius @xmath89 , the so - called corotation radius , at which the pattern speed is equal to the disk rotational rate ( i.e. @xmath90 ) , the wave within @xmath89 will propagate slower than the rotation of the disk , in fact decreasing the mechanical energy of the disk , resulting in a so - called _ negative energy wave _ ( e.g. , @xcite ) . such negative energy wave can interact , at the corotation radius , with the _ positive energy wave _ that develops at @xmath91 . the negative energy wave can increase its amplitude by _ losing _ energy , which in turn can increase the amplitude of the positive energy wave , thus feeding the instability . see , e.g. , @xcite for an expressive description of the phenomenon . + the papaloizou - pringle instability gives rise to an ( @xmath92 ) non - axisymmetric structure on a dynamical time scale ( i.e. over a time period comparable to the rotation period ) . such non - axisymmetric structure can persist for much longer than the dynamical time scale , resulting in strong gw emission . the papaloizou - pringle instability and the resulting non - axisymmetric structure have been observed in 3d relativistic simulations of bh - torus systems @xcite . these simulations indicate that bh - accretion disk systems subject to the papaloizou - pringle instability emit gws in the @xmath86hz frequency range that may be detectable up to @xmath73mpc with advanced ligo / virgo @xcite . strong magnetic fields present in the accretion disk can enhance the instability for thick disks ( and may suppress it for thin disks ) @xcite . nevertheless , we note that the papaloizou - pringle instability was found in simulations for initially axisymmetric tori @xcite . such complete axial symmetry may not develop in compact binary mergers . numerical simulations of compact binary mergers have not yet shown signs of the development of the papaloizou - pringle instability ( e.g. , @xcite ) . * suspended accretion * in order for accretion around a central , rotating bh to continue for the time scales of long grbs , it has been suggested that accretion may be suspended " ( i.e. slowed down or temporarily halted ; e.g. , @xcite ) . such suspended accretion would be achieved through magnetic fields , which transfer some of the rotational energy of the bh to the disk . a fraction of this rotational energy could then be radiated away through gws @xcite . in order to emit gws , the accretion disk needs to lose its axial symmetry , e.g. , through fragmentation . gw radiation from suspended accretion has been suggested to carry away as much as @xmath93m@xmath14c@xmath53 in the sensitive frequency band of ligo / virgo - like interferometers over a duration comparable to the duration of long grbs ( @xmath76s ) @xcite . suspended accretion , nevertheless , requires highly ordered accretion , which may be difficult to achieve due to the development of mri heating or disk turbulence ( e.g. , @xcite ) . simulations to date ( e.g. , @xcite ) have not provided support for suspended accretion , even in the presence of @xmath94 g magnetic fields . in very rapidly rotating stars , infalling matter may fragment even before the formation of a bh - torus system . it has been suggested that , in analogy with observed phenomena in star formation , the collapsing core may fragment and form two or more compact objects @xcite . such fragmentation was observed in relativistic numerical simulations of approximate pre - supernova cores @xcite . such core fragmentation would give rise to strong , characteristic gw emission , similarly to the case of binary mergers . nevertheless , the rotation rate necessary for such a fragmentation seems to be difficult to achieve with current stellar models @xcite . highly magnetized neutron stars ( magnetars ) are thought to be the engine behind soft gamma repeaters ( sgrs ) and anomalous x - ray pulsars ( axps ) @xcite . some magnetars occasionally produce so - called _ giant flares _ that resemble short grbs ( and references therein ) . giant flares are much less energetic than typical short grbs that are at cosmological distances , hence they are only detectable from the milky way or nearby galaxies , up to @xmath95mpc with current instruments . the lack of excess short - grb population from the direction of these galaxies indicates that magnetars may only be responsible for a small sub - population of short grbs @xcite . giant flares are much more common ( with much weaker gamma emission ) than other types of grbs , as indicated by the rate of observed galactic and nearby giant flares . due to their high rate , they may be detectable with gw observatories even though they are much weaker sources of gws than other , extragalactic grbs . giant flares ( as well as sgr activity or abrupt changes in the ns spin period ; @xcite ) are thought to be the result of so - called _ starquakes _ : the tectonic activity ( cracking ) of the ns crust @xcite , which is followed by the reconfiguration of the ns s magnetic fields @xcite . starquakes induce seismic vibrations in the ns , causing the observed quasi - periodic oscillations ( qpos ) in the x - ray tails of giant flares . alternatively , it has been suggested @xcite that qpos can not be driven by oscillations in the ns crust due to the quick ( @xmath75s ) dissipation of the vibration via alfvn waves into the neutron star interior . nevertheless , the similarities between the statistical properties of sgr events and earthquakes may provide further evidence for the crustal origin of sgr events @xcite . ns seismic vibrations result in the emission of gws @xcite , although this emission may be weak @xcite . in the optimal scenario for gw emission , magnetic reconfiguration can liberate @xmath96erg of crustal elastic energy , which is the upper limit for the energy radiated away via gws @xcite . this energy can be even higher if the ns is of strange quark matter @xcite . quasi - periodic oscillations ( qpos ) with various frequencies from @xmath97hz and durations up to @xmath73s have been observed in every giant flare x - ray afterglow so far . some of these qpos may be connected to the seismic modes of the nss . of special interest are qpos around @xmath73hz , which fall into the most sensitive frequency band of ligo / virgo . possible gw emission of @xmath98erg at this frequency would be detectable out to @xmath1kpc with advanced ligo / virgo ( given 10@xmath99 sensitivity improvement compared to initial detectors ) @xcite . it is possible that gws are emitted for a significantly longer time ( days to months ) than the observed electromagnetic qpo @xcite , making longer - term gw searches necessary . a fraction of the energy from magnetic reconfiguration may excite ns fundamental quadrupolar fluid modes ( _ f modes _ ; e.g. , @xcite , although see @xcite ) . if ns @xmath56 modes are excited , they are damped by gws on a very short time scale ( @xmath100ms ; @xcite and references therein ) , which is shorter than most other potential damping mechanisms . therefore , most of the energy in the @xmath56 modes may be radiated away via gws @xcite . ns @xmath56 modes oscillate at @xmath101khz ( for stiff eos ) , not too far from the most sensitive band of advanced ligo / virgo @xcite . given the most extreme case in which @xmath102erg being transferred into @xmath56 modes that oscillate at @xmath75khz , the resulting gw could be detected up to @xmath103mpc with advanced ligo / virgo ( given 10@xmath99 sensitivity improvement compared to initial detectors ) @xcite . stacking " the gw signals from multiple events can extend this distance even farther @xcite . if the excitation energy is comparable to the observed giant flare energy ( @xmath104erg ) and if not all energy is radiated away through gws , only galactic sources would be detectable with advanced gw detectors . furthermore , recent magnetohydrodynamic simulations of magnetized nss indicate that only a small fraction of the released energy is converted to @xmath56 modes , which would make the detection of @xmath56-mode gws less likely @xcite . while the commonly favored model for the central engine behind grbs is a rapidly accreting bh @xcite , there are observational indications @xcite that a fraction of short grbs may originate from highly magnetized , rapidly spinning protoneutron stars , so called _ millisecond protomagnetars _ @xcite . further , outflows driven by such protomagnetars could also lead to the creation of long grbs @xcite . the emergence of a millisecond protomagnetar from various cosmic events is therefore a plausible step in the creation of some grbs . the possibility that some grbs are driven by rapidly spinning protomagnetars opens the door to alternative grb progenitors . while traditionally considered progenitors , i.e. ( i ) the core collapse of massive stars @xcite and ( ii ) ns - ns mergers , can themselves lead to the creation of millisecond protomagnetars , other cosmic events , such as ( iii ) the accretion - induced collapse of white dwarfs @xcite , or the merger of two white dwarfs @xcite , are also plausible progenitors . millisecond protomagnetars may emit a sizable fraction of their rotational energy via gws @xcite , although they may also lose angular momentum via magnetic processes ( e.g. , @xcite ) . fast rotation can result in the emergence of rotational instabilities , especially if the protoneutron star is differentially rotating ( see also section [ section : protoneutronstar ] ) . if a millisecond magnetar indeed drives a grb , it needs to maintain fast rotation for durations comparable to the duration of the grb , potentially allowing sufficient time for rotational instabilities to emerge and result in gw emission . if a significant fraction of the rotational energy ( @xmath51m@xmath52c@xmath53 ; e.g. , @xcite ) is converted into gravitational radiation , the resulting gw signal could be detected to a distance of @xmath105mpc with advanced detectors ( see also section [ section : protoneutronstar ] ) . the rotation rate of millisecond magnetars , and therefore the emitted gw strength and frequency , likely depends on the cosmic event they originate from . accretion - induced collapse of white dwarfs that accrete mass and angular momentum from a non - degenerate companion may produce the most rapidly rotating protomagnetars . due to continued accretion , such white dwarfs will be rapidly rotating upon collapse , resulting in a rapidly rotating protomagnetar , and a strong gw signal . on the other hand , the remnant of a binary white dwarf merger will likely collapse to a magnetar with a significant delay ( @xmath106yr ) , allowing the remnant to lose most of its angular momentum via stellar winds @xcite . binary white dwarf mergers are therefore not likely to be significant sources of gws . this section addresses the question : what can the detection of gw signals add to our understanding of the physics of grbs and their progenitors ? detectable electromagnetic radiation , our main source of information on grbs , is observable only from relatively large distances ( @xmath107 cm ) from the central engine . on the other hand , gws are created right at the central engine , and can convey information about it without being distorted or absorbed by matter on their way to the observer . here we outline some of the questions of interest that could be addressed through observing the gw signature of grbs . 1 . * progenitors of short grbs * as compact binary mergers will be detectable from hundreds of mpc with the advanced ligo - virgo network @xcite , within these distances it will be possible to confirm or rule out compact binaries as grb progenitors . while identification based on electromagnetic signals is not always straightforward @xcite , the detection of gws will be likely essential to unambiguously reveal the nature of the progenitor @xcite . by connecting each measured source with the presence or absence of detectable gw emission , and by exploring the connections between observed electromagnetic counterparts and their binary progenitors , the central engine will be probed in unprecedented ways . for example , observing grbs in coincidence with those binaries for which numerical simulations predict the formation of accretion disks would indicate a connection between accretion disks and grbs . already , analyses of gw data from initial detectors targeting gamma - ray bursts have helped rule out binary mergers as the central engines of selected events @xcite . * population prospects * the populations of short grbs @xcite , as well as compact binaries @xcite , are highly uncertain . the observation of gws from compact binary coalescences could be an effective way of determining the source population . due to the very large field - of - view of gw detectors and the very weak beaming of gw emission , they can be used efficiently to locate practically all sources within their well - determined horizon distance of hundreds of megaparsecs . as source distances can also be determined using gws ( see next subsection ) , the binary coalescence population could also be mapped as a function of redshift , although only with the sensitivity of third - generation gw detectors @xcite . * cosmological parameters * the detection of a binary coalescence gw signal from an observed short grb could be used to determine the distance and redshift of the source @xcite . in fact , the determination can be done independently of the cosmological distance ladder . distances of short grbs can be reconstructed to a @xmath108 precision for @xmath109mpc for ns - ns , and for @xmath110gpc for ns - bh binaries using the advanced ligo - virgo detector network @xcite . reconstructed source distances can then be used to accurately reconstruct the luminosity of short grbs , as well as to test hubble relation @xcite , with one year of observation with advanced detectors potentially allowing for a @xmath111 precision @xcite . further , in combination with measurements of the cosmic microwave background , reconstructed binary merger distances could be used to constrain the dark energy equation of state @xcite . * jet angular structure * the angular structure of relativistic jets from short grbs is poorly constrained @xcite . the rate of binary mergers detected through gws , together with the rate of short grbs for which a binary progenitor can be confirmed with gws , could be used to determine the opening angle of short grbs . + further , the polarization of the inspiral gw signals in principle could be used to characterize the viewing angle of observed grbs @xcite . the gw polarization from a binary inspiral depends on the viewing angle compared to the rotational axis of the binary . towards the rotational axis , the gw signal is circularly polarized , while the polarization becomes elliptical for off - axis observers , eventually becoming linear for observers in the equatorial plane . for sufficiently strong gw signals from the inspiral phase , reconstructing the eccentricity of the gw polarization could provide information on the opening angle . with a large number of measurements , the angular structure of the jet could be mapped as well @xcite . these correlations can be compared to similar constraints that might follow from the detailed multiband light curve of the blastwave @xcite . nevertheless , due to the relatively weak change in gw polarization with angle , the gw polarization may only be identified for large ( @xmath112 ) angular differences , and/or for very high snr ( @xmath113 ) , making the utility of gw polarization limited . * neutron star equation of state * nuclear forces have a profound influence on the structure of nss ( e.g. , @xcite ) . they determine , among other things , the relation between ns mass and radius , or the ns mass limit ( for the rotating and non - rotating cases ) . the eos of matter is poorly constrained at ns densities @xcite . the observation of nss with masses up to @xmath114m@xmath14 @xcite imply a stiff eos ( a ns with soft eos would not be able to support this much mass ) . + the evolution of ns - ns / bh - ns mergers strongly depends on the nss nuclear eos . therefore , the gw signal from binary coalescences can be used to determine / constrain the eos . for instance , if a ns is tidally disrupted during a binary merger , the orbital frequency at which tidal disruption occurs can be used to determine the radius of the ns . this , together with the ns mass reconstructed from the inspiral gravitational waveform @xcite , can be used to constrain the nuclear eos @xcite . + the tidal deformation of nss in a binary system can affect the gravitational waveform even prior to the merger phase . the gw energy spectrum during the last few orbits of a ns - ns binary prior to merger can be used to determine the compactness ratio ( mass / radius ) of the nss @xcite . for these last few orbits of the binary , the gw frequency is at a high but still reasonably sensitive frequency for ligo - type detectors , making the analysis of these last few orbits feasible . + the tidal deformation of a ns in the inspiral phase can be described by one parameter ( the so - called _ love number _ @xcite ) , which is effectively the ratio of the star s induced quadrupole moment to the quadrupole moment of the perturbing tidal gravitational field of the binary companion . flanagan & hinderer @xcite showed that the nuclear eos of nss can be constrained even through the early inspiral phase due to the effect of tidal deformation on the waveform . read et al . @xcite simulated the inspiral phase of ns - ns binaries , showing that the ns eos can be constrained ( the ns radius can be determined to within @xmath115 km precision , which , together with the ns mass , would rule out a part of the eos parameter space ) with advanced gw detectors for a source at a distance of @xmath116mpc . bh - ns mergers can similarly be used to gain information on the ns eos @xcite . + for some ns - ns mergers , collapse to a bh is delayed and a hypermassive ns is formed ( see section [ subsubsection : mergerphase ] ) . during the merger , non - axisymmetric oscillation modes of the hypermassive ns are excited , resulting in gw emission @xcite . the waveform of the resulting gravitational radiation depends on the nuclear eos . consequently , the observed waveform can be used to significantly constrain the eos @xcite . the gw signature of hypermassive - ns oscillations could be detected with snr@xmath117 out to @xmath118mpc ( depending on the eos ) , which may be sufficient to constrain the eos ( at such distances , the source can be confirmed with high accuracy via gws from the inspiral phase ) @xcite . + further , it is possible that the inner part of a ns becomes strange quark matter at high densities . a star made of strange quark matter can be self - bound , a marked difference from hadronic nss that are gravitationally bound @xcite . the gw emission of a strange quark star is substantially different from that of hadronic nss @xcite , hence gws could be used to determine whether ns interiors may contain strange quark matter @xcite . 6 . * magnetic fields in protoneutron stars * magnetic fields are suspected to play a significant role in the origin of short grbs @xcite , and are known to vary between individual pulsars ( e.g. , from their spindown rate @xcite ) . in the inspiral phase , internal magnetic fields in nss ( for realistic field strengths ) have negligible effect on the dynamics of the binary coalescence @xcite . while magnetic tension in the ns can reduce tidal deformation , this effect will be too weak to be detectable with planned gw detectors for realistic pre - merger magnetic strengths ( @xmath119 g ) @xcite . for bh - ns binaries , magnetic fields seem not to alter the dynamics and therefore the gravitational waveform @xcite . + magnetic fields can significantly affect the post - merger behavior of ns - ns binaries , if a hypermassive ns is formed after the merger . upon merger , magnetic fields will be amplified via kelvin - helmholtz instabilities in the shear layer between the merging nss @xcite , or later via differential rotation @xcite . even if magnetic fields are small prior to merger , the amplified fields can be strong enough to affect the evolution of the merger remnant and therefore gw emission . magnetic fields in the hypermassive ns compete with gws in dissipating angular momentum from the hypermassive ns . the decrease of angular momentum eventually results in the delayed collapse " ( @xmath1ms@xmath120s , depending on the mass and the eos ) of the hypermassive ns into a bh ( e.g. , @xcite ) . + by probing the magnetic properties of a large number of nss in binary mergers , one could determine , e.g. , the highest achievable magnetic field frozen in a stationary ns ( @xcite ; e.g. , via the time difference between binary merger and grb ) . * accretion disks * the large - amplitude , quasiperiodic gw emission from instabilities in the accretion disk , probably together with the detection of the earlier inspiral phase from the binary , can be utilized to reconstruct properties of the accretion disk , even if no electromagnetic signal is observed from the source . beyond reconstructing the mass and lifetime of the accretion disk , the development of the instabilities is sensitive to the properties of the accretion disk ( see section [ section : accretiondiskinstabilities ] ) , i.e. the development of the instabilities by itself can already be informative . 1 . * progenitors of long grbs * it is believed that the origin of at least some long grbs are the core collapses of massive stars . this could be confirmed if the gws in coincidence with a long grb progenitor were detected . nevertheless , if no gw emission is detected in coincidence with long grbs , this does not rule out the core - collapse model , although may constrain core - collapse dynamics . if gws are detected in coincidence with a long grb , the internal evolution of a core collapse , which may largely depend on the progenitor , could also be examined . for example , the formation of a protomagnetar or a bh - torus system , or the fragmentation of the massive stellar core may be differentiated via gws . neutron star equation of state * the mass , radius and rotation rate of nss have a profound effect on their potential gw emission through rotational instabilities . detecting the gw signature of these instabilities can be used to constrain the mass - radius relation , rotation , and therefore the eos , of nss ( e.g. , @xcite ) . + further , rotational instabilities in nss result in qualitatively different gw emission for conventional and strange quark nss @xcite . whether a ns is composed of strange quark matter could be inferred from the detected gw signature of the ns s rotational instability . * neutron star internal physics * the evolution of the gravitational waveform from protoneutron star rotational instabilities strongly depends on the nuclear eos at ns densities , as well as the physical parameters of the protoneutron star . differential rotation inside a protoneutron star , temperature ( e.g. , @xcite ) , viscosity and neutrino cooling @xcite may all leave their mark on the evolution of rotational instabilities and the resulting gw emission . further , if a hypermassive ( or supramassive ; see section [ subsubsection : mergerphase ] for their definition ) ns collapses into a bh due to either losing angular momentum or accretion , the rotational frequency prior to collapse provides information on the ns eos @xcite . this rotational frequency may be inferred from the gw signal from the rotating ns @xcite . * accretion physics * because gw emission from massive stellar collapses can be produced by the accretion onto the central object following the collapse , the gw signal therefore can provide insight into accretion physics : ( i ) the mechanisms that transport angular momentum ; ( ii ) instabilities ; and ( iii ) interaction with the black hole . + for bh - torus systems the strength of viscosity ( and possibly the strength of other processes through which the torus loses angular momentum ) can be determined , as these are competing effects for angular momentum loss : the total loss of angular momentum can be inferred from the rate of change of the gw frequency , while the loss through gws can be inferred from the gw signal amplitude . + accretion of matter by the protoneutron star results in extended gw emission due to the intake of angular momentum @xcite . as the time scale and nature of accretion is likely different for accretions from supernova fallback material or material from a companion star , the extended gw signal from rotational instabilities will carry important information on the accretion mechanism . 5 . * magnetic fields in neutron stars * sufficiently strong magnetic fields inside a protoneutron star can be competing with gws in radiating angular momentum away . as ns spindown may be reflected in the gw frequency , measuring the gw amplitude and ns spindown may provide information on the strength and nature of magnetic fields present in the ns . the detection of gws in coincidence with a giant flare from a magnetar could provide information on the processes that lead to flaring in magnetars ( i.e. tectonic activity @xcite or global reconfiguration @xcite vs origin from the magnetosphere @xcite ) . non - detection , nevertheless , does not necessarily rule out these models . upon detection , the frequency of gws from nss excited by , e.g. , starquakes , could be used to infer the ns mass , radius and eos @xcite , or potentially even the strength of magnetic fields inside the ns @xcite . the detection of millisecond magnetars gw signature in coincidence with a grb would confirm their presence and role as the central engine behind some grbs . further , the gw signature of millisecond magnetars may indicate their origin ( e.g. , typical rotation rates may differ for different mechanisms ; see section [ section : millisecondmagnetars ] ) . advanced gw detectors will also be able to determine , out to @xmath121mpc , whether the progenitor of a short grb was a binary merger of a ns with another ns or a bh @xcite . the confirmation of a compact binary merger would rule out the other scenarios in which millisecond magnetars can form , e.g. , white dwarf accretion - induced collapses , or white dwarf binary mergers . the success of detection of gws from grb progenitors depends on the strength of the emitted gws , the sensitivity of gw observatories , as well as the observation strategies used to separate gw signals from the background . the search strategy , e.g. , the use of multiple messengers , can also add to the information one can extract from the source . various transient gw search strategies exist , aiming to find the gw signature of grbs . these strategies differ mainly in their prior assumptions on the gravitational waveform . in the most general case , one can look for so - called gw bursts with minimal assumptions on the waveform , namely defining a maximum signal duration and bandwidth @xcite . these generic burst searches can be modified to include additional information about the source . for instance if one considers an accretion - type emission , circular ( on - axis ) or linear ( off - axis ) polarizations can be required from gw signals , increasing detection sensitivity for signals satisfying these conditions . other model - dependent assumptions include , e.g. , gw signals rapidly evolving characteristic frequency @xcite , such as in the case of compact binary mergers . for compact binary coalescences , the gravitational waveform for a given mass configuration can be accurately calculated for the initial inspiral and final ringdown phases , and to some extent the merger phase @xcite . for such systems with properties in a highly limited parameter space , it is beneficial to use a signal waveform template bank and perform matched filtering based analysis ( e.g. , @xcite ) . below we review the gw search strategies targeting the progenitors of grbs . we indicate the prospects of such searches through presenting projected sensitivity estimates based on observational results with initial detectors . advanced detectors will be @xmath122 more sensitive than initial detectors , the exclusion distances therefore will be @xmath122 greater , corresponding to a @xmath251000-fold improvement in achievable source rate constraints . further , advanced detectors will be sensitive in a wider frequency band . additionally , the advanced detector network will eventually be larger , with the inclusion of kagra and probably ligo india . these improvements will further increase the achievable sensitivity of the advanced detector network @xcite , and will further improve signal parameter reconstruction ( e.g. , source direction ; @xcite ) @xcite . search strategies for gw transients can be divided into four main categories based on their use of other messengers . 1 . * untriggered * searches , in which no information is used from other messengers . the analyses use gw data alone to find astrophysical sources ( e.g. , @xcite ) . externally - triggered * searches , in which one specifically looks for gw signals from sources confirmed via other messengers . for instance one can search for gw signals from the progenitors of detected grbs , using their location , time , and other parameters ( e.g. , @xcite ) . * electromagnetic follow - up * searches , in which gw signal candidates are used to trigger electromagnetic follow - up searches with other telescopes ( e.g. , @xcite ) . * multimessenger * searches , in which one uses ( typically sub - threshold ) signal candidates from multiple observatories of different messengers in a joint search ( e.g. , @xcite ) . in the following we briefly review some of the gw search strategies and past gw searches for grbs , organized along the above categories . untriggered transient searches provide a generic way to identify plausible gw sources without relying on the detection of other messengers . while grbs are highly beamed ( e.g. , @xcite ) , the beaming of gw emission from grb progenitors is weak ( e.g. , @xcite ) . for realistic grb beaming angles , there will likely be more observations of gws from off - axis grb progenitors than gw observations triggered by grbs @xcite . to help quantify the source population detectable with untriggered ( and other ) searches , it is useful to characterize the sensitivity of a gw observatory to a gw source with the so - called _ horizon distance _ @xmath123 . the horizon distance is defined as the distance at which a gw source from optimal orientation ( such that emission towards the earth is the strongest ) and optimal location ( i.e. the most sensitive direction of the gw detector ) is observed with a given snr ( typically chosen to be @xmath124 for untriggered searches ; see also section [ subsubsection : extrig ] ) by a single gw detector . compared to this optimal orientation and location , the orientation and direction - averaged distance at which the source produces the @xmath124 is @xmath125 ( both direction and orientation - averaging contribute a factor @xmath126 ) . for a ns - ns binary with 1.35 m@xmath14 mass for each ns , the projected horizon distance for untriggered searches with advanced detectors ( upon non - detection ) is @xcite @xmath127 where @xmath83 is the so - called chirp mass ( i.e. the horizon distance only depends on this specific combination of the masses of the two compact objects ) . for higher - mass binaries , the horizon distance is somewhat lower than the prediction of eq . [ eq : dh ] ( see @xcite ) . external triggers in gw searches provide additional information and increased search sensitivity for gw analyses @xcite . for example an external grb trigger reduces the temporal and spatial extent in which one has to search for a gw signal @xcite . a detected external trigger can also be used to set constraints on the gw signal . for example a short grb implies that the gw signal is probably a binary merger . chen & holz @xcite estimates that for grb beaming factor @xmath128 ( corresponding to opening angle @xmath129 ) , the number of triggered and untriggered detections of gws from grb progenitors would be comparable . further , even for greater beaming factors ( i.e. smaller opening angles ) , externally triggered searches will enable the detection of some additional , particularly interesting , sources . the extra information in external triggers can be taken into account in the horizon distance by effectively lowering the snr threshold ( i.e. by keeping the false - alarm rate constant ) . the results of chen & holz imply that the sensitivity of externally triggered searches is greater than that of untriggered searches by a factor @xmath130 ( for inspiral searches ; this factor is similar for other types of searches as well ) , i.e. @xmath131 the same sensitivity increase was found by dietz et al . @xcite . while this increased horizon distance means that many more sources fall within observable reach , the actual number of detected sources will be decreased by the beaming of the electromagnetic ( or other ) signal , as well as the sky coverage of the available electromagnetic telescopes . on the other hand , all triggered sources will be detected face - on , while untriggered sources can take any orientation , which adds to the sensitivity of triggered searches compared to untriggered searches ( for binary mergers , the gw amplitude is @xmath126 times greater face - on than averaged over all directions ) . chen & holz @xcite find that the externally - triggered observation rate is a sizable addition , although less than , the untriggered gw observation rate , while kelley et al . @xcite find that grb - triggered searches present no appreciable addition to the total rate @xcite . while both of these studies approximate gw background as purely gaussian , real gw background features a non - gaussian tail that may increase the importance of externally triggered and follow - up searches . previous grb - triggered gw searches include searches for both unmodeled gw bursts and compact binary mergers . for example the short grbs 070201 and 051103 had electromagnetic positions overlapping nearby galaxies ( andromeda and m81 , respectively ) . no binary merger counterpart was found in either case , ruling out a binary progenitor in those galaxies , favoring an sgr model ( a binary progenitor from a more distant galaxy is not ruled out , albeit it is unlikely ) @xcite . nevertheless , it is worth noting that , given their expected rate of occurrence , a binary merger within the distance of andromeda or m81 ( @xmath132mpc ) would be extremely unlikely . for grb - triggered gw searches with initial ligo / virgo , see also , e.g. , @xcite . several gw searches aimed to identify gws in coincidence with sgr flares @xcite . most recently abadie _ et al . _ @xcite used 1279 flares from six magnetars as triggers , aiming to identify gw signals from neutron - star @xmath56-mode ringdowns or other gw - producing mechanisms . the search set constraints on the energy of gw emission from these flares comparable to some giant flares electromagnetic energies , and an order of magnitude below previously existing limits . for a nearby magnetar ( sgr 0501 + 4516 ) located at @xmath133kpc that emitted a large number of flares during the observation period , the obtained upper limit on @xmath56-mode gw emission was @xmath134 erg . the upper limit for gw emission at @xmath135 hz , i.e. within most sensitive frequency band of ligo / virgo , was @xmath136 erg , a promising limit compared to some theoretical upper limits on gw emission from sgrs @xcite . with advanced ligo / virgo , one can expect , beyond more stringent limits on @xmath56-mode emission , constraints on the maximum emissible _ total _ energy from other ns oscillation modes ( we note that these upper limit are constraining only if a substantial fraction of total energy goes into gws , which is debated ; e.g. , @xcite ) . with the rise of a global gw detector network , it became possible to reconstruct the direction of a gw signal candidate , albeit with a relatively large uncertainty . such reconstruction enables the use of gw signal candidates in triggering electromagnetic follow - up observations @xcite . there are various scientific advantages of such a follow - up search : 1 . as electromagnetic observations can not continuously cover the whole sky with high sensitivity , gws can be used to guide these telescopes and point towards the right direction at the right time . often the most interesting and strongest emission from an electromagnetic transient occurs in a short period after its onset . since gw emission is , in many cases , expected to precede the onset of electromagnetic emission , telescopes have the chance to commence follow - up observation in a very early emission stage , or even catch the onset of the electromagnetic event . 3 . gw signal candidates will often have too low significance to be unambiguously identified as extraterrestrial signals . the observation of an electromagnetic follow - up event can enhance the significance of the joint observation , making detection more probable . 4 . similarly to other multimessenger searches , information from the different messengers can enhance the information ( and therefore science ) that one can extract from the source . electromagnetic follow - up searches of gw event candidates is one of the most promising and one of the fastest - evolving areas of gw astrophysics ( see for recent observational development with the ligo - virgo gw detectors ) . see section [ section : emcounterpart ] below for a more detailed description of the potentially detectable electromagnetic counterparts and search strategy . the sensitivity of electromagnetic follow - up searches will differ from the sensitivity of externally triggered searches due to ( i ) the limited reach of electromagnetic transient observations and ( ii ) the finite sky area that is scanned by observatories following up the gw event candidate . taking ( i ) into account in defining the horizon distance , one gets @xmath137 where @xmath138 is the reach of the electromagnetic follow - up observatory . note that @xmath139 refers to on - axis sources , while follow - up observations are the most interesting for off - axis sources ( due to the greater source rate ) . for off - axis directions , gw emission is typically weaker by a factor @xmath126 . the rate of follow - up observations further depends on sky coverage of the follow - up observatories , as well as the beaming of the electromagnetic signal . for a survey of the electromagnetic detectability of gw event candidates , see @xcite . multimessenger analyses generalize the idea behind externally triggered searches by combining information from sub - threshold signal candidates of various messengers into one , more powerful analysis . while in many cases each type of sub - threshold messenger would individually have too low significance to be identified as an astrophysical signal of interest , the combination of individual significances can greatly enhance the analysis potential of identifying astrophysical events ( see section [ chapter : gwhensearch ] for a multimessenger search ) . the _ astrophysical multimessenger observatory network _ ( amon ) initiative @xcite plans to combine sub - threshold triggers of multiple messengers in a low - latency coherent multi - messenger analysis . amon is planned to combine multiple messengers to identify candidates that would be sub - threshold events for individual observatories . the extracted information would be used , besides claiming detection itself , to initiate electromagnetic follow - up observations , which can add to the number of coincidentally observable messengers . the sensitivity of multimessenger searches depends on the false alarm statistics of the different detectors included in the search , as well as the types of messengers . transient messengers with amplitude signatures , such as gws or high - statistics photon signals , will have a relatively well defined horizon distance , beyond which the detection of such a messenger is highly unlikely . other , discrete messengers , such as high - energy neutrinos , will have no such horizon distance , since there is a finite probability of detecting at least one neutrino even from a very distant source with expected signal strength @xmath140 neutrino . these distant sources can significantly contribute to the total number of detectable multimessenger sources , especially in the case in which their presence can be confirmed with other messengers ( see , e.g. , @xcite ) . for such discrete messengers , the gw - multimessenger horizon distance is comparable to that of externally triggered searches : @xmath141 the rate of detected multimessenger sources will be limited by beaming , and the probabilistic nature of observing at least 1 discrete messenger ( e.g. , high - energy neutrino @xcite ) . this section outlines the observable electromagnetic emission of grb progenitors , focusing on counterparts other than gamma rays , and their utility in gw searches . due to the highly beamed emission of most grbs . ] , the majority of grbs are off - axis , i.e. their prompt emission can not be observed from the earth . nevertheless , weaker , off - axis electromagnetic emission from grb progenitors could be detected during gw follow - up observations , if the electromagnetic emission is present beyond a few minutes after the prompt emission ( i.e. the delay between detecting a gw signal candidate and pointing a telescope towards the expected source location ) . most compact object mergers observed by ligo / virgo will not be accompanied by observable grbs due to the small gamma - ray beaming angle . for this reason , the last few years have seen extensive investigation into rapid follow - up search strategies for more isotropic , but more difficult to identify , electromagnetic counterparts . follow - up searches aim to minimize the time between the identification of a gw event candidate and the start time of the observation with electromagnetic telescopes in the required directions . the time delay of sending a request for follow - up search to electromagnetic telescopes after a gw event was @xmath25tens of minutes for initial searches ( mostly due to the manual confirmation of event selection ) . this delay will likely be reduced to @xmath25minutes for searches with advanced detectors @xcite . the uncertainty in direction reconstruction of gw event candidates with the planned advanced detector network will likely be in the @xmath25tens of square degrees range ( e.g. , @xcite ) . with such an uncertainty , telescopes with large fields of view will need to be utilized , focusing on electromagnetic emission that is ongoing for hours - days so the telescopes can scan through the interesting sky area . existing galaxy catalogs can also be used to largely decrease the sky area that needs to be mapped ( see section [ subsection : hostgalaxies ] ) . we note here that , besides gw observations , the observation of the prompt gamma - ray emission itself has a position resolution of a few degrees , i.e. comparable to the resolution of gw observations . afterglow observations are therefore also important for the direction reconstruction of observed grbs . beyond direction reconstruction , grb afterglows also promise to carry important information connected to gw radiation . corsi & mszros @xcite pointed out that the observed x - ray plateau in the afterglow of some grbs may be connected to the formation of a highly magnetized millisecond pulsar that can lose angular momentum through gws during a 10@xmath142s period ( see also @xcite ) . several promising emission models have been identified which predict longer - lived , weakly ( or not at all ) beamed emission that can be detected on a distance scale comparable to the sensitivity range of advanced gw detectors . in the following list of these models , we focus on optical counterparts of compact binary coalescences . for a survey of the detectability of some of these emission models , see , e.g. , @xcite . 1 . * electromagnetic remnants from compact binary mergers * ns - ns and ns - bh mergers can eject energetic ( sub - relativistic or relativistic ) outflows . these energetic outflows can interact with matter in the interstellar medium , creating a long - lasting radio signal @xcite . radio remnants from dynamically ejected sub - relativistic material can appear months to years after the binary merger , and can last for several years @xcite . a deep radio survey can identify these radio remnants out to @xmath143mpc @xcite . mildly / ultra - relativistic outflows may create even brighter radio emission , on the time scale of weeks @xcite . the detectability of the radio transient strongly depends on circum - merger density , and therefore the signal may be lower for binaries that left their galaxy prior to merger . + an interesting case is shocks generated within the nss upon the merger of a ns - ns binary @xcite , which can plausibly drive ultra - relativistic outflows . these shock waves , besides heating the nss , result in shock breakout from the nss surface , driving a nearly omnidirectional , ultra - relativistic outflow @xcite . the outflow decelerates in the ambient medium , resulting in a bright x - ray flare seconds after the merger that could be detectable with , e.g. , swift xrt @xcite . the flare gradually changes to optical and radio as the outflow decelerates , potentially detectable with , e.g. , pan - starrs @xcite and evla @xcite . * macronovae / kilonovae * matter ejected from compact binary encounters may produce electromagnetic optical or near - infrared transients that are suitable for follow - up searches of gw signal candidates @xcite . matter ejected by binary mergers ( tidal tails or accretion disk outflows ; e.g. , @xcite ) , initially at nuclear densities , expands and undergoes r - process nucleosynthesis , producing heavier , radioactive elements . the decay of such elements produces an isotropic emission that lasts for days , named macronova " @xcite or kilonova " @xcite . this is an appealing model , because it suggests that virtually all ligo detected ns - ns or bh - ns mergers would have detectable optical counterparts . however , finding these counterparts in the large gw directional uncertainty region would likely require a combination of powerful survey telescopes and follow - up spectroscopy , as used by palomar transient factory , and pan - starrs , or the future lsst @xcite . 3 . * x - ray / optical afterglow * the relativistic outflow from a grb central engine interacts with matter in the circumburst medium , emitting a long - lasting electromagnetic afterglow in a wide range of frequencies ( e.g. , @xcite ) . at x - ray and optical frequencies ( lasting from minutes to a day ) , afterglow emission is the strongest shortly after the prompt grb in the direction of the relativistic outflow ( i.e. on - axis ) , and rapidly decays afterwards @xcite . on - axis x - ray emission is detectable ( e.g. , with the swift xrt telescope ) over a period of @xmath75day after the prompt burst for source distances out to the horizon distance of gw telescopes @xcite . off - axis x - ray and optical emission , far from the axis of rotation , peaks at a significantly lower flux , making the detection of off - axis x - ray less useful in multimessenger searches @xcite . for radio frequencies around @xmath1ghz , on - axis emission for long grbs peaks at 3 - 6 days in the source rest frame after the prompt burst @xcite , while off - axis emission peaks on a similar time scale . * magnetic interaction between bh & ns * for ns - bh binaries , once the bh enters the magnetosphere of the ns , it can interact with the ns magnetic field @xcite . for strong enough ns magnetic fields ( @xmath144 g ) , such an interaction results in the copious emission of electromagnetic radiation , via a mechanism analogous to the blandford - znajek mechanism @xcite . the electromagnetic radiation may resemble an extremely short ( @xmath145 ) grb , and could be detected from distances of @xmath146mpc using the swift or fermi satellites @xcite . although this mechanism produces a short electromagnetic emission , it is expected to be less beamed than the standard " gamma - ray emission . * magnetic interaction between ns & ns * the magnetic interaction of nss in ns - ns binaries can drive an electromagnetic ( x - rays or gamma - rays ) signal within the last few seconds prior to merger through electric dissipation @xcite . as the non - magnetic ns moves through the magnetosphere of the other , highly magnetized ns ( @xmath147 g ) , the induced electromagnetic force across the non - magnetic ns sets up a circuit connecting the two stars . some of this electric potential dissipates either in the surface layer of the magnetic ns or in the space between the two stars , depending on the resistivity of the space between the two stars . a part of the dissipated energy is carried away through electromagnetic emission , which can reach up to @xmath98erg@xmath50s@xmath148 in the last @xmath7s prior to merger @xcite . 6 . * neutron star flares induced by tidal crust cracking * nss in inspiraling binary systems are subject to tidal deformation . when this deformation exceeds a critical level , the crust of the ns cracks , possibly resulting in a ns flare similar to observed flares @xcite ( see section [ subsection : magnetars ] on magnetar giant flares ) . such flare would represent an isotropic electromagnetic counterpart for binary mergers . observed short - grb precursors may be created by crust cracking @xcite . many gw sources , and in particular grbs , are expected to be copious emitters of neutrinos @xcite . astrophysical neutrino emission is expected within two main sub - groups based on two distinct emission mechanisms , with two distinct energy ranges . mev energy neutrinos ( with energies @xmath149mev ) are produced in the extremely hot , dense central regions of core - collapse supernovae and probably grbs . high - energy neutrinos ( with energies @xmath150gev ) are expected to be emitted by shock accelerated particles in relativistic outflows driven by the central engine of the grb ( e.g. , @xcite ) . so far only mev energy astrophysical neutrinos have been confirmed , and in one instance , for supernova 1987a @xcite . one of the advantages of joint gw - neutrino searches @xcite is that gw and neutrino detectors ( mev and high energy ) continuously observe the whole sky ( one high - energy neutrino detector observing half the sky ) , recording signal candidates without the need to `` point '' the detector in a particular direction . such full sky coverage is of special importance for multimessenger searches as the sky coverage for each messenger needs to overlap for a joint search . further , while electromagnetic follow - up searches require low - latency response from electromagnetic telescopes , gw - neutrino observations can be performed with high latency without loss of information . nevertheless , fast analysis of joint gw - neutrino observations can also enable the electromagnetic follow - up of joint event candidates . in core - collapse supernovae , most of the released gravitational binding energy is emitted in a burst of @xmath25mev neutrinos ( e.g. , ) . both mev neutrino @xcite and gw emissions @xcite are expected to commence near core bounce within a few milliseconds ( @xmath151 ms ) . this temporal correlation is orders of magnitude tighter than correlation with electromagnetic signals @xcite , and can greatly enhance the sensitivity of a search for joint gw - mev neutrino sources . mev neutrino emission peaks within a fraction of a second after bounce , while emission continues for up to tens of seconds as the protoneutron star cools and contracts . the joint detection of gws and mev neutrinos could provide constraints on the core - collapse supernova mechanism as well as information on the properties of matter at high energies and densities @xcite . for example the neutrino spectrum from core - collapse supernovae depend both on the nuclear eos as well as the spin of the core @xcite . one can break this degeneracy using the additional information available in the gw channel , inferring information on both the eos and the spin of the core . a potential secondary collapse due to hadron - quark phase transition in the protoneutron star during its post - bounce evolution @xcite would also result in characteristic neutrino and gw emissions @xcite . long and short grb central engines are also probably strong emitters of mev neutrinos @xcite . as the progenitors of long grbs are likely the progenitors of core - collapse supernovae as well , emission from long grbs may commence similarly to the case of core - collapse supernovae . in the core - collapse scenario , neutrino emission from the core ceases upon black hole formation , while a newly formed accretion disk will become a significant source , as a large fraction of the accretion energy can be carried away by neutrinos @xcite . compact binary mergers the likely progenitor of short grbs are expected to emit mev neutrinos @xcite from the accretion disk around the central black hole . in the case of ns - ns binary mergers , a hypermassive ns may form prior to bh formation , and further produce neutrino @xcite as well as gws @xcite emission . mev neutrino emission from compact binary mergers is expected to commence at the beginning with the merger phase , in coincidence with the gw burst due to the merger . this gw burst signal is preceded by the much longer inspiral phase , and followed by the ringdown phase . such temporal coincidence can be utilized in a joint search . nevertheless , for typical distances of binary mergers , mev neutrinos would be difficult to detect @xcite . several large - scale mev neutrino detectors are currently in operation . these include super - kamiokande ( japan ) @xcite , kamland ( japan ) @xcite , lvd ( italy ) @xcite , borexino ( italy ) @xcite and baksan ( russia ) @xcite . further , the icecube high - energy neutrino detector @xcite is also capable of detecting bursts of mev neutrinos @xcite , albeit without the ability to reconstruct the source direction . recognizing the importance of multimessenger observations , super - kamiokande , lvd , borexino and icecube are members of the supernova early warning system ( snews ) @xcite . these observatories send real - time triggers of detected supernova candidate events , which are distributed to the astronomer community to allow for low - latency follow - up electromagnetic ( or other ) searches . the most sensitive current mev neutrino detectors ( super - kamiokande and icecube ) are able to detect the mev neutrino signal of a supernova from up to @xmath25100 kpc @xcite . the expected event number within this distance , however , is relatively small outside of the milky way . planned megaton detectors , such as lbne @xcite or hyper - kamiokande @xcite , could detect supernovae from up to @xmath8510 mpc @xcite with a supernova rate of @xmath115 per year . multimessenger searches of mev neutrinos and gws could further increase these detectable supernova rates @xcite , and would provide increased confidence in a detected signal . this can be especially important when no electromagnetic counterpart is detected . non - thermal , high - energy ( @xmath152gev ) neutrinos are thought to be created within relativistic outflows driven by the central engine @xcite . emission from early grb afterglows is also plausible ( e.g. , @xcite ) . the emission mechanism is likely similar for both long and short grbs , the former expected to be the stronger emitter . core - collapse supernovae with rapidly rotating cores @xcite , magnetars @xcite and millisecond protomagnetars @xcite are also thought to emit high - energy neutrinos . while recent limits obtained with the icecube detector constrain some emission models @xcite , the standard fireball picture for neutrino emission remains viable @xcite . as their production mechanism is connected to that of gamma photons , high - energy neutrinos are probably beamed similarly to that of gamma ray emission . this decreases the rate of detectable joint sources similarly to the case of electromagnetic emission . for a joint detection , an especially interesting class of sources are grbs with little or no detectable electromagnetic counterpart , which are therefore only detectable via gws and neutrinos . such sources include so - called _ choked _ grbs . in the core - collapse supernova scenario , gamma - ray emission is observable only if the relativistic outflow from the central engine , that is responsible for the production of gamma - rays , breaks out of the star @xcite . the outflow , in order to advance , needs to be driven by an ( active ) central engine . if the breakout time of the outflow is longer than the duration of the active central engine , the outflow stalls , producing no observable gamma - ray emission @xcite . high - energy neutrinos are produced within the outflow and can escape through the stellar envelope , producing similar neutrino emission as for `` successful '' grbs . other interesting sources include low - luminosity ( ll ) grbs , that can frequently go unobserved if their luminosity falls below the threshold of gamma - ray telescopes . ll grbs , while weaker neutrino sources than their high - luminosity ( hl ) counterparts , are thought to be much more frequent , making their overall neutrino flux comparable or even surpass the flux from conventional hl grbs @xcite . besides being interesting for sources not observed through their electromagnetic emission , gw - high - energy neutrino observations can provide information on the internal structure of the progenitor , as well as the properties and dynamics of the relativistic outflow ( e.g. , @xcite ) . a potentially interesting subclass of grbs for joint gw - high energy neutrino observations is low - luminosity grbs @xcite that may be a distinct population from high - luminosity grbs @xcite . it was suggested that more massive stars may form a black hole after collapse , and produce a high - luminosity , high - lorentz factor outflow , while less massive stars may form a neutron star after collapse , and drive an outflow with low luminosity and lorentz factor @xcite . more recently , bromberg et al . @xcite found that estimated jet breakout times for low - luminosity grbs are much longer than the observed durations , in marked difference with high - luminosity grbs , suggesting that low - luminosity grbs may be choked " , i.e. their relativistic outflow stalls before breaking out of the stellar envelope ( see also @xcite ) . low - luminosity grbs are thought to be an order of magnitude more common than their high - luminosity counterparts , with lower electromagnetic luminosity and smaller beaming angle @xcite , making them a promising source type for multimessenger detections . joint searches for gws and high - energy neutrinos are well - suited for multimessenger analyses , since both messengers have typically sub - threshold significances . for this purpose baret @xcite developed a multimessenger analysis method that combines the significance and other information from the two messengers . the location of potential or confirmed host galaxies of grbs can be utilized in gw searches to enhance sensitivity as well as in event selection or localization . long grbs are found to occur in star - forming regions of distant galaxies , in accordance with their association to the deaths of massive stars . consequently , the rate of long grbs is correlated to the blue luminosity of galaxies . long grbs are actually found in the very brightest regions of their host galaxies , which are significantly fainter and more irregular than typical host galaxies of core - collapse supernovae , suggesting that they are associated with extremely massive stars , and galaxies with limited chemical evolution @xcite . the distribution of short grbs seem to follow that of old ( @xmath75gyr ) stellar populations within normal star forming as well as elliptical galaxies @xcite , in stark contrast with the distribution of long - grb host galaxies @xcite . it is highly unlikely that short and long grbs are drawn from the same underlying population @xcite . nevertheless , the rate of short grbs in a given galaxy is correlated with the optical light of their host , and to less extent with the blue luminosity of the galaxy @xcite , indicating that short - grb progenitors have a wide age distribution of several gyr . as neutron stars may experience an initial kick at birth @xcite , binary mergers can take place far away from the star - forming region where they originate from . for kick velocities of @xmath153 and inspiral times of several gyr , binaries could travel large distances before their merger @xcite , traveling far outside their host galaxies . typical predicted distances are @xmath154kpc @xcite , the predicted distance distribution being well - matched by observed short - grb distributions @xcite . magnetars @xcite form another possible progenitor type of short grbs , making up less than @xmath25one - third of their population @xcite . they are weaker and softer gamma - ray emitters than other short grbs , making them identifiable mostly at smaller distances , within the milky way and the large magellanic cloud . magnetars are thought to be created in supernova explosions @xcite . as they can receive an initial kick velocity during the supernova explosion similarly to radio pulsars , only a fraction of them may be near its respective supernova remnant , in accordance with observations @xcite . magnetars are observed to be mostly young ( @xmath2510@xmath155 yr ) objects @xcite , therefore they did not travel far from the star forming region of their respective supernovae . consequently , extragalactic magnetar flares can be expected to occur within the star - forming regions of galaxies . the expected distribution of grbs can be used to enhance gw searches through identifying preferential gw source directions ( e.g. , @xcite ) . a priori information on source distribution has been used in multiple gw - grb searches . the galaxy distribution plays a significant role in electromagnetic follow - up searches . since most gw signal candidates have a reconstructed direction with an uncertainty of a few degrees ( which may be scattered over a larger area on the sky ) , the sky area of the possible gw source direction is greater than the field - of - view of most astronomical instruments used for follow - up observations . to decrease the surveyed area on the sky , the directions in which electromagnetic follow - up is performed is down - selected using the directions of galaxies overlapping with gw sky area of the possible gw source direction @xcite . in multimessenger searches , the distribution of galaxies can be used to increase the significance of astrophysical signals as well as to reject background events whose directions do not overlap with the direction of a galaxy ( e.g. , @xcite ) . the significance of event candidates can be enhanced by weighing events with the a priory probability distribution of multimessenger sources ( e.g. , based on the luminosity and distance of galaxies ) . galaxy sky locations are especially important for gw sources that are detectable only within @xmath73mpc . on this distance scale , existing galaxy catalogs are more complete ( e.g. , @xcite ) , and the number of galaxies is small within the sky area defined by the directional uncertainty of gw observations ( @xmath25 few degrees ; @xcite ) . the use of galaxy direction on this scale could therefore significantly add to the sensitivity of these searches . galaxy catalogs can still aid searches for sources at farther distances , although the sensitivity gain significantly reduces with distance @xcite . in this review we presented an overview of the major directions in which gravitational - wave ( gw ) astronomy has the potential of advancing our understanding of grbs and their progenitors . we discussed the different promising directions that can lead to gw emission from grb progenitors that is detectable with second - generation gw observatories . we also discussed some of the major scientific questions that could be addressed upon the detection of gws from these sources . second generation gw observatories , commencing their operation in the next few years ( e.g. , @xcite ) , will be sensitive enough to reach distances for which the detection of gw signals from grb progenitors becomes possible . the gw signal expected from compact binary coalescences , the likely progenitors of short grbs , is fairly well understood as a function of the ( often unknown ) source properties . the detection of gws can provide estimates or constraints on these source properties . we presented a detailed overview of the questions that can be addressed about short - grb progenitors upon the detection of their gw signature . the core - collapse of massive stars , the likely progenitor of long grbs , has less constrained theoretical gw emission models . this is partly due to the lack of our detailed understanding of the progenitor ( e.g. , how cores maintain their high rotation rate prior to collapse ) , and partly due to the complex dynamics of the collapsing core and the resulting compact object ( protoneutron star or black hole ) and accretion disk , whose understanding requires further theoretical and numerical studies . nevertheless , the core - collapse dynamics may lead to gw signals that are detectable on distance scales comparable to the lower end of the long - grb distance scale ( @xmath156mpc ) . for example rotational instabilities in differentially rotating protoneutron - stars , as well as global instabilities in tori around black holes , both lead to rotating , non - axisymmetric systems with potentially strong gw emission . starquakes in highly - magnetized neutron stars , likely responsible for producing a fraction of short grbs , result in seismic oscillations in the neutron star , which in turn may produce gws detectable with second generation gw observatories . the detection ( or non - detection ) of such gws coincident with the flaring activity of neutron stars can be informative , e.g. , on the properties of matter at nuclear densities , as well as on the properties of astrophysical neutron stars . the authors gratefully acknowledge the excellent comments , help , and suggestions of alessandra corsi , jolien creighton , riccardo desalvo , raymond frey , kunihito ioka , jonah kanner , kostas kokkotas , koutarou kyutoku , ilya mandel , brian metzger , kohta murase , ehud nakar , tsvi piran , maurice van putten , luciano rezzolla , stephan rosswog , yuichiro sekiguchi , peter shawhan and masaru shibata . the authors especially thank richard oshaughnessy for his detailed comments throughout the preparation of the manuscript . ib & sm are grateful for the generous support of columbia university in the city of new york and the national science foundation under cooperative agreement phy-0847182 . ros & pb are supported by nsf award phy-0970074 and the uwm research growth initiative . ros & pb appreciate the hospitality of the ucsb kitp , supported by the nsf with award nsf phy11 - 25915 . 100 url # 1#1urlprefix[2][]#2 mszros p 2012 _ astroparticle physics _ issn 0927 - 6505 http://www.sciencedirect.com/science/article/pii/s0927650512000692 t a 2007 assessing millisecond proto - magnetars as grb central engines _ revista mexicana de astronomia y astrofisica , vol . 27 _ ( _ revista mexicana de astronomia y astrofisica , vol . vol 27 ) pp 8090 ( _ preprint _ ) s a and hawley j f 1997 instability , turbulence , and enhanced transport in accretion disks _ iau colloq . 163 : accretion phenomena and related outflows _ ( _ astronomical society of the pacific conference series _ vol 121 ) ed wickramasinghe d t , bicknell g v and ferrario l p 90 j , marka s , shawhan p s _ et al . _ 2007 looc up : locating and observing optical counterparts to unmodeled pulses in gravitational waves _ american astronomical society meeting abstracts _ ( _ bulletin of the american astronomical society _ vol 38 ) pp 910+ p s , ligo scientific collaboration and virgo collaboration 2010 looc - up : seeking optical counterparts to gravitational - wave signal candidates _ american astronomical society meeting abstracts # 215 _ ( _ bulletin of the american astronomical society _ vol 42 ) pp 406.06+ p s 2012 rapid alerts for following up gravitational wave event candidates _ society of photo - optical instrumentation engineers ( spie ) conference series _ ( _ society of photo - optical instrumentation engineers ( spie ) conference series _ vol 8448 ) ( _ preprint _ ) n , aussel h , burke b e _ et al . _ 2002 pan - starrs : a large synoptic survey telescope array _ society of photo - optical instrumentation engineers ( spie ) conference series _ ( _ society of photo - optical instrumentation engineers ( spie ) conference series _ vol 4836 ) ed tyson j a and wolff s pp 154164 i , kelley l z and ramirez - ruiz e 2012 towards improving the prospects for coordinated gravitational - wave and electromagnetic observations _ iau symposium _ ( _ iau symposium _ vol 285 ) ed griffin r e m , hanisch r j and seaman r pp 358360 ( _ preprint _ )

astronomers do not have a luxury of testing their ideas of , e.g. , galaxy and star formation in a purpose designed laboratory . instead , testing grounds are provided by observations and numerical simulations . the latter in effect represent experiments with a given set of physical laws included . clearly , the more physics is included in the simulations the more realistic the latter should be . presently , two basic physical processes gravity and hydrodynamics are modelled well by a variety of methods and codes . for example , one reliable and widely used numerical method of modelling gas dynamics is smoothed particle hydrodynamics ( sph ) @xcite . it has been used in the various fields of astrophysics , and is especially powerful when resolving high density regions is key . the method is grid - less and fully lagrangian in nature , facilitating modelling of arbitrary geometry systems . for reviews of the sph method see , for example , @xcite . radiation transfer and interaction with matter is another basic process operating in astrophysical systems . starting from the pioneering ideas of @xcite , great efforts have been expended to model radiation matter interactions in sph . when gaseous systems modelled are optically thick , photons scatter or get absorbed and re - emitted multiple times before escaping the system @xcite . this situation is well approximated by the diffusion approximation . several authors have already incorporated this radiation transfer scheme in sph @xcite . the diffusion approximation is usually complemented by a flux limiter method to model optically thin regions where photons stream freely rather than diffuse ( e.g. , * ? ? ? most recently @xcite implemented the diffusion radiation scheme in the cosmological code gadget @xcite supplementing it by the variable eddington tensor as a closure relation . ray tracing methods is a principally different approach to radiative transfer , where the radiative transfer equation is solved along chosen directions ( rays ) . most applications only consider rays that start at discrete sources of radiation field , such as bright stars , etc . , essentially neglecting the diffuse radiation field . as far as the density estimation is concerned , some authors used density defined at the locations of the nearest neighbours along the rays @xcite , whereas others approximated the density field using the nodes of the tree structure one way or another @xcite . the most recent developments use the sph density field directly , e.g. , c.f . codes sphray @xcite and traphic @xcite . @xcite developed a method to transport radiation on adaptive random lattices , and showed that the algorith is very efficient for cosmological re - ionisation problems . monte - carlo methods are similar in spirit and yet substantially different from the ray tracing codes . in the former , the idea is to discretise the radiation field into `` packets '' , choose directions and emission time of these packets stochastically to obey proper physical constraints , and then propagate the packets through matter in accord with radiation transfer equations . the main problem for the method is stochastic noise caused by photon packet statistics ( for an early sph application see * ? ? ? @xcite presents some new interesting ideas about using a smoothing kernel in a monte - carlo simulation . most of these applications were tailored to photo - ionisation problems in the field of star formation or cosmology , where the density field can be considered static as rays propagate through it ( the static diffusion limit ; see * ? ? ? * ) , and where radiation pressure effects can be omitted . there are also a number of radiation transfer codes working with a grid - based hydrodynamics codes rather sph ( e.g. , * ? ? ? we do not discuss these methods here . here we present and test a new photon packet based radiation transfer scheme combined with an sph code . the method uses the sph density field directly , thus preserving the `` native '' sph resolution . the previous efforts did not consider the radiation pressure effects , studying instead radiative heating / cooling and photo - ionisation . here we study the radiation pressure effects instead . we assume that gas equation of state is known ( e.g. , isothermal or polytrophic ) and consider only the radiation pressure forces . this allows us to thoroughly test precision of our approach and numerical noise effects . there is no fundamental difficulty in including the heating / cooling and photoionisation processes in our scheme , and we shall extend our method in that direction in the near future . another defining characteristic of our new method is that the photon field is evolved in the same time - dependent way , although on shorter time steps , as gas dynamics , and therefore the method is intrinsically time - dependent . radiative transfer equation along a ray @xcite is @xmath0 where @xmath1 is the specific radiation intensity , @xmath2 is the opacity coefficient , @xmath3 is gas density and @xmath4 is emissivity of the gas . in general @xmath1 and @xmath4 are functions of direction , radiation frequency , position and time . the first term on the right hand side represents removal of radiation from the beam by absorption and scattering , whereas the last term describes local emission of radiation into the beam s direction . in monte carlo methods , the radiation field is sampled via photon packets . in the simplest reincarnation of the method , both terms on the right hand side of equation [ perenos ] are treated stochastically . the photon mean free path , @xmath5 , is calculated as @xmath6 . a random number , @xmath7 , uniformly distributed over @xmath8 $ ] , is generated . a photon is allowed to travel distance @xmath9 , at which point it interacts with gas by passing its energy and momentum to gas _ completely_. the photon is then re - emitted according to physics of a chosen set of radiation processes , and followed again until it escapes the system . our radiation transfer scheme is slightly modified from this . firstly , instead of using discontinuous photon jumps , we explicitly track packet s trajectory in space as @xmath10 where @xmath11 and @xmath12 are the current and initial photon locations , and @xmath13 const is the photon propagation speed ( not necessarily the speed of light ; see [ sec : prompt ] ) . secondly , photon momentum , @xmath14 , and photon energy @xmath15 are reduced _ continuously _ due to absorption or scattering as @xmath16 thus , the first term on the right hand side of equation [ perenos ] is treated continuously rather than stochastically . this reduces the statistical noise significantly . photons are discarded when their momentum drops below a small ( @xmath17 ) fraction of their `` birth momentum '' @xmath18 . however , the re - emission term , i.e. , the last one in equation [ perenos ] is modelled similarly to classical monte - carlo methods in that it is stochastic in nature . this is unavoidable in monte carlo methods due to the need to employ a finite number of photon packets ( e.g. , * ? ? ? in practice , at every photon s time step @xmath19 , the photon momentum absorbed , @xmath20 , is calculated according to equation [ absorption_only ] . as stated in the introduction , we concentrate in this paper on the radiation pressure effects , and assume that radiation absorbed from the beam is completely re - emitted on the spot in a new random direction . this is the case for pure scattering of radiation or for local radiative equilibrium between radiation and gas . in these cases the amount of momentum ( energy ) to be re - emitted in the interaction considered is exactly @xmath20 . probability of re - emitting a new photon with momentum ( energy ) @xmath18 ( @xmath21)is defined as @xmath22 . a random number , @xmath7 , uniform on @xmath8 $ ] , is generated . if @xmath23 , a new photon with momentum @xmath18 and a random direction of propagation is created . this conserves photon field s energy and momentum in a time - average sense . this scheme can be adapted to allow for non - equilibrium situations when radiative cooling is not equal radiative heating . multi - frequency radiation transfer is also straight forward if tedious to implement . we have done some of these developments already and will report it in a future paper . the radiation transfer approach described in [ sec : smc ] embodies the classical radiation transfer equations and must be correct on that basis . nevertheless , we tested our approach against the discrete monte - carlo method on a simple test problem where the gas density and opacity are fixed and independent of the radiation field . this setup does not require using gas dynamics at all , and hence no sph particles were used to model the density field . in the test , photons are emitted at a constant rate at the midplane of a plane - parallel slab @xmath24 $ ] . the main parameter of the problem is the total optical depth of the gas between @xmath25 and @xmath26 , @xmath27 . we ran the radiation transfer part of the code until a steady state photon packet distribution was achieved . figure [ fig : slab_rt ] shows the ratio of the local radiation energy density to the radiation flux out of the slab as a function of @xmath28-coordinate inside the slab . for convenience , the speed of light is set to unity . two tests are shown , one for @xmath29 ( left panel ) and the other for @xmath30 ( right panel ) . black solid curves show the same quantity computed with the standard monte carlo technique , whereas the red dashed curves are computed with the modified method described above . for the latter time - dependent calculation , we averaged over a number of snapshots . the disagreement is quite small everywhere except for the region near @xmath25 . the difference in that region is simply due to the finite time resolution of the red curves . namely , these curves are obtained by averaging snapshots of the simulation data . the time step between the snapshots is long enough for the photons introduced at times between the snapshots to diffuse away from their starting @xmath25 location , broadening the peak . the monte carlo curve , on the other hand , is obtained by following individual photons progress out of the slab , and therefore the information about the initial location of the photon is retained and is evident in the peak at @xmath25 . these tests demonstrate that the radiation transfer is modelled properly with our approach at least in the simple situations considered here . we now discuss coupling the radiation transfer scheme to sph . a photon mean free path is given by @xmath31 where @xmath2 is the gas opacity coefficient and @xmath3 is gas mass density _ at the photon s current location . _ we require an accurate method of local gas density estimation , and therefore we use directly the density field representation of the sph method . in the sph , gas density at a position @xmath32 is given by the superposition of smoothed contributions from individual sph particles : @xmath33 where @xmath34 is the sph kernel , @xmath35 is the location , @xmath36 is the smoothing radius of particle @xmath37 , and @xmath38 is the contribution of particle @xmath37 to the density at @xmath39 . the summation goes over all the `` neighbours '' sph particles that contribute to @xmath3 at @xmath39 , i.e. , have non - zero values of @xmath34 at this location . note that in this approach there is no expectation of the number of gas neighbours of a photon , @xmath40 , to be constant or limited in any other way . in particular , having @xmath41 is perfectly feasible and simply means that the photon propagates through an `` empty '' patch of space . having determined @xmath5 at the photon s location , we find the decrement in photon s momentum , @xmath42 , at that location according to equation [ absorption_only ] . this decrement is then passed to the photon s sph neighbours to enforce conservation of momentum . if @xmath43 , there is a question of what fraction of @xmath42 should be passed to a particular neighbour @xmath37 . we assume that the neighbours contribute to the interaction with a photon directly proportionally to their density @xmath44 at the photon s location ( see equation [ eqn : density ] ) . hence the momentum passed to neighbour @xmath37 is @xmath45 as we limit the photon time step to ensure that @xmath46 ( see equation [ dtgamma ] ) , this approach yields physically correct result . the momentum transferred to the sph particle @xmath37 from interactions with different photon packets @xmath47 is additive , and is used to define the radiation pressure force on that particle . we sum all the interactions that the sph particle experienced during its time step @xmath48 , and then define the radiation pressure force on particle @xmath37 by @xmath49 due to limitations on photon propagation time steps @xmath19 ( see below ) , an sph particle may interact several times with a particular photon @xmath47 . there is however ho paradox here , as this simply means that the radiation transfer equation is integrated in multiple points along the photon s trajectory within each sph particle , increasing the precision of the method over traditional monte - carlo approaches . this formulation also allows us to enforce an exact pair - wise conservation of momentum in the interactions of matter and radiation . the energy transfer can be calculated in exactly same manner except @xmath42 is replaced by @xmath50 . external radiation sources such as stars or accreting compact objects emit photon packets with a given momentum @xmath18 . the rate of packet emission is given by @xmath51 where @xmath52 is the luminosity of the source and @xmath53 is the speed of light . the radiation field can be chosen to be isotropic or beamed / restricted to a range of directions . there are several issues to consider when deciding how far the photon can travel in a single flight @xmath54 . first of all , this distance should be much smaller than the typical sph smoothing length , @xmath55 , in the region , or else the photon will `` skip '' interactions with some sph particles altogether . this constraint is important in both optically thin and thick regimes . in the optically thick case , an additional constraint need to be placed to insure that photons do not propagate in one step by more than a fraction of their mean free path , @xmath5 . in the opposite case photons would `` diffuse '' though optically thick regions in an unphysical way , i.e. , too quickly . these constrains are combined by requiring @xmath56\ ; , \label{dtgamma}\ ] ] where @xmath57 is a small dimensionless number . in practice we use @xmath58 . as discussed below in [ sec : prompt ] , in the `` prompt escape '' regime , it is possible to reduce the photon propagation speed below the speed of light , which then allows us to integrate photon trajectory on longer time steps without compromising the physics of the problem . following @xcite , radiation hydrodynamics of a problem can be divided into three different regimes . let @xmath59 be a characteristic gas velocity , such as the sound speed or the bulk gas velocity , whichever is greater . define @xmath60 and optical depth of the system @xmath61 , where @xmath62 is the geometric size of the system . the three limiting regimes @xcite are @xmath63 @xmath64 @xmath65 in the first regime a typical photon leaves the system in a single flight . in the second case the photon scatters or get absorbed and re - emitted approximately @xmath66 times before leaving the system . for what follows it is important that in the first two regimes photons escape from the system on timescale , @xmath67 , much shorter than the matter distribution can alter significantly , i.e. , @xmath68 here @xmath69 is the geometric size of the system . in contrast , in the dynamic diffusion limit the photon diffusion time is larger than the crossing time of the system , @xmath70 . equation [ tesc ] shows that the exact value of the speed of light is irrelevant in the free streaming and the static diffusion limits ; it is so high it can be considered infinite . to the contrary , in the dynamic diffusion limit the exact value of speed of light is important as it defines the time scale on which radiation from the system leaks out , and that time scale is long . therefore , we combine the free streaming limit and the static diffusion limit of @xcite into the `` prompt escape '' regime given by the equation [ tesc ] . in this limit it is numerically convenient and physically permissible to reduce the photon propagation speed @xmath71 below @xmath53 , as long as the system still satisfies equation [ tesc ] with @xmath53 replaced by @xmath71 . we found that this speeds up calculations in which gas dynamics is important , although the scaling is not as efficient as the factor @xmath72 . the reason for that is that although photon time step ( equation [ dtgamma ] ) is indeed longer by the factor of @xmath72 , the number of photon packets for a given @xmath14 is correspondingly higher ( see equation [ ngamma ] and note that @xmath73 ) . this radiation - gas momentum transfer method has been implemented in the sph / n - body code gadget @xcite that is widely used for cosmological simulations . for the tests presented here the cosmological options of the code are turned off . gadget uses a barnes - hut tree to speed up calculation of gravitational forces and for finding neighbours . as photon packets have no mass associated with them , we turn off the gravity calculation for these particles . they are also not included in building the barnes - hut tree . to find sph neighbours of a photon , we first find all sph particles that are inside a sphere with size @xmath74 which is chosen to be much larger than the mean sph particle smoothing length . we then further select only those sph particles @xmath37 that contain the photon within their smoothing length @xmath36 . after calculating the radiation pressure force for a given sph particle @xmath37 , the corresponding radiative acceleration @xmath75 is added to the hydrodynamical and gravitational accelerations that the particle experiences . we have also added the radiation pressure acceleration to the time step criteria for the sph particles , as described in @xcite . this ensures that sph particle time steps are appropriately short for particles with large radiation pressure accelerations . in [ sec : slab_transfer ] we tested the radiation transfer methods for the density field given by a simple analytical function ( a constant ) . in [ sec : interactions ] we presented a way to model radiation transfer in an arbitrary density field represented by the sph particles . it is a logical step forward in complexity of the tests to now repeat the slab test of [ sec : slab_transfer ] in a self - consistent sph density field . to accomplish this , we consider a non self - gravitating accretion disc in orbit around a massive ( @xmath76 in code units ) central source . the disc is assumed to be locally isothermal , with internal energy @xmath59 scaling as @xmath77 , where @xmath69 is radius , and @xmath78 is the internal energy at the inner edge of the disc , @xmath79 . this scaling yields a constant ratio of the vertical disc scale height , @xmath80 , to radius . the outer radius of the disc is @xmath81 . the disc was then relaxed for a large number of orbits without any radiation field , which yields a gaussian vertical density profile @xcite . keeping this density profile fixed , we then introduced photon sources in the disc midplane and allowed the photons to propagate out of the disc keeping the opacity coefficient @xmath82 fixed everywhere in the disc . figure [ fig : slab_transfer ] shows the resulting photon energy density distribution within the slab ( solid curve ) in exactly same manner as in the fixed density tests shown before in figure [ fig : slab_rt ] . as before , the radiation field has been averaged over several snapshots to reduce statistical noise . the respective monte - carlo result obtained for the same density distribution is shown in figure [ fig : slab_transfer ] with the red dashed curve . there is a very good agreement between the curves , except for the peak region in the red curve . time sampling differences in the two simulations explain the discrepancy in the curves ; see discussion of figure [ fig : slab_rt ] in [ sec : slab_transfer ] . this test suggests that the radiation transfer part of our approach is working as expected . below we shall move on to tests that involve gas dynamics . young massive stars produce most of their radiation in the uv wavelengths . dust in the interstellar gas has a very large absorption opacity for uv photons , with @xmath83 up to a few hundred @xmath84 g. if these photons are absorbed and re - emitted in the infrared , where roseland opacity is orders of magnitude smaller than @xmath83 , in many applications one can effectively assume @xmath85 . in this case it is sufficient to consider only the radiation pressure from the uv photons emitted by the young stars , and neglect the re - radiated component . finally , one can approximate the large uv opacity by an infinitely large one , @xmath86 . photon packet propagation is then trivial : packets travel in straight lines with constant momentum @xmath87 until they encounter an sph particle(s ) , at which point they are absorbed and their momentum is transferred to that particle(s ) . this `` on the spot '' absorption method can also be used to model _ fast _ gas winds from massive stars or luminous black holes in the momentum driven regime . in the latter case the cooling time in the outflow is short . shocked outflow gas cools very quickly , and hence its thermal pressure can be neglected . the momentum transferred to the ambient medium provides the push to drive the shell out . if the wind velocity is much higher than the velocity of the expanding shell and the speed of sound in the ambient gas , one can also neglect the mass outflow from the source . this is justified as the mass flux in the wind is small compared with that of the ambient gas being driven out . the `` on the spot '' approximation presents a convenient test ground of our code since it is possible to derive exact analytical solutions in the simplest cases . in particular , we consider a single radiation source embedded in an infinite initially uniform isothermal medium . gravity is turned off for simplicity . in practice we set up periodic boundary conditions for a cubic box with dimensions , @xmath62 , of unity on a side . these boundary conditions are appropriate and do not affect our results since the radiation force effects are contained to a small region within the box during the simulations . gas internal energy is fixed at @xmath88 , hence sound speed @xmath89 . the total mass of the gas inside the box is @xmath76 . the unit of time is @xmath90 . @xmath91 sph particles is used in these tests . the initial condition is obtained by relaxing the box without the radiation source to a state of nearly constant gas density over hundreds of dynamical times for the simulation box . we first consider a case with a relatively small source luminosity @xmath92 in the code units . in this case a quasi - steady state should be set up quickly . we consider that state here . figure [ fig : onthespot ] shows the density profiles obtained in four different runs at dimensionless time @xmath93 , when the sph quantities approach a quasi steady state . theoretical expected density profile is given by @xmath94 for @xmath95 and @xmath96 const @xmath97 for @xmath98 . here @xmath99 is the size of the cavity opened up by the radiation . this is obtained by requiring a force balance between the momentum flux from the source , @xmath100 , and the external pressure of the gas @xmath101 : @xmath102^{1/2}\;. \label{rcav}\ ] ] the expected discontinuous density profile is shown with the solid ( black ) line ( we have neglected a slight density increase due to evacuation of the gas from the cavity , since @xmath103 , the box size ) . figure [ fig : onthespot_v ] shows instantaneous velocities of sph particles corresponding to the respective panels in figure [ fig : onthespot ] . theoretically expected result for velocity in the steady state is @xmath104 everywhere , of course . figures [ fig : onthespot ] and [ fig : onthespot_v ] show that at relatively large values of the photon packet momentum , corresponding to a lower packet injection rate @xmath105 , the method is inaccurate . for @xmath106 , no cavity is opened at all around the source . moreover , sph particle velocities have random components at a 30% fraction of the sound speed , even at regions far behind the expected discontinuity at @xmath107 . in contrast , tests with @xmath108 and @xmath109 produce nearly identical results with velocity fluctuations below a few percent level at @xmath110 . physically , if @xmath111 , interaction of an sph particle with a single photon packet may accelerate the particle to velocity exceeding the sound speed , leading to a significant numerical noise . in the tests presented here , a typical sph particle momentum is @xmath112 , as @xmath113 . therefore , the two runs with @xmath114 do very poorly , as expected . these numerical experiments show that the minimum accuracy requirement _ in optically thick _ limit is @xmath115 for further quantitative analysis of the errors , consider now the width @xmath116 of the transition layer . this layer is defined as the region in which sph density goes from zero to unity . first , note that on average , a gas parcel within volume @xmath117 situated on the inner face of the transition region receives one photon packet every @xmath118 seconds . during this time , sph particles inside the parcel experience no radiation pressure `` pushes '' . we assume that in the absence of radiation forces , the volume is quickly accelerated by the pressure gradient force to @xmath119 . there are @xmath120 sph particles within the volume , where @xmath121 is the typical number of sph neighbours ( usually chosen to be around 40 ) . due to the spherical symmetry of the problem , the parcel travels inwards to the radiation source a distance @xmath122 before it is turned back by the arrival of the next photon packet . finally , since the volume element is located on the inner boundary of the cavity evacuated by the radiation pressure , it will expand when moving into the cavity . hence the appropriate smoothing length of the element will be larger than that far from the cavity , and in general we expect @xmath123 in that region . using the latter estimate in equation [ deltat1 ] , we obtain for the thickness of the transition layer , @xmath124^{1/3}\;. \label{deltar}\ ] ] the transition layer thickness calculated in this way is indicated in figure [ fig : onthespot ] with dotted ( @xmath125 ) and dashed ( @xmath126 ) lines . it appears to be a good estimate for the three runs out of the four shown in the figure . in the run with the highest photon packet resolution , i.e. , the smallest @xmath14 , equation [ deltar ] is an under - estimate . this is simply because a density discontinuity in sph can not be narrower than the width of the kernel , i.e. , the minimum smoothing length , which is about @xmath127 for all the four tests . inserting equations [ rcav ] and [ ngamma ] into equation [ deltar ] , and requiring @xmath128 to not exceed the characteristic smoothing length of the problem leads to the same requirement of the packet s momentum to be small enough compared with the typical sph momentum ( equation [ psph ] ) . summarising the results of these tests , we see that we can use the condition [ psph ] as the photon packet resolution requirement in an optically thick limit . in addition , stochastic noise arguments appeared useful when determining the width of the transition layer in figure [ fig : onthespot ] . keeping to the model of a single source radiating photons isotropically into an infinite , constant density , isothermal medium , we performed a further test whereby the luminosity of the source was significantly higher , @xmath129 . this time the focus of our attention is the evolution of the radius of the expanding shell with time , well before it reaches the equilibrium cavity size . this test is closely related to the ram - pressure driven wind test ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , but with the important distinction that in our method the photon ` gas ' is massless . this however should be a good approximation to a low density but high velocity outflow . the other important aspect of our model is the isothermal condition , corresponding physically to a shocked gas region that is allowed to cool very quickly , i.e. when the cooling time in the shocked gas is far shorter than the dynamical time . this tends to occur when the temperature is less than @xmath91 k , so that line cooling from ions dominates @xcite . the picture then is that of a momentum - conserving ` snowplow ' phase of an expanding gaseous bubble as it sweeps up the ambient gas into an increasingly massive shell . this model has three zones : ( i ) an evacuated cavity through which free - streaming photons are being emitted from the central star ( ii ) a shell of shocked ism gas at temperature @xmath130 bounded by an outer shock front and ( iii ) the undisturbed uniform ism at temperature @xmath130 . the initial conditions for this test and numerical setup are same as in [ sec : steady ] . the choice of the higher luminosity is motivated by a requirement that the mach number be high so that the pressure from the ambient gas did not have a significant effect on the evolution of the shell during the run . to aid this , the gas internal energy was set to a lower value of @xmath131 . the analytical solution to a momentum - conserving bubble can be derived by considering the thin - shell approximation , whereby the swept - up gas is assumed to be concentrated in an infinitely thin shell that is being driven by the impinging wind . in reality the shell gets thicker the more gas it sweeps up , but at all times the thickness of the shell , @xmath132 , is much less than the radius @xcite : @xmath133 where @xmath134 is the mach number and @xmath69 is the radius of the shell . the zeroth order accuracy analytical solution is found by equating the rate of change of momentum of the expanding shell to the momentum flux from the stellar luminosity : @xmath135 = \frac{l}{c}\;.\ ] ] the time evolution of the radius of the shell is therefore given by , @xmath136 we assumed here that the velocity of the particles carrying the momentum outflow from the source , @xmath71 , is much larger than the shell velocity , @xmath137 . this solution can be further improved by taking account of the finite speed of the wind ( photon packets in our code ) . in this case there is a time delay of @xmath138 before the wind particles strike the ambient gas . in addition , external pressure from the ambient medium provides some restoring force , slowing down the expansion of the bubble . factoring these effects into equation [ eq : shell ] leads to a more accurate description of the shell : @xmath139 = \frac{l}{c } \left ( t - \frac{r}{v_{\gamma}}\right ) - 4\pi r^{2 } \rho c_{s}^2 t\ ] ] which is integrated numerically to obtain a solution for @xmath69 . figure [ fig : expanding_shell ] shows the results of the test with @xmath140 . the solid black line shows the time evolution of the shock front ( identified by the peak in the density distribution ) in the sph simulation . the analytical result given by equation [ ranalytical1 ] is shown with the red dotted line . the green dotted line in figure [ fig : expanding_shell ] is a numerical solution of the more accurate equation [ eq : shell_numerical ] . the difference between the two expected solutions ( red and green ) is rather minimal since the external pressure is small compared with the ram pressure of the outflow and the photon packet s velocity is large , @xmath141 . figure [ fig : expanding_shell ] shows an excellent agreement between the expected 1d solution and the simulation s result . a simple but useful test of gas dynamics is provided by optically thin gas illuminated by a constant radiation flux . clearly the radiation force in this case must be a constant independent of position or time . in the stochastic approach of our method , however , the force can in principle vary spuriously due to statistical fluctuations , which may lead to numerical artefacts . our goal here is to investigate their nature , magnitude and scaling with the number of photon packets used . as in [ sec : onspot ] , the initial condition is a uniform density box with a side equal to 1 . the sph particles are allowed to interact with each other hydrodynamically only ( i.e. , self - gravity is turned off ) . in the tests reported here , a given number of photons @xmath142 propagates in the positive @xmath143-direction . the momentum of photons is set by requiring the time and volume averaged radiative acceleration be equal to a parameter @xmath144 . in the limit of an infinite number of photons and an absence of any numerical artefacts , all the sph particles should be accelerated with acceleration @xmath144 , so that their @xmath143-component of velocity is @xmath145 . in reality different sph particles interact with different number of photons due to stochastic fluctuations , and hence their accelerations and @xmath146 are not exactly equal . we define the mach number of spurious sph velocity fluctuations as a measure of the error in this test : @xmath147^{1/2}\ ; , \label{macherr}\ ] ] where @xmath148 is the total number of the sph particles , and @xmath149 is the average particle velocity at time @xmath150 . the quantity in the square brackets is the velocity dispersion , of course . we can build a simple `` theory '' to estimate the error in these tests . we shall assume that velocity dispersion is driven by the stochastic fluctuations in the number of photon - sph interactions in different regions of the box . we shall also assume that fluctuations in these will be uncorrelated on a timescale of order the sound crossing time of the box , @xmath151 . during this time the mean number of the interactions between photons and a parcel of gas with size @xmath152 ( where @xmath55 is the smoothing length ) is @xmath153 where @xmath154 is the volume average density of photons in the box . the parcel is going to be accelerated to velocity @xmath155 during this time . each photon then accelerates the parcel by a velocity increment of @xmath156 as the number of photon - parcel interactions fluctuates by @xmath157 , we expect that random velocity fluctuations will be of the order of @xmath158 figure [ fig : slab_box_error ] compares the measured error and the one predicted by equation [ dispv ] for three runs . in the runs the sph particle number is fixed at @xmath159 , @xmath160 , @xmath161 , and the number of photon packets inside the box is varied from @xmath162 ( upper curve ) , through 300 ( middle , dashed curve ) to @xmath163 ( lower , dashed curve ) . the respective straight lines show the `` theoretical '' estimate given by equation [ dispv ] . the agreement is reassuringly good . one can also note that the error indeed saturates rather than grows with time , except for a few initial sound crossing times . this is due to the stochastic fluctuation becoming uncorrelated on dynamical time , as we assumed above . we also ran a large number of additional tests all with the same setup described in this section , but varying the number of sph particles up to @xmath164 , mean radiation pressure acceleration @xmath144 ( increasing it to @xmath165 and @xmath166 ) , and also varying the photon velocity @xmath71 by a factor of 2 in either direction . the resulting errors and their scalings were compared with that predicted by equation [ dispv ] , confirming the excellent agreement of the expected and measured errors further . equation [ dispv ] shows that , at the same @xmath144 ( equivalently , the same radiation flux ) , the velocity error scales as @xmath167 the scaling of @xmath168 with the number of photon packets as @xmath169 could be intuitively expected on the basis of poisson statistics of random fluctuations . the scaling of equation [ dispv ] with the smoothing length @xmath55 , which is proportional to @xmath170 , shows another important point . if a larger number of sph particles is used , resulting in a higher spatial resolution within the simulation volume , then the number of photon packets should also be increased ( if one wishes to keep the velocity errors within a given limit ) . consider optically thin gas in the vicinity of a very massive point mass , e.g. , a black hole , radiating exactly at the eddington limit . obviously , gas particles should experience no net force from the point source . in particular , if sph particle velocities are zero initially , they should remain zero , and any gas motion is an evidence of problems in the numerical methods used . for ease of analysis , we consider a self - gravitating gas cloud , with a polytrophic index @xmath171 . in the absence of external forces the cloud settles into a well known equilibrium configuration for polytropic stars . this equilibrium state is obtained by relaxing ( that is evolving ) the isolated cloud for many dynamical times to provide a noise - free initial condition . gravity is switched on in this test . however , as our focus is on testing the radiation pressure force , the gravitational force of the gas acting on the point mass is switched off . the radiation source is hence fixed in its initial location , which simplifies a quantitative analysis of the results below . the mass unit for the test is the mass of the point source , @xmath172 . the mass of the cloud is @xmath173 . the polytropic constant is such that the equilibrium size of the cloud , @xmath174 , is approximately 4 units of length . 25,000 sph particles are used to model the cloud . the tidal force from the central point is of the order of @xmath175 , where @xmath69 is distance to the point mass . there are two distinct regimes , then . in the first the tidal force acting on the cloud exceeds the self - gravity of the cloud , @xmath176 , whereas in the opposite regime the self - gravity of the cloud dominates over the tidal force . the former regime occurs when @xmath177 , the separation between the point mass and the centre of the cloud , is relatively small , @xmath178 . the case of a negligible tidal force , @xmath179 , occurs when the cloud is far away from the black hole , e.g. , @xmath180 . in these tests the cloud is initially positioned at @xmath25 and the emitter is at @xmath182 . at this separation , the tidal force is approximately 10 times greater than self - gravity of the cloud . thus , a @xmath183% error in the radiation pressure force is larger than the self - gravity holding the cloud together . these tests are hence expected to be very sensitive to numerical deficiencies of our method . we performed two tests in this setup . in the first , labelled lrz10 ( low resolution , @xmath181 ) , the dimensionless photon momentum is @xmath184 , whereas in the second , named hrz10 ( higher resolution ) , @xmath185 . the mass of the sph particles is the same in both tests , @xmath186 . the typical momentum of sph particles , defined as @xmath187 , is @xmath188 in code units . therefore , in both of these runs @xmath189 . figure [ fig : z10_standard ] shows the column density profile of the cloud at time @xmath190 for both runs . the left panel shows run lrz10 and the right panel shows hrz10 . time @xmath190 corresponds to about 3 free - fall times at the given source - cloud separation . in both cases there is a certain deformation of the cloud . as expected , the lower resolution test produces poorer results than the higher resolution test . a more detailed error analysis is described in [ sec : error_analysis ] below . in this test the separation of the cloud and the source is set to @xmath191 , thus we label the test z30 . the photon packet momentum is @xmath192 in the test . the parameters of the gas cloud are exactly the same as in tests lrz10 and hrz10 , but because the cloud is farther away from the point mass , typical sph momentum is a little lower , @xmath193 . thus in terms of ratio @xmath194 test z30 is very similar to lrz10 . however , since the cloud is farther away , the self - gravity of the cloud actually exceeds the tidal force by a factor of about 3 . therefore , one expects intuitively that the same level of stochastic error in the radiation pressure force will result in smaller perturbations as the radiation pressure force itself is smaller . this is indeed borne out by the test . by the time @xmath195 , which is again about three free - fall time scales from the clouds initial location , no visible deformation of the cloud has occurred . we therefore do not show the column density for test z30 . to quantify the accuracy of the tests , we calculate the centre of mass velocity of the cloud , @xmath196 as a function of time . using this in place of @xmath197 in equation [ macherr ] , we define the mean mach number of spurious velocity fluctuations in the cloud . figure [ fig : z10_standard_err ] shows these quantities as a function of time for lrz10 ( red curves ) and z30 ( black ) runs . the mach number of velocity fluctuations are shown with thick solid lines . the centre of mass velocities are shown normalised to @xmath198 , where @xmath199 , the circular keplerian velocity at the appropriate separation @xmath200 for the test . the motion of the cloud s centre of mass characterises the systematic error of the method . both runs demonstrate that the velocity of the centre of mass of the cloud is between 1 to 2 % of the keplerian velocity after about 3 free - fall times . this translates into inaccuracy in the calculation of the radiation pressure force versus point mass gravity , averaged over all sph particles , of a fraction of a percent . this accuracy should be sufficient for simulations in which gravity forces are calculated to a similar fractional precision . furthermore , we found that this offset , although always small for a large enough number of photon packets , actually depends on the random number generator used to approximate the isotropic photon field . the results can thus be improved by using quasi - random rather than pseudo - random sequences for photon packet s angular distribution . we shall investigate this issue in the future . the mach number of the fluctuations , instead , quantifies the difference in the radiative accelerations received by different parts of the cloud . we again use the simple logic of random stochastic fluctuations in the number of photon packets @xmath201 interacting with an sph particle with smoothing length @xmath55 , as explained in [ sec : slab ] . thin solid curves in figure [ fig : z10_standard_err ] show the mach number of velocity errors predicted by this simple argument . evidently , the errors are again explainable by the poisson noise estimates . in this test the initial condition is the same optically thin cloud as used in [ sec : static ] , but the point mass source is located exactly at the centre of the cloud , and allowed to accrete gas particles that are separated from it by less than @xmath202 . the radiation field in this test is halved to @xmath203 . while the analytical solution to this problem ( for the given initial density profile ) is not known to us , we note that the setup is exactly identical to a non - radiating sink particle with mass @xmath204 placed into the centre of the cloud and allowed to accrete the gas within the sink radius @xmath205 . the desired control result is thus obtained by running the hydro - gravity part of the code only with the sink particle mass @xmath204 . photon momentum is set to @xmath206 . figure [ fig : accretion1 ] presents both the radiation hydrodynamical simulation and the control hydro run with @xmath204 at dimensionless time @xmath207 . the lower panel shows the sph particle radial velocities normalised by the local free - fall velocity @xmath208 with @xmath204 . the upper panel shows the density profile of the cloud . both velocity and density profiles of the rhd simulation and the control run are nearly identical , although one can notice a larger amount of scatter in the @xmath209 test . the results continue to be very similar at latter times as well , and the accretion rate histories are also very similar . the radiation transfer method thus performs reasonably well . to quantify the errors of the tests , we estimate their magnitude based on the poisson noise arguments in the same manner as in [ sec : slab ] . in particular , we can apply equations [ dv1 ] and [ dispv ] , except that the characteristic time is now not @xmath151 but the local dynamical time , @xmath210 with @xmath211 ( and @xmath212 in the code units ) . thus , the number of photon packets passed through a given particle is estimated as @xmath213 . the stochastic velocity fluctuations due to the finite number of photon packets is then @xmath214 to compare the expected fluctuations given by equation [ dv_accretion ] with those actually incurred , we ran a poorer resolution test with @xmath108 , i.e. , ten times larger than the test shown in figure [ fig : accretion1 ] . the errors ( scatter ) are then obviously larger . figure [ fig : accretion_errors ] shows the results in black dots for two snapshots at times indicated . the red lines show the control run , as before , but this time velocity curves in the lower panel of the figure include the expected errors . in particular , the upper red sequence of dots is @xmath215 , whereas the lower red dots is @xmath216 , where @xmath168 is calculated with equation [ dv_accretion ] . evidently , the scatter in the radiation transfer simulation ( black dots ) is very similar in magnitude to what is predicted by equation [ dv_accretion ] . at the earlier time snapshot , shown on the left side of figure [ fig : accretion_errors ] , the measured velocity scatter is somewhat larger than our prediction . we believe this is due to the fact that equation [ dv_accretion ] does not include the poisson noise resulting from the initial distribution of photons packets at @xmath217 . the latter is generated by sampling a uniform random distribution in both emission time and directions . in conclusion , these tests demonstrate the potential of the method for accretion and radiation pressure problems , and that equation [ dv_accretion ] once again provides a reliable estimate of the stochastic errors of the results . a new method for implementation of radiation pressure force in an sph code has been presented . radiation is modelled via a time - dependent monte - carlo approach . photon packet s momentum is transferred to sph particles via direct photon - to - gas interactions that are calculated as photons propagate through the field of sph particles . the sph density field is used to calculate the sph density at photon packet s positions , thus maximising the accuracy of the method . as a result , local and global momentum conservation in the interactions between radiation and gas is achieved . several tests have been presented to check the consistency and accuracy of the method , as well as error scalings with parameters of the problem . it was found that the main deficiency of the method , as for any monte carlo method , is the poisson noise resulting from a finite number of photon packets used . in order to reduce velocity fluctuations to subsonic levels , photon packet s momentum @xmath14 should be chosen to be below the typical sph particle momentum , @xmath218 , where @xmath219 is the characteristic sph particle velocity . we shall now try to generalise and summarise the error estimates and various constraints on performance of the code . we first assume an optically thin case . let @xmath220 be the radiation flux at a given location in the gas . the flux is related to the photon number density @xmath221 , photon momentum @xmath14 and photon velocity @xmath71 through @xmath222 as the energy carried by the photon packet is @xmath223 . the rate at which photons are passing through sph particles is @xmath224 the momentum passed to the sph particle by one photon packet is @xmath225 and thus the velocity change @xmath226 is independent of sph mass @xmath227 ( in the optically thin approximation ) . therefore , at time @xmath150 the velocity fluctuations can be estimated as @xmath228^{1/2 } \sim \frac{\kappa}{h } \left[\frac{{\cal f}_{\rm rad } p_\gamma t}{c}\right]^{1/2}\ ; , \label{dvoft}\ ] ] now , let us define radiation acceleration time @xmath229 , where @xmath230 is the radiative acceleration acting on the sph particle , and @xmath231 is the radiation pressure force on the sph particle . during this characteristic time , the radiation pressure force would accelerate the sph particle from rest to velocity @xmath232 in the absence of other forces . using equation [ dvoft ] , the mach number of the fluctuations , defined by equation [ macherr ] , can be shown to be @xmath233^{1/2}\;.\ ] ] finally , introducing the `` optical depth '' of the sph particle as @xmath234 , we arrive at @xmath235^{1/2 } \left(\frac{t}{t_{\rm rad}}\right)^{1/2}\ ; , \label{merrfinal}\ ] ] where @xmath236 . this was derived in the optically thin limit , i.e. , when @xmath237 . in the limit @xmath238 , the momentum passed from one photon to an sph particle is reduced from the expression given by equation [ dmom ] to @xmath239 where @xmath240 is the number of sph neighbours for the photon . as the latter is always at least one when a photon packet interacts with gas , we arrive at the following estimate for the optically thick case , @xmath241^{1/2 } \left(\frac{t}{t_{\rm rad}}\right)^{1/2}\ ; , \label{merrthick}\ ] ] both the optically thin limit ( equation [ merrfinal ] ) and the optically thick limit ( equation [ merrthick ] ) demonstrate the importance of choosing photon packet momentum , @xmath87 to be significantly smaller than that of an sph particle to guarantee a reasonable accuracy . we also notice that applicability of the method to a particular problem depends on the desired level of accuracy . for example , tidal disruption of a gaseous cloud near a luminous super - massive black hole is a highly dynamic process . to get an insight in the overall dynamics of the process , it is sufficient to simulate the system for a few dynamical times at the peri - centre of the orbit . a sufficiently small photon packet s momentum , @xmath242 , should provide a sufficient accuracy in this case . however if one is interested in a secular evolution of a system , then the required precision is much higher , and careful tests should be done to establish the required value of @xmath14 , or equivalently the photon injection rate @xmath243 . let us now estimate the required number of photon packets for a given luminosity of the system @xmath244 , subject to the constraint @xmath245 , @xmath246 . number of packets emitted per unit time is @xmath247 where we exploited the fact that @xmath248 . @xmath249 here is the total gas mass of the system . the minimum time that the simulation should last for is the dynamical time , @xmath250 . during this time the number of photon packets emitted is @xmath251 , and thus @xmath252 where @xmath253 is the radiation energy emitted during the system light crossing time and @xmath254 is the total gas energy . if there are no additional sources of radiation in the cloud , such as bright stars , then the luminosity of the system is of the order of @xmath255 where @xmath256 is the cooling time . in that case the ratio is @xmath257 many astrophysically interesting problems are in the regime where the cooling time is comparable with dynamical time of the system , in which case the ratio becomes @xmath258 from these expressions it is obvious that the number of photons does not need to exceed the number of sph particles while satisfying the @xmath259 condition for these particular cases . however , if there is a very bright point source dominating the luminosity of the system , then the number of photon packets needs to be correspondingly higher . for example , in the tests presented in [ sec : static ] , photons produced during the duration of the simulations outnumbered sph particles by factors of @xmath260 . our radiation transfer method unfortunately becomes very expensive in optically thick medium , i.e. , when the optical depth of the system , @xmath261 , is very high . this is due to two factors . first of all , photons scatter or get absorbed / re - emitted @xmath262 ( that is many ) times before they exit the system . they thus `` hang around '' for longer before exiting the system , hence increasing computational cost of the simulation . secondly , they require smaller time steps as we now show . recall that in [ sec : prompt ] we argued that it is numerically permissible and beneficial to use photon packets with velocity @xmath71 smaller than the velocity of light in the following case . in moderately optically thick , non relativistic plasmas , photons pass through the system much quicker than dynamical time , and hence exact value of @xmath71 is unimportant . the photon time step can then be increased inversely proportional to @xmath71 . however , this is possible only as long as @xmath263 or else using too small a value for @xmath71 would incorrectly put the radiation transfer into the dynamic diffusion limit ( see [ sec : prompt ] ) . further , in the optically thick case the photon time step is limited to @xmath264 , where @xmath5 is the mean free path ( equation [ dtgamma ] ) . comparing this to dynamical time of the system , @xmath265 where we used equation [ vgamma ] to constrain @xmath71 . this expression changes to @xmath266 if one sets @xmath267 . obviously , in both of these cases the time step becomes very small as @xmath261 increases , and the code becomes quite inefficient . in problems with a very high optical depth combining the dynamic monte carlo and the diffusion approximation in very optically thick regions would be optimal , although it is not clear whether such `` on the fly '' method could be devised . a new time - dependent algorithm to model radiation transfer in sph simulations was presented . in the present paper we concentrated on the radiation pressure effects , i.e. , the momentum transfer between the radiation and the gas , assuming a given equation of state for the gas . we performed a number of tests of the code . our method is grid - less and can be applied to arbitrary geometries . the main disadvantage of the method , as with any photon packet based schemes , is the poisson noise due to a finite number of photons . on the other hand , these stochastic errors can be readily estimated and controlled by increasing the number of photon packets . the method is best applicable to optically thin or moderately optically thick systems , as following photon trajectories become very expensive at high optical depths . extension of the approach to include energy exchange between the gas and radiation will be presented in a future paper . multi - frequency radiation transfer can also be naturally added . the authors express gratitude to volker springel who provided the code gadget-3 , helped with the numerical implementation and provided useful comments on the paper . theoretical astrophysics research at the university of leicester is supported by a stfc rolling grant . h. j. g. l. m. , cassinelli j. p. , 1999 , introduction to stellar winds , introduction to stellar winds , by henny j. g. l. m. lamers and joseph p. cassinelli , pp . isbn 0521593980 . cambridge , uk : cambridge university press , june 1999 .

for a general metric measure space we give sufficient conditions to assure that the group of isometries and the group of measure - preserving isometries are lie groups . as a particular case , spaces satisfying the _ riemannian curvature - dimension condition _ , @xmath0spaces , fulfill the conditions . for certain classes of spaces it has been shown that the full group of isometries , @xmath2 , is a lie group . for example , myers and steenrod proved this fact for riemannian manifolds in @xcite , fukaya and yamaguchi for alexandrov spaces with curvature bounded by above and by below see @xcite and @xcite , and cheeger , colding , and naber in the case of ricci limit spaces in @xcite and @xcite . in contrast , there are examples of metric spaces for which @xmath2 is not a lie group . assume that @xmath3 is a fully supported measure in the remaining of the text . the main result of this article is : [ th:03 ] let @xmath4 be a locally compact , length metric measure space . assume a. @xmath5 has @xmath3-@xmath6 unique euclidean ( metric ) tangent cones , and that b. for every @xmath7 with @xmath8 there exists a unique optimal dynamical plan @xmath9 and it is induced by a map . then the group of measure - preserving isometries , @xmath10 , and the group of isometries , @xmath2 , are lie groups with respect to the compact - open topology . by a theorem of van danzig and van der waerden @xcite and lemma [ cl:01 ] we can conclude that @xmath2 and @xmath10 are compact if @xmath5 is compact . the essential property of assumption @xmath11 in theorem [ th:03 ] is that the set of metric spaces that might appear as tangent cones is compact , and that the isometry groups of the tangent cones are lie groups with a uniform positive lower bound on the size of its subgroups . consequently we can relax hypothesis @xmath11 by only requiring such a behavior of the tangent cones @xmath3-almost everywhere . we remark that the hypothesis of a fully supported measure is a natural requirement for metric measure spaces . nonetheless , we can drop this hypothesis in the theorem by demanding instead a geodesically convex support and that assumption @xmath12 holds for arbitrary @xmath13 rather than just for absolutely continuous final measures . for a comment on the choice of topology on the groups see remark [ re:01 ] . the so - called _ curvaturedimension conditions _ generalize the notion of a lower ricci curvature bound to metric measure spaces . the use of optimal mass transport theory to state the conditions is characteristic to all of them . the conditions are crucial refinements made from an intintal condition introduced first independently by lott and villani and by sturm in @xcite , and in @xcite , @xcite . the current shape of the curvature - dimension conditions is result of the works of various researchers . some important contributors are l. ambrosio , k. bacher , m. erbar , k. kuwada , n. gigli , a. mondino , t. rajala , g. savar , and k.t . sturm , for a historical recount one can consult for example @xcite and @xcite . we study spaces satisfying the _ riemannian curvature - dimension condition _ , the _ reduced curvature - dimension condition _ , and the _ measure contraction property_. we write for short @xmath14-condition , @xmath15-condition , and @xmath16 . among these conditions @xmath0-spaces reproduce behavior characteristic of riemannian manifolds better than the others , for instance see @xcite . on the other hand , the other two conditions allow spaces with finsleriann geometries @xcite . the @xmath16 is the weakest condition among them all . the inclusion relation between these spaces can be written as @xmath17 riemannian manifolds with weighted volume measure and alexandrov spaces with the hausdorff measure are examples of spaces that satisfy the hypotheses of theorem [ th:03 ] . furthermore , due to works of mondino - naber@xcite , and of rajala - sturm @xcite it s known that @xmath0-spaces also meet the assumptions of the main theorem . [ th:01 ] let @xmath18 , @xmath19 , and @xmath20 be an @xmath14-space . then the group @xmath10 is a lie group . [ th:02 ] let @xmath18 , @xmath19 , and @xmath20 be an @xmath14-space . then the group @xmath2 is a lie group . examples of @xmath0-spaces are alexandrov and ricci limit spaces with the hausdorff measure , and generalized cone constructions over @xmath0-spaces ( @xcite , @xcite , @xcite , @xcite , @xcite ) . however , it is not known whether the class of @xmath0-spaces is strictly bigger than that of ricci limit spaces . another class of examples of @xmath0-spaces comes from limits of weighted manifolds with a lower bound on the bakry - emery ricci tensor . hence the present work is a generalization of previous results . a question that arises is under which conditions the automorphism groups are lie groups for spaces satisfying different curvature - dimension conditions . as corollary of theorem [ th:03 ] we obtain that [ co:02 ] the groups @xmath2 and @xmath10 are lie groups for essentially non - branching @xmath15-spaces and essentially non - branching @xmath1-spaces that have @xmath3-@xmath6 euclidean tangents . the corollary follows from the results of @xcite , and @xcite where the existence and uniqueness of optimal geodesic maps is proved on these classes of spaces . furthermore , we also show that the @xmath1-condition by itself is not enough to guarantee a lie group behavior of such groups . this is concluded from lemma [ lm : mcp ] . [ co:03 ] there exists an @xmath21 space for which neither @xmath10 nor @xmath2 are lie groups . corollaries [ th:02 ] , [ co:02 ] and [ co:03 ] are new in the literature . corollary [ th:01 ] is a new result as well and it was also proved by guijarro and santos - rodrguez in @xcite which appeared during the completion of this manuscript . we describe now the basic ideas of the proof of theorem [ th:03 ] . a remarkable result of gleason@xcite and yamabe@xcite in the early 1950 s asserts that a locally compact , topological group is not a lie group if and only if every neighborhood of the identity has a non - trivial subgroup . and the remarks below . ] if a group has this property we say that it has the small subgroup property , @xmath22 , for short . supposing that @xmath2 is not a lie group , by using a blow up argument , we show that the assumption of the @xmath3-@xmath6 infinitesimal euclideanity and the @xmath22 imply the existence of non - trivial isometries with large measure of the fixed point set . this is shown in proposition [ pr:01 ] . the existence of a single non - trivial isometry with a fixed point set of positive measure violates the assumption of the uniqueness of the optimal geodesic maps . this is in contradiction with the assumptions , thus we conclude that @xmath2 is a lie group . finally , calling into play the closed subgroup theorem we show that @xmath10 is a lie group . the use of blow - up arguments is common in the proofs done for alexandrov and ricci limit spaces . the delicate point is to guarantee a non - trivial convergence ( or a convergence at all ) of subgroups of isometries acting on the sequence of scaled spaces . in comparison to alexandrov spaces it is not known that in our spaces geodesics do not branch . whereas in the case of ricci limit spaces a crucial step in the works of cheeger - colding and of colding - naber depends on the connectedness properties of the regular set . these are difficult results to obtain and there s no reason to expect that they hold in the general setting of theorem [ th:03 ] . therefore we proceed in a different manner , giving a short argument that use optimal transport tools and measure properties of the regular set . in the next section we summarize definitions and results that will be used . in section [ se : mms ] we find proposition [ pr:01 ] . the rest of the work needed to conclude the proof of theorem [ th:03 ] , and the proof itself is in section [ se : pr ] . in the end of the manuscript we present an example of an @xmath1-space where @xmath10 is not a lie group . the author would like to express his gratitude to jrgen jost for his crucial support and valuable advise and to rostislav matveev , and jim portegies for their significant involvement in this work . he thanks fernando galaz - garcia , martin kell , and tapio rajala for the observations and discussions that led to some of the remarks after theorem [ th:03 ] , as well as fabio cavalletti and andrea mondino for bringing @xcite to his attention . the author is thankful to nidhi kaihnsa and ruijun wu for commenting on earlier versions of the manuscript . this work was supported by the imprs program of the max planck institute for mathematics in the sciences and partially by conacyt . we set notation and compile definitions and results used in the paper . the text is mainly self - contained , however , we provide references for some more elaborate definitions to maintain brevity . readers familiar with the @xmath0-condition can skip the first and last subsection probably just taking a look at the remarks . in what follows we denote by @xmath20 a metric measure space , that is , a triple @xmath20 where * @xmath23 is a @xmath24 and , * @xmath3 is a _ non - negative borel measure finite on every bounded set_. we assume @xmath25 . a pointed metric measure space ( @xmath26 ) , @xmath27 , is a @xmath28 together with a base point @xmath29 . in the text a geodesics is a map , @xmath30 \to x$ ] , such that : @xmath31 where @xmath32 . we write @xmath33 for the space of all geodesics on @xmath5 endowed with the `` @xmath34 '' distance . a metric space is called a _ geodesic space _ if for every given pair of points @xmath35 there exists a geodesic , @xmath36 , that joins @xmath37 and @xmath38 . for @xmath39 $ ] define the _ evaluation map _ , @xmath40 , as @xmath41 for @xmath42 . the _ restriction map _ , @xmath43 , is defined as @xmath44 for @xmath45 , \gamma \in { \mathrm{geo}}(x)$ ] and the real function @xmath46 . we deal with group actions on sequences of @xmath26 . the pointed gromov - hausdorff ( @xmath47 ) and pointed equivariant gromov - hausdorff convergence ( @xmath48 ) provide a canonical type of convergence in this framework . for definitions we refer to @xcite and @xcite , @xcite . any @xmath47-limit of a sequence of scaled spaces , @xmath49 for @xmath50 , is called a ( metric ) . we denote the set of all tangent cones of @xmath5 at @xmath37 by @xmath51 . the existence or uniqueness of tangent cones is not guaranteed in a general setting . however , the set of points of @xmath5 with unique euclidean tangent cones is called the @xmath52 @xmath53 of @xmath5 , written as @xmath54 . in detail , @xmath55 we say that @xmath20 has _ @xmath3-@xmath6 euclidean tangents _ if the set of numbers @xmath56 is finite and if @xmath57 . for a fixed @xmath58 , the @xmath59-@xmath52 @xmath53 , @xmath60 , is the set @xmath61 , where for a given @xmath62 the set @xmath63 is defined as all points @xmath29 for which there exists a @xmath64 such that @xmath65 above @xmath66 is the gromov - housdorff distance and @xmath67 is the ball of radius @xmath68 around @xmath69 . note that @xmath70 and that for every @xmath71 the measure @xmath72=0 if @xmath5 has @xmath3-@xmath6 euclidean tangents . we write @xmath73 and @xmath74 for the support and the _ pushforward measure _ of a measure @xmath75 . two @xmath28 @xmath76 , @xmath77 are isomorphic if there exists an isometry @xmath78 we use the word @xmath79 to make reference to usual metric isometries . in contrast , we refer to maps satisfying ( [ def : iso ] ) as _ measure - preserving isometries_. particularly , we note that an isometry is defined on the whole space @xmath80 and does not necessarily satisfy ( [ def : iso ] ) . by definition @xmath20 is always isomorphic to @xmath81 . this induces a canonical equivalence class of isometric metric measure spaces where * only the support of the measure is relevant*. we assume that @xmath82 . this is a natural restriction in the class of isomorphisms of @xmath28 . we endow the groups @xmath2 and @xmath10 with the compact - open topology , making them topological groups ( @xcite pp.46 ) . [ re:01 ] we explain and motivate our choice of topology for @xmath10 . for locally compact metric spaces it s natural to endow @xmath2 with the compact - open topology since the structure under study is of pure metric nature . in addition , in this context , the rigidity of the isometries assures that pointwise convergence implies convergence w.r.t . the compact - open topology . being second - countable . in turn , second - countability of @xmath2 is concluded form the fact that @xmath5 is a locally compact metric space . consult for instance @xcite pp.46.][footnote ] alternatively , on @xmath28 and @xmath10 there is an additional structure of interest , namely , the measure structure . however , as we explain below , once again the rigidity of the measure - preserving isometries guarantee that any reasonable choice of topology on @xmath10 coincides with the compact - open topology . observe that topology that only considers the measure structure is too coarse for our purposes since it does nt see metric properties . therefore , a logical way to proceed is to couple a measure - wise and a metric - wise convergence . however , as we explained , the weakest metric convergence , the pointwise convergence , coincides with the compact - open convergence . on the other hand , in lemma ( [ cl:01 ] ) we prove that for a locally compact @xmath28 compact - open convergence of measure - preserving isometries implies the weak convergence of the pushforward measures . this motivates our choice of topology . denote by @xmath83 the identity component of @xmath84 , that is , the largest connected set containing the identity element @xmath85 . as definition we say that @xmath84 is a lie group if and only if the identity component @xmath83 is a lie group and if @xmath86 is discrete . see remark [ rm : car ] ] recall that for a connected , locally compact metric space @xmath2 is a topological group with respect to the compact - open topology . [ th : da - wa ] let @xmath87 be a connected , locally compact metric space . then @xmath2 is locally compact with respect to the compact - open topology . furthermore if @xmath5 is compact , then @xmath2 is compact . a topological group has the @xmath88 , @xmath22 , if every neighborhood of the identity contains a non - trivial subgroup . if there exists a neighborhood of the identity with no non - trivial subgroup we say that @xmath84 has the @xmath89 @xmath90 , @xmath91 for short . below we cite an outstanding result that characterizes lie groups in terms of the @xmath91 . [ th : gl - ya ] let @xmath84 be a locally compact , topological group . then @xmath84 is a lie group if and only if it has the no small subgroups property . in @xcite yamabe generalizes gleason s theorem to the infinite dimensional case , however , @xmath84 is assumed to be connected . we present an argument , due to an undisclosed russian mathematician , which shows that we can also consider non - connected groups . an equivalent way of stating theorem [ th : gl - ya ] is : assuming the same hypothesis , then there exists an open subgroup @xmath92 such that for every neighborhood of @xmath93 there exists a normal subgroup @xmath94 that makes @xmath95 a lie group @xcite . if @xmath84 has the @xmath91 the only small normal subgroup is @xmath85 itself , thus making @xmath96 a lie group which by definition means that @xmath97 is discrete . however , this implies that @xmath98 is discrete since @xmath99 is discrete and @xmath100 . [ rm : car ] we make another observation regarding the cardinality of @xmath86 . in principle the group of components might be uncountable . fortunately , we can also discard this behavior . assume that @xmath84 is a second - countable lie group . by definition @xmath86 is discrete which is equivalent to @xmath83 being open . in turn , this implies that the quotient map is open and it follows that @xmath86 is second - countable since by assumption @xmath84 is second - countable . a second - countable space is separable , and discrete separable spaces are countable . to end , recall that in a locally compact , connected metric space @xmath2 is second - countable . the following result due to cartan will be useful to us . [ th : ca ] let @xmath101 be a closed subgroup of a lie group @xmath84 . then @xmath101 is an embedded lie group . curvature - dimension conditions require a type of convexity condition of an entropy functional defined on the space of probability measures of a metric measure space . the different versions of the curvature - dimension conditions arise from different versions of the convexity type condition . however , all curvature - dimension conditions are compatible with the smooth framework and are stable under pointed measured - gromov - hausdorff convergence . optimal transport theory provides the appropriate tools to formalize the ideas . let @xmath102 be the space of probability measures defined on @xmath87 and @xmath103 the subspace of measures with finite second moments . for @xmath104 the wasserstein squared distance is defined as @xmath105 the infimum taken over all measures @xmath106 with first and second marginals equal to @xmath107 and @xmath108 respectively . if there exists a measurable function @xmath109 such that the measure @xmath110 is a minimum we call @xmath111 an _ optimal map_. given @xmath104 define the set @xmath112 as the set of all measures @xmath113 such that the pushforward @xmath114 realizes the minimum in ( [ eq : w - d ] ) . a measure @xmath115 is called an _ optimal geodesic plan _ and if such a measure @xmath113 is the lift of an optimal map we call it an _ optimal geodesic map_. next we define for @xmath116 and @xmath117 the distortion coefficient , @xmath118\times { \mathbb r}^+\to { \mathbb r}\cup\infty$ ] , as : @xmath119 we ve decided not to include the definition of the @xmath0-condition since it s rather technical and we do nt use it explicitly . alternatively , we present the results that show that @xmath14-spaces satisfy the assumptions of theorem [ th:03 ] . nevertheless , before let s recall that the condition was introduced in @xcite and further developed in @xcite and @xcite to which the reader is referred to for a comprehensive discussion on the subject . the condition itself couples a curvature - dimension condition with an infinitesimal riemannian behavior known as infinitesimally hilbertianity . [ th : gi - ra - st ] let @xmath18 , and @xmath120 $ ] and @xmath20 be an @xmath14 space . then for every @xmath104 , @xmath121 there exist a unique optimal geodesic plan @xmath9 . furthermore , there exists @xmath122 with @xmath123 such that the map @xmath124 is injective for all @xmath125 . the existence of @xmath126 euclidean tangents in @xmath0-spaces was proved in @xcite , later the uniqueness was shown by mondino and naber as a byproduct of the rectifiability of @xmath14-spaces as metric spaces ( compare with @xcite for rectifiability as @xmath28 ) . a weaker version of their result is enough for us . [ th : mo - na ] let @xmath127 , @xmath128 and @xmath20 be an @xmath14 space . then @xmath129 has @xmath3-@xmath6 euclidean ( metric ) tangents , i.e. @xmath57 . in the last section we will work with ohta s definition of the @xmath1-condition . intuitively , it requires contraction properties of measures as the @xmath0-condition but only for starting @xmath130 measures . the condition is stable w.r.t . to @xmath131-convergence @xcite . [ def : mcp ] for @xmath132 , @xmath133 , or @xmath134 and @xmath135 a @xmath28 , @xmath20 , has the @xmath136-measure contraction property , @xmath137 , if the following conditions are satisfied . for every point @xmath138 and a measurable set @xmath139 with @xmath140 ( and @xmath141 if @xmath142 ) there exists a probability measure @xmath113 on @xmath33 such that @xmath143 , @xmath144 , and @xmath145.\ ] ] below we show the existence of isometries with large fixed point set , @xmath146 , if @xmath2 has the small subgroup property . define for @xmath147 , @xmath138 , and a subgroup @xmath148 : @xmath149 for fixed @xmath150 , the function @xmath151 is continuous in @xmath68 and @xmath37 as long as every closed ball in @xmath5 is the closure of its interior , as noted in @xcite . this holds true when @xmath5 is a length space , for example . we do not assume a fully supported measure in the coming proposition . [ pr:01 ] let @xmath152 be a @xmath28 where every closed ball coincides with the closure of its interior . assume that @xmath25 and that @xmath5 has @xmath3-@xmath6 euclidean tangents . if @xmath2 has the small subgroups property , then for every @xmath29 , @xmath153 , and @xmath154 there exist a non - trivial subgroup @xmath155 such that for every @xmath156 @xmath157 we assume that @xmath158 since the inequality above trivially holds true if @xmath159 . we will argue by contradiction . the strategy is the following : assuming that inequality ( [ eq : resultado ] ) does nt hold we will find for every @xmath160 a quadruple @xmath161 with the following properties : @xmath162 the existence of such a family of quadruples leads to a contradiction as we explain now . observe that if for every @xmath160 there exists a quadruple as above , then for a sequence @xmath163 we can pick a converging subsequence @xmath164 , denoted in the same way , such that the scaled spaces converge in the @xmath165-sense to @xmath166 where @xmath167 , and @xmath168 is a non - trivial subgroup satisfying @xmath169 . this creates the contradiction , since there does not exist a non - trivial subgroup of euclidean isometries , @xmath170 , such that @xmath171 . therefore , in order to complete the proof it s enough to show the existence of such a family of quadruples as was just argued . we proceed to construct such a family . suppose that ( [ eq : resultado ] ) does not hold . that is , there exist @xmath138 , @xmath153 , and @xmath154 such that for every non - trivial subgroup @xmath172 there exists an @xmath173 where @xmath174 note that necessarily@xmath175 . take @xmath176 and choose small enough @xmath177 so that @xmath178 , and @xmath179 the @xmath3-@xmath6 euclidean tangents of @xmath5 , together with the continuity from below of the measure and the fact that @xmath180 make possible the choice of such a @xmath181 . indeed , since for @xmath182 it holds that @xmath183 we can write @xmath184 as a countable union of sets . now just notice that @xmath185 choose @xmath186 big enough and take @xmath187 . inequality ( [ eq : regular grande ] ) combined with the @xmath188 assumption imply that for every non - trivial @xmath189 there exist @xmath190 such that @xmath191 on the other hand , we can find a non - trivial small subgroup @xmath192 because of the small subgroups property of @xmath2 . in particular , because of inequality ( [ fixed smaller regular ] ) , there exists @xmath193 and @xmath194 such that @xmath195 denote by @xmath196 . by construction it follows that @xmath197 finally , the continuity of @xmath198 and the intermediate value theorem imply that there exists @xmath199 such that @xmath200 , and @xmath201 ( see figure [ figura 1 ] ) . hence for @xmath160 we have found a quadruple @xmath202 satisfying ( [ eq : condiciones ] ) which finishes the proof . ( -0.2,0 ) ( 4.2,0 ) node[right]@xmath203 ; ( 0,-.2 ) ( 0,4.2 ) ; ( 1,0 ) node[below ] @xmath204 ; ( 3.8,0 ) node[below ] @xmath181 ; plot ( x_,x _ ) node[right ] @xmath205 ; ( 0,1 ) to[out=20,in=180 ] ( 2,2 ) ; ( 2,2 ) to[out=0,in=230 ] ( 4.2,3 ) ; ( 4.2,3 ) node[right ] @xmath206 ; ( 1,0 ) ( 1,1.69 ) ; ( 3.8,0 ) ( 3.8,2.6 ) ; the result can be interpreted in the following manner . suppose that @xmath20 satisfies the hypotheses of the proposition [ pr:01 ] . then for every @xmath207 there exist an isometry @xmath208 with @xmath209 . we start this section by proving a lemma that shows that the uniqueness of optimal geodesic maps is sufficient to guarantee that non - trivial isometries have fixed - point sets of measure zero . [ lm:02 ] let @xmath20 be a @xmath28 such that for every @xmath7 with @xmath121 there exists a unique optimal geodesic plan @xmath9 . furthermore , assume that @xmath113 is concentrated on a set of geodesics , @xmath122 , such that the map @xmath210 is injective . let @xmath175 be an isometry of @xmath5 . then @xmath211 . we proceed by contradiction . suppose that there exist @xmath212 , and a set @xmath213 with positive measure . let @xmath214 and define the probability measures @xmath215 and @xmath216 . we denote by @xmath9 the unique geodesic plan between @xmath107 and @xmath108 . let @xmath122 be the set where @xmath113 is concentrated and where @xmath217 is injective . set : @xmath218 @xmath219 is the subset of geodesics of @xmath220 that end in @xmath221 and @xmath222 is the projection of @xmath223 onto the set @xmath224 . we have that none of these sets are empty , that the measures @xmath225 and that @xmath226 . the last fact is a consequence of the injectivity of @xmath227 . we define now the measure @xmath228 as @xmath229{\kern-.6pt\bigwedge\kern-.6pt } { \rule[-\textheight/2]{1ex}{\textheight } } } { \textheight}}{0.5ex}}\stackon[1pt]{{f^{-1}}}{\tmpbox}})_{\#}\pi|_{\gamma_2},\ ] ] where the bijection of @xmath33 , @xmath230 , induced by some @xmath231 is written as @xmath232 . the measure @xmath233 is a symmetric analog of @xmath113 but @xmath234 . indeed , note that @xmath235 because @xmath236 by construction . we claim that @xmath237 is also a dynamical plan . this would contradict the hypothesis of the uniqueness of @xmath113 and finish the proof of the lemma . we proceed to verify the claim . we need to show that @xmath233 minimizes @xmath238 . the minimum taken over all measures @xmath239 such that @xmath240 for @xmath241 . we check that the pushforwards of @xmath233 under the evaluation map are as above . for this we observe that for @xmath231 , @xmath242 , and @xmath243 $ ] @xmath244 and that @xmath245{\kern-.6pt\bigwedge\kern-.6pt } { \rule[-\textheight/2]{1ex}{\textheight } } } { \textheight}}{0.5ex}}\stackon[1pt]{g^{-1}}{\tmpbox}}$ ] . next we compute the pushforward of @xmath233 under @xmath246 : @xmath247{\kern-.6pt\bigwedge\kern-.6pt } { \rule[-\textheight/2]{1ex}{\textheight } } } { \textheight}}{0.5ex}}\stackon[1pt]{f^{-1}}{\tmpbox}})_{\#}\pi|_{\gamma_2 } \\ & = & ( f)_\ # ( e_t)_{\#}\pi|_{\gamma_1}+(f^{-1})_\ # ( e_t)_{\#}\pi|_{\gamma_2}.\end{aligned}\ ] ] then @xmath248 since @xmath249 . as for the other pushforward we have that @xmath250 . to finish we see that @xmath237 by showing that the value of @xmath251 is the minimum of the functional . @xmath252{\kern-.6pt\bigwedge\kern-.6pt } { \rule[-\textheight/2]{1ex}{\textheight } } } { \textheight}}{0.5ex}}\stackon[1pt]{f^{-1}}{\tmpbox}})_{\#}\pi|_{\gamma_2 } \right)(\gamma)\\ & = & \int_{{\mathrm{geo}}(x ) } l\circ\hat{f}(\gamma)\cdot \chi_{\gamma_1 } ( \gamma ) { \mathrm{d}}\pi ( \gamma ) + l\circ { \savestack{\tmpbox}{\stretchto { \scaleto { \scalerel*[\widthof{\ensuremath{f^{-1}}}]{\kern-.6pt\bigwedge\kern-.6pt } { \rule[-\textheight/2]{1ex}{\textheight } } } { \textheight}}{0.5ex}}\stackon[1pt]{f^{-1}}{\tmpbox}}(\gamma ) \cdot \chi_{\gamma_2}(\gamma){\mathrm{d}}\pi ( \gamma)\\ & = & \int_{{\mathrm{geo}}(x ) } l(\gamma ) \cdot ( \chi_{\gamma_1}+\chi_{\gamma_2})(\gamma){\mathrm{d}}\pi ( \gamma ) = \int_{{\mathrm{geo}}(x ) } l(\gamma ) { \mathrm{d}}\pi(\gamma).\end{aligned}\ ] ] [ rmk : weaken ] the hypothesis in lemma [ lm:02 ] can be we weaken . we may require the existence of the unique geodesic plan only for final measures satisfying @xmath253 rather than for an arbitrary @xmath254 . we can repeat the proof choosing as final measure @xmath255 where @xmath256 is a sufficiently small ball . consistently with theorem [ th : gi - ra - st ] we get the next result . recall that for @xmath19 it holds that @xmath257 . [ co:01 ] let @xmath20 be an @xmath258-space . if @xmath259 for some @xmath260 then @xmath261 . next we show that @xmath10 is a closed subgroup of @xmath2 . [ cl:01 ] let @xmath20 be a connected , locally compact @xmath28 . then @xmath10 is a locally compact closed subgroup of @xmath2 with respect to the compact - open topology . we show that @xmath10 is closed . the local compactness of @xmath10 follows from the fact that @xmath10 is a closed subgroup of a locally compact group . let @xmath262 be a converging sequence w.r.t . the compact - open topology with limit @xmath263 . it is easy to see that @xmath264 is an isometry . thus , to finish the proof , it remains to check that @xmath265 . this follows from the regularity of the measure , as we argue below . indeed , since the measures @xmath266 are all equal , they trivially converge weakly to @xmath3 . on the other hand we will show that the pushforward of @xmath3 under @xmath267 weakly converges to the measure @xmath268 . therefore , @xmath265 by the uniqueness of the limit . by using the definition of the pushforward and the continuity of @xmath269 , it is enough to verify that for every bounded continuous function with bounded support , @xmath270 , it holds that @xmath271 to show that @xmath272 . after the following observation it is clear that this last equality holds . assuming that @xmath273 is as above we can construct an @xmath3-integrable function , @xmath84 , such that @xmath274 for all @xmath29 and make use of the dominated convergence theorem . take for example the multiple of the characteristic function @xmath275 , where @xmath276 is a bound on @xmath273 and @xmath277 and @xmath278 are such that @xmath279 . the existence of such a pair @xmath280 is guaranteed because @xmath273 has bounded support , and because @xmath281 converges uniformly in compact subsets . the integrability of @xmath84 follows from @xmath3 being finite on bounded sets . we give the proof of the main result . compare theorem [ th:03 ] to theorem 4.5 in @xcite . we begin by recalling that a locally compact , complete length space is a geodesic space . then @xmath20 satisfies the assumptions of lemma [ lm:02 ] and proposition [ pr:01 ] . for now we focus on the isometry group @xmath2 . lemma [ lm:02 ] asserts that the fixed point set of any non - trivial isometry has measure zero . since there does not exist a non - trivial isometry with large fixed point set @xmath2 has the no small subgroup property , this is the contrapositive to proposition [ pr:01 ] . finally , because @xmath2 is locally compact , the @xmath91 is equivalent to @xmath2 being a lie group as stated in theorem [ th : gl - ya ] . the fact that @xmath10 is a lie group is concluded by using the closed subgroup theorem . we are able to use the closed subgroup theorem since we have proved that @xmath282 is a closed subgroup in lemma [ cl:01 ] . we show that the @xmath1-condition is not strong enough to guarantee that the group of measure - preserving isometries is a lie group . we also see that for a geodesic and compact @xmath28 with finite measure @xmath10 might fail to be a lie group . we start by presenting a well - known example to develop intuition about the connection between @xmath10 not being a lie group , @xmath10 having small subgroups , and the branching of geodesics . denote the circle of radius @xmath68 by @xmath283 . the hawaiian earring , @xmath284 , is the space we obtain after gluing the circles @xmath285 by identifying one point of every circle , see figure [ fg : ex ] . endow @xmath284 with the arc - length distance @xmath286 and the 1-dimensional hausdorff measure @xmath287 . this makes @xmath288 a compact , geodesic metric measure space with finite measure . observe that @xmath289 where the compact - open topology coincides with the product topology . hence @xmath2 is totally disconnected but not discrete . by definition , @xmath2 is not a lie group since @xmath290 is not discrete . ( 8.5,2)(8,1.8 ) ; ( 8.5,2)(10.5,2)(10.9,2 ) ; ( 11.4,2)(10.9,2 ) ; ( 8.5,2.13)(8,1.8 ) ; ( 8.5,2.13)(9,2.2)(10.5,2 ) ; ( 11.4,1.8)(10.9,2 ) ; ( 8.5,2.26)(8,1.8 ) ; ( 8.5,2.26)(9,2.4)(10.5,2 ) ; ( 11.4,1.9)(10.9,2 ) ; ( 8.5,2.4)(8,1.8 ) ; ( 8.5,2.4)(9,2.6)(10.5,2 ) ; ( 11.4,1.95)(10.9,2 ) ; ( 8.5,1.9)(8,1.8 ) ; ( 8.5,1.9)(9,1.85)(10.5,2 ) ; ( 11.4,1.85)(10.9,2 ) ; ( 8.5,1.8)(8,1.8 ) ; ( 8.5,1.8)(9,1.7)(10.5,2 ) ; ( 11.4,2.1)(10.9,2 ) ; ( 11.4,2.05)(10.9,2 ) ; ( 8.5,1.8 ) ( 8.5,2.4 ) ; ( 11.4,1.8 ) ( 11.4,2.1 ) ; consider the segment @xmath296\times \{0\ } \subset { \mathbb r}^2 $ ] and define the set @xmath297 by replacing the segment @xmath298\times \{0\ } \subset y^{n-1 } $ ] with a 2-dimensional diamond - shaped set @xmath299 for @xmath300 , see figure [ fg : ex ] . the @xmath301-th diamond @xmath299 is chosen to such that it does not intersect @xmath302 , for @xmath303 . specifically , the sets @xmath299 are the 2-dimensional symmetric diamond - shaped sets of width @xmath304 , height @xmath305 , and center @xmath306 given by @xmath307 for @xmath308 , and @xmath309 . we then give @xmath310 an @xmath28-structure and define @xmath311 the existence of the limit follows from the compactness of @xmath1-spaces since lemma [ lm : mcp ] shows that @xmath310 satisfies the @xmath21-condition for every @xmath312 . we fix the distance , @xmath313 , coming from the @xmath314-norm on @xmath297 for every @xmath315 . to fix a measure define @xmath316 on @xmath302 by @xmath317^{-1 } \cos^2(x)\ , \chi|_{d_k}(x)\qquad \hfill \text{for } k\in { \mathbb n},\ ] ] and @xmath318 the characteristic function of the set @xmath224 . denote by @xmath319 and by @xmath320 . we set on @xmath297 the measure @xmath321 defined as @xmath322 in words , @xmath321 has a 2-dimensional contribution from @xmath323 with constant density for fixed @xmath37-coordinate , and a 1-dimensional contribution from @xmath324 which is absolutely continuous w.r.t . the 1-dimensional hausdorff measure . we fix some notation before the next lemma . given an @xmath301-tuple @xmath325 consider the the projected @xmath326-diamond space , @xmath327 , that we obtain from the sequence @xmath328 for some @xmath329 . that is , the @xmath326-diamond space @xmath327 is the space we obtain by replacing the @xmath330th - diamond with a segment inside @xmath310 . we write @xmath331 , and define the height as @xmath332 for @xmath333 and @xmath334 . we also define @xmath335 the set @xmath336 consists of euclidean geodesics that go from @xmath337 to @xmath338 . for @xmath339 and @xmath340 define @xmath341 as the geodesic obtained after gluing @xmath342 with @xmath343 and reparametrizing . the image of @xmath344 is the union of a line segment going from the left vertex of @xmath302 to @xmath345 with its reflection over @xmath346 . define @xmath347 as the set of all such geodesics . because of the stability of the @xmath1-condition it is enough to show that @xmath310 @xmath349 for every @xmath312 and every pair of @xmath350-tuples @xmath351 as above . we proceed to check the condition borrowing key ideas from the proof of @xcite . fix @xmath352 and @xmath350-tuple . definition [ def : mcp ] requires that for every @xmath353 and @xmath139 with @xmath140 we give @xmath354 such that @xmath355 , @xmath356 and , @xmath357 we will define such a @xmath113 as the uniformly distributed probability measure on a carefully chosen set @xmath358 . we start by noting that we can analyze separately the sets @xmath359 for fixed @xmath360 . the simplification can be made because we will pick geodesics with projection @xmath361 , for @xmath362 . the choice implies that the first coordinate contributes to the dilatation of the measure @xmath363 a factor equal to @xmath203 . because of this we are able to reduce the analysis to estimating separately the dilatation of the sets @xmath364 for every @xmath365 . accordingly , to verify the @xmath21-condition , it is enough to provide a @xmath366 such that @xmath367 , @xmath368 , and @xmath369,\ , \gamma \in \gamma.\ ] ] we fix now an @xmath370 . suppose the exists some @xmath373 for which @xmath374 . we can assure that the density @xmath375 is independent of @xmath376 for @xmath377 by transporting the measure , @xmath378 , along geodesics whose restriction to @xmath302 is exactly the set @xmath347 . this implies that there is no contribution to the contraction of the measure coming from the @xmath38-coordinate . hence the density @xmath379 changes as if we were transporting measure inside the projected @xmath326-diamond space @xmath380 and therefore the density of @xmath378 would satisfy inequality ( [ eq : mcp ex ] ) if the @xmath326-diamond spaces @xmath381 . however , if there exists @xmath382 such that @xmath383 we can repeat the argument and estimate the density @xmath379 as if we were transporting the measure in an @xmath384-diamond space . finally we just realize that for every initial @xmath385 we can keep finding such @xmath330 s until we have projected to an @xmath386-diamond space @xmath387 for @xmath372 . we have that the segment @xmath388 because of@xcite , and in @xcite it is shown that @xmath389 satisfies the @xmath21-condition , this covers the cases of @xmath390 so we move to @xmath391 . we assume that @xmath392 because of symmetry and fix @xmath393 and @xmath394 . we can conclude from the preceding claim that the only situation left to check is that of @xmath395 and @xmath396 . the way in which we transport the measure is the following . we start by expanding the measure uniformly from @xmath397 to a set @xmath398 with the same relative height as @xmath364 . then we transport the measure from @xmath398 to @xmath399 without changing the relative height of the set @xmath400 with respect to @xmath401 . we continue through @xmath402 and expand again keeping the heights ratio constant from @xmath403 to @xmath364 . the image of a transporting geodesic is the union of segments of straight lines described below , see figure [ fg : ex ] . to define @xmath220 first choose any set @xmath404 such that @xmath405 for @xmath406 . write @xmath407 , @xmath408 , and @xmath409 for the times at which the @xmath37-coordinate of any geodesic @xmath410 is equal to @xmath411 , @xmath399 , and @xmath403 . geodesics in @xmath412 have the same length . now define @xmath220 as the set of all geodesics satisfying the following : @xmath413 , @xmath414 , @xmath415 , and @xmath416 , where @xmath417 is the subset of geodesics of @xmath418 that cross through @xmath364 . we estimate the density of the corresponding measure , for @xmath419 we have that @xmath420 for @xmath421 . the shape of the diamond @xmath422 allows to estimate @xmath423 . we can bound the time when the geodesics reach @xmath411 by @xmath424 , and the length of the geodesics is necessarily @xmath425 . moreover in @xcite the estimate @xmath426 for all @xmath427 \times ( 0 , \pi/2 + 1/4)$ ] is proved . putting inequalities together we obtain inequality ( [ eq : mcp ex ] ) for @xmath428 $ ] . to finish note that for @xmath429 $ ] , @xmath378 is uniformly distributed in the @xmath38-coordinate . thus , its density is equal to the one of the transport in the projected space @xmath389 , which is a @xmath430-space . this shows that inequality ( [ eq : mcp ex ] ) is satisfied for @xmath429 $ ] . observe that the automorphism groups of @xmath431 are @xmath432 . thus we obtain corollary @xmath433 , confirming that the measure contraction property is not strong enough to guarantee that the isometry group or the measure - preserving isometry group are lie groups .

the presented work takes place in a more general framework : the modelling of unsteady mixed flows in any kind of closed pipe taking into account the cavitation problem and air entrapment . we are interested in flows occurring in closed pipes with non uniform sections , where some parts of the flow can be free surface ( it means that only a part of the pipe is filled ) and other parts are pressurized ( it means that the pipe is full ) . the transition phenomenon between the two types of flows occurs in many situations such as storm sewers , waste or supply pipes in hydroelectric installations . it can be induced by sudden changes in the boundary conditions as failure pumping . during this process , the pressure can reach severe values and may cause damages . the simulation of such a phenomenon is thus a major challenge and a great amount of works was devoted to it these last years ( see @xcite,@xcite,@xcite,@xcite , for instance ) . the classical shallow water equations are commonly used to describe free surface flows in open channels . they are also used in the study of mixed flows using the preissman slot artefact ( see for example @xcite ) . however , this technic does not take into account the depressurisation phenomenon which occurs during a water hammer except in recent works @xcite . on the other hand the allievi equations , commonly used to describe pressurized flows , are written in a non - conservative form which is not well adapted to a natural coupling with the shallow water equations . a model for the unsteady mixed water flows in closed pipes and a finite volume discretisation have been previously studied by two of the authors @xcite and a kinetic formulation has been proposed in @xcite . we propose here the * pfs*-model which tends to extend naturally the work in @xcite in the case of a closed pipe with non uniform section . for the sake of simplicity , we do not deal with the deformation of the domain induced by the change of pressure . we will consider only an infinitely rigid pipe . the paper is organized as follows . section 2 is devoted to the derivation of the free surface model from the @xmath0d _ incompressible _ euler equations which are written in a suitable local reference frame ( following @xcite ) in order to take into account the local effects produced by the changes of section and the slope variation . the construction of the free surface model is done by a formal asymptotic analysis . seeking for an approximation at first order gives the model called * fs*-model . in section [ sectionpfmodel ] , we adapt the derivation of the * fs*-model to derive the pressurized model , called * p*-model , from the @xmath0d _ compressible _ euler equations . writing the source terms of these two models , * p * and * fs*-model , into a unified form and using the same couple of conservative unknowns as in @xcite , we propose in section [ sectioncoupling ] a model for mixed flows , that we call * pfs*-model . we state some mathematical properties of this model . section [ pfsroe ] is devoted to the extension of the vfroe scheme described in @xcite that was used for the case of uniform pipes . in section [ sectionwb ] , we show how to construct a convection matrix in order to get an exactly well - balanced scheme . several numerical tests are presented in section [ numericaltests ] . the classical shallow water equations are used to describe physical situations like rivers , coastal domains , oceans and sedimentation problems . these equations are obtained from the incompressible euler system ( see e.g. @xcite ) or from the incompressible navier - stokes system ( see for instance @xcite ) by several techniques ( e.g. by direct integration or asymptotic analysis ) . we adapt here the derivation in @xcite to get a new unidirectional shallow water model . we start from the @xmath0d incompressible euler equations where we neglect the acceleration following the @xmath1-axis supposing the existence of a privileged main flow axis . we write then the euler equations in the local serret - frenet reference frame in order to take into account the local effects produced by the changes of section and the slope variation . then we derive a shallow water model by a formal asymptotic analysis ( done in subsection [ sectionfsmodelasymptotic ] ) . let us consider the cartesian reference frame @xmath2 . in the corresponding coordinate system @xmath3 , the @xmath0d incompressible euler system writes : @xmath4 where @xmath5 denotes the velocity with components @xmath6 , @xmath7 is the isotropic pressure tensor , @xmath8 the density of the fluid at atmospheric pressure @xmath9 and @xmath10 is the exterior strength of gravity . we close classically system ( [ e3dicartesian ] ) using a kinematic law for the evolution of the free surface : _ any free surface particle is advected by the fluid velocity _ @xmath11 and on the wet boundary , we assume the no - leak condition @xmath12 where @xmath13 is the outward unit normal vector to the wet part of the pipe ( see f. [ supplementaire ] ) . we set the pressure @xmath14 to @xmath15 at the free surface . we define the domain @xmath16 of the flow at time @xmath17 as the union of sections @xmath18 ( assumed to be simply connected compact sets ) orthogonal to some plane curve @xmath19 lying in @xmath20 to follow the privileged main flow axis . we choose the parametrization @xmath21 in the cartesian reference frame @xmath2 where @xmath22 follows the vertical direction ; @xmath23 is then the elevation of the point @xmath24 over the plane @xmath25 ( see f. [ oxoz ] ) . we define a local reference frame as follows : we introduce the curvilinear variable defined by : @xmath26 where @xmath27 is an arbitrary abscissa . we set @xmath28 and we denote by @xmath29 the @xmath30-coordinate of any fluid particle @xmath31 in the serret - frenet reference frame @xmath32 at point @xmath24 with @xmath33 the tangent vector @xmath34 , the normal and @xmath30 the binormal vector ( see f. [ oxoz ] and f. [ oyoz ] for the notations ) . @xmath30 is normal to @xmath19 in the vertical plane @xmath20 . then , at each point @xmath35 , @xmath36 is defined by the set : @xmath37,\ , y \in [ \alpha(x , z),\beta(x , z)]\right\}\ ] ] where @xmath38 denotes the radius , @xmath39 the physical water height at section @xmath36 . we denote @xmath40 ( respectively @xmath41 y - coordinate of the left ( respectively right ) boundary point of the domain at altitude @xmath29 , @xmath42 ( see f. [ oyoz ] ) . we denote also @xmath43 by @xmath44 which is the @xmath29-coordinate of the water level . [ oxoz ] in the sequel , we will use a curvilinear map which will be an admissible transformation under the geometrical hypothesis on the domain : @xmath45 : : let @xmath46 be the algebraic curvature radius of the plane curve @xmath47 . we assume that : @xmath48 following the work in @xcite , we write system ( [ e3dicartesian ] ) in the serret - frenet reference frame @xmath32 at point @xmath49 by the transformation @xmath50 using the divergence chain rule lemma that we recall here : [ lemma ] let @xmath51 be a @xmath52 diffeomorphism and + @xmath53 the jacobian matrix of the transformation with determinant @xmath54 . then , for any vector field @xmath55 , one has : @xmath56 and , for any scalar function @xmath57 , one has : @xmath58 where @xmath59 stands for the transpose of the matrix @xmath60 . let @xmath61 be the components of the velocity vector in the @xmath62 coordinates defined as @xmath63 where @xmath64 is the matrix @xmath65 where we denote by @xmath66 the angle @xmath67 in the @xmath68 plane . using lemma [ lemma ] , the incompressible euler system in the variables @xmath62 reads : @xmath69 where @xmath70 is the determinant of the transformation and @xmath71 the interested reader can find the details of the calculus in @xcite . we have denoted by @xmath72 the derivative with respect to the space variable @xmath73 of any function @xmath74 . on the wet boundary , the no - leak condition reads : @xmath75 [ remarkadmissibilitycondition ] notice that @xmath76 is the algebraic curvature of the axis at point @xmath77 and the function @xmath78 depends only on the variables @xmath79 . moreover , under the hypothesis @xmath45 , we have @xmath80 in @xmath81 . consequently , @xmath82 defines a diffeomorphism and thus the performed transformation is admissible . in this section , we perform a formal asymptotic analysis on system ( [ e3dicurvilinear ] ) . according to the work in @xcite , the shallow water equations can be obtained from the incompressible navier - stokes equations with particular boundary conditions . here , we perform this analysis directly on the incompressible euler system in order to get @xmath83 for some small parameter @xmath84 . let us introduce the usual small parameter @xmath85 where @xmath86 ( the height ) and @xmath87 ( the length ) are two characteristics dimensions along the @xmath30 and @xmath33 axis respectively . moreover , we assume that the characteristic dimension along the @xmath88 axis is the same as for the @xmath22 axis . we introduce the other characteristics dimensions @xmath89 for time , pressure and velocity respectively and the dimensionless quantities as follows : @xmath90 @xmath91 in the sequel , we set @xmath92 and @xmath93 ( i.e. we only consider laminar flows ) . under these hypotheses , we have @xmath94 . thus , the rescaled system ( [ e3dicurvilinear ] ) reads : @xmath95 where @xmath96 @xmath97 @xmath98 is the froude number along the @xmath33 axis and the @xmath30 or @xmath34 axis where @xmath31 is any generic variable equal to @xmath87 or @xmath86 . formally , when @xmath99 vanishes , system ( [ e3dicurvilinearrescaled ] ) reduces to : @xmath100 let us introduce the conservative variables @xmath101 and @xmath102 representing respectively the wet area and the discharge defined as : @xmath103 where @xmath104 is the mean value of the velocity : @xmath105 we integrate the preceding system ( [ e3dicurvilinearrescaledepsilonvanishes1]-[e3dicurvilinearrescaledepsilonvanishes2]-[e3dicurvilinearrescaledepsilonvanishes3 ] ) along the cross - section with the approximation @xmath106 and @xmath107 . then , returning to the physical variables , the free surface model , that we call * fs*-model , reads : @xmath108 where @xmath109 and @xmath110 are respectively the classical term of hydrostatic pressure and the pressure source term defined by : @xmath111 which are obtained from the integration of the pressure term @xmath112 in equation ( [ e3dicurvilinearrescaledepsilonvanishes1 ] ) with @xmath113 ( obtained from equation ( [ e3dicurvilinearrescaledepsilonvanishes3 ] ) ) . in these formulas @xmath114 is the width of the cross - section at position @xmath73 and at height @xmath29 . the additional term @xmath115 is defined by @xmath116 . it is the @xmath29-coordinate of the center of mass : @xmath117 in system ( [ fsmodel ] ) , we may add a friction term @xmath118 to take into account the dissipation of energy . we have chosen this term @xmath119 as the one given by the manning - strickler law ( see e.g. @xcite ) : @xmath120 the term @xmath121 is defined by : @xmath122 , @xmath123 is the strickler coefficient of roughness depending on the material , @xmath124 is the hydraulic radius and @xmath125 is the perimeter of the wet surface area ( length of the part of the channel s section in contact with the water ) . in this section , we present a new set of unidirectional shallow water like equations to describe pressurized flows in closed non uniform water pipes . this model is constructed to be coupled in natural way with the * fs*-model ( [ fsmodel ] ) . starting from the @xmath0d compressible euler equations in cartesian coordinates , @xmath126 @xmath127 where @xmath5 and @xmath128 denotes the velocity with components @xmath6 and the density respectively . @xmath129 is the scalar pressure and @xmath10 the exterior strength of gravity . we define the pressurized domain of the flow as the continuous extension of @xmath81 ( see subsection [ subsectionframework ] ) defined by some plane curve @xmath19 with parametrization @xmath21 in the cartesian reference frame @xmath2 ; we recall that @xmath23 is then the elevation of the point @xmath35 over the plane @xmath25 ( see f. [ oxoz ] ) . the curve may be , for instance , the axis spanned by the center of mass of each orthogonal section @xmath130 to the main mean flow axis , particularly in the case of a piecewise cone - shaped pipe . notice that we consider only the case of infinitely rigid pipes , thus the sections @xmath131 are only @xmath132-dependent . we then write equations ( [ euler3d_mass_conservation]-[euler3d_momentum_conservation ] ) in the @xmath62 coordinates introduced in subsection [ subsectionframework ] . as we want a unidirectional model , we suppose that the mean flow follows the @xmath73-axis . to this end , we neglect the second and third equation for the conservation of the momentum . + by a straightforward computation , the mass and the first momentum conservation equation in the @xmath62 coordinates reads : @xmath133 applying the same asymptotic analysis developed in subsection [ sectionfsmodelasymptotic ] , equations ( [ euler3d_mass_conservation]-[euler3d_momentum_conservation ] ) read : @xmath134 we choose the linearized pressure law : @xmath135 ( see e.g. @xcite ) in which @xmath8 represents the density of the fluid at atmospheric pressure @xmath9 , @xmath136 is some function set to zero and @xmath137 the water compressibility coefficient ( equal to @xmath138 in practice ) . the sonic speed is then given by @xmath139 and thus @xmath140 . + for @xmath141 , @xmath142 is the outward unit vector at the point @xmath143 in the @xmath144-plane and @xmath145 stands for the vector * @xmath146 * ( as displayed on f. [ oyoz ] ) . following the section - averaging method performed in subsection [ sectionfsmodelasymptotic ] , we integrate system ( [ e3dcrescaledcurvilinear ] ) over the cross - section @xmath144 . noting the averaged values over @xmath144 by the overlined letters ( except @xmath147 ) , and using the approximations @xmath148 the shallow water like equations read : @xmath149 where @xmath150 is the velocity in the @xmath151-plane . we denote by @xmath152 the area of the cross - section of the pipe at position @xmath73 . the integral terms appearing in ( [ stv_mass ] ) and ( [ stv_momentum ] ) vanish , as the pipe is infinitely rigid , i.e. @xmath153 ( see @xcite for the dilatable case ) . it follows the non - penetration condition ( see f. [ restriction ] ) : @xmath154 omitting the overlined letters ( except @xmath147 ) , we introduce the conservative variables @xmath155 and dividing equations ( [ stv_mass])-([stv_momentum ] ) by @xmath8 we get : @xmath156 as introduced previously for the * fs*-model in section ( [ sectionfsmodelasymptotic ] ) , we may introduce the friction term @xmath157 given by the manning - strickler law ( see e.g. @xcite ) : @xmath158 where @xmath159 is defined by : @xmath160 , @xmath123 is the strickler coefficient of roughness depending on the material and @xmath161 is the hydraulic radius where @xmath125 is the perimeter of the wet surface area ( length of the part of the channel s section in contact with the water , equal to @xmath162 in the case of circular pipe ) . [ [ section ] ] this choice of variables is motivated by the fact that this system is formally close to the * fs*-model ( [ fsmodel ] ) where the terms @xmath163 , @xmath164 , @xmath115 are respectively the counterparts of @xmath165 , @xmath166 , @xmath167 in system ( [ pmodel ] ) . let us remark that the term @xmath147 is continuous through the change of state ( pressurized to free surface or free surface to pressurized state ) when the same curve plane is chosen ( in practice , the main axis of the pipe ) . then , we are motivated to connect `` continuously '' system ( [ fsmodel ] ) and ( [ pmodel ] ) through transition points ( through the change of state ) by defining a continuous pressure law . it leads to a `` natural '' coupling between the pressurized and free surface model as we will see in section [ sectioncoupling ] . the formulations of the * fs*-model ( [ fsmodel ] ) and * p*-model ( [ pmodel ] ) are very close to each other . the main difference comes from the pressure law . in order to build a coupling between the two models , we have to define a pressure that ensures its continuity through transition points in the same spirit of @xcite . as pointed out in the previous section , we will use the common couple of unknowns @xmath168 and the same plane curve @xmath19 ( see remark [ remarkrestrictionplanecurve ] ) to get a continuous model for mixed flows . [ remarkrestrictionplanecurve ] the plane curve with parametrization @xmath21 is chosen as the main pipe axis in the axisymmetric case . actually this choice is the more convenient for pressurized flows while the bottom line is adapted to free surface flows . thus we must assume small variations of the section ( @xmath169 small ) or equivalently small angle @xmath170 as displayed on f. [ restriction ] . we introduce a state indicator @xmath171 ( see f. [ depression ] ) such that : @xmath172 next , we define the _ physical wet area _ @xmath173 by : @xmath174 and a modified pressure law ( see f. [ depression ] ) which ensures its continuity through the change of state by : @xmath175 * * indeed , when a change of state occurs we have : @xmath176 which ensures the continuity of the pressure . * the flux gradient @xmath177 is discontinuous through the change of state since @xmath178 finally , from the * p*-model ( [ pmodel ] ) , the * fs*-model ( [ fsmodel ] ) , the definition of @xmath171 ( [ definitione ] ) , the definition of @xmath173 ( [ definitionphysicalwetarea ] ) and the pressure law ( [ pfsmodelpressureinflux ] ) , the * pfs*-model for unsteady mixed flows can be simply expressed into a single formulation as : @xmath179 where @xmath180 , @xmath181 , and @xmath182 denotes respectively the friction , the pressure source and the geometry source term defined as follows : @xmath183 and @xmath184 stands for @xmath185 . @xmath186 represents the @xmath29-coordinate of the water level : @xmath187 [ rquerecoveringbothmodels ] setting @xmath188 in system ( [ pfs ] ) , we obtain obviously the free surface model ( [ fsmodel ] ) . for all pressurized states , when @xmath189 , the pressure law ( [ pfsmodelpressureinflux ] ) reads , for instance , in the case of circular pipe : @xmath190 which is not exactly the pressure law of the * p*-model ( [ pmodel ] ) . indeed , the derivation of the * p*-model is done with the linearized pressure law ( [ linearizedpressurelaw ] ) ( see section [ sectionpfmodel ] ) with @xmath191 . thus , the property of the continuity of models ( [ pmodel])-([fsmodel ] ) through a change of state is obtained if and only if @xmath136 is chosen as @xmath192 which is the hydrostatic pressure corresponding to a full section . [ depression ] ( top ) , pressurized state with overpressure @xmath193 ( bottom left ) , pressurized state with depression @xmath194 ( bottom right).,title="fig : " ] the * pfs*-model ( [ pfs ] ) satisfies the following properties : [ thmpfsmodel ] 1 . 2 . the right eigenvalues of system ( [ pfs ] ) are given by : @xmath195 with @xmath196 where @xmath197 is the width of the free surface ( see f. [ oyoz ] ) . + then , system ( [ pfs ] ) is strictly hyperbolic on the set : @xmath198 3 . for smooth solutions , the mean velocity @xmath199 satisfies @xmath200 the quantity @xmath201 is called the total head . the still water steady state reads : @xmath202 5 . it admits a mathematical entropy @xmath203 which satisfies the entropy relation for smooth solutions @xmath204 notice that the total head and @xmath205 are defined continuously through the transition points . the term @xmath206 is also called `` corrective term '' since it allows to write the equations ( [ thmpfsequationforu ] ) and ( [ thmpfsentropy ] ) with ( [ thmpfsmathematicalentropy ] ) . * proof of theorem [ thmpfsmodel ] : * the results ( [ thmpfsequationforu ] ) and ( [ thmpfsentropy ] ) are obtained in a classical way . indeed , equation ( [ thmpfsequationforu ] ) is obtained by subtracting the result of the multiplication of the mass equation by @xmath207 to the momentum equation . then multiplying the mass equation by @xmath208 and adding the result of the multiplication of equation ( [ thmpfsequationforu ] ) by @xmath209 , we get : @xmath210 we see that the term @xmath211 is identically @xmath15 since we have @xmath212 when the flow is free surface whereas @xmath213 when the flow is pressurized . moreover , from the last inequality , when @xmath212 , we have the classical entropy inequality ( see @xcite ) with @xmath205 : @xmath214 while in the pressurized case , it is : @xmath215 finally , the entropy for the * pfs*-model reads : @xmath216 let us remark that the term @xmath217 makes @xmath205 continuous through transition points and it permits also to write the entropy flux under the classical form @xmath218 . in this section , we adapt the vfroe scheme described in @xcite . the new terms appearing in the * pfs*-model related to the curvature and the section variation are upwinded in the same spirit of @xcite . the numerical scheme is adapted to discontinuities of the flux gradient occurring in the treatment of the transitions between free surface and pressurized states . the spatial domain is a pipe of length @xmath87 . the main axis of the pipe is divided in cells @xmath220,\ 1\leq i\leq n$ ] . @xmath221 denotes the timestep and we set @xmath222 . + the discrete unknowns are @xmath223 . for the sake of simplicity , the boundary conditions are not treated ( the interested reader can find this treatment in details in @xcite ) . we propose to extend the finite volume discretisation @xcite to the * pfs*-model using the upwinding of the new source terms : the curvature and section variation of the pipe . in what follows , we do not write the @xmath171 dependency . first , following leroux _ et al . _ @xcite we use piecewise constant functions to approximate @xmath224 @xmath225 as well as the term @xmath226 and the cross section area @xmath152 . adding the equations @xmath227 , @xmath228 and @xmath229 , the * pfs*-model can be written under a non - conservative form with the variable @xmath230 : @xmath231 where @xmath232 and @xmath233 where we have written the pressure source term due to the geometry @xmath234 as @xmath235 . for instance , for a circular cross - section pipe we have : @xmath236 let @xmath237 be an approximation of the mean value of @xmath238 on the mesh @xmath239 at time @xmath240 . since the values of @xmath241 are known , integrating the above equations over @xmath242x_{i-1/2},x_{i+\frac{1}{2}}[\times [ t_n , t_{n+1 } [ $ ] , we can write a finite volume scheme as follows : @xmath243 with @xmath244 @xmath245 is the exact or an approximate solution to the riemann problem at interface @xmath246 associated to the left and right states @xmath247 and @xmath248 . let us also remark that the term @xmath249 does not appear explicitly in this formulation since @xmath184 , @xmath250 and @xmath251 are null on @xmath242x_{i-1/2},x_{i+\frac{1}{2}}[$ ] but contributes to the computation of the numerical flux . the computation of the interface quantities @xmath252 will depend on two types of interfaces located at the point @xmath253 : the first one is a non transition point , when the flow on both sides of the interface is of the same type . the second one is a transition point , when the flow changes of type through the interface . we recall the approach used in @xcite and adapt it here to the new terms . according to the type of interface , we have to solve two different linearised riemann problems . expanding the term @xmath254 in the non - conservative equations ( [ pfsncv ] ) , the quasilinear formulation of the * pfs*-model ( [ pfs ] ) reads : @xmath255 with @xmath256 the convection matrix defined by @xmath257 where @xmath258 and @xmath259 denotes the speed of the water . @xmath260 is then @xmath261 for the pressurized flow or @xmath262 for the free surface flow . let us remark that , since @xmath263 , the pressure source term @xmath264 does not appear in equation ( [ fnc ] ) . to compute the interface quantities denoted by @xmath265 for the left hand side and @xmath266 for the right hand side ( see figure [ qamp ] below ) , we solve the following linearised riemann problem : @xmath267 with @xmath268 and @xmath269 where @xmath270 is some approximate state of the left @xmath271 and the right @xmath272 state . [ remarkchoiceoftilded ] we will see in section [ sectionwb ] that the classical approximation @xmath273 of the roe matrix @xmath274 defined by @xmath275 is not a suitable choice to preserve the still water steady state . however , we propose in section [ sectionwb ] a new approximation of @xmath276 which maintains it perfectly . we have then @xmath277 . the eigenvalues of the matrix @xmath276 are @xmath278 , @xmath279 , @xmath280 , @xmath281 , @xmath282 and the associated right eigenvectors : @xmath283 @xmath284 ) . the lines @xmath285 corresponds to the characteristic lines @xmath286 , for @xmath287 . ] we denote @xmath14 the transition matrix associated to the right eigenvectors of @xmath288 and @xmath289 its inverse . setting @xmath290 , the solution of the riemann problem are constant states connected by shocks propagating along the characteristic lines @xmath286 . the jump associated to the eigenvectors @xmath291 is then equal to @xmath292 . in particular , the discharge is continuous through the line @xmath293 since the fifth component of vectors @xmath294 , @xmath295 and @xmath296 are null . it follows that the equation associated to the wet area @xmath297 remains conservative . thus , for instance in the subcritical case ( when @xmath298 ) , we have : @xmath299 where @xmath300 is the upwinded source term @xmath301 . the friction term can also be upwinded in the same way . writing the friction term under a conservative form @xmath302 ( for some arbitrary @xmath303 ) allows us to write the `` static '' slope @xmath224 as a `` dynamic '' one as follows : @xmath304 that we denote again @xmath224 . thus , the upwinding of the dynamic slope @xmath305 reads : @xmath306 which is equal to : @xmath307 since @xmath297 and @xmath209 are constant on each cells . the terminology `` dynamic '' and `` static '' slope is used since one is @xmath308-dependent while the other is only @xmath132-dependent . in the case of a transition point , we assume that the propagation of the interface ( pressurized - free surface or free surface - pressurized ) has a constant speed @xmath309 during a time step . the half line @xmath310 is then the discontinuity line of @xmath311 . + let us now consider @xmath312 and @xmath313 the ( unknown ) states respectively on the left and on the right hand side of the line @xmath310 with @xmath314 . both states @xmath315 and @xmath316 ( resp . @xmath317 and @xmath318 ) correspond to the same type of flow . thus it makes sense to define the averaged matrices in each zone as follows : * for @xmath319 , we set @xmath320 for some approximation @xmath321 which connects the state @xmath271 and @xmath322 ( see remark [ remarkchoiceoftilded ] ) . * for @xmath323 , we set @xmath324 for some approximation @xmath321 which connects the state @xmath325 and @xmath272 ( see remark [ remarkchoiceoftilded ] ) . then we formally solve two riemann problems and use the rankine - hugoniot jump conditions through the line @xmath310 which writes : @xmath326 with @xmath327 . according to the left ( @xmath316 , @xmath328 ) and right unknowns ( @xmath318 , @xmath329 ) at the interface @xmath330 and the sign of the speed @xmath309 , we have to deal with four cases : * pressure state propagating downstream , * pressure state propagating upstream , * free surface state propagating downstream , * free surface state propagating upstream . we can next consider two couples of `` twin cases '' : pressure state is propagating downstream ( or upstream ) as shown in the figure [ plus1 ] and free surface state propagating downstream ( or upstream ) as shown in the figure [ plus2 ] . moreover , for all existing transition case , the upwinded altitude term @xmath331 in @xcite are replaced by @xmath332 . [ [ pressure - state - propagating - downstream ] ] pressure state propagating downstream : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + it is the case when on the left hand side of the line @xmath333 , we have a pressurized flow and on the right hand side we have a free surface flow : the speed @xmath309 of the transition point being positive . following song @xcite ( see also @xcite ) , an equivalent stationary hydraulic jump must occur from a supercritical to a subcritical condition and thus the characteristics speed satisfies the inequalities : @xmath334 where @xmath261 is the sound speed for the pressure flow , @xmath335 , @xmath336 , and @xmath337 are defined by the same formula obtained in the case of a non transition point but according to @xmath338 and @xmath339 . therefore , only the characteristic lines drawn with solid lines are taken into account . indeed they are related to incoming waves with respect to the corresponding space - time area @xmath340 . conversely , the dotted line @xmath341 , for instance , related to the free surface zone but drawn in the area of pressurized flow is a `` ghost wave '' and is not considered . thus @xmath342 and @xmath315 , @xmath316 are connected through the jumps across the characteristics @xmath343 and @xmath344 . eliminating @xmath309 in the rankine - hugoniot jump relations ( [ rh1])-([rh2 ] ) , we get @xmath316 as the solution to the nonlinear system : @xmath345 finally , we obtain : @xmath346 [ [ free - surface - state - propagating - downstream ] ] free surface state propagating downstream : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + on the left hand side of the line @xmath333 we have a free surface flow while on the right hand side , we have a pressurized flow ( the speed @xmath309 of the transition point being positive ) . following song @xcite again , the characteristic speed satisfies the inequalities : @xmath347 there are two incoming characteristic lines with respect to the free surface area @xmath340 ( actually three with @xmath343 ) and they can connect the given left state @xmath315 with any arbitrary free surface state @xmath328 . thus only one characteristic line ( @xmath348 ) gives any information ( it is the equation ( [ slchp1 ] ) above ) as an incoming characteristic line with respect to the pressurized zone @xmath349 . from the jump relations through the characteristic @xmath350 , and after the elimination of @xmath309 in the rankine - hugoniot jump relations ( [ rh1]),([rh2 ] ) we get another equation , namely equation ( [ slchp2 ] ) above . it remains to close the system of four unknowns @xmath351 . firstly , we use a jump relation across the transition point ( with speed @xmath309 ) for the total head @xmath352 , from equation ( [ thmpfsequationforu ] ) , which writes : @xmath353 finally , we use the relation : @xmath354 we have then to solve the nonlinear system : @xmath355 the states @xmath328 et @xmath329 are then obtained by the following identities : @xmath356 \dsp ap = am + \frac{g\,\at_l\,\psi_l^r}{\ut_l^2-\ct_l^2}\\[5 mm ] \dsp qm = qp = qmp = q_l + \frac{g\,\at_l\,\psi_l^r}{2\,\ct_l } \;+ \\ \hspace{4.8cm}\dsp + \;\frac{\ut_l^2-\ct_l^2}{2\,\ct_l}\,(a^- - a_l ) - \frac{\ut_l - \ct_l}{2\,\ct_l}\,(q^- - q_l ) \end{array}\ ] ] finally , the update state @xmath357 are obtained by the same relation as in the case of a non transition point . using equations ( [ fvscheme ] ) we update the values of @xmath358 with a standard stability condition of courant - friedrich - levy controlling the time step size @xmath221 . to update the state @xmath171 in the cell @xmath239 ( see f. [ updatea ] ) , we use a discrete version of the state indicator @xmath171 defined by ( [ definitione ] ) equal to @xmath360 for a pressurized flow and @xmath15 otherwise . following @xcite , after the computation of the wet area @xmath361 we predict the state of the flow in the cell @xmath239 by the following criterion : * if @xmath362 then : + if @xmath363 then @xmath364 , else @xmath365 , * if @xmath366 : + if @xmath367 then @xmath365 , else @xmath368 . indeed , if @xmath367 it is clear that the state of the flow in the cell @xmath239 becomes pressurized , on the other hand if @xmath369 in a mesh previously pressurized , we do not know _ a priori _ if the new state is free surface ( @xmath370 and the value of the wetted area is less than @xmath371 ) or pressurized ( in depression , with @xmath372 and the value of the wetted area is equal to @xmath373 : see remark [ remarkdepressionoverpressurestates ] and f. [ rhocriterion ] ) . + so far as we do not take into account complex phenomena such that entrapment of air pockets or cavitation and keeping in mind that the cfl condition ensures that a transition point crosses at most one mesh at each time step , we postulate that : 1 . if the state of the flow in the cell @xmath239 is free surface at time @xmath240 , its state at time @xmath374 is only determined by the value of @xmath361 and it can not become in depression . 2 . if the state of the flow in the cell @xmath239 is pressurized at time @xmath240 and if @xmath363 , it becomes free surface if and only if at least one adjacent cell was free surface at time @xmath240 . this is exactly the discrete version of the continuous @xmath375 criterion in remark [ remarkdepressionoverpressurestates ] and displayed on f. [ rhocriterion ] . [ cols="^,^ " , ] [ remarkdepressionoverpressurestates ] as we do not take into account complex phenomena such that entrapment of air pockets , each connected component of the pressurized area is simply connected ( see f. [ rhocriterion ] ) . moreover , for each depression area @xmath256 , its closure @xmath376 is a strict subset of the pressurized set . it follows that when @xmath377 on each pressurized area , we observe a depression as displayed on f. [ rhocriterion ] . moreover , we may also use a visual depression indicator given by the function @xmath375 : the case @xmath212 corresponds to a free surface state while @xmath378 to an overpressure state and @xmath379 to a depression state . on f. [ rhocriterion ] , we draw the behavior of the interface speed @xmath309 in the @xmath380-plane and the graph of the function @xmath375 at fixed time @xmath381 . this section is devoted to the construction of an exactly well - balanced scheme in the sense that it maintains perfectly the still water steady states . this scheme , noted ewbs , is obtained by a suitable definition of the convection matrix . the numerical approximation of the * pfs*-model ( [ pfsmodelpressureinflux ] ) reads : @xmath382 where @xmath383 stands for @xmath384 and @xmath385 . for instance , in the subcritical case , the interface quantities reads : @xmath386 where the upwinded source term reads : @xmath387 and @xmath388 stands for @xmath389 with @xmath390 given by @xmath391 and the approximation of @xmath392 to be specified . starting from a discrete state @xmath393 at time @xmath240 such that : ( h ) : : let @xmath394 such that : @xmath395 , @xmath396 and @xmath397 satisfy the discrete still water steady state equation ( according to equation ( [ thmpfsequationforu ] ) ) : @xmath398 we will say that : [ definitionwbs ] 1 . 2 . the numerical scheme ( [ numericalschemea]-[numericalschemeq])-([definitionoftildebthetasq ] ) for some approximations of the terms @xmath399 is @xmath400 well - balanced ( also denoted by @xmath400-wb ) if : @xmath401 with @xmath402 , @xmath403 the well - balanced order of the numerical scheme ( [ numericalschemea ] ) and ( [ numericalschemeq ] ) respectively . 3 . the numerical scheme ( [ numericalschemea]-[numericalschemeq])-([definitionoftildebthetasq ] ) for some approximations of the terms @xmath399 is exactly well - balanced ( also denoted by ewb ) if : @xmath404 we will denote by @xmath400-wbs the @xmath400 well - balanced scheme and ewbs the exactly well - balanced scheme . ( sf ) : : in the rest of this paper , we assume @xmath405 constant , the radius @xmath406 and @xmath224 are given affine functions , the angle @xmath407 is constant which implies that the jumps across the interface @xmath246 : @xmath408 , @xmath409 are constant and @xmath410 is null . then under this simplified framework , we show that : [ thmwb ] 1 . 2 . * the numerical scheme ( [ numericalschemea]-[numericalschemeq])-([definitionoftildebthetasq ] ) with the classical choice @xmath411 and non constant section @xmath152 is not well - balanced in the sense of definition [ definitionwbs ] ( we have @xmath412 ) . * for constant section and @xmath413 , the numerical scheme ( [ numericalschemea]-[numericalschemeq])-([definitionoftildebthetasq ] ) with ( [ demisomme ] ) is ewb . 3 . under a suitable choice of @xmath414 , the numerical scheme ( [ numericalschemea]-[numericalschemeq])-([definitionoftildebthetasq ] ) is ewb . the following section deals with the numerical scheme with @xmath392 defined as ( [ demisomme ] ) where we show the first point of theorem [ thmwb ] . the second point is studied in section [ subsectionewbs ] where a convenient definition of @xmath392 leads to an ewbs the simpler choice of definition for the convection matrix @xmath273 is the one obtained by the approximation of the mean value of the roe matrix @xmath415 . this approximation is given by @xmath416 that we call `` the classical approximation '' . thus , defining @xmath392 as follows : @xmath417 provides the classical approximation . but , it is not suitable to preserve the still water steady state : we will see that the numerical scheme ( [ numericalschemea]-[numericalschemeq])-([definitionoftildebthetasq ] ) with ( [ demisomme ] ) defines a non well - balanced scheme in the sense of definition ( [ definitionwbs ] ) since @xmath412 . to this end , let us assume * ( h ) * and * ( sf ) * at time @xmath240 , equations ( [ interfacequantitiessubcriticalcase ] ) read for every @xmath418 : @xmath419 denoting @xmath420 by @xmath421 , we have : @xmath422 where @xmath423 denoting then @xmath424 and observing that @xmath425 and @xmath426 we deduce : @xmath427 it follows that : @xmath428 then , we denote @xmath429 by @xmath430 with @xmath431 and @xmath432 where @xmath433 and @xmath434 as the term @xmath435 is at least of order @xmath436 since @xmath437 we have : @xmath438 it follows that : @xmath439 for constant section @xmath152 and @xmath413 , it is easy to see that the numerical scheme ( [ numericalschemea]-[numericalschemeq])-([definitionoftildebthetasq ] ) with ( [ demisomme ] ) is ewb . although the scheme ( [ numericalschemea]-[numericalschemeq])-([definitionoftildebthetasq ] ) with ( [ demisomme ] ) have an order @xmath412 for non constant section , the still water steady state for the pressurized case is very well maintained for great value of the sonic speed @xmath261 . but it is not the case for the free surface numerical scheme ( see f. [ statiosl ] with @xmath440 , @xmath441 , uniform pipe with diameter @xmath360 ) . we plot in f. [ comparaisonmeanmatrixstationnarymatrixc30 ] and f. [ comparaisonmeanmatrixstationnarymatrixc100 ] a still water steady state for two values of @xmath261 for a given @xmath442 ( with @xmath443 ): the absolute error obtained is @xmath444 for @xmath445 and @xmath446 for @xmath447 . indeed , let us consider , for the sake of simplicity , the case @xmath448 . at the continuous level , the still water steady state equation reads : @xmath449 with the hypothesis * ( h ) * , in particular using equation ( [ discretesteadystateequation ] ) , we write : @xmath450 given @xmath451 , for great value of @xmath261 , we can approximate @xmath452 by : @xmath453 then , replacing the right hand - side of @xmath452 in ( [ diffainterface ] ) , we have @xmath454 as we can see @xmath455 is always equal to @xmath360 but the constant @xmath456 plays the role of a smoothing term which helps the scheme ( [ numericalschemea]-[numericalschemeq])-([definitionoftildebthetasq ] ) with ( [ demisomme ] ) to stabilize rapidly towards the equilibrium . physically , @xmath261 is approximatively @xmath457 ( for a pressurized flow without air ) , thus @xmath458 . since @xmath459 , this feature is not observed for the free surface numerical scheme . this section is devoted to the construction of an ewbs . we have seen in subsection [ subsectionsteadystateswithclassicalapproximation ] that the classical approximation of with @xmath414 ( [ demisomme ] ) is not appropriate to preserve the still water steady state . thus , we have to find a suitable definition for @xmath414 to obtain an exactly well - balanced scheme . for this purpose , let us assume * ( sf ) * and start with : at the discrete level , the still water steady state is perfectly maintained ( see f. [ discretrepresentationdomain ] ) : it exists @xmath394 such that for every @xmath418 , if @xmath460 and @xmath461 a1 : : : @xmath462 , a2 : : : @xmath463 , a3 : : : @xmath464 , then , for all @xmath465 the conditions a1 , a2 and a3 holds . . ] the condition a2 is satisfied if and only if @xmath466 the condition a3 is satisfied if and only if @xmath467 the condition a1 is satisfied for pressurized flows if and only if @xmath468 the condition a1 is satisfied for free surface flows if and only if @xmath469 for circular cross - section pipe , @xmath452 is computed by : @xmath470 with @xmath471 is the angle displayed on f. [ computationomega ] . ] thus , the discrete still water steady state is perfectly maintained if and only if , @xmath472 is the solution of the non - linear system : @xmath473 where we have replaced the expression of @xmath452 in ( [ conditiona2]-[conditiona3 ] ) by ( [ conditiona1pf ] ) for pressurized by ( [ conditiona1fs ] ) for free surface flows : @xmath474 with @xmath475 . for circular pipe , @xmath476 is given by ( [ forcircularpipe ] ) . finally , the numerical scheme ( [ numericalschemea]-[numericalschemeq])-([definitionoftildebthetasq ] ) with @xmath477 as the solution of the non linear system ( [ snl ] ) defines an exactly well - balanced scheme . for uniform pipe and pressurized flow , the previous system simply writes : @xmath478 the solution is easily obtained by : @xmath479 let us also remark that using the relation @xmath480 , we have : @xmath481 it follows that @xmath392 can be expressed as the mean value of ( [ atildebyai ] ) and ( [ atildebyai+1 ] ) as follows : @xmath482 for small @xmath451 , we have @xmath483 it follows that the scheme ( [ numericalschemea]-[numericalschemeq])-([definitionoftildebthetasq ] ) with ( [ demisomme ] ) is the zero order approximation of the solution given by the ewbs . the same analysis shows that the free surface numerical scheme with ( [ demisomme ] ) is also the zero order approximation of the solution given by the ewbs . [ [ section-1 ] ] on f. [ comparaisonmeanmatrixstationnarymatrixc30 ] , f. [ comparaisonmeanmatrixstationnarymatrixc100 ] , f. [ statiosl ] , we display the still water steady pressurized and free surface state computed by the ewbs and the scheme with the approximation ( [ demisomme ] ) . the steady state with the ewbs is numerically well preserved while as pointed out before ( see subsection [ subsectionsteadystateswithclassicalapproximation ] ) the classical approach is not convenient . we also display an unsteady simulation on f. [ testnonstationnaire ] where the results of the two methods are very well reproduced . the previous sections deals with the well - balanced property of the numerical scheme ( [ numericalschemea]-[numericalschemeq])-([definitionoftildebthetasq ] ) for free surface and pressurized flows . to use the well - balanced scheme developed in subsections ( [ subsectionsteadystateswithclassicalapproximation ] ) and ( [ subsectionewbs ] ) , we start from the discrete representation of a mixed still water steady state ( as displayed on f. [ discretrepresentationdomain ] ) . assume that there exists @xmath484 such that all cells @xmath239 on the left hand side of the interface @xmath485 are free surface while the other are pressurized : @xmath486 the interface @xmath485 is such that the speed of propagation of the interface is null : @xmath487 ( see f. [ wzero ] ) . therefore , @xmath488 and @xmath489 . for example on f. [ wzero ] , we apply the free surface numerical scheme on the left hand side of the interface and the pressurized one on the right cells . as the ewbs preserves the pressurized and free surface still water steady state , it also preserves the mixed still water steady state . the numerical validation for pipes with constant section and slope has been previously studied by two of the authors in @xcite and thus are not presented in this paper . since experimental data for mixed flows in any pipe are not available , we focus on the behavior of our method for several circular cross - section contracting and expanding pipe . notice that , the equivalent pipe method is not relevant for the mixed flows as pointed out by @xcite for instance . the mixed flow case is numerically performed on a water hammer test . starting from an horizontal free surface still water steady state , the water hammer occurs immediately after the increase of the upstream piezometric head while the downstream discharge is set to @xmath15 . the prescribed hydrograph produces a travelling wave which produces a pressurized state propagating from upstream to downstream end . physically an trapped air pocket may appear : it is not taken into account in the * pfs*-model . actually , the trapped air pockets vanish or move ; some parts of these pockets undergo condensation / vaporisation and others parts move and lead to a two phase flow . consequently the sound speed decreases . as our model does not take into account these phenomena , the value of @xmath261 is assumed to be constant . moreover we should have to deal with the entrapment of air bubbles which have a non negligible effect ( see @xcite for instance ) . the numerical experiments are performed in the case of a @xmath490 @xmath143 long closed circular pipe at altitude @xmath491 with @xmath15 slope which corresponds to the elevation and slope of the main pipe axis ( we have @xmath492 ) . the manning roughness coefficient is @xmath493 . the simulation starts from a steady state as a free surface flow with a discharge @xmath494 . the upstream boundary condition is a prescribed hydrograph ( see f. [ boundarycondition ] ) while the downstream discharge is kept constant to @xmath495 ( as displayed on f. [ boundarycondition ] ) . we compare then the results obtained for uniform , contracting an expanding pipes . for each test , the parameters are the same except the downstream diameter : the upstream diameter is kept constant to @xmath496 . the contracting pipe is chosen for @xmath497 and the expanding one for @xmath498 ( where @xmath256 denotes the downstream diameter ) . let us recall that the zero water level corresponds to the main pipe axis . the piezometric head is defined as @xmath499 : @xmath500 results are then represented on f. [ x50 ] . the sudden raise of the upstream piezometric level produces a pressurized state with a travelling wave . a water hammer is then observed since the downstream discharge is null . a careful analysis of the flow which is performed by the variable @xmath171 or equivalently by @xmath375 ( see remark [ remarkdepressionoverpressurestates ] and f. [ rhocriterion ] ) shows that after this transition point , the flow is pressurized but in depression which starts approximatively at time @xmath501 for the contracting pipe , @xmath502 for the uniform pipe and @xmath503 for the expanding one . we display the piezometric line for different times around the depression time ( see f. [ depress ] ) and also the graph of the function @xmath375 which confirm that the observed times correspond exactly to a depression state for each pipes ( see f. [ depressx ] ) . we also observe a little smoothing effect and absorption due to the first order discretisation type . we have derived a free surface and a pressurized model which have been coupled using a common set of variables and a suitable pressure law . we have thus obtained a mathematical model for unsteady mixed flows in non uniform water pipes , that we have called * pfs*-model . this model takes into account the local perturbation of the section and of the slope . the * pfs * model is numerically solved by a vfroe scheme using the interfacial upwind to include the source terms into the numerical fluxes . we have shown that the classical approximation of the convection matrix ( the roe matrix approximation ) is not suitable to preserve the still water steady state ( except for the pressurized case where the value of @xmath261 helps the scheme to maintain this state ) . moreover , we have proposed a manner to obtain an exactly well - balanced scheme . as mentioned in @xcite this numerical method with the classical approximation of the convection matrix , for constant section , reproduces correctly laboratory tests for uniform pipes and can deal with multiple points of transition between the two types of flows . as pointed out before , due to the lack of experimental data for non uniform pipes , we have only shown the behavior of the piezometric line which seems reasonable ( at less no major difference was observed ) . as a well - known feature on approximate godunov scheme , the upwinding of the source terms may introduce stationary waves with a vanishing denominator when critical flows occurs i.e. @xmath504 . moreover , in its actual form , the presented numerical scheme is not able to deal with drying and flooding area . nevertheless , it may be possible to introduce a cut - off function to avoid division by zero for each problems : critical stationary waves , drying area and flooding area . but it is not the better choice that we can propose , since , truncation of the wet area induces a loss of water mass leading to the non - conservativity of the mass . nevertheless , at the present time , we are interested in a mathematical kinetic formulation of the * pfs * model and the construction of a numerical kinetic scheme that avoids all these drawbacks @xcite . the next step is to take into account the air entrainment which may have non negligible effects on the behaviour of the piezometric head . a first approach has been derived in the case of perfect fluid and perfect gas modelised as a bilayer model based on the * pfs*-model @xcite . nguyen trieu dong . sur une mthode numrique de calcul des coulements non permanents soit surface libre , soit en charge , soit partiellement surface libre et partiellement en charge . , 2:149158 , 1990 . f. kerger , p. archambeau , s. erpicum , b. j. dewals , and m. pirotton . improved one - dimensional numerical simulation of transient mixed flow in water pipe . in _ proceedings of 4th int . conf . on advanced computational methods in engineering liege , belgium . _ , 2008 . f. kerger , s. detrembleur , p. archambeau , s. erpicum , b. j. dewals , and m. pirotton . an experimental analysis of effects induced by moving bodies in shallow water . in _ proceedings of 2nd int . junior researcher and engineer workshop on hydraulic structures , pisa , italy . _ , 2008 . p.l . roe . some contributions to the modelling of discontinuous flow . in b. e. engquist , s. osher , and r. c. j. somerville , editors , _ large - scale computations in fluid mechanics . proceedings of the fifteenth ams - siam summer seminar on applied mathematics held at scripps institution of oceanography , la jolla , calif . , june 27-july 8 , 1983 _ , volume 22 of _ lectures in applied mathematics _ , pages 163193 . american mathematical society , 1985 .

in this paper we demonstrate that the profile of angular distribution of electrons emitted in the process of a single - photon ionization from a strongly polarizable target ( in particular , a metal cluster anion ) exposed to a laser field can be noticeably modified due to the action of the field @xmath2 of alternating dipole moment @xmath3 induced at the residue by the laser field . the field @xmath2 acts on the escaping electron and , being dependent on the polarization vector of the laser field , brings additional dependence of the cross section on the emission angle . the degree of this dependence is determined by the magnitude of @xmath3 which , in turn , is proportional to the residue s dynamic polarizability . therefore , for a target with large polarizability the effect can be very pronounced even for laser fields of low intensities . the effect of the laser - field induced dipole moment in atomic photoionization process was discussed in @xcite ( where one finds also the references to the earlier works by the authors ) and , more recently , in @xcite . both papers were devoted to the study of _ multiphoton _ detachment of electrons from atomic negative ions . in @xcite the process was analyzed by means of the perturbation theory . numerical analysis was carried out for halogen and alkali metal anions , and it demonstrated , that the induced dipole moment influences the magnitude of total cross section of the many - photon detachment . the paper @xcite dealt with the process of a _ strong - field _ many - photon detachment of an atomic anion . the non - perturbative adiabatic theory , developed in @xcite , was modified to account for the field of the induced dipole moment . the numerical calculations , performed for rb@xmath4 and cs@xmath4 , showed the increase of the detachment rate and essential modification of the angular distribution . in our paper we investigate the influence of the induced dipole moment on the angular and spectral distributions of photoelectrons in the process of _ single - photon _ ionization ( or detachment ) occurring in the _ weak - field _ regime . numerical results are presented for sodium cluster anions na@xmath5 with @xmath6 . the corresponding residue clusters are characterized by large values of the dynamic dipole polarizability @xmath7 , which in the vicinity of plasmon resonances , i.e. at @xmath8 ev , can be as large as @xmath9 a.u . , which by far exceed the maximum values of @xmath7 of individual atoms ( e.g. , @xmath10 a.u . for cs at the photon energy 0.47 ev @xcite ) . we demonstrate that dramatic changes to the profile of angular distribution occur in vicinities of the plasmon resonances and for comparatively low intensities @xmath11 w/@xmath1 of the laser field . these are exactly the ranges used in recent experimental studies @xcite of the photodetachment from various metal cluster anions carried out by means of angle resolved photoelectron spectroscopy . therefore , the experimental test of the predicted effects seems to be feasible . the atomic system of units , @xmath12 is used throughout the paper . consider the process of photoionization of a strongly polarizable target ( a cluster , a fullerene , a complex molecule , etc ) by spatially homogenous linearly polarized laser field @xmath13 we assume , that the field wavelength @xmath14 greatly exceeds the characteristic size @xmath15 of the target , @xmath16 . thus , the process can be treated within the dipole - photon limit . the laser field ionizes the target and polarizes the residue . therefore , in addition to other fields ( the static potential of the residue , the long - range polarizational potential ) the escaping photoelectron feels the field due to the oscillating dipole moment @xmath3 induced at the residue : @xmath17 to simplify the consideration we assume , that the residue is spherically symmetric , so that its tensor of polarizability contains only a scalar part which is notated as @xmath7 . then , the moment @xmath3 reads @xmath18 where @xmath19 . the field of the induced dipole can be written as follows : @xmath20 with @xmath21 . this field is dependent on the direction of @xmath22 , therefore , the angular distribution of photoelectrons acquires additional dependence on the polarization of the laser field . the degree of this dependence is determined by the magnitude of @xmath3 , which is proportional to the residue dynamic dipole polarizability . hence , for a target with large polarizability the modification of the angular distribution due to the induced dipole moment can become noticeable even for a laser field of a comparatively low intensity . let us clearly state the conditions which we impose on the parameters of the laser field . these include the following inequalities : @xmath23 the first condition implies that the laser field strength is much smaller than the strength @xmath24 of the internal field in the target . this condition allows one to ignore the laser field when constructing the wavefunction of the bound electron with the binding energy @xmath25 . numerical results presented in section [ numericalresults ] refer to the laser field strength @xmath26 v / cm . the second condition states that the energy @xmath27 of the escaping electron is large compared to the ponderomotive shift due to the wiggling in the laser field . owing to this condition we do not `` dress '' the photoelectron wavefunction with the laser field . the last inequality in ( [ conditions ] ) indicates that the photon energy exceeds the ionization threshold of the bound electron . therefore , the process of a _ single - photon _ ionization is allowed for which the energy conservation law reads @xmath28 combining the conditions ( i ) and ( iii ) one demonstrates that the following inequality is met : @xmath29 where @xmath30 is a so - called keldysh parameter , and @xmath31 . the limit @xmath32 defines a weak field regime ( see , e.g. @xcite ) . thus , in this paper we investigate the influence of the induced dipole moment @xmath3 on the angular and spectral distributions of photoelectrons in the process of _ single - photon _ ionization ( or detachment ) occurring in the weak field regime . earlier the role of @xmath3 was studied in the process of many - photon detachment ( i.e. , when @xmath33 ) from atomic negative ions . it was done in the weak - field @xcite and in the strong - field @xcite limits . let us note , that the weak field condition ( ii ) does not immediately imply the applicability of the perturbative approach with respect to @xmath2 . indeed , from ( [ problem.2])-([problem2.3a ] ) follows that @xmath34 exceeds @xmath35 by a factor @xmath36 . for a strongly polarizable target this factor can be large enough in a wide range of radial distances @xmath37 , and this might lead to a non - perturbative treatment of the action of @xmath2 . to carry out a quantitative analysis we first postulate that under certain conditions the field ( [ problem2.3a ] ) can be considered as a time - dependent perturbation which modifies the wavefunction of the escaping electron . the criterion of applicability of this approach will be formulated in the course of calculations . the matrix element @xmath38 , which describes a dipole - photon transition of the electron from the initial bound state @xmath39 to the final state @xmath40 with the asymptotic momentum @xmath41 and energy @xmath42 , can be written in the following form : @xmath43 the right - hand side of this equation represents the power series in @xmath2 . the structure of the terms of the series is illustrated by diagrams in figure [ diagram.fig ] . ( indicated with the dashed vertical lines ) . the double line represents the initial bound state . the solid line stands for the escaping electron whose wavefunction is `` dressed '' with all static fields . ] the zeroth order term @xmath44 describes the direct process of photoionization standing for the matrix element of the operator @xmath45 : @xmath46 where @xmath47 we assume that the wavefunction @xmath48 ( whose asymptotic form is the plane wave + the incoming wave ) corresponds to the state `` dressed '' with all _ static _ fields . the delta - function on the right - hand side of ( [ m0.1 ] ) expresses the energy conservation law ( [ m0.3 ] ) . the terms @xmath49 with @xmath50 are the corrections to the transition amplitude due to the @xmath51-times interaction of the photoelectron with @xmath2 . since this field explicitly depends on @xmath52 ( see ( [ problem2.3a ] ) ) then in each act of the interaction the electron energy is changed ( increased or decreased ) by @xmath53 . as a result , not all terms on the right - hand side of ( [ problem.3 ] ) represent the correction to the direct amplitude @xmath44 , which implies the validity of ( [ m0.3 ] ) . for example , evaluating @xmath54 ( the second diagram in figure [ diagram.fig ] ) one finds , that it contains the terms proportional to @xmath55 and to @xmath56 , which conflict with the conservation law ( [ m0.3 ] ) . therefore , the term @xmath54 must be ignored when constructing the correction to @xmath44 . the lowest - order correction originates from the amplitude @xmath57 ( the last diagram in figure [ diagram.fig ] ) , which contains the terms proportional to @xmath58 . applying the standard technique of the time - dependent perturbation theory , one derives the following expression for this contribution : @xmath59 where @xmath60 here @xmath61 is the green s function of the photoelectron and @xmath62 is the unit vector of the laser field polarization . hence , within the second second order perturbation theory in @xmath2 , the photoionization amplitude is written as : @xmath63 using the amplitude ( [ ampl.1 ] ) one can write the following expression for the spectra - angular distribution : @xmath64 here @xmath65 stands for the differential cross section calculated without the influence of the induced dipole moment . within the dipole - photon approximation the differential cross section @xmath66 can be presented in the well - known general form ( see , e.g. , @xcite ) : @xmath67 where @xmath68 is the total cross section of photoionization , @xmath69 stands for the angle between @xmath41 and @xmath70 , @xmath71 is the legendre polynomial of the second order . the quantity @xmath72 stands for the angular asymmetry parameter @xcite . the factor @xmath73 defines the profile of the angular distribution . equation ( [ cs.3 ] ) indicates , that the profile is modified by the factor @xmath74 , when the field ( [ problem2.3a ] ) is taken into account . the correction term @xmath75 is defined as follows : @xmath76 formula ( [ cs.3 ] ) was obtained by means of the perturbation theory . thus , it is implied that the absolute value of the correction term satisfies the condition @xmath77 . if otherwise , the field @xmath2 must be accounted for in all orders of the perturbation series ( [ problem.3 ] ) . it follows from ( [ ampl.3 ] ) , that @xmath75 is proportional to the squared induced dipole moment , @xmath78 . therefore , for a target of a small polarizability , one has to apply very intensive laser fields in order to achieve a noticeable correction to the photoelectron angular distribution , i.e. @xmath79 . on the contrary , for a strongly polarizable target ( e.g. , a metallic cluster or / and its anion ) a strong effect can be expected for laser fields of a moderate strength . in section [ numericalresults ] we demonstrate , that dramatic change of the angular distribution profile can occur for @xmath80 v / cm . it is important to note , that the correction term ( [ cs.5 ] ) depends on the ratio of the amplitudes @xmath81 and @xmath82 . thus , one can expect that it is less sensitive to the approximation chosen to describe the photoelectron wavefunction than the cross section @xmath83 . therefore , to estimate the influence of the target polarization one can calculate @xmath75 within the framework of a comparatively simple approximation , which is described below . in this section we derive an approximate expression for the correction term @xmath75 using a simple , although realistic and physically clear , model . to start with let us assume , that the ionized target is a spherically - symmetric anion . then , firstly , the ground state wavefunction @xmath84 does not depend on the angles of the position vector : @xmath85 . secondly , the wavefunction @xmath86 of the photoelectron does not contain the contribution of a long - range coulomb field of the residue . these assumptions , being not of a principal nature , allow one to simplify the intermediate algebra . making use of the partial wave expansion @xmath87 where @xmath88 are the spherical harmonics , @xmath89 are the scattering phaseshift , and @xmath90 are the radial wavefunctions , one writes the amplitude @xmath82 in the following form : @xmath91 another simplification is based on the approximation that the field @xmath2 acts on the escaping electron only outside the target , i.e. at the distances @xmath92 . additionally , one can assume , that the wavefunction @xmath93 is concentrated in the interior of the target , i.e. at @xmath94 . therefore , when evaluating the right - hand side of ( [ ampl.3 ] ) one takes into account that @xmath95 . this allows one to relate @xmath81 to the amplitude @xmath82 . to do this let us approximate the green s function @xmath96 as follows : @xmath97 here @xmath98 is the largest / smallest of @xmath99 , @xmath100 stands for the irregular solutions of the radial schrdinger equation for the energy @xmath42 . the quantity @xmath101 is the exact radial green s function corresponding to the multipolarity @xmath102 . the first equality in ( [ longdistance.2 ] ) states the partial - wave expansion of the green s function . the approximate equality is based on the assumption @xmath103 . using ( [ longdistance.2 ] ) in ( [ ampl.3 ] ) one notices that the radial integral over @xmath104 reduces to that from ( [ m0_born.1 ] ) . therefore , the amplitude @xmath81 acquires the form : @xmath105 with @xmath106 here the integration is carried out over the spatial region @xmath107 . the correction term ( [ cs.5 ] ) is related to @xmath108 as @xmath109 . let us comment on the evaluation of the right - hand side of ( [ longdistance.8 ] ) . introducing the expansion ( [ m0.1a ] ) and representing the green s functions @xmath110 and @xmath111 in terms of the partial - wave series similar to ( [ longdistance.2 ] ) , one carries out the angular integration by means of the standard methods ( see , e.g. , @xcite ) and arrives at : @xmath112 { { \rm im}}\ , b_{ll^{\prime}}. \label{longdistance1.5}\end{aligned}\ ] ] the kronecker symbols @xmath113 , which reduce the allowed values of the orbital momenta to @xmath114 and @xmath115 , reflect the dipole selection rules applicable to each of the vertices in the last diagram in figure [ diagram.fig ] . the ratio @xmath116 reduces to @xmath117 with the help of the recurrence relation for the legendre polynomials . the quantities @xmath118 stand for the following radial integrals : @xmath119 for any realistic target the exact evaluation of these integrals can be done by numerical means only , implying the computation of the regular @xmath90 and the irregular @xmath120 functions as well as of the green s functions @xmath121 and @xmath122 . to carry out the approximate integration one can substitute these quantities with the leading terms of their asymptotic representations . for @xmath90 and @xmath120 this results in @xmath123 and @xmath124 . the leading terms in the expansions of the radial green s functions are : @xmath125 here @xmath126 , @xmath127 , @xmath128 and @xmath129 are the elements of the scattering matrix . for the real argument @xmath130 this quantity reads as @xmath131 . by using the asymptotic formulae one carries out analytical evaluation of the integrals from ( [ longdistance1.5a ] ) . the final result ( the leading term ) for @xmath132 is as follows : @xmath133 _ { l=1,3 \atop l^{\prime}=0,2 } , \label{functionb3.3}\end{aligned}\ ] ] where @xmath134 , @xmath135 and @xmath136 . it follows from the approximations formulated above , that formula ( [ functionb3.3 ] ) is valid if the photoelectron momentum @xmath137 and the ground state energy @xmath138 satisfy the conditions : @xmath139 letting @xmath140 in ( [ functionb3.3 ] ) one finds the plane - wave born limit of @xmath132 . the presence of oscillatory terms on the right - hand side of ( [ functionb3.3 ] ) is physically clear . indeed , in the limit @xmath141 the wavelength of the photoelectron is less than the size of the target , which is supposed to have a well - pronounced edge ( e.g. , the edge of an ionic core in a cluster ) . thus , the electron experiences a diffraction at the edge when leaving the core . this diffraction leads to the oscillatory character of the cross section . this phenomenon was discussed previously in connection with various processes involving metallic clusters and fullerenes : the radiative electron capture by metallic clusters @xcite , the electron scattering @xcite , the polarizational bremsstrahlung @xcite . carrying out obvious summations in ( [ longdistance1.5 ] ) , one writes correction factor as follows : @xmath142 where @xmath143 formula ( [ correctionfactor2.1 ] ) stresses the proportionality of @xmath75 to the laser field intensity ( which is proportional to @xmath144 ) and exhibits explicitly its dependence on the emission angle @xmath69 , which enters , as well as in ( [ cs.4a ] ) , via the second order legendre polynomial . as a function of @xmath69 the modulus of @xmath75 attains its maximum either at @xmath145 ( and @xmath146 ) or at @xmath147 , depending on the sign of the product @xmath148 and on the relative magnitudes of @xmath149 and @xmath150 . for further reference let us define the quantity @xmath151 which allows one to estimate the laser field strength needed to achieve a noticeable correction to the angular distribution ( see the discussion in section [ numericalresults ] ) . substituting ( [ correctionfactor2.1 ] ) into ( [ cs.3 ] ) , taking into account ( [ cs.4a ] ) and expressing @xmath152 via @xmath71 and @xmath153 , one derives the following formula for the differential cross section within the lowest order of perturbation theory in @xmath2 : @xmath154 the coefficients @xmath155 are related to the asymmetry parameter @xmath156 and to @xmath157 via @xmath158 , \quad a_4 = { 18 \over 35}\,e_0 ^ 2\,\beta\,\kappa_2 . \label{modification2.9}\end{aligned}\ ] ] the right - hand side of ( [ modification2.3 ] ) suggests that not only the profile of the angular distribution can be changed , but the magnitude of the cross section , integrated over the emission angles , is scaled by the factor @xmath159 which can be greater or lower than one depending on the signs and magnitudes of @xmath156 and @xmath157 . if the action of @xmath2 is ignored then @xmath160 , @xmath161 and @xmath162 , and equation ( [ modification2.3 ] ) reduces to ( [ cs.4a ] ) . numerical analysis of the influence of the induced dipole moment on the profile of the angular distribution was carried out for the process of photodetachment of sodium cluster anions na@xmath163 . the results presented below in this section refer to @xmath164 and @xmath165 . the corresponding neutral clusters are spherically - symmetric which makes applicable the theory described in section [ framework ] . the parameters of sodium clusters and their anions , used in our numerical calculations , are summarized in table [ nan.table ] . let us comment on these data . the ionic core radius was calculated using the standard relation @xmath166 with the wigner - seitz radius @xmath167 set to 4 a.u . the values of electron affinities @xmath168 are taken from a recent work by kostko @xcite which contains , in particular , an excellent collection of reference data on various metal clusters . the other data presented in the table refer to the polarizability of the core and the parameters of the plasmon peak . we considered the dynamic dipole polarizability @xmath7 of the clusters within the framework of the resonance plasmon approximation ( see , e.g. , @xcite ) : @xmath169 here @xmath170 is the static polarizability , @xmath171 is the surface plasmon energy , and @xmath172 is the linewidth of the plasmon resonance . due to the spill - out effect the static polarizability of a metallic cluster of a radius @xmath15 exceeds the classical value @xmath173 , which characterizes a metallic ball of the same radius . in the table we present the ratio @xmath174 calculated within the rpa @xcite and the lda @xcite ) as well as the experimentally measured values @xcite . the @xmath171 values were obtained with the help of the relation proposed in @xcite to account for the size - dependence of the plasmon energy : @xmath175 , where @xmath176 ev is the mie plasmon energy in a sodium metallic sphere , and @xmath177 takes into account the spill - out effect . @r|rrrrrrrr & & & & & & + & ( a.u . ) & ( ev ) & ( a.u . ) & @xcite & @xcite & @xcite & ( ev ) & ( ev ) + na@xmath178 & 8.00 & 0.915 & 512 & 1.48 & 1.44 & 1.72 & 2.64 & 0.2 + na@xmath179 & 10.86 & 1.337 & 1280 & 1.41 & 1.37 & 1.67 & 2.81 & 0.4 + na@xmath180 & 13.68 & 1.509 & 2560 & 1.10 & 1.32 & 1.60 & 2.90 & 0.7 + na@xmath181 & 15.48 & 1.719 & 3712 & 1.24 & 1.22 & - & 2.95 & 0.4 + na@xmath182 & 18.06 & 1.847 & 5888 & 1.22 & 1.20 & - & 2.99 & 0.2 + [ nan.table ] the size dependence of the width @xmath172 of the plasmon resonance for neutral sodium clusters with @xmath183 up to several hundreds was analyzed in @xcite within the rpa scheme and in @xcite within a semiclassical approach . these two methods , being consistent for @xmath184 , deviate noticeably for lower @xmath183 values . the rpa calculations reproduce the @xmath185 scaling law which , is inadequate when compared to the experimental data for @xmath186 ( see the discussion in @xcite ) . the semiclassical calculations @xcite seem to be consistent with the experimental data in the whole region @xmath187 . the values of @xmath172 , presented in the table , were deduced from figure 1 in @xcite . it is seen from the table , that for all @xmath183 the plasmon energy exceeds the electron affinity . this fact justifies the choice of cluster anions in a view of the present study . indeed , the inequality @xmath188 leads to the additional enhancement of @xmath75 owing to the pronounced increase of the modulus of the dynamic polarizability in the region @xmath189 . to conclude the discussion on the parameters of the chosen targets let us note , that both criteria from ( [ functionb3.3a ] ) can be easily met for na@xmath163 . the first inequality fails only in the narrow near - threshold region @xmath190 but is well fulfilled for @xmath189 . the validity of the second condition one proves by introducing @xmath191 . and for na@xmath192 as indicated . + in each graph the upper panel represents the dependence of @xmath193 on @xmath53 ; the middle panel - dependences @xmath194 and @xmath195 , eqs . ( [ correctionfactor2.3])-([correctionfactor2.1a ] ) ; the lower panel - dependence of @xmath196 , eq . ( [ correctionfactor2.1b ] ) . parameters @xmath171 and @xmath172 are as indicated in table [ nan.table ] . the data refer to the @xmath197 ratio equal to 1.45 for na@xmath198 and to 1.30 for na@xmath180 . , title="fig : " ] + and for na@xmath192 as indicated . + in each graph the upper panel represents the dependence of @xmath193 on @xmath53 ; the middle panel - dependences @xmath194 and @xmath195 , eqs . ( [ correctionfactor2.3])-([correctionfactor2.1a ] ) ; the lower panel - dependence of @xmath196 , eq . ( [ correctionfactor2.1b ] ) . parameters @xmath171 and @xmath172 are as indicated in table [ nan.table ] . the data refer to the @xmath197 ratio equal to 1.45 for na@xmath198 and to 1.30 for na@xmath180 . , title="fig : " ] but for na@xmath199 and na@xmath200 . for both clusters the ratio @xmath197 is set to 1.20 . , title="fig : " ] + but for na@xmath199 and na@xmath200 . for both clusters the ratio @xmath197 is set to 1.20 . , title="fig : " ] the results of calculation of the quantities relevant to the correction term @xmath75 are presented in figures [ na08_40.fig ] and [ na58_92.fig ] . for each anion the calculations were performed for the photon energies above the ionization threshold @xmath168 and in the vicinity of the plasmon peak . the latter manifests itself as a resonance in the dependence of @xmath193 on @xmath53 , as it is seen from the upper panel in each graph . the dynamic polarizabilities were calculated via ( [ numerical.1 ] ) with the parameters @xmath171 and @xmath172 as indicated in table [ nan.table ] . given the differences between the calculated and experimentally measured values of the static polarizabilities @xmath170 , as well as the absence of the experimental data for na@xmath181 and na@xmath182 , for the ratio @xmath197 we used the values close to the theoretical data ( see the figures captions ) . the dependences of the coefficients @xmath149 and @xmath150 on @xmath53 are presented in the middle panels . they were computed from ( [ correctionfactor2.3 ] ) and ( [ functionb3.3 ] ) . to calculate the right - hand side of the latter formula beyond the plane - wave born approximation one must know the scattering phaseshifts @xmath201 and @xmath202 . to determine the phaseshifts we solved numerically the phase equation ( see , e.g. , @xcite ) for an electron moving in a local potential . the model potential , which was used to describe the interaction between the photoelectron and the neutral residue , accounted for the following three terms : ( a ) the short - range potential due to the coulomb interaction with the core s electrons and the ionic jellium core , ( b ) the exchange - correlation term , which was treated within the gunnarsson - lundqvist approximation @xcite , ( c ) the polarization potential was considered in the form @xmath203 for @xmath37 and @xmath204 if otherwise . inclusion of the polarization term is important for an adequate description the low - energy electron scattering from metallic clusters ( e.g. , @xcite ) . a more rigorous treatment accounts for the many - body correlations , which may be considered , for example , within the dyson equation scheme @xcite . however , for the purpose of our paper it is sufficient to treat the polarization potential in the simple form described above . comparing the middle panels in the graphs from figures [ na08_40.fig ] and [ na58_92.fig ] one notices that the behaviour of @xmath194 ( and of @xmath195 as well ) exhibits some general trends along with the specific features determined by the dynamic and geometrical properties of an individual target . common to all clusters is the existence of extrema for @xmath194 at @xmath189 , which is due to the factor @xmath193 , see ( [ correctionfactor2.3 ] ) . the extremum can be clearly defined ( see the curves @xmath149 for na@xmath198 and @xmath150 for na@xmath182 which have distinct maxima ) or less pronounced ( both curves for na@xmath181 ) or barely seen ( the case of na@xmath180 ) . the variation in the shape of the extremum , as well as the behaviour of the curves in the regions below and above the plasmon resonance , can be understood by analyzing the @xmath53-dependence of the right - hand sides in ( [ correctionfactor2.3 ] ) and ( [ functionb3.3 ] ) . the non - oscillatory part of the latter depends on the photon energy mainly through the factor @xmath205 . this leads to @xmath206 . therefore , a symmetric resonance profile of @xmath193 is distorted due to the factor @xmath207 . for a powerful and narrow resonance ( as in the case of na@xmath182 ) the distortion is comparatively small , whereas for a wide resonance the disproportion between the low- and the high-@xmath53 shoulders may completely change the behaviour of the curve in the vicinity of the resonance , as it happens for na@xmath180 . in either case the factor @xmath207 results in a faster decrease of @xmath208 with @xmath53 beyond the resonance . in the region @xmath209 the curves @xmath194 , being enhanced by the factor , exhibit a non - monotonous oscillatory character ( most clearly seen in na@xmath180 , na@xmath181 and na@xmath182 graphs ) . this is due to the second and the third terms on the right - hand side of ( [ functionb3.3 ] ) , which contain the oscillating factors @xmath210 and @xmath211 , dependent on the cluster size , and the factors @xmath212 , @xmath213 , dependent on the phaseshifts . the solid curves in the middle panels of each graph represent the dependences @xmath195 defined in ( [ correctionfactor2.1a ] ) . for a given photon energy the quantity @xmath214 stands for the maximum value of @xmath215 within the interval @xmath216 $ ] . it was already mentioned , that the correction term must satisfy the condition @xmath77 . thus , taking into account that @xmath217 , one can introduce a critical value @xmath218 , which defines the upper boundary for the laser field strength consistent with the perturbative approach adopted in this paper : @xmath219 the corresponding intensity of the laser field ( in w/@xmath1 ) can be calculated from the relation @xmath220 , where @xmath221 is measured the atomic units . in the limit @xmath222 a strong inequality @xmath223 is valid , so that the photoionization process is not affected by the induced dipole moment . the influence of the latter can become pronounced for @xmath224 resulting in @xmath79 . @l|crrccccc & @xmath225 & & & @xmath226 & @xmath227 & @xmath228 & @xmath229 & @xmath230 + & ( ev ) & ( @xmath231 ) & ( @xmath231 ) & ( @xmath231 ) & ( @xmath232 a.u . ) & ( @xmath233w/@xmath1 ) & ( @xmath232 a.u . ) & ( @xmath232 ev ) + na@xmath178 & 2.64 & -0.470 & -0.031 & 0.501 & 1.41 & 0.66 & 17.5 & 1.44 + na@xmath179 & 2.73 & 0.021 & 0.022 & 0.043 & 4.82 & 7.67 & 30.8 & 15.7 + na@xmath180 & 2.72 & 0.022 & -0.100 & 0.078 & 3.58 & 4.23 & 37.0 & 8.64 + na@xmath181 & 2.89 & 0.151 & -0.160 & 0.231 & 2.08 & 1.43 & 44.9 & 2.61 + na@xmath182 & 2.98 & 0.128 & -1.589 & 1.461 & 0.83 & 0.23 & 50.1 & 0.39 + [ nan.table2 ] dependences @xmath218 are plotted in the lower panels in figures [ na08_40.fig ] and [ na58_92.fig ] . for all clusters this function attains its minimum in the point @xmath234 in the vicinity of the plasmon resonance . the values of @xmath225 , as well as of other quantities calculated in this point , are presented in table [ nan.table2 ] . it is instructive to compare the critical field @xmath235 with the strength @xmath229 of the internal anionic field . the latter can be related to the ionization potential @xmath168 through @xmath236 . it is seen that @xmath229 exceeds @xmath235 by the order of magnitude . the strong inequality @xmath237 , being in accordance with the first condition from ( [ conditions ] ) , justifies the approximation adopted in this paper to neglect the modification of the ground state orbitals due to the laser field . the last column in the table contains the values of ponderomotive energy of the photoelectron in the critical laser field . comparing the ponderomotive shifts and the kinetic energies @xmath238 ev , one notices that the latter is larger by several orders of magnitude . therefore , the second condition in ( [ conditions ] ) is also well fulfilled . hence , one can expect , that for the intensities @xmath239 w/@xmath1 the modification of the spectral - angular distribution due to the action of the field of the induced dipole moment can be quite pronounced . to carry out the corresponding quantitative estimates one can use the formulae obtained within the framework of perturbation theory . of the angular distribution ( [ profile.1 ] ) calculated for different sodium cluster anions and for the three values of asymmetry parameter @xmath156 ( as indicated ) , and for several values of the ratio @xmath240 ( see the legend in the graph @xmath241 for na@xmath242 ) . the data refer to the photon energies @xmath225 indicated in table [ nan.table2 ] . thick black curves in each graph represent the field free distribution @xmath243 , eq . ( [ profile.0 ] ) . the length of line segment connecting the origin and a given point on the curves is equal to the value of @xmath244 or @xmath243 ( in absolute units ) in the corresponding direction . the emission angle @xmath69 is measured with respect to @xmath22 , which is shown in the graphs with @xmath245 . in each graph the spacing between ticks is equal to 1 for both directions . , title="fig : " ] + of the angular distribution ( [ profile.1 ] ) calculated for different sodium cluster anions and for the three values of asymmetry parameter @xmath156 ( as indicated ) , and for several values of the ratio @xmath240 ( see the legend in the graph @xmath241 for na@xmath242 ) . the data refer to the photon energies @xmath225 indicated in table [ nan.table2 ] . thick black curves in each graph represent the field free distribution @xmath243 , eq . ( [ profile.0 ] ) . the length of line segment connecting the origin and a given point on the curves is equal to the value of @xmath244 or @xmath243 ( in absolute units ) in the corresponding direction . the emission angle @xmath69 is measured with respect to @xmath22 , which is shown in the graphs with @xmath245 . in each graph the spacing between ticks is equal to 1 for both directions . , title="fig : " ] + of the angular distribution ( [ profile.1 ] ) calculated for different sodium cluster anions and for the three values of asymmetry parameter @xmath156 ( as indicated ) , and for several values of the ratio @xmath240 ( see the legend in the graph @xmath241 for na@xmath242 ) . the data refer to the photon energies @xmath225 indicated in table [ nan.table2 ] . thick black curves in each graph represent the field free distribution @xmath243 , eq . ( [ profile.0 ] ) . the length of line segment connecting the origin and a given point on the curves is equal to the value of @xmath244 or @xmath243 ( in absolute units ) in the corresponding direction . the emission angle @xmath69 is measured with respect to @xmath22 , which is shown in the graphs with @xmath245 . in each graph the spacing between ticks is equal to 1 for both directions . , title="fig : " ] + of the angular distribution ( [ profile.1 ] ) calculated for different sodium cluster anions and for the three values of asymmetry parameter @xmath156 ( as indicated ) , and for several values of the ratio @xmath240 ( see the legend in the graph @xmath241 for na@xmath242 ) . the data refer to the photon energies @xmath225 indicated in table [ nan.table2 ] . thick black curves in each graph represent the field free distribution @xmath243 , eq . ( [ profile.0 ] ) . the length of line segment connecting the origin and a given point on the curves is equal to the value of @xmath244 or @xmath243 ( in absolute units ) in the corresponding direction . the emission angle @xmath69 is measured with respect to @xmath22 , which is shown in the graphs with @xmath245 . in each graph the spacing between ticks is equal to 1 for both directions . , title="fig : " ] + of the angular distribution ( [ profile.1 ] ) calculated for different sodium cluster anions and for the three values of asymmetry parameter @xmath156 ( as indicated ) , and for several values of the ratio @xmath240 ( see the legend in the graph @xmath241 for na@xmath242 ) . the data refer to the photon energies @xmath225 indicated in table [ nan.table2 ] . thick black curves in each graph represent the field free distribution @xmath243 , eq . ( [ profile.0 ] ) . the length of line segment connecting the origin and a given point on the curves is equal to the value of @xmath244 or @xmath243 ( in absolute units ) in the corresponding direction . the emission angle @xmath69 is measured with respect to @xmath22 , which is shown in the graphs with @xmath245 . in each graph the spacing between ticks is equal to 1 for both directions . , title="fig : " ] the calculated profiles of the angular distributions are presented in figure [ profiles.fig ] . in the present paper we have not made an attempt to compute the angular asymmetry parameter @xmath246 for sodium anion clusters ( neither have we found its values in the literature ) . instead , for each target we carried out the calculations for the three values , @xmath247 , aiming to demonstrate that the modification of the profile takes place for _ any _ @xmath156 from the allowable interval @xmath248 $ ] . in each graph from figure [ profiles.fig ] the thick solid curve represents the unperturbed profile @xmath249 the emission angle @xmath69 is measured with respect to the vector @xmath22 , which is indicated in the graphs with @xmath245 . other curves correspond to the modified profiles @xmath250 ( see ( [ cs.3 ] ) , ( [ correctionfactor2.1 ] ) and ( [ correctionfactor2.1b ] ) ) calculated for several values of the ratio @xmath251 . for each anion the data refer to the photon energy @xmath225 . figure [ profiles.fig ] demonstrates , that even for comparatively low intensities ( corresponding to @xmath252 ) the changes in the profiles are already visible . for @xmath253 they become much more pronounced . it is seen that variation of the profile depends , both qualitatively and quantitatively , on the value of @xmath156 and on the choice of the target anion which enters via the coefficients @xmath254 and @xmath255 . in the case @xmath245 the field - free profile @xmath243 is proportional to @xmath256 yielding @xmath257 as the minimum and @xmath258 as the maximum values within the interval @xmath259 $ ] . the modified profile can be written as @xmath260 with @xmath261 and @xmath262 . one easily shows , that for @xmath263 the function @xmath264 acquires additional extremum . however , this inequality can not be met for the considered targets ( see the values of @xmath254 , @xmath255 and @xmath265 in table [ nan.table2 ] ) and for @xmath253 . therefore , the pattern of the modified angular distribution is similar to that obtained without the influence of the induced dipole . namely , the strongest emission occurs in the directions perpendicular to the field @xmath22 and there is no emission for @xmath145 and @xmath266 . on the whole , the distortion increases with @xmath267 , attaining the largest values at @xmath147 . the trend of the largest distortion depends on the sign of @xmath254 : if @xmath268 ( as in the cases of all anions but na@xmath178 ) then @xmath269 whereas negative @xmath254 leads to a shrinking of the modified profiles ( as it happens for na@xmath242 ) . for @xmath270 the unperturbed angular distribution is isotropic . the modified profile reads as @xmath271 , with @xmath272 and @xmath273 defined as above . the way the profile evolves with the increase of @xmath267 to a great extent depends on the sign of @xmath255 . for @xmath274 the emission along @xmath275 dominates over the emission in the perpendicular directions ( see the case of na@xmath276 ) , whereas for negative @xmath255 the perpendicular direction becomes preferable . the graphs from the central column in figure [ profiles.fig ] demonstrate , that for @xmath277 the distortion can become dramatic both in form and in magnitude . in the limit @xmath241 the modified profile behaves as @xmath278 . the additional extrema appear if the following two conditions , @xmath279 and @xmath280 , are met . these conditions can be satisfied in the limit @xmath281 for all considered anions but na@xmath276 . most explicitly it is seen in na@xmath192 and na@xmath200 graphs , where the curves with @xmath282 represent the four - petal angular distributions with the maximum emission at @xmath283 and @xmath284 . the same pattern characterizes na@xmath285 , although in this case it is hardly visible due to a strong decrease in the magnitude of @xmath286 for @xmath282 . ( see ( [ spectra.1 ] ) ) calculated for different sodium cluster anions and for the three values of asymmetry parameter @xmath156 ( as indicated in the top left graph ) . the data refer to the laser field intensity @xmath287 w/@xmath1 . ] finally , let us briefly discuss the influence of the induced dipole moment on the cross section integrated over the emission angles . recalling ( [ modification2.3 ] ) and ( [ modification2.9 ] ) , one finds the following formula for the ratio of the cross sections with , @xmath288 , and without , @xmath289 , account for the correction : @xmath290 figure [ spectra.fig ] presents the ratios @xmath291 versus @xmath53 . for each anion the ratio was calculated using the dependences @xmath149 and @xmath149 , presented in figures [ na08_40.fig ] and [ na58_92.fig ] , and for three values of the asymmetry parameter ( as indicated ) . the intensity of the laser field was set to @xmath287 w/@xmath1 , which corresponds to the field strength @xmath292 a.u . comparing the latter value to the critical field @xmath218 ( see figures [ na08_40.fig ] and [ na58_92.fig ] ) one notices , that for all targets and in the whole range of @xmath53 the parameter @xmath293 is less than one . thus , formula ( [ spectra.1 ] ) is applicable for the quantitative estimate of the correction due to the induced moment . as one can see from figure [ spectra.fig ] , the largest deviation ( up to @xmath294 ) of @xmath288 from @xmath289 occurs for na@xmath285 and na@xmath200 in the vicinities of the plasmon resonances , where @xmath267 reaches its maximum values : @xmath295 for na@xmath285 and @xmath296 for na@xmath200 . in the case of na@xmath200 a strong inequality @xmath297 is valid ( see figure [ na58_92.fig ] or / and table [ nan.table2 ] ) . therefore , the magnitude of the correction term from the right - hand side of ( [ spectra.1 ] ) becomes sensitive to the choice of the asymmetry parameter . as a result , the ratio @xmath298 strongly varies with @xmath156 . on the contrary , for na@xmath285 the relation @xmath299 is valid , leading to a weak dependence of the ratio on @xmath156 . for the anions na@xmath192 and na@xmath199 the largest values of @xmath267 equal to 0.05 and 0.15 , correspondingly . hence , the intensity @xmath300 is small compared to @xmath301 . as a result , the deviation of @xmath288 from the field - free value @xmath289 is less pronounced . in summary , we have demonstrated that the dipole moment , induced at the residue by the incoming laser field , can strongly influence the photoionization process . the additional time - dependent long - range field due to the induced moment , acting on the escaping electron , modifies the spectral - angular and spectral distributions of photoelectrons . the effect itself , as well as its quantitative treatment , is quite sensitive to the choice of the target and to the parameters of the laser field . we have demonstrated , that for a strongly polarizable target ( e.g. , metallic cluster anions ) strong modifications of the characteristics of a single - photon ionization process in the vicinity of the plasmon resonance ( typically , @xmath302 ev ) can be achieved by applying laser fields of comparatively low intensities @xmath0 w/@xmath1 . from a theoretical viewpoint , the advantage of this regime is that the field of induced dipole moment can be treated perturbatively , whereas the influence of the laser field on the photoelectron ( a ponderomotive effect ) and on the ground state of the target can be ignored . we have shown , that even in this weak - field regime one can expect dramatic changes in the profile of angular distribution of photoelectrons as well as in the @xmath53 dependence of the photoionization cross section . to a great extent , these changes depend on the parameters of the target , like its radius , the static and dynamic polarizabilities , the energy of the plasmon resonance peak and the affinity ( or , the ionization potential in the case of a neutral target ) . additional variety can be obtained by changing the intensity and the frequency of the laser field . the analysis , which we have carried out , is based on simple models adopted for the description of the photodetachment process ( the plasmon resonance approximation ) and of the interaction of the escaping electron with the static field of the residue ( via the lda + long - range polarization potential ) . we have also restricted ourselves to the case of spherically - symmetric targets . a more accurate quantitative treatment must include more sophisticated approaches , which take into account the many - body correlation effects intrinsic to both the photodetachment and the electron escape . however , these will leave untouched the physics behind the basic phenomenon discussed in our paper . it is possible to study the predicted phenomena by means of modern experimental techniques . the latter include the production and manipulation of the beams of cluster anions of various types and sizes , the energy and angle resolved photoelectron spectroscopy , the availability of the laser fields of various intensity , frequency and pulse duration @xcite . we believe , that with all these components at hand , soon it will become possible to test the theory against experimental data . we are grateful to roman polozkov for the help in numerical calculation of the ground states of sodium clusters , and to steven manson and bernd von issendorff for the helpful discussions . this work was supported by the european commission within the network of excellence project excell ( project number 515703 ) . 99 p.a . golovinski , i.yu . kiyan , 1990 , _ sov . - uspekhi _ * 33 * , 453 . chernov , i.yu . kiyan , h. helm , b.a . zon , _ phys . rev . _ a * 71 * , 033410 ( 2005 ) . gribakin , m.yu . kuchiev , _ phys . rev . _ a * 55 * , 3760 ( 1997 ) . o. kostko , c. bartels , j. schwbel , c. hock , b. v. issendorff , _ j. phys . series _ * 88 * , 012034 ( 2007 ) . o. kostko , 2007 , _ photoelectron spectroscopy of mass - selected sodium , coinage metal and divalent metal cluster anions . _ inaugural - dissertation ( universitt freiburg ) . gordon w. f. drake ( ed . ) , 2006 , _ springer handbook of atomic , molecular and optical physics _ ( germany : springer ) .

quantum open systems have been a subject of fundamental interest for many years . problems in this category include those of understanding how the irreversibility of time appears in system dynamics , why macroscopic systems can be treated with classical mechanics instead of quantum mechanics , how wave functions collapse as a result of measurements done with macroscopic instruments , and why and how quantum systems approach a thermal equilibrium state through interaction with their environments.@xcite theories of quantum open systems have also been used to construct models of practical interest , in particular to account for line shapes in epr , nmr@xcite and laser spectra,@xcite to evaluate chemical reaction rates@xcite and electron and charge transfer rates@xcite in chemical physics , and to explore the lifetimes of quantum entanglement states in quantum information theory.@xcite the phenomena mentioned above arise from the unavoidable interaction of a system with its environment . in the quantum mechanical case , dissipative systems are often modeled as main systems coupled to heat - bath degrees of freedom at finite temperature . this coupling gives rise to thermal fluctuations and dissipation that drive the systems toward the thermal equilibrium state . the heat - bath degrees of freedom are then reduced using such methods as the projection operator method and the path integral method . the projection operator approach is effective if the interaction between the system and the bath is weak . if one further assumes that the correlation time of the noise arising from the interaction with the bath is very short ( the markovian assumption ) , equations of motion for the density matrix elements can be derived , and these can be solved numerically . the most commonly used equations of this kind are the quantum master equations@xcite and the redfield equation.@xcite it has been proven , however , that quantum master equations and the redfield equation do not satisfy the necessary positivity condition.@xcite careful analyses addressing this problem have been carried out by several researchers.@xcite as a method to preserve positivity , the rotating wave approximation ( rwa ) , which modifies the interaction between the system and the heat bath , has been applied.@xcite however , this may alter the thermal equilibrium state as well as the dynamics of the original total hamiltonian . then , building on these results , it was shown that the violation of the positivity condition results from the markovian assumption . specifically , it was found that even if the dissipation process is markovian , the fluctuation process may not be , because it must satisfy the fluctuation - dissipation theorem.@xcite the time convolution - less ( tcl ) master equation has a wider range of applicability than the quantum master equations and redfield equation , because it allows the system - bath interaction to be non - perturbative and fluctuation and dissipation to be non - markovian.@xcite in order for a non - perturbative treatment to be possible , however , the system hamiltonian and the bath interactions of the tcl equation must commute.@xcite thus , the tcl equation can not be used to treat many important problems involving molecules , atoms , and spins driven by a time - dependent laser or magnetic field . in addition , because of the factorized nature of the system - bath interaction , in the case of a non - commuting excitation , the tcl equation can not be used to calculate nonlinear response functions of the system operator involved in the optical multidimensional spectrum.@xcite path integral monte carlo simulations do not have the limitations of any of the approaches discussed above , and for this reason , they are capable of incorporating imaginary path integrals and unfactorized initial conditions more easily , but this approach is computationally heavy , because the number of paths to be evaluated grows rapidly with time , while sampling fails due to the phase cancellation of wave functions.@xcite much effort has been made to extend the applicability of this method.@xcite because this approach can easily incorporate the semi - classical approximation in the bath , it may have an advantage in the study of polyatomic systems treated in multi - dimensional coordinates , but applications to this point incorporating full quantum dynamics have been limited to relatively small systems . many of the above - mentioned limitations can be overcome with the hierarchal equations of motion ( heom ) for the reduced density matrix , which are derived by differentiating the reduced density matrix elements defined by path integrals.@xcite this approach was introduced to investigate the connection between the phenomenological stochastic liouville equation and the dynamical hamiltonian theory , and was originally limited to the case in which the spectral distribution function takes the drude form ( i.e. , the ohmic form with a lorentzian cutoff ) and the bath temperature is high.@xcite however , with the inclusion of low temperature corrections terms , this temperature limitation has been eliminated.@xcite in addition , with the extension of the dimension of the hierarchy , this approach is capable of treating a great variety of spectral distribution functions.@xcite this formalism is valuable because it can treat not only strong system - bath coupling but also quantum coherence between the system and bath , which is essential to calculate nonlinear response functions . the system - bath coherence becomes particularly important if the bath interaction is regarded as non - markovian , as was found from nonlinear optical measurements in the late 1980s , when laser technology reached the femto - second time scale , which is much shorter than the noise correlation time of environmental molecules.@xcite the heom approach has been used to study such problems , which include multi - dimensional spectroscopy.@xcite recently , it was shown that system - bath coherence also plays an important role in calculations of quantum measures involving concurrence@xcite and non - markovianity@xcite under multiple external perturbations.@xcite because the heom approach is computationally heavy , a variety of methods have been developed to study dissipative dynamics in realistic situations . @xcite it has been applied to the study of multi - dimensional vibrational spectroscopies,@xcite photosynthetic antenna systems,@xcite fermion systems , @xcite quantum ratchets,@xcite resonant tunneling diodes,@xcite and dissociation of tightly bounded electron - hole pairs.@xcite while the applicability of the heom approach continues to expand , the basic nature of the hierarchy elements has not been thoroughly explored . the purpose of this paper is to investigate the role of correlated initial equilibrium states in the heom formalism . until this time , the heom have been derived by assuming a factorized initial state , @xmath0 \exp[-\beta \hat h_b]$ ] , at inverse temperature @xmath1 , where @xmath2 and @xmath3 are the system and bath hamiltonians , respectively , while the true thermal equilibrium state of the system is given by @xmath4 $ ] , where @xmath5 is the system - bath interaction . the difference between the factorized and correlated initial states becomes large for strong @xmath5 . analysis based on an analytic solution of a brownian oscillator system indicates that even if we start from a factorized initial state , the system reaches the true equilibrium state , @xmath6\}$ ] , through transient phenomena arising from the factorized initial state , for example phenomena known as initial sweeping.@xcite with the heom approach , we have run the heom program until all of the hierarchy elements reach the steady state and then used these elements as the initial conditions of the correlated thermal equilibrium state . the accuracy of this method has been confirmed by analyzing multi - dimensional spectra obtained with it.@xcite nevertheless , it would be interesting to derive the heom starting from a correlated initial thermal state in order to obtain an analytically derived expression for the system - bath coherence in the heom formalism . moreover , with a simple generalization , we can also derive the heom in imaginary time , which corresponds to the inverse temperature . we show that the imaginary - time heom is convenient for obtaining correlated thermal equilibrium states and the thermodynamic variables of the reduced system . the organization of the paper is as follows . in sec . ii we present a model hamiltonian and its influence functional with correlated initial states . in sec . iii , we derive the heom from the density matrix elements using the influence functional formalism with the correlated initial states given in sec . ii . in sec . iv , we derive the imaginary - time heom , which is convenient for evaluating correlated thermal equilibrium states and the thermodynamic quantities of the system . in sec . v , to confirm the validity and numerical efficiency of our approach , we report the results of numerical integrations of the heom carried out over real time and imaginary time for a spin - boson system and compare their results . thermodynamic variables and expectation values for the spin - boson system are also calculated as a demonstration . section vi is devoted to concluding remarks . we consider a situation in which the system interacts with a heat bath that gives rise to dissipation and fluctuation in the system . to illustrate this , let us consider a hamiltonian expressed as @xmath7 where @xmath8 is the hamiltonian of the system , denoted by a , defined by the creation and annihilation operators @xmath9 and @xmath10 . the bath degrees of freedom are treated as an ensemble of harmonic oscillators , @xmath11 with the momentum , position , mass , and frequency of the @xmath12th bath oscillator given by @xmath13 , @xmath14 , @xmath15 and @xmath16 , respectively . the system - bath interaction is given by @xmath17 where @xmath18 is the system part of the interaction , and @xmath19 is the coupling constant between the system and the @xmath12th oscillator . the heat bath can be characterized by the spectral distribution function , defined by @xmath20 and the inverse temperature , @xmath21 , where @xmath22 is the boltzmann constant . note that if @xmath9 and @xmath10 respectively represent the creation and annihilation operators of spin states , the above hamiltonian is the spin - boson hamiltonian , @xcite which has been studied with various approaches.@xcite now , let us introduce the fermion coherent state @xmath23 , which satisfies @xmath24 and @xmath25 , where @xmath26 and @xmath27 are grassmann numbers ( g - numbers ) . @xcite note that here we consider a two - level system , but extension to a multi - level system is also straightforward.@xcite in practice , we can treat a g - number system in the same manner as a c - number system , as long as we maintain the time order of the operators in the integral . in the path integral representation , the time propagator of the wave function ( the feynman propagator ) for the total system is expressed as @xmath28 } \int_{{\bold x}(0)={\bold x}_0}^{{\bold x}(t)={\bold x } } d[{\bold x}(\tau ) ] \nonumber \\ & & \times { \rm e}^{\frac{\rm i}{\hbar } s_a[\phi ^{\dag},\,\phi ; \,t ] + \frac{i}{\hbar } \int_0^t d\tau \left [ \frac12{m \dot { \bold x}^2(\tau)}- \frac12 m \omega^2 { \bold x}^2(\tau ) + v ( \tau){\bold x } ( \tau ) \right ] } , \label{gtot}\end{aligned}\ ] ] where @xmath29 is the normalization constant , @xmath30 $ ] represents a functional integral over a set of grassmann variables , and @xmath31\equiv \pi_j \int d[x_j(\tau)]$ ] denotes path integrals over the bath oscillator coordinates , with @xmath32 , @xmath33 , and @xmath34 . here , the action for the system s hamiltonian , @xmath35 , is denoted by @xmath36=\int^t_{0 } d\tau l_a({\phi}^{\dag},{\phi})$ ] , with the lagrangian @xmath37 the thermal equilibrium state can also be expressed in the path integral representation by making the replacement @xmath38 in eq .. we thereby obtain @xmath39 } \int_{\bar { \bold x}(0)={\bold x}'_0}^{\bar { \bold x}(\beta\hbar)={\bold x}_0 } d[\bar { \bold x}(\tau ' ) ] \nonumber \\ & & \times { \rm e}^{-\frac{\rm 1}{\hbar } \bar s_a[\bar \phi ^{\dag},\,\bar \phi ; \,\beta \hbar ] -\frac{1}{\hbar } \int_0^{\beta \hbar } d\tau ' \left [ \frac12{m \dot { \bar { \bold x}}^2(\tau')}+ \frac12 m \omega^2 \bar { \bold x}^2(\tau ' ) - { \bar v } ( \tau ' ) \bar { \bold x } ( \tau ' ) \right ] } , \label{geqtot}\end{aligned}\ ] ] where @xmath40 is the normalization constant , the g - numbers @xmath41 form the coherent representations of the operators @xmath42 for the equilibrium distribution , @xmath43 $ ] is the euclid action of the system obtained from @xmath36 $ ] through the replacement @xmath44 , and @xmath45 . the total density matrix elements is then given by @xmath46 the heat - bath degrees of freedom can be eliminated by integrating over the bath coordinates as @xmath47 . the reduced density operator is then expressed in the coherent representation of the g - numbers as@xcite @xmath48 where @xmath49 } \int_{\bar \phi ( 0)=\phi'_0}^{\bar \phi ( \beta \hbar)=\phi_0 } d[{\bar \phi}^{\dag}(\tau ' ) \bar \phi ( \tau ' ) ] \int_{\phi'(0)=\phi_0'}^{\phi'(t)=\phi ' } { d[{\phi ' } ^{\dag } ( \tau ) \phi ' ( \tau ) ] } \nonumber \\ & & \times \bar \rho_0^{eq}[\bar \phi ^\dag , \bar \phi ; \beta \hbar ] { \rm e}^{\frac{\rm i}{\hbar } s_a[\phi ^{\dag},\,\phi ; \,t ] } f[\bold v ; t , \beta\hbar ] { \rm e}^ { - \frac{\rm i } { \hbar } s_a^{\dag}[{\phi ' } ^{\dag},\,\phi ' ; \,t]}. \label{rhotot1}\end{aligned}\ ] ] here , @xmath50 $ ] is the thermal equilibrium distribution of the system @xmath51 itself , defined by the euclid action , @xmath52 $ ] , and @xmath53 $ ] with @xmath54 is the influence functional for correlated initial states.@xcite employing the counter path illustrated in fig . 1 , we can express the influence functional in the path integral representation as ( see appendix a ) @xmath55 = \int d\bold x \int_{{\bold x}'(t)={\bold x}}^{{\bold x}(t)={\bold x } } d[\tilde { \bold x}(s ) ] e^{-\frac{i}{\hbar } \int_c ds\left [ \frac12{m \dot { \tilde { \bold x}}^2(s)}- \frac12 m \omega^2\tilde { \bold x}^2(s ) + \tilde v_c(s)\tilde { \bold x } ( s ) \right ] } , \label{eq : f_x}\end{aligned}\ ] ] where @xmath56\equiv \pi_j \int d[x_j(\tau ) ] \int d[\bar x_j(\tau ' ) ] \int d[x_j'(\tau)]$ ] , with the contour path defined by @xmath57 and @xmath58 with @xmath59 , @xmath60 and @xmath61 , respectively . the path integral used here to derive the heom is expressed in terms of an influence functional . the calculation of the influence functional for a heat bath consisting of harmonic oscillators is analogous to that of the generating functional for a brownian oscillator system if we regard the system operator in the system - bath interaction @xmath62 as an external force acting on the bath . @xcite then , the influence functional can be calculated analytically and is found to be @xmath63=\exp\left\{{\tilde \phi}[\tilde v_c ; t , \beta\hbar]\right\}$ ] , where the influence phase is expressed as ( see appendix a ) @xmath64 = \frac1{\hbar^2 } \int_c ds '' \int_{c ' } ds ' \tilde v_c(s '' ) \tilde v_{c'}(s ' ) l ( s '' - s ' ) . \label{eq : phita}\end{aligned}\ ] ] here , @xmath65 represents the counter path for @xmath66 that follows @xmath67 along @xmath68 under the condition @xmath69 and @xmath70 . \label{eq : ltitau}\end{aligned}\ ] ] after dividing the contour of the integral in eq . into @xmath71 , @xmath72 and @xmath73 , we have ( see appendix b ) @xmath74 & = & - \frac1{\hbar^2 } \int_0^t dt '' \int_0^{t '' } dt ' v^\times(t '' ) \left[-i l_1 ( t''-t ' ) v^\circ(t ' ) + l_2(t''-t ' ) v^\times(t ' ) \right ] \nonumber \\ & & + \frac{i}{\hbar^2 } \int_0^t dt '' \int_0^{\beta \hbar } d \tau ' v^\times(t '' ) l ( t '' + i \tau ' ) \bar v ( \tau ' ) \nonumber \\ & & + \frac1{\hbar^2 } \int_0^{\beta \hbar } d \tau '' \int_0^{\tau '' } d \tau ' \bar v ( \tau '' ) \bar v ( \tau ' ) \bar l(\tau '' - \tau ' ) , \label{eq : phitb}\end{aligned}\ ] ] where we define @xmath75 , @xmath76 and @xmath77 the functionals @xmath78 and @xmath79 represent the commutator and anticommutator of @xmath62 . this form of the influence functional has been used to analytically study quantum brownian systems.@xcite the first term in eq . represents a commonly used influence functional derived from the factorized initial conditions . @xcite the collective bath oscillator coordinate , @xmath80 , is regarded as a driving force for the system through the interaction @xmath81 . the time - dependent kernels are then represented by @xmath82\rangle$ ] and @xmath83 , respectively , where @xmath84 is the heisenberg representation of the operator @xmath85 . @xcite the function @xmath86 is analogous to the classical correlation function of the bath induced noise @xmath87 and corresponds to the fluctuations . the dissipation corresponding to @xmath88 is related to @xmath89 through the quantum version of the fluctuation - dissipation theorem , @xmath90 = \hbar \omega \coth(\beta \hbar \omega/2 ) \bar c [ \omega]/2 $ ] , which insures that the system evolves toward the thermal equilibrium state for finite temperatures.@xcite the second term in eq . consists of the cross - terms between the real - time and imaginary - time integrals that describe the correlation between the initial equilibrium state and the dynamical state at time @xmath91 . this term represents the contribution of the correlated initial conditions and can be regarded as the non - markovian effects with respect to both real and imaginary times . the last term describes the influence of the heat bath on the thermal equilibrium state of the system . in the following sections , in order to derive the imaginary - time hoem , we consider the full equilibrium state of the system , @xmath92 $ ] , by including the last term in @xmath50 $ ] . for @xmath93 , we have@xcite @xmath94 @xmath95 where @xmath96 are the matsubara frequencies . we thus have @xmath97 \cos(\omega t ) \nonumber \\ + i \frac{2}{\beta\hbar } \int_{0}^{\infty } d\omega j(\omega ) \sum_{k=1}^{\infty } \frac{2\nu_k}{\nu_k^2+\omega^2 } \sin ( \nu_k \tau ) \sin(\omega t ) , \label{eq : lfourier}\end{aligned}\ ] ] and @xmath98 . \label{eq : lbardef } \end{aligned}\ ] ] for the case @xmath99 , we use the definition given in eq . to obtain @xmath100 @xmath101 \cos(\omega t).\end{aligned}\ ] ] we assume that the spectral density @xmath102 has an ohmic form with a lorentzian cutoff and write@xcite @xmath103 where the constant @xmath104 represents the width of the spectral distribution of the collective bath modes and is the reciprocal of the correlation time of the noise induced by the bath . the parameter @xmath105 is the system - bath coupling strength , which represents the magnitude of damping . with eq . for @xmath93 , we obtain @xmath106 } \right\ } e^{-\gamma t } \nonumber \\ & & + \sum_{k=1}^{\infty } \textcolor{red } { c_k ' \left[\cos ( \nu_k \tau ) -i \sin ( \nu_k \tau ) \right ] } e^{-\nu_k t } , \label{eq : ltitau2}\end{aligned}\ ] ] where @xmath107 @xmath108 and @xmath109 . at @xmath110 , the above equation reduces to @xmath111 where @xmath112 , @xmath113 , and @xmath114 for @xmath115 , while at @xmath99 , we have @xmath116 and @xmath117 with @xmath118 . here , we choose @xmath119 so as to satisfy @xmath120 , where @xmath121 represents the characteristic frequency of the system . under this condition we can apply the approximation @xmath122 with negligible error at the desired temperature , @xmath123 . we define the equilibrium distribution function of the system under the influence of the heat bath through the replacement of the last term of @xmath124 $ ] ( as expressed in eq . ) appearing @xmath53 $ ] with @xmath50 $ ] . we then obtain @xmath125=\frac{1}{z_b } { \rm exp}&&\left[- \frac1{\hbar } \bar { s_a}[\bar \phi ^ { \dag } , \bar \phi ; \beta \hbar ] \right.\nonumber \\ & & \left . + \sum_{k=0}^{\infty } \frac{\bar c_k}{\hbar^2 } \int_{0}^ { \beta \hbar } d \tau '' \int_{0}^{\tau '' } d \tau ' \bar v(\tau '' ) \bar v ( \tau ' ) \cos \left ( \nu_k ( \tau''- \tau')\right ) \right ] . \label{eq : pci}\end{aligned}\ ] ] the influence functional @xmath126 $ ] is redefined through this replacement as @xmath127 & & = \exp \left [ \left ( -\frac{i}{\hbar } \right)^2 \int_{0}^t dt '' \operatorname{e } ^ { - \gamma t '' } % \left\ { v^{\times}(t '' ) \left ( \int_{0}^{t '' } dt ' \operatorname{e } ^ { \gamma t ' } \theta ( t ' ) % + -i \bar \theta ( \beta \hbar ) \right ) % + \xi(t '' ) % \right\ } \right ] \nonumber \\ & & \times \exp \left [ \left ( -\frac{i}{\hbar } \right)^2 \int_{0}^t dt''\sum\limits_{k= 1}^{k } \operatorname{e } ^ { - \nu_k t '' } v^{\times}(t '' ) \left ( \int_{0}^{t '' } dt ' \operatorname{e } ^ { \nu_k t ' } \psi_k ( t ' ) % + -i \bar \psi_k ( \beta \hbar ) \right ) \right ] \nonumber \\ & & \times \exp \left [ - \int_{0}^t dt '' \xi(t '' ) \right ] , \label{eq : influencefcoh}\end{aligned}\ ] ] where @xmath128 @xmath129 } \right\ } , \label{eq : thetabar}\end{aligned}\ ] ] and for @xmath130 , @xmath131 , \label{eq : psi_k}\end{aligned}\ ] ] @xmath132 , } \label{eq : psi_kbar}\end{aligned}\ ] ] and @xmath133 note that in the high temperature limit , @xmath134 , the noise correlation function reduces to @xmath135 . this indicates that the heat bath oscillators interact with the system in the form of gaussian markovian noise.@xcite the equations of motion for the reduced density operator can be derived by evaluating the time derivative of the wavefunctions on the left - hand and right - hand sides and the influence functional.@xcite if we consider the auxiliary matrix defined by @xmath136 } \int_{\bar \phi ( 0)=\phi'_0}^{\bar \phi ( \beta \hbar)=\phi_0 } d[{\bar \phi}^{\dag}(\tau ' ) \bar \phi ( \tau ' ) ] \int_{\phi'(0)=\phi_0'}^{\phi'(t)=\phi ' } { d[{\phi ' } ^{\dag}(\tau ) \phi ' ( \tau ) ] } \nonumber \\ & & \times \bar \rho^{eq } [ \bar \phi ^\dag , \bar \phi ; \beta \hbar ] { \rm e}^{\frac{\rm i}{\hbar } s_a[\phi ^{\dag},\,\phi ; \,t ] } f_{j_1 , \cdots , j_k}^{(n ) } [ \bold v ; t , \beta\hbar ] { \rm e}^ { - \frac{\rm i } { \hbar } s_a^{\dag}[{\phi ' } ^{\dag},\,\phi ' ; \,t ] } \label{eq : rhon_jk}\end{aligned}\ ] ] where @xmath137 & & = \left\ { \textcolor{red } { -\frac{i}{\hbar } } % color \int_{0}^{t } dt ' \operatorname{e } ^ { -\gamma ( t - t ' ) } \theta ( t ' ) % + -i \operatorname{e } ^ { -\gamma t } \bar \theta ( \beta \hbar ) \right\ } ^{n } \nonumber \\ & & \times \prod_{k=1}^k \left\ { \textcolor{red } { -\frac{i}{\hbar } } % color \int_{0}^{t } dt ' \operatorname{e } ^ { -\nu_k ( t - t')}\psi_k ( t ' ) % + -i \operatorname{e } ^ { -\nu_k t } \bar \psi_k ( \beta \hbar)\right\ } ^{j_k } \nonumber \\&&\times f_{ci}[\bold v ; t , \beta\hbar ] \label{auxfv}\end{aligned}\ ] ] for nonnegative integers @xmath138 . among the @xmath139 , only @xmath140 has a physical meaning , and the others are introduced for computational purposes . differentiating @xmath141 with respect to @xmath91 , we obtain the following hierarchy of equations in operator form : @xmath142 \hat{\rho}_{j_1 , \dots , j_k}^{(n)}(t ) \notag\\ & & -\frac{i}{\hbar}\hat v^{\times } \hat{\rho}_{j_1 , \dots , j_k}^{(n+1)}(t ) - \frac{i}{\hbar } \sum_{k=1}^k \hat v^{\times } \hat{\rho}_{j_1 , \dots , j_k+1,\dots , j_k}^{(n)}(t ) \notag\\ & & - \frac{i n}{\hbar } \hat{\theta } \hat{\rho}_{j_1,\dots , j_k}^{(n-1 ) } ( t ) - \sum_{k=1}^k \frac{i j_k}{\hbar } \hat{\psi}_k \hat{\rho}_{j_1,\dots , j_k-1,\dots , j_k}^{(n)}(t ) , \label{eq : heom}\end{aligned}\ ] ] where @xmath143 is the liouvillian of @xmath2 , and the relaxation operators @xmath144 and @xmath145 are obtained through the replacement @xmath146 and @xmath147 in eqs . and , where @xmath148 and @xmath149 for any operand operator @xmath150 and @xmath151 , and @xmath152 + \sum_{k=1}^k \frac { c_k'}{\nu_k } \right\ } \hat{v}^\times\hat{v}^\times . \label{heomt3}\end{aligned}\ ] ] the above expression is identical to the heom with a factorized initial state and can be truncated in the same manner as in the factorized case for large @xmath153 , where @xmath121 is the characteristic frequency of the system.@xcite if we add the counter term to the hamiltonian , we have an additional term in eq .. @xcite while the terms from the correlated initial state @xmath154 and @xmath155 do not appear in eq . , they define the hierarchy elements for the correlated initial equilibrium state . to demonstrate this point , we consider the initial states of the density operators , obtained by setting @xmath110 in eqs . and : @xmath156 \left(-i \bar \theta ( \beta \hbar ) \right)^{n } \prod_{k=1}^k \left ( -i \bar \psi_k ( \beta \hbar ) \right)^{j_k } \bar \rho [ \bar \phi ^\dag , \bar \phi ; \beta \hbar ] . \nonumber \\ \label{eq : correlatedheom}\end{aligned}\ ] ] here , @xmath157 and @xmath158 & & = { \rm exp}\left [ - \frac1{\hbar } \bar { s_a}[\bar \phi ^ { \dag } , \bar \phi ; \tau ] \right ] % { \rm exp}\left [ c_0''\int_{0}^{\tau } d \tau '' \int_{0}^{\tau '' } d \tau ' \bar v(\tau '' ) \bar v ( \tau ' ) \right ] \nonumber \\ & & \times { \rm exp}\left [ \sum_{k=0}^{\infty } \frac{\bar c_k}{\hbar^2 } \int_{0}^{\tau } d \tau '' \int_{0}^{\tau '' } d \tau ' \bar v(\tau '' ) \bar v ( \tau ' ) \cos ( \nu_k \tau'')\cos ( \nu_k \tau ' ) \right ] \nonumber \\ & & \times { \rm exp}\left [ \sum_{k=1}^{\infty } \frac{\bar c_k}{\hbar^2 } \int_{0}^{\tau } d \tau '' \int_{0}^{\tau '' } d \tau ' \bar v(\tau '' ) \bar v ( \tau ' ) \sin ( \nu_k \tau'')\sin ( \nu_k \tau ' ) \right],\end{aligned}\ ] ] and we have @xmath92= z_{b } \bar \rho [ \bar \phi ^\dag , \bar \phi ; \beta \hbar]$ ] . this defines the correlated equilibrium initial conditions of eq .. in the next section , we derive the equations of motion to evaluate these hierarchy elements . the thermal equilibrium state @xmath159 $ ] at time @xmath110 and inverse temperature @xmath160 can be obtained by considering the imaginary - time derivative of the reduced density matrix elements given in eq.([eq : pci ] ) . this is expressed as @xmath161 } ( \tau ) & & = - \hat h_a \hat { \bar \rho}_{k^1,\dots , k^{m}}^{\,[m : l ] } ( \tau ) + \frac{1}{\hbar } \sum\limits_{k^{m+1 } = 0}^{k } \bar c_{k^{m+1 } } \cos(\nu_{k^{m+1 } } \tau ) \hat v \hat { \bar \rho}_{k^1,\dots , k^{m+1}}^{\,[m+1:l ] } ( \tau ) \nonumber \\ & & + \frac{1}{\hbar } \sum\limits_{k^{m+1 } = 0}^{k } \bar c_{k^{m+1 } } \sin(\nu_{k^{m+1 } } \tau ) \hat v \hat { \bar \rho}_{k^1,\dots , k^{m+1}}^{\,[m+1:l+1 ] } ( \tau ) \nonumber \\ & & + \frac{1}{\hbar}\sum\limits_{h = 1}^{m - l}\cos(\nu_{k^h } \tau ) \hat v \hat { \bar \rho}_{k^1,\dots , k^{h-1},k^{h+1},\dots , k^{m}}^{\,[m-1:l ] } ( \tau ) \nonumber \\ & & + \frac{1}{\hbar}\sum\limits_{h = m - l+1}^{{\color{red}{m}}}\sin(\nu_{k^h } \tau ) \hat v \hat { \bar \rho}_{k^1,\dots , k^{h-1},k^{h+1},\dots , k^{m}}^{\,[m-1:l-1 ] } ( \tau ) , \label{eq : imheom}\end{aligned}\ ] ] where @xmath162}(\tau)$ ] is the density operator defined in path integral form as @xmath163 } ( \phi _ 0^\dag , \phi ' _ 0 ; \tau ) & & = \int_{\bar \phi ( 0)=\phi'_0}^{\bar \phi ( \tau)=\phi_0 } d[\bar \phi ^{\dag}(\tau ) \bar \phi ( \tau ) ] \prod_{g=1}^{m - l } \left ( \int_0^{\tau } d\tau_g \cos(\nu_{k^g}\tau_g ) \bar v(\tau_g ) \right ) \nonumber \\ & & \times\prod_{g'=m - l+1}^{m } \left ( \int_0^{\tau } d\tau_{g ' } \sin(\nu_{k^{g'}}\tau_{g ' } ) \bar v(\tau_{g ' } ) \right ) \bar \rho[\bar \phi ^\dag , \bar \phi ; \tau ] . \label{eq : imheomelm}\end{aligned}\ ] ] note that the first product in eq . contains @xmath164 factors , and the second contains @xmath165 factors . thus , the expression there is @xmath164th order in @xmath166 and @xmath165th order in @xmath167 . also , @xmath168 } ( \tau)$ ] and @xmath169 } ( \tau)$ ] in eq . denote the hierarchy elements defined by eq . without the index @xmath170 for @xmath171 . note that any exchange of suffixes @xmath172 and @xmath173 in eq . that merely results in the permutation of two cosine factors or two sine factors leaves the total integral unchanged , while one that results in the arguments of a sine and cosine being exchanged will generally cause the total integral to change . to truncate the hierarchy equations , we choose some large value of @xmath174 and set the elements at @xmath175th order to 0 . we thus obtain a closed set of equations up to @xmath176th order . to illustrate the structure of the hierarchy given in eq . , here we write out the equations up to second order . the hierarchy starts from the zeroth - order equation , which is that for the thermal equilibrium state density matrix : @xmath177 } ( \tau ) & & = - \hat h_a \hat { \bar \rho}^{\,[0:0 ] } ( \tau ) \nonumber \\ & & + \frac{1}{\hbar } \sum\limits_{k^{1 } = 0}^{k ' } \bar c_{k^{1 } } \cos(\nu_{k^{1 } } \tau ) \hat v \hat { \bar \rho}_{k^{1}}^{\,[1:0 ] } ( \tau ) + \frac{1}{\hbar } \sum\limits_{k^{1 } = 0}^{k ' } \bar c_{k^{1 } } \sin(\nu_{k^{1 } } \tau ) \hat v \hat { \bar \rho}_{k^1}^{\,[1:1 ] } ( \tau).\end{aligned}\ ] ] then , the first order consists of two equations , @xmath178 } ( \tau ) & & = - \hat h_a \hat { \bar \rho}_{k^1}^{\,[1:0 ] } ( \tau ) + \frac{1}{\hbar}\cos(\nu_{k^1 } \tau ) \hat v \hat { \bar \rho}^{\,[0:0 ] } ( \tau ) \nonumber \\ & & + \frac{1}{\hbar } \sum\limits_{k^{2 } = 0}^{k ' } \bar c_{k^{2 } } \cos(\nu_{k^{2 } } \tau ) \hat v \hat { \bar \rho}_{k^1,k^{2}}^{\,[2:0 ] } ( \tau ) + \frac{1}{\hbar } \sum\limits_{k^{2 } = 0}^{k ' } \bar c_{k^{2 } } \sin(\nu_{k^{2 } } \tau ) \hat v \hat { \bar \rho}_{k^1,k^{2}}^{\,[2:1 ] } ( \tau),\end{aligned}\ ] ] @xmath179 } ( \tau ) & & = - \hat h_a \hat { \bar \rho}_{k^1}^{\,[1:1 ] } ( \tau ) + \frac{1}{\hbar}\sin(\nu_{k^1 } \tau ) \hat v \hat { \bar \rho}^{\,[0:0 ] } ( \tau ) \nonumber \\ & & + \frac{1}{\hbar } \sum\limits_{k^{2 } = 0}^{k ' } \bar c_{k^{2 } } \cos(\nu_{k^{2 } } \tau ) \hat v \hat { \bar \rho}_{k^1,k^{2}}^{\,[2:1 ] } ( \tau)+ \frac{1}{\hbar } \sum\limits_{k^{2 } = 0}^{k ' } \bar c_{k^{2 } } \sin(\nu_{k^{2 } } \tau ) \hat v \hat { \bar \rho}_{k^1,k^{2}}^{\,[2:2 ] } ( \tau),\end{aligned}\ ] ] and the second order consists of three equations , @xmath180 } ( \tau ) & & = - \hat h_a \hat { \bar \rho}_{k^1,k^{2}}^{\,[2:0 ] } ( \tau ) + \frac{1}{\hbar } \cos(\nu_{k^1 } \tau ) \hat v \hat { \bar \rho}_{k^{2}}^{\,[1:0 ] } ( \tau)+ \frac{1}{\hbar } \cos(\nu_{k^2 } \tau ) \hat v \hat { \bar \rho}_{k^{1}}^{\,[1:0 ] } ( \tau ) \nonumber \\ & & + \frac{1}{\hbar } \sum\limits_{k^{3 } = 0}^{k ' } \bar c_{k^{3 } } \cos(\nu_{k^{3 } } \tau ) \hat v \hat { \bar \rho}_{k^1,k^2,k^3}^{\,[3:0 ] } ( \tau ) + \frac{1}{\hbar } \sum\limits_{k^{3 } = 0}^{k ' } \bar c_{k^3 } \sin(\nu_{k^{3 } } \tau ) \hat v \hat { \bar \rho}_{k^1,k^2,k^{3}}^{\,[3:1 ] } ( \tau),\end{aligned}\ ] ] @xmath181 } ( \tau ) & & = - \hat h_a \hat { \bar \rho}_{k^1,k^{2}}^{\,[2:1 ] } ( \tau ) + \frac{1}{\hbar } \cos(\nu_{k^1 } \tau ) \hat v \hat { \bar \rho}_{k^{2}}^{\,[1:1 ] } ( \tau ) + \frac{1}{\hbar } \sin(\nu_{k^2 } \tau ) \hat v \hat { \bar \rho}_{k^1}^{\,[1:0 ] } ( \tau ) \nonumber \\ & & + \frac{1}{\hbar } \sum\limits_{k^{3 } = 0}^{k ' } \bar c_{k^{3 } } \cos(\nu_{k^{3 } } \tau ) \hat v \hat { \bar \rho}_{k^1,k^2,k^3}^{\,[3:1 ] } ( \tau ) + \frac{1}{\hbar}\sum\limits_{k^{3 } = 0}^{k ' } \bar c_{k^3 } \sin(\nu_{k^{3 } } \tau ) \hat v \hat { \bar \rho}_{k^1,k^2,k^{3}}^{\,[3:2 ] } ( \tau),\end{aligned}\ ] ] @xmath182 } ( \tau ) & & = - \hat h_a \hat { \bar \rho}_{k^1,k^{2}}^{\,[2:2 ] } ( \tau ) + \frac{1}{\hbar } \sin(\nu_{k^1 } \tau ) \hat v \hat { \bar \rho}_{k^{2}}^{\,[1:1 ] } ( \tau)+ \frac{1}{\hbar } \sin(\nu_{k^2 } \tau ) \hat v \hat { \bar \rho}_{k^{1}}^{\,[1:1 ] } ( \tau ) \nonumber \\ & & + \frac{1}{\hbar } \sum\limits_{k^{3 } = 0}^{k ' } \bar c_{k^{3 } } \cos(\nu_{k^{3 } } \tau ) \hat v \hat { \bar \rho}_{k^1,k^2,k^3}^{\,[3:2 ] } ( \tau ) + \frac{1}{\hbar } \sum\limits_{k^{3 } = 0}^{k ' } \bar c_{k^3 } \sin(\nu_{k^{3 } } \tau ) \hat v \hat { \bar \rho}_{k^1,k^2,k^{3}}^{\,[3:3]}(\tau).\end{aligned}\ ] ] from the definition , the initial conditions are set as @xmath183}(0)= \bf i$ ] , where @xmath184 is the unit operator , with all other hierarchy elements set to zero . the calculated elements @xmath162}(\beta \hbar)$ ] must be normalized after the integration over imaginary time is carried out by dividing by @xmath185}(\beta \hbar)\}$ ] . a significant difference between the real - time heom , given in eq . , and the above imaginary - time heom is that the former contain damping terms proportional to @xmath104 and @xmath186 , whereas the latter contain sinusoidal terms . the imaginary - time heom readily yield the desired quantities , as they are solved by integrating over the pre - determined interval from @xmath99 to @xmath187 , in contrast to the situation for the real - time heom , in which the integration must be carried out until convergence to the steady state is realized . any equilibrium expectation value of the system can be easily evaluated from @xmath183}(\beta \hbar)$ ] . moreover , we can evaluate the imaginary - time correlation functions@xcite from eq . in the same manner that the real - time correlation functions are evaluated from the real - time heom.@xcite the correlated initial states for the real - time heom can be constructed from the hierarchy elements of the imaginary - time heom . the relations between the real - time and imaginary - time heom elements are similar to the relations between the expectation value of the collective bath oscillator coordinate and the real - time heom elements.@xcite here , we present the relations between the two sets of elements up to second order in the system - bath interaction : @xmath188 } ( \beta\hbar ) , \label{eq : rho000}\end{aligned}\ ] ] @xmath189 } ( \beta\hbar ) + \sum_{k=1}^{k ' } \left(\textcolor{red}{c_k''}\hat { \bar \rho}_{k}^{\,[1:0 ] } ( \beta\hbar ) \textcolor{red}{+i c_k ' } \hat { \bar \rho}_{k}^{\,[1:1 ] } ( \beta\hbar ) \right ) \right ] , \label{eq : rho100}\end{aligned}\ ] ] @xmath190 } ( \beta\hbar ) \textcolor{red}{-i c_k ' } \hat { \bar \rho}_{k}^{\,[1:1 ] } ( \beta\hbar ) \right ) , \label{eq : rho0k0}\end{aligned}\ ] ] @xmath191 } ( \beta\hbar ) + c_0''\sum_{k=1}^{k ' } \left(\textcolor{red}{c_k '' } \hat { \bar \rho}_{0k}^{\,[2:0 ] } ( \beta\hbar ) \textcolor{red}{+i c_k'}\hat { \bar \rho}_{0k}^{\,[2:1 ] } ( \beta\hbar ) \right ) \right . \nonumber \\ \left . + \sum_{k=1}^{k ' } \sum_{k'=1}^{k ' } \left(\textcolor{red}{c_k''c_{k ' } '' } \hat { \bar \rho}_{kk'}^{\,[2:0 ] } ( \beta\hbar ) \textcolor{red}{+2i c_k''c_{k'}'}\hat { \bar \rho}_{kk'}^{\,[2:1]}(\beta\hbar ) \textcolor{red}{- c_{k } ' c_{k'}'}\hat { \bar \rho}_{kk'}^{\,[2:2]}(\beta\hbar ) \right ) \right],\end{aligned}\ ] ] @xmath192 } ( \beta\hbar ) \textcolor{red}{-2 i c_k ' c_{k ' } ' } \hat { \bar \rho}_{k , k'}^{\,[2:1 ] } ( \beta\hbar ) \right . \nonumber \\ & & \left . \textcolor{red}{- c_k ' c_{k'}'}\hat { \bar \rho}_{k , k'}^{\,[2:2 ] } ( \beta\hbar ) \right ) . \label{rho011}\end{aligned}\ ] ] the elements @xmath193 are obtained by setting @xmath194 in eq .. in practice , in order to evaluate the heom elements in the case of correlated initial conditions from the imaginary - time heom , the cutoff , @xmath195 , must be comparable to the cutoff of used for the real - time heom given in eq . , @xmath119 . if we only need the equilibrium distribution , eq . , however , we may choose @xmath195 even slightly smaller than @xmath196 . the equilibrium reduced density matrix has been evaluated from various approaches.@xcite equation allows us to calculate the same matrix elements using the reduced equation of motion approach . moreover , this approach allows us to evaluate the correlated initial conditions of the real - time heom . in this section , we report the results of numerical simulations that demonstrate the applicability and validity of the imaginary - time heom , given in eq . , for the system hamiltonian @xmath197 we chose the system parameters as @xmath198 and @xmath199 or @xmath200 , and the bath parameters as @xmath201 , @xmath202 , and @xmath203 for the system - bath interaction @xmath204 . we truncated the hierarchy by setting @xmath205 } ( \tau)=0 $ ] for @xmath206 in the imaginary - time heom , while we truncated by setting @xmath207 for @xmath208 with @xmath209 in the real - time heom . the 4th - order runge - kutta method was used for both the imaginary - time and real - time integrations with time steps of @xmath210 and @xmath211 , respectively . the real - time heom were integrated from the factorized initial conditions @xmath212 and @xmath213 at @xmath110 , and steady states were realized between @xmath214 and @xmath215 . first , we verified the accuracy of the imaginary - time heom by comparing the equilibrium state obtained from them with the steady state distributions obtained from the real - time heom for the temperatures @xmath216 , and @xmath217 with @xmath218 and @xmath219 . we found that the steady - state calculated from the real - time heom , @xmath220 , deviates from the equilibrium state calculated from the imaginary - time heom , @xmath221 } ( \beta\hbar)/ z_{a}$ ] by less than @xmath222 of difference at @xmath223 . the difference between the two results become larger for larger @xmath1 and for deeper hierarchy elements , because we solved the two kinds of heom using different truncation schemes . other than this difference , however , the imaginary - time heom results are consistent with the real - time heom results . this also indicates that the steady - state elements obtained from eq . indeed represent the correlated thermal equilibrium state defined by eq .. note that we must chose @xmath224 in order to accurately calculate the real - time heom elements for the correlated initial conditions from the imaginary - time heom . then , in order to obtain a better accuracy for deeper hierarchy elements in the imaginary - time heom , we used a small time step in the numerical integrations . for this reason , the computational costs for the real - time and imaginary - time heom were comparable . however , if we merely needed the equilibrium distribution @xmath221 } ( \beta \hbar)$ ] to one percent accuracy , we could use a smaller cuttoff @xmath225 and/or a larger time step for the imaginary - time heom and thereby reduce the computational costs to less than @xmath226 of that for the results reported here . .hierarchy elements calculated from the imaginary - time heom and real - time heom for several values of the inverse temperature , @xmath1 . [ cols="^,^,^,^,^,^,^,^",options="header " , ] although with the real - time heom , we can calculate only the probability distribution , with the imaginary - time heom we are able to calculate thermodynamic variables via the partition function of the reduced system , @xmath185}(\beta \hbar)\}$ ] . note that the total partition function can be expressed as @xmath227 , where the partition function of the bath is given by @xmath228 because we consider an infinite number of oscillators , however , the partition function of the bath can not be determined . for this reason , we consider the system part , @xmath229 , only . we calculated the helmholtz free energy , @xmath230 , the entropy , @xmath231 , the internal energy , @xmath232 , the heat capacity , @xmath233 , and the susceptibility , @xmath234 , from @xmath229 for several values of @xmath1 . to obtain these quantities , we numerically integrated the imaginary - time heom for fixed @xmath218 and @xmath199 to obtain @xmath229 for @xmath1 satisfying @xmath235 at steps of @xmath236 . for the susceptibility , we also calculated the free energy for @xmath237 in order to evaluate the derivative with respect to @xmath238 at @xmath239 . the quantities mentioned above obtained using the imaginary - time heom are compared in fig . 1 with the corresponding quantities for a system characterized by the canonical distribution , @xmath240\}$ ] , with the same hamiltonian @xmath2 ( with @xmath199 ) and inverse temperature @xmath1 . this corresponds to the partition function of the system in the case that the total partition function takes the factorized form @xmath241\ } tr_b\ { \exp[-\beta \hat h_b]\}$ ] . the thermodynamic quantities are then given by @xmath242 , @xmath243 , @xmath244 , @xmath245 - @xmath246 $ ] , and @xmath247 . also , note that the susceptibility for finite @xmath238 is expressed as @xmath248/2 $ ] . the superscript `` 0 '' on these quantities indicates that these are calculated using the conventional statistical physics approach , which is equivalent to assuming a factorized thermal equilibrium state . as seen in fig . 2 , in both cases of the spin - boson and factorized spin system , the entropy and internal energy decrease with the inverse temperature , while the heat capacities of both systems exhibit maxima at inverse temperatures near @xmath249 , where the thermal excitation energy becomes comparable to the excitation energy . the entropy in the spin - boson case is larger than that in the factorized case at lower temperatures because the spin - boson system involves more degrees of freedom , due to the presence of the system - bath interaction . it is also seen that the internal energy is systematically lower in the spin - boson case than in the factorized case . this indicates that the bath absorbs some of the system energy through the interaction . the degree to which the system energy is absorbed by the bath increases as @xmath1 approaches the thermal excitation energy of the system and , as a result , the heat capacity of the spin - boson system becomes smaller than that of the factorized spin system near the peak position at @xmath249 . compared with the other thermodynamic variables , the difference between the susceptibilities in the two cases is small . this is because the system - bath interaction has the same form as the magnetic excitation , and the effects of @xmath238 are suppressed by the strong system - bath interaction . it is important to note here that those states regarded as the thermal equilibrium states in the two cases compared above are different . in the conventional treatment , the thermal equilibrium state of the system corresponds to the case of a factorized partition function , while in the present treatment of the spin - boson system , we consider the thermal equilibrium state of the total system . although the difference between the equilibrium thermodynamic quantities for the spin - boson system and the factorized spin system are rather minor in the static case considered in fig . 2 , the difference becomes significant when we study the dynamics of the system , because in this case , the positivity condition is often violated in the conventional treatment . this may indicate that treatments based on the canonical distribution are inherently incompatible with dynamical states . by utilizing the hierarchy elements , we can calculate expectation values of the system and bath . for example , the expectation value of the system energy , @xmath250 , is obtained from eq . as @xmath251 using the first element of the hierarchy , the expectation value of the system - bath interaction , @xmath252\}$ ] , is evaluated as @xmath253 where @xmath254 and @xmath255 are obtained from eqs . and , respectively . in fig . 2 , we present the expectation values @xmath250 and @xmath256 and the internal energy of the system , @xmath257 , as obtained by numerically integrating eq . for various coupling strengths , @xmath105 , at @xmath258 and @xmath259 , with a step size of @xmath260 . at those temperatures , the system part of the energy increases linearly , while the interaction part decreases linearly as a function of the coupling strength , @xmath105 , but the rates of decrease and increase are smaller for lower temperatures , because the thermal activity of the bath is lower in this regime . we should mention that the internal energy , @xmath257 , contains the system part of the interaction energy but not the bath part . the system part and bath part of the interaction energies are calculated as @xmath261 and @xmath262 , respectively . the internal energy decreases as a function of @xmath105 because the system part of the interaction energy also decreases as a function of @xmath105 . both the bath and system parts of the interaction energy decrease as function of @xmath105 , but the bath part of the contribution is much larger than the system part , because the bath contains many degrees of freedom . in this paper , we derived the real - time and imaginary - time heom starting from the influence functional formalism with a correlated thermal initial state . it was shown that the thermal equilibrium state calculated from the imaginary - time heom is equivalent to the steady state solution of the real - time heom . because the imaginary - time heom is defined in terms of integrals carried out over the definite time interval from @xmath263 to @xmath264 and because the elements of the imaginary - time heom are real , we were able to calculate the hierarchy elements more easily in this case than in the case of the real - time heom . moreover , using the imaginary - time heom , we were able to calculate the partition function , and from this , we could directly obtain several thermodynamic quantities , namely , the free energy , entropy , internal energy , heat capacity , and susceptibility of the system in the dissipative environment . the expectation values of not only the system energy but also the system - bath interaction energy were also evaluated from the hierarchy elements obtained from the real- and imaginary - time heom . we found that for the purpose of studying equilibrium properties , rather than dynamical behavior , the imaginary - time heom is more usueful than the real - time heom . in this paper , we derived the heom for a system in the energy eigenstate representation , but extension to a system in the coordinate space representation ( or the wigner representation ) is straightforward.@xcite this extension will be helpful for identifying the pure quantum effects , because we can compare the quantum results with the classical results obtained from the classical limit of the imaginary - time heom.@xcite the author is grateful to professor hermann grabert , professor gerhard stock , professor heinz - peter breuer , and dr . lothar mchenber and their group members at the freiburg institute for advanced studies for their hospitality . financial support from the humboldt foundation , the university of freiburg , and a grant - in - aid for scientific research ( a26248005 ) from the japan society for the promotion of science are acknowledged . because the extension to a many oscillator system is straightforward , we start from a single oscillator bath , described by the hamiltonian @xmath265 we then consider the density matrix elements with the three source terms @xmath266 , @xmath267 , and @xmath268 defined by@xcite @xmath269 e^ { \frac{i}{\hbar}\int_0^{t } d\tau \left[\frac12{m \dot { x}^2}- \frac12 m \omega^2 x^2 + v(\tau ) x \right ] } \nonumber \\ & & \times \int _ { \bar x(0)=x_0'}^{\bar x(\beta \hbar)=x_0 } d[\bar x(\tau ' ) ] e^{-\frac1{\hbar } \int_0^{\beta \hbar } d\tau ' \left[\frac12{m \dot { \bar x}^2}+ \frac12 m \omega^2 \bar x^2 - { \bar v}(\tau ' ) \bar x \right ] } \nonumber \\ & & \times \int_{x'(0)=x_0'}^{x'(t)=x ' } d[x'(\tau ) ] e^{- \frac{i}{\hbar}\int_0^{t } d\tau \left[\frac12{m { \dot x}'^2}- \frac12 m \omega^2 x'^2 + v'(\tau)x ' \right ] } . \label{eq : gv}\end{aligned}\ ] ] in order to evaluate @xmath270 , we start from the feynman propagator for eq . , expressed as @xmath271 e^ { \frac{i}{\hbar}\int_0^{t } d\tau \left[\frac12{m \dot { x}^2}- \frac12 m \omega^2 x^2 + v(\tau ) x \right ] } \nonumber \\ & & = \sqrt { \frac{m \omega}{2 \pi i \hbar \sin ( \omega t ) } } e^{\frac{i}{\hbar}s(x , x_0;t;v ) } , \label{eq : green}\end{aligned}\ ] ] where @xmath272 . \label{eq : s_harm}\end{aligned}\ ] ] the equilibrium distribution , @xmath273 , is obtained from eq . by replacing @xmath274 with @xmath275 . this yields @xmath276 e^{-\frac1{\hbar } \int_0^{\beta \hbar } d\tau ' \left[\frac12{m \dot { \bar x}^2}+ \frac12 m \omega^2 \bar x^2 - { \bar v}(\tau ' ) \bar x \right ] } \nonumber \\ & & = \frac1{2\sinh\left(\frac{\beta \hbar \omega}2 \right ) } \sqrt{\frac1{2 \pi \langle x^2\rangle } } \exp \left [ - \frac1{2 \langle x^2 \rangle } \left ( \frac{x+x'}{2 } - \bar r[{\bar v } ; \beta \hbar ] \right)^2 \right . \nonumber \\ & & - \left . \frac1{2\hbar^2 } \langle p^2 \rangle ( x - x')^2 + \frac{i}{\hbar } \bar p[{\bar v } ; \beta \hbar ] ( x - x ' ) + { \bar \phi}[{\bar v } ; \beta \hbar ] \right ] , \label{eq : rhoqqt}\end{aligned}\ ] ] where @xmath277 @xmath278 @xmath279 = \frac{1}{\hbar } \int_{0}^{\beta\hbar } d \tau ' { \bar v } ( \tau ' ) \bar l ( \tau ' ) , \label{eq:<barr>}\end{aligned}\ ] ] @xmath280 = \frac{i m } { \hbar } \int_{0}^{\beta\hbar } d \tau ' { \bar v } ( \tau ' ) \dot { \bar l } ( \tau ' ) , % \left[\coth \left(\frac{\beta \hbar\omega}2 \right ) \sinh \omega \tau + \cosh \omega \tau \right ] \label{eq:<barp>}\end{aligned}\ ] ] and @xmath281 = \frac1{\hbar^2 } \int_{0}^{\beta\hbar } d \tau '' \int_{0}^{\tau '' } d \tau ' { \bar v } ( \tau '' ) { \bar v } ( \tau ' ) \bar l ( \tau '' - \tau ' ) . \label{eq : phita2}\end{aligned}\ ] ] here , we have @xmath282 . % & & = \frac{\hbar}{2m\omega } \frac{\cosh\left ( \frac{\beta \hbar\omega}2 - \omega \tau ' \right ) } { \sinh\left ( \frac{\beta \hbar\omega}2 \right)}. \label{eq : p1}\end{aligned}\ ] ] note that the partition function for the oscillator itself , @xmath283 , can be obtained from eq . as @xmath284 using the counter path , we can express the total density matrix , eq . , as@xcite @xmath285 e^{-\frac{i}{\hbar } \int_c ds \left [ \frac12{m \dot { \tilde x}^2(s)}- \frac12 m \omega^2\tilde x^2(s ) + \tilde v_c(s)\tilde x(s ) \right ] } , \label{eq : gj33}\end{aligned}\ ] ] where @xmath286\equiv \int d[x(\tau ) ] \int d[\bar x(\tau ' ) ] \int d[x'(\tau)]$ ] and the contour paths are defined by eqs . and . we can obtain the full density matrix elements by simply replacing the integral @xmath287 in eqs.- with the contour integral @xmath288 . in the wigner representation , we have the distribution @xmath289 after the normalization , this can be expressed as@xcite @xmath290 \right)^2 - \frac1{2 \langle p^2\rangle } ( p-\tilde p[\tilde v_c ; t , \beta\hbar])^2 + { \tilde \phi}[\tilde v_c ; t , \beta\hbar ] \right ] , \label{eq : wignerh}\end{aligned}\ ] ] where @xmath291 = -\frac{i}{\hbar } \int_{c } ds ' \tilde v_c(s ' ) l(s'),\end{aligned}\ ] ] @xmath292 = \frac{i m}{\hbar } \int_{c } ds ' \tilde v_c(s ' ) { \dot l}(s'),\end{aligned}\ ] ] @xmath293 & = & - \frac1{\hbar^2 } % \frac1{2 m \hbar \omega } \int_{c } ds '' \int_{c ' } ds ' \tilde v_c ( s '' ) \tilde v_{c ' } ( s ' ) l(s''- s ' ) . % \nonumber \\ & \times & % \left [ \coth\left ( \frac{\beta \hbar\omega}2 \right ) % \cos\left(\omega(s - s')\right ) % + i \sin \left(\omega ( s - s')\right ) \right ] . \label{eq : p2}\end{aligned}\ ] ] here , @xmath294 is the analytically continued matsubara green function obtained from eq . through the replacement @xmath295 , given by @xmath296 and @xmath65 represents the counter path for @xmath66 that follows @xmath67 along @xmath68 under the condition @xmath69 . for the bath hamiltonian appearing in eq . with the interaction @xmath297 , eq . is expressed as eq .. by tracing out @xmath298 and @xmath299 , we obtain the influence functional for correlated initial conditions , @xmath63=\exp\left\{{\tilde \phi}[\tilde v_c ; t , \beta\hbar]\right\}$ ] . 20 a. o. caldeira and a. j. leggett , physica * 121a * , 587 ( 1983 ) . a. j. leggett , s. chakravarty , a. t. dorsey , m. p. a. fisher , a. garg , and z. zwerger , rev . phys . * 59 * , 1 ( 1987 ) . u. weiss , _ quantum dissipative systems _ ( world scientific , singapore , 2008 ) 3rd ed . r. kubo , m. toda and n. hashitsume , _ statistical physics _ ( springer - verlag , 1985 ) vol . * 2*. a. g. redfield , adv . . reson . * 1 * , 1 ( 1965 ) . r. kubo , adv . 15 * , 101 ( 1969 ) . s. mukamel , _ principles of nonlinear optical spectroscopy _ ( oxford university press , new york , 1995 ) . p. hanggi , p. talkner and m. borkovec , rev . phys . * 62 * , 252 ( 1990 ) . r. a. marcus , rev . phys . * 65 * , 599 ( 1993 ) . v. may and o. khn , _ charge and energy transfer dynamics in molecular systems _ ( wiley - vch , berlin , 2003 ) . h. p. breuer and f. petruccione , _ the theory of open quantum systems _ ( oxford university press , new york , 2002 ) . e. b. davies , _ quantum theory of open systems _ ( academic press , 1976 ) . v. gorini , a. frigerio , m. verri , a. kossakowski , and e. c. g. sudarshan , rep . math . phys . * 13 * , 149 ( 1978 ) . h. spohn , rev . phys . * 52 * , 569 ( 1980 ) . p. pechukas , phys . lett . * 73 * , 1060 ( 1994 ) . k. f. f. romero , p. talkner and p. hanggi , phys . rev . * 69 * , 052109 ( 2004 ) . a. frigerio , j. t. lewis and j. v. pul , adv . appl . math . * 2 * , 456 ( 1981 ) . a. frigerio , j. t. lewis and j. v. pul , j. approx . theor . * 45 * , 310 ( 1985 ) . y. tanimura , j. phys . jpn . * 75 * , 082001 ( 2006 ) . r. egger and c. h. mak , phys . rev . b*50 * , 15210 ( 1994 ) . n. makri , j. math . phys . * 36 * , 2430 ( 1995 ) . j. cao , l. w. ungar and g.a . voth , j. chem . phys * 104 * , 4189 ( 1996 ) . n. makri and d. e. makarov , j. chem . * 102 * , 4600 ( 1995 ) ; * 102 * , 4611 ( 1995 ) . m. thorwart , p. reimann and p. hnggi , phys . e * 62 * , 5808 ( 2000 ) . e. bukhman and n. makri , j. phys . a * 111 * , 11320 ( 2007 ) . v. jadhao and n. makri , j. chem . phys . * 129 * , 161102(2008 ) . y. tanimura , phys . * a41 * , 6676 ( 1990 ) . a. ishizaki and y. tanimura , j. phys . . jpn . * 74 * , 3131 ( 2005 ) . r. x. xu , p. cui , x. q. li , y. mo , and y. j. yan , j. chem . phys . * 122 * , 041103 ( 2005 ) . p. han , r. x. xu , b. li , j. xu , p. cui , y. mo and y. j. yan , j. phys . chem . b * 110 * , 11438 ( 2006 ) . y. tanimura and s. mukamel , j. phys . soc . jpn . * 63 * , 66 ( 1994 ) . m. tanaka and y.tanimura , j. phys . . jpn . * 78 * , 073802 ( 2009 ) . m. tanaka and y.tanimura , j. chem . phys . * 132 * , 214502 ( 2010 ) . y. tanimura , j. chem . phys . * 137 * , 22a550 ( 2012 ) . c. kreisbeck and t. kramer , j. phys . lett . * 3 * , 2828 ( 2012 ) . j. ma , z. sun , x. wang , and f. nori , phys . rev . a * 85 * , 062323 ( 2012 ) . ding , r .- x . xu , y .- j . yang , j. chem . phys . * 136 * , 224103 ( 2012 ) . h. liu , l. zhu , s. bai , and q. shi , j. chem . phys . * 140 * , 134106 ( 2014 ) . y. tanimura and r. kubo , j. phys . . jpn . * 58 * , 1199 ( 1989 ) . y. tanimura , t. suzuki and r. kubo , j. phys . * 58 * , 1850 ( 1989 ) . y. tanimura and y. maruyama , j. chem . phys . * 107 * , 1779 ( 1997 ) . t. kato and y. tanimura , j. chem . phys . * 120 * , 260 ( 2004 ) . a. ishizaki and y. tanimura , j. chem . phys . * 125 * , 084501 ( 2006 ) . a. ishizaki and y. tanimura , j. phys . a * 111 * , 9269 ( 2007 ) . a. sakurai and y. tanimura , j. phys . a * 115 * , 4009 ( 2011 ) . a. g. dijkstra and y. tanimura , new j. phys . * 12 * , 055005 ( 2010 ) . l. chen , r. zheng , q. shi and y .- j . yan , j. chem . phys . * 132 * , 024505 ( 2010 ) . l. chen , r. zheng , y. jing , q. shi , j. chem . phys . * 134 * , 194508 ( 2011 ) . b. hein , c. kreisbeck , t. kramer , and m. rodriguez , new j. phys . * 14 * , 023018 ( 2012 ) . c. kreisbeck , t. kramer and a. aspuru - guzik , j. phys . b , * 117 * , 9380 ( 2013 ) . h .- breuer , e .- laine , and j. piilo , phys . rev . lett . * 103 * , 210401 ( 2009 ) . a. g. dijkstra and y. tanimura , phys . lett . * 104 * , 250401 ( 2010 ) . a. g. dijkstra and y. tanimura , phil . a * 370 * , 3658 ( 2012 ) . a. g. dijkstra and y. tanimura , j. phys . * 81 * , 063301 ( 2012 ) . q. shi , l. p. chen , g. j. nan , r .- x . yan , j. phys . chem . * 130 * , 084105 ( 2009 ) . j. hu , r .- x . yan , j. chem . 133 * , 101106 ( 2010 ) . b. l. tian , j. j. ding , r .- x . yan , j. chem . phys . * 133 * , 114112 ( 2010 ) . j. zhu , s. kais , p. rebentros , a. aspuru - guzik , j. phys . b * 115 * , 1531 ( 2011 ) . c. kreisbeck , t. kramer , m. rodriguez , b. hein , j. chem . theor . comp . * 7 * , 2166 ( 2011 ) . j. strmpfer , k. schulten , j. chem . theor . comp . * 8 * , 2808 ( 2012 ) . j. m. moix and j. cao , j. chem . phys . * 139 * , 134106 ( 2013 ) . l. zhu , h. liu , and q. shi , n. j. phys . * 15 * , 095020 ( 2013 ) . a. ishizaki and g. r. fleming , proc . 106 * , 17255 ( 2009 ) . j. strmpfer , k. schulten , j. chem . phys . * 131 * , 225101 ( 2009 ) . j. strmpfer and k. schulten , j. chem . 134 * , 095102 ( 2011 ) . a. g. dijkstra and y. tanimura , new j. phys . * 14 * , 073027 ( 2012 ) . j. s. jin , s. welack , j. luo , x - q li , p. cui , r .- x . xu and y.j . yan , j. chem . phys . * 126 * , 134113 ( 2007 ) . j. s. jin , x. zheng , and y. j. yan , j. chem . * 128 * , 234703 ( 2008 ) . x. zheng , j. jin , s. welack , m. luo , and y- . j. chem . phys . * 130 * , 164708 ( 2009 ) . a. kato and y. tanimura , j. phys . b , * 117*,13132 ( 2013 ) . a. sakurai and y. tanimura , j. phys . jpn , * 82 * , 033707 ( 2013 ) . a. sakurai and y. tanimura , new j. phys . * 16 * , 015002 ( 2014 ) . y. yao , w. yang , y. zhao , j. chem . 140 * , 104113 ( 2014 ) . y. tanimura and s. mukamel , phys . e * 47 * , 118 ( 1993 ) . k. okumura and y. tanimura , phys . e * 53 * , 214 ( 1996).y . tanimura and k. okumura , j. chem . phys . * 106 * , 2078 ( 1997).y . suzuki and y. tanimura , phys . e * 59 * , 1475 ( 1999 ) . j. cao and g. a. voth , j. chem . phys . * 100 * , 5093 ( 1994 ) l. zhu , h. liu , w. xie and q. shi , j. chem . phys . * 137 * , 194106 ( 2012 ) . j. m. moix , y. zhao , and j. cao , phys . b*85 * , 115412 ( 2012 ) . c. k. lee , j. moix , and j. cao , j. chem phys . * 136 * , 204120 ( 2012 ) . y. tanimura and p. g. wolynes , rev . a*43 * , 4131 ( 1991 ) . y. tanimura and p. g. wolynes , j. chem . phys . * 96 * , 8485 ( 1992 ) . t. steffen and y. tanimura , j. phys . 69 * , 3115 ( 2000 ) . y. tanimura and t. steffen , j. phys . * 69 * , 4095 ( 2000 ) .

we have recently given a combinatorial model for the irreducible characters of a complex semisimple lie group @xmath1 , and , more generally , for the demazure characters @xcite . this model was defined in the context of the equivariant @xmath2-theory of the generalized flag variety @xmath3 . our character formulas were derived from a chevalley - type formula in @xmath4 . our model was based on enumerating certain saturated chains in the bruhat order on the corresponding weyl group @xmath5 . this enumeration is determined by an alcove path , which is a sequence of adjacent alcoves for the affine weyl group @xmath6 of the langland s dual group @xmath7 . alcove paths correspond to decompositions of elements in the affine weyl group . our chevalley - type formula was was expressed in terms of a certain @xmath8-matrix , that is , in terms of a collection of operators satisfying the yang - baxter equation . this setup allowed us to easily explain the independence of our formulas from the choice of an alcove path . there are other models for chevalley - type formulas in @xmath4 and for the characters of the irreducible representations of @xmath1 with highest weight @xmath0 , such as the littelmann path model . littelmann @xcite showed that the characters can be described by counting certain continuous paths in @xmath9 . these paths are constructed recursively , by starting with an initial one , and by applying certain root operators . by making specific choices for the initial path , one can obtain special cases which have more explicit descriptions . for instance , a straight line initial path leads to the lakshmibai - seshadri paths ( ls paths ) . these were introduced before littelmann s work , in the context of standard monomial theory @xcite . they have a nonrecursive description as weighted chains in the bruhat order on the quotient @xmath10 of the corresponding weyl group @xmath5 modulo the stabilizer @xmath11 of the weight @xmath0 ; therefore , we will use the term ls chains when referring to this description . ls paths were used by pittie and ram @xcite to derive a @xmath12-chevalley formula . recently , gaussent and littelmann @xcite , motivated by the study of mirkovi - vilonen cycles , defined another combinatorial model for the irreducible characters of a complex semisimple lie group . this model is based on ls - galleries , which are certain sequences of faces of alcoves for the corresponding affine weyl group . according to @xcite , for each ls - gallery there is an associated littelmann path and a saturated chain in the bruhat order on @xmath10 . our alcove path model from @xcite and ls - galleries , which were developed independently , can be related to each other in the case of regular weights . the main goal of this paper is to extend the model introduced in @xcite from semisimple lie algebras to complex symmetrizable kac - moody algebras . instead of alcove paths ( that make sense only in finite types ) , the present paper is based on @xmath0-chains , which are chains of positive roots defined by certain interlacing conditions . these new objects extend the notion of a reflection ordering @xcite , and have many interesting combinatorial properties , some of which are investigated in this paper . the extension from the finite to the infinite case is nontrivial since the description in terms of the affine weyl groups and alcoves is no longer available in the latter case . compared to the geometric approach in @xcite ( involving the equivariant @xmath2-theory of @xmath3 ) , the generalization in this paper is carried out purely in the context of representation theory , and , in this process , new features of our model are added . for instance , we define root operators in our model , and study their properties . this allows us to show that our model satisfies the axioms of an admissible system of stembridge @xcite . thus , we easily derive character formulas , a littlewood - richardson rule for decomposing tensor products of irreducible representations , as well as a branching rule . the approach via admissible systems was already applied to ls chains in ( * ? ? ? * section 8) . stembridge s approach has the advantage of making a part of the proof independent of a particular model for weyl characters , by using a system of axioms for such models . the root operators in our model define certain colored graph structures which we call semiperfect crystals . in the case of a special @xmath0-chain , we show that our model gives rise to ls paths ( which are also defined in the kac - moody setup ) via a certain process of passing to the limit . it then follows that the semiperfect crystals for the special @xmath0-chain coincide with kashiwara s crystals ( called here perfect crystals ) of the corresponding irreducible integrable highest weight modules @xcite . furthermore , the independence of our model from the choice of a @xmath0-chain was proved in @xcite in the finite case . hence , in this case , we know that our semiperfect crystals are perfect crystals for any choice of a @xmath0-chain . we briefly discuss the relationship between our model and the littelmann path model ( in the kac - moody setup ) . the former is a discrete counterpart of the latter without being simply a translation of it into a different language . let us explain . both models are based on certain choices ( of a @xmath0-chain and of an initial path , respectively ) . however , in its full generality , the path model is a recursive one , as explained above , whereas ours is a nonrecursive one . the definition of root operators in our model always involves reflecting a single segment in a @xmath0-chain , whereas several segments of a path might be reflected in the path model ( cf . section [ sec : finite ] ) . furthermore , we have a generalization of the left - key construction of lascoux - schtzenberger ( relevant to demazure characters , cf . remark [ lskey ] ) in our model for any choice of a @xmath0-chain , whereas , in the path model , a similar generalization exists only for ls paths . although ls paths can be derived from our model , as explained above , the two are not equivalent either . indeed , certain information is lost in the process of passing to the limit , namely the order in which certain hyperplanes are crossed , when a path is deformed into one which passes through the intersection of these hyperplanes . ( in some cases , such as the one related to the @xmath12-chevalley formula in @xcite , the lost information needs to be recovered , but this is done by a nontrivial lift from @xmath10 to @xmath5 ; the @xmath12-chevalley formula in @xcite based on our model obviates this lift . ) the main advantage of our model is related to its combinatorial nature . this model is computationally efficient because it uses simple labels for the vertices of an irreducible crystal , namely certain finite sets of positive integers ( called admissible subsets ) ; these can be efficiently constructed . although a @xmath0-chain might be an infinite object , we do not need to store it , since we always work with a finite set of objects ( cf . remark [ finiteness ] ) . more precisely , all we need is a procedure for comparing two elements in a @xmath0-chain , namely two pairs , each consisting of a positive root and a nonnegative integer ( see proposition [ constr - lambdachain ] for a simple procedure of this type ) . let us also note that our model is equally simple for both regular and nonregular highest weights @xmath0 . indeed , instead of working with chains of cosets in @xmath10 ( as in the case of ls chains ) or with possibly lower dimensional faces of alcoves ( as in the case of ls - galleries ) , we always work with chains in @xmath5 and , in the finite case , with alcoves too . compared to ls chains and ls - galleries , we also eliminate the need for making extra choices after the initial choice ( of an initial path , of a sequence of alcoves / lower dimensional faces of them , or of a @xmath0-chain ) . unlike ls - galleries , our model extends to the kac - moody case . finally , it leads to an extensive generalization of the combinatorics of irreducible characters from lie type @xmath13 ( where the combinatorics is based on young tableaux ) to arbitrary type . for example , an alcove path analog of schtzenberger s evacuation is given in @xcite . the study of the combinatorics of our model will be continued in future publications . let us now present our combinatorial formula for characters . fix a complex symmetrizable kac - moody algebra @xmath14 ; see @xcite . let @xmath15 be the associated set of positive real roots . for a root @xmath16 , let @xmath17 be the corresponding coroot . let @xmath0 be a dominant integral weight , and let @xmath18 be the associated irreducible representation of @xmath14 with the highest weight @xmath0 . we define a @xmath0-chain ( of roots ) @xmath19 as a map @xmath20 , @xmath21 , from a totally ordered index set @xmath22 to positive roots , that satisfies the conditions below ; given @xmath23 and @xmath16 , we use the notation @xmath24 and @xmath25 . 1 . for a root @xmath16 , the number @xmath26 of occurrences of @xmath27 equals @xmath28 . given any triple of roots @xmath29 such that @xmath30 and @xmath31 for some integer @xmath32 , as well as @xmath23 such that @xmath33 , we have @xmath34 it turns out that , in the finite case , condition ( 2 ) is equivalent to a condition stating that any triple of roots @xmath29 with @xmath35 satisfies the following interlacing property : there is exactly one element from the set @xmath36 between any two consecutive @xmath37 s , as well as before the first @xmath37 ; and there are no @xmath27 s or @xmath38 s after the last @xmath37 . note that , according to condition ( 1 ) , the index set @xmath22 is always a countable or a finite set . however , it is not always order - isomorphic to a subset of @xmath39 . for example , it may contain infinite intervals . [ exist - lc ] for any dominant integral weight @xmath0 , there exists a @xmath0-chain . for a weight @xmath0 , there are usually many @xmath0-chains . we will give an explicit construction of a @xmath0-chain ( in proposition [ constr - lambdachain ] ) . for @xmath16 , let @xmath40 denote the associated reflection . the reflections @xmath40 generate the weyl group @xmath5 . the covering relations @xmath41 , for @xmath42 , @xmath43 , define the bruhat order on @xmath5 . these covering relations are labelled by the corresponding positive roots @xmath27 . let us fix a @xmath0-chain @xmath19 . let us say that a finite subset @xmath44 of the index set @xmath22 is admissible if the roots @xmath45 are labels of an increasing saturated chain in the bruhat order starting at the identity element , i.e. , we have : @xmath46 for @xmath16 and @xmath47 , let @xmath48 be the affine reflection given by @xmath49 . for a @xmath0-chain , let @xmath50 be the associated sequence of nonnegative integers defined by @xmath51 . let us define the weight @xmath52 of an admissible subset @xmath53 by @xmath54 for a dominant integral weight @xmath0 and any @xmath0-chain @xmath19 , the character @xmath55 of the irreducible representation @xmath18 is given by @xmath56 where the sum is over the admissible subsets @xmath57 of the index set @xmath22 . [ th : char_formula ] let @xmath14 be the lie algebra of type @xmath58 . let us fix a choice of simple roots @xmath59 , and let @xmath60 be the first fundamental weight . in this case , there is only one @xmath0-chain @xmath61 ( assuming that @xmath62 ) . the index set @xmath22 has 3 admissible subsets : @xmath63 . the subset @xmath64 is not admissible because the reflection @xmath65 does not cover the identity element . in this case , theorem [ th : char_formula ] gives the following expression for the character of @xmath67 : @xmath68 we will define a colored graph structure on the collection of admissible subsets . from this , we will deduce a rule for decomposing tensor products of irreducible representations and a branching rule . the general outline of the paper follows . in section [ sec : notation ] , we review basic notions related to roots systems for complex symmetrizable kac - moody algebras and fix our notation . in section [ sec : admissible ] , we discuss crystals and give stembridge s axioms of admissible systems . in sections [ sec : lambdachains]-[sec : folded ] , we define our combinatorial model . in section [ sec : lambdachains ] , we discuss @xmath0-chains . in section [ sec : chains - foldings ] , we define and study folding operators ; we use them to construct more general chains of roots from a @xmath0-chain , which we call admissible foldings . in section [ sec : folded ] , we study combinatorial properties of admissible subsets and admissible foldings . in section [ sec : rootoperators ] , we define root operators on admissible subsets / foldings . in section [ sec : mainthm ] , we prove that our combinatorial model satisfies stembridge s axioms . this enables us to derive a character formula for complex symmetrizable kac - moody algebras , a littlewood - richardson rule for decomposing tensor products of irreducible representations , as well as a branching rule . in section [ sec : ls ] , we explain the connection between our model and ls chains . in section [ sec : finite ] , we discuss the way in which the present model specializes to the one in our previous paper @xcite in the finite case . we are grateful to john stembridge for the explanation of his work , and to v. lakshmibai and peter magyar for helpful conversations . in this section , we briefly recall the general setup for complex symmetrizable kac - moody algebras and their representations . we closely follow ( * ? ? ? * section 1 ) , and refer to @xcite for more details . let @xmath69 be a finite - dimensional real vector space with a nondegenerate symmetric bilinear form @xmath70 , and let @xmath71 be a crystallographic root system of rank @xmath72 with simple roots @xmath73 . by this , we mean that @xmath74 is the set of real roots of some complex symmetrizable kac - moody algebra . the finite root systems of this type are the root systems of semisimple lie algebras . note that , in the infinite case , it is possible for the simple roots to span a proper subspace of @xmath69 ; indeed , it can happen that the bilinear form is degenerate on the span of the simple roots . given a root @xmath27 , the corresponding _ coroot _ is @xmath75 . the collection of coroots @xmath76 forms the _ dual root system . _ for each root @xmath27 , there is a reflection @xmath77 defined by @xmath78 more generally , for any integer @xmath32 , one can consider the affine hyperplane @xmath79 and let @xmath48 denote the corresponding reflection , that is , @xmath80 the _ weyl group _ @xmath5 is the subgroup of @xmath81 generated by the reflections @xmath82 for @xmath83 . in fact , the weyl group @xmath5 is a coxeter group , which is generated by the _ simple reflections _ @xmath84 corresponding to the simple roots @xmath85 , subject to the _ coxeter relations _ : @xmath86 here the relations of the second type correspond to the distinct @xmath87 in @xmath88 for which the dihedral subgroup generated by @xmath89 and @xmath90 is finite , in which case @xmath91 is half the order of this subgroup . the weyl group is finite if and only if @xmath74 is finite . an expression of a weyl group element @xmath92 as a product of generators @xmath93 which has minimal length is called a _ reduced decomposition _ for @xmath92 ; its length @xmath94 is called the _ length _ of @xmath92 . for @xmath95 , we say that @xmath96 _ covers _ @xmath92 , and write @xmath97 , if @xmath98 , for some @xmath99 , and @xmath100 . the transitive closure `` @xmath101 '' of the relation `` @xmath102 '' is called the _ bruhat order _ on @xmath5 . let us note that @xmath74 can be characterized by the following axioms : 1 . @xmath73 is a linearly independent set . @xmath103 for all @xmath104 . @xmath105 for all distinct simple roots @xmath106 and @xmath107 . @xmath108 . let @xmath109 be the set of positive roots , that is , the set of roots in the nonnegative linear span of the simple roots . then @xmath74 is the disjoint union of @xmath15 and @xmath110 . we write @xmath111 ( respectively , @xmath112 ) for @xmath16 ( respectively , @xmath113 ) , and we define @xmath114 to be @xmath115 ( respectively , @xmath116 ) . we also use the notation @xmath117 . let us choose a set of _ fundamental weights _ @xmath118 , that is , @xmath72 vectors in v such that @xmath119 . the lattice of ( integral ) _ weights _ @xmath120 is the lattice generated by the fundamental weights . the set @xmath121 of _ dominant weights _ is given by @xmath122 if we replace the weak inequalities above with strict ones , we obtain the strongly dominant weights . it is known that every @xmath5-orbit in @xmath69 has at most one dominant member . the ( integral ) tits cone @xmath123 is defined to be the union of all @xmath5-orbits of dominant integral weights , or , equivalently , @xmath124 we have @xmath125 in the finite case , but not otherwise . we now define a ring @xmath8 that contains the characters of all irreducible representations of the corresponding kac - moody algebra . in the finite case , one may simply take @xmath8 to be the group ring of @xmath120 , but in general more care is required . first , we choose a height function @xmath126 , that is , a linear map assigning the value 1 to all simple roots . second , for each @xmath127 , let @xmath128 denote a formal exponential subject to the rules @xmath129 for all @xmath130 . we now define the ring @xmath8 to consist of all formal sums @xmath131 with @xmath132 satisfying the condition that there are only finitely many weights @xmath0 with @xmath133 and @xmath134 , for all @xmath135 . the ring @xmath8 contains the formal power series ring @xmath136 $ ] . in particular , if @xmath137 has constant term 1 , then @xmath138 has a multiplicative inverse in @xmath8 . for each @xmath139 with a finite @xmath5-stabilizer , we define @xmath140 where @xmath141 . it is not hard to check that @xmath142 is a well - defined member of @xmath8 . moreover , if @xmath0 is dominant , then @xmath143 if and only if @xmath0 is strongly dominant . in that case , @xmath144 has constant term 1 , and @xmath142 is invertible in @xmath8 . since the scalar product is nondegenerate , we may select @xmath145 so that @xmath146 for all @xmath104 . this given , for each @xmath139 we define @xmath147 it is easy to show that @xmath148 , and hence @xmath55 , do not depend on the choice of @xmath149 . by the kac - weyl character formula @xcite , @xmath55 is the character of the irreducible integrable module of highest weight @xmath0 for the corresponding kac - moody algebra . this section closely follows ( * ? ? ? * section 2 ) . we refer to this paper for more details . a _ crystal _ is a 4-tuple @xmath150 satisfying axioms ( a1)-(a3 ) below , where * @xmath151 is a set whose elements are called objects ; * @xmath152 and @xmath153 are maps @xmath154 ; * @xmath155 are bijections between two subsets of @xmath151 . a crystal is called an _ admissible system _ if it satisfies axioms ( a0 ) and ( a4 ) . a crystal is called _ semiperfect _ if it satisfies axioms ( a4 ) and ( a5 ) . for all real numbers @xmath156 , there are only finitely many objects @xmath157 such that @xmath158 . axiom ( a0 ) implies that the generating series @xmath159 is a well - defined member of @xmath8 . for each @xmath160 , we call @xmath161 , @xmath162 , and @xmath163 the _ weight _ , _ depth _ , and _ rise _ of @xmath157 . @xmath164 , @xmath165 . we define the depth and rise in the direction @xmath106 by @xmath166 and @xmath167 . in fact , we will develop the whole theory in terms of @xmath168 and @xmath169 rather than @xmath162 and @xmath170 . @xmath155 is a bijection from @xmath171 to @xmath172 . we let @xmath173 denote the inverse map . @xmath174 , @xmath175 . hence , we also have @xmath176 . the maps @xmath177 and @xmath155 act as raising and lowering operators that provide a partition of the objects into @xmath106-strings that are closed under the action of @xmath177 and @xmath155 . for example , the @xmath106-_string through _ @xmath157 is ( by definition ) @xmath178 where @xmath179 and @xmath180 . we define partial orders on @xmath151 , one for each @xmath181 , by @xmath182 any assignment of elements @xmath183 of a totally ordered set to pairs @xmath184 with @xmath185 is called a _ timing pattern _ for @xmath151 . a timing pattern is called _ coherent _ if the following two conditions are satisfied for all pairs @xmath184 such that @xmath185 and @xmath186 : * @xmath187 ; * for all @xmath188 , all integers @xmath189 , and all @xmath190 , there is an object @xmath191 such that @xmath192 and @xmath193 if and only if there is an object @xmath194 such that @xmath195 and @xmath196 . the following axiom ensures the existence of a certain sign - reversing involution used to cancel the negative terms in the kac - weyl character formula . there exists a coherent timing pattern for @xmath151 . note that , compared to @xcite , here we let the timing pattern take values in any totally ordered set , and we reverse the total order previously considered . however , these minor changes , dictated by our needs , are easily taken care of by minor changes in the corresponding proofs in @xcite . we say that @xmath151 has a maximum object if it does so with respect to the partial order generated by all partial orders @xmath197 , for @xmath104 . @xmath151 has a maximum object . semiperfect crystals are admissible systems . it suffices to show that axiom ( a0 ) is a consequence of the other axioms . let @xmath198 be the maximum object of @xmath151 and let @xmath199 . any other object of @xmath151 is obtained from @xmath198 by applying lowering operators @xmath155 . each of these operators decreases weight by @xmath106 , and thus they decrease height by 1 . it follows that we may have at most @xmath200 elements of height @xmath201 , where @xmath72 is the number of simple roots . @xcite[thmst ] if @xmath151 is an admissible system , then @xmath202 in particular , if @xmath151 is a semiperfect crystal with maximal object @xmath203 , then @xmath204 . given @xmath205 , let @xmath206 denote the root subsystem of @xmath74 with simple roots @xmath207 . following @xcite , we let @xmath208 , @xmath209 , and @xmath210 denote the corresponding weyl group , weight lattice , and character ring . provided that we use the height function inherited from @xmath74 ( in which case @xmath211 ) , it is easy to see that any admissible system @xmath151 can also be viewed as an admissible system relative to @xmath206 using only the operators @xmath177 and @xmath155 for @xmath212 . given @xmath213 , we let @xmath214 denote the weyl character ( relative to @xmath206 ) corresponding to @xmath0 . the following branching rule is given in @xcite . @xcite[branching ] if @xmath151 is an admissible system and @xmath205 , we have @xmath215 finally , note that one can define on @xmath151 the structure of a directed colored graph by constructing arrows @xmath216 colored @xmath181 for each @xmath217 . also note that the most important source of crystals are integrable representations of quantum groups @xmath218 corresponding to kac - moody algebras @xmath219 . it is well - known that such representations can be encoded by combinatorial data @xmath150 satisfying axioms ( a1)-(a3 ) , i.e. , by crystals ( cf . @xcite , see also e.g. @xcite ) . these crystals are usually called _ perfect crystals _ , while the corresponding directed colored graphs are known as _ crystal graphs_. fix a dominant weight @xmath0 . throughout this paper , we will use the term `` sequence '' for any map @xmath220 from a totally ordered set @xmath22 to some other set . we will use the notation @xmath221 , and , if @xmath22 is finite or countable , the usual notation @xmath222 , where @xmath223 . given an element @xmath224 and an index @xmath225 , we also define the following counting functions : @xmath226 assuming that the corresponding cardinalities are finite . if the elements of a sequence are positive roots and @xmath27 is such a root , then we set @xmath25 . [ lambdachain ] a @xmath0-_chain ( of roots ) _ is a sequence of positive roots @xmath19 indexed by the elements of a totally ordered set @xmath22 , which satisfies the conditions below . 1 . the number of occurrences of any positive root @xmath27 , i.e. , @xmath227 , is @xmath28 . 2 . given any triple of roots @xmath29 such that @xmath30 and @xmath31 for some integer @xmath32 , as well as @xmath23 such that @xmath33 , we have @xmath34 note that finding a @xmath0-chain amounts to defining a total order on the set @xmath228 such that condition ( 2 ) above holds , where @xmath229 for any @xmath230 in @xmath22 . one particular example of such an order can be constructed as follows . fix a total order on the set of simple roots @xmath231 . for each @xmath230 in @xmath22 , let @xmath232 , and define the vector @xmath233 in @xmath234 . the map @xmath235 is injective . indeed , assume that @xmath236 for @xmath230 and @xmath237 . if @xmath238 , the root system @xmath239 would contain two proportional positive coroots @xmath240 , which is not possible . also , the fact that @xmath241 implies that @xmath242 . hence , we can define a total order on @xmath22 by @xmath243 iff @xmath244 in the lexicographic order on @xmath234 . we are now ready to prove the existence of @xmath0-chains ( cf . lemma [ exist - lc ] ) . [ constr - lambdachain ] given the total order on @xmath22 defined above , the sequence @xmath19 defined by @xmath229 for @xmath230 is a @xmath0-chain . throughout this proof , we work with coroots only and modify the above notions accordingly . let us fix a triple of positive coroots @xmath29 such that @xmath245 for some fixed integer @xmath246 . let @xmath247 and @xmath248 we define a map @xmath235 for @xmath249 to @xmath250 as above . this map is not injective , because we have @xmath251 for @xmath252 . however , we can still define a total order @xmath249 if , in addition to setting @xmath243 whenever @xmath244 in the lexicographic order , we impose @xmath253 for @xmath252 . hence , we can define the sequence @xmath254 by @xmath255 for @xmath256 . the remainder of the proof consists in showing that the finite sequence @xmath257 , where @xmath258 has the induced total order , is a concatenation of pairs @xmath259 and @xmath260 ( in any order ) . hence , given @xmath261 such that @xmath262 , we have @xmath263 . on the other hand , it is easy to see that @xmath264 . thus , condition ( 2 ) in the definition of a @xmath0-chain is verified in the case when @xmath27 and @xmath37 are both positive or both negative roots . the remaining case is checked in a similar way . for each @xmath181 in @xmath88 , let us denote by @xmath265 , @xmath266 , @xmath267 the coefficients of @xmath268 in @xmath27 , @xmath269 , @xmath37 , respectively ; clearly , @xmath270 . also , let @xmath271 , @xmath272 , and @xmath273 , where @xmath274 . the interlacing condition to be verified is straightforward if @xmath275 or @xmath276 . assume @xmath277 . in this case , the mentioned condition is checked based on the following three claims about the finite sequence @xmath257 : 1 . if an entry @xmath27 is followed by @xmath269 , or vice versa , there is an entry @xmath37 in - between ; 2 . between two entries @xmath27 , or two entries @xmath269 , there is an entry @xmath37 ; 3 . the sequence can not start with @xmath37 , but must end with @xmath37 . for the first claim , let @xmath278 , which means @xmath279 . if the inequality is strict , then the rational number @xmath280 is strictly in - between the previous two ones ; therefore , we have @xmath281 , which proves the claim . otherwise , all three numbers are equal . we can now repeat our reasoning above with @xmath282 and @xmath283 replaced by @xmath265 and @xmath266 for @xmath284 . if equality still holds , we let @xmath285 etc . at some point , we must have strict inequalities ; indeed , otherwise @xmath27 and @xmath269 would be proportional , which is impossible . for the second claim , we can assume that we have two consecutive entries @xmath27 with corresponding indices @xmath286 and @xmath287 . furthermore , we can assume that we have an entry @xmath269 with corresponding index @xmath288 such that @xmath289 it is straightforward to check that we have @xmath290 therefore , we have @xmath291 , which proves the claim . for the last claim , note that the first three entries in the sequence @xmath257 are @xmath292 in some order , and that the corresponding vectors in @xmath234 have their first components equal to @xmath293 . the fact that the sequence can not start with @xmath37 follows by an argument similar to the proof of the first claim above . on the other hand , the sequence must end with @xmath37 because the largest value of the first component in the vectors involved is @xmath294 , and this value only appears in @xmath295 for @xmath296 . the following immediate consequence of condition ( 2 ) in the definition of a @xmath0-chain will be used several times below . [ countroots ] for any two positive roots @xmath30 and @xmath23 such that @xmath33 , we have @xmath297 we conclude this section by showing that , in the case of a finite root system , we can simplify the definition of a @xmath0-chain . [ equivalence ] in the case of a finite root system , condition ( 2 ) in definition [ lambdachain ] is equivalent to the following interlacing condition . 1 . for each triple of positive roots @xmath298 with @xmath35 , the finite sequence @xmath299 , where @xmath300 has the induced total order , is a concatenation of pairs @xmath259 and @xmath301 ( in any order ) it is easy to see that the interlacing condition ( 2@xmath302 ) is equivalent to the following one : given any @xmath23 and any triple of positive roots @xmath298 with @xmath35 , we have @xmath303 clearly , condition ( 2 ) in definition [ lambdachain ] implies ( [ basiccount ] ) ; indeed , one just has to set @xmath304 in the former . assume that @xmath27 , @xmath38 , and @xmath37 are all positive roots and @xmath305 . given that ( [ basiccount ] ) holds , we check condition ( 2 ) in this case using induction on @xmath32 , which starts at @xmath306 . given @xmath246 , we know that @xmath307 , for @xmath308 , are coroots , as elements of the @xmath309-string through @xmath310 . the induction step consists in writing @xmath311 , and in combining ( [ basiccount ] ) with the induction hypothesis . now assume that @xmath38 and @xmath37 are positive roots , but @xmath27 is a negative root . we check condition ( 2 ) in this case in a similar way , using induction which starts at @xmath304 . given an arbitrary @xmath312 , we know that @xmath313 , for @xmath308 , are coroots , as elements of the @xmath309-string through @xmath314 . we now write @xmath315 the induction step is completed by combining ( [ basiccount ] ) with the induction hypothesis ( in ( [ eq1 ] ) ) and condition ( 2 ) in the case already verified ( in ( [ eq2 ] ) ) . for the rest of our construction ( sections [ sec : chains - foldings][sec : mainthm ] ) , let us fix a dominant integral weight @xmath0 and fix an arbitrary @xmath0-chain @xmath19 . we will use the notation @xmath316 for the reflection @xmath317 . we start by associating to our fixed @xmath0-chain the closely related object @xmath318 where @xmath149 is a fixed dominant weight satisfying @xmath146 for all @xmath104 . here , as well as throughout this article , we let @xmath319 be greater than all elements in @xmath22 . we use operators called _ folding operators _ to construct from @xmath320 new objects of the form @xmath321 here @xmath322 is a root , @xmath323 , any given root appears only finitely many times in @xmath324 , and @xmath325 is in the @xmath5-orbit of @xmath149 . more precisely , given @xmath324 as above and @xmath225 in @xmath22 , we let @xmath326 and we define @xmath327 where @xmath328 let us now consider the set of all @xmath324 that are obtained from @xmath320 by applying folding operators ; we call these objects the _ foldings _ of @xmath320 . clearly , @xmath329 is an involution on the set of foldings of @xmath320 . in order to describe this set , let us note that the folding operators commute . indeed , if @xmath324 is as in ( [ defgamma ] ) , @xmath243 , @xmath330 , and @xmath331 , then we have @xmath332 so @xmath333 . this means that every folding @xmath324 of @xmath320 is determined by the set @xmath334 . more precisely , if @xmath335 , then @xmath336 . we call the elements of @xmath57 the _ folding positions _ of @xmath324 , and write @xmath337 . throughout this paper , we will use @xmath57 and @xmath337 interchangeably . for instance , according to the above discussion , we have @xmath338 where @xmath339 denotes the symmetric difference of sets . hence , it makes sense to define the folding operator @xmath329 on @xmath57 ( compatibly with the action of @xmath329 on @xmath340 ) by @xmath341 . using the notation above and the fact that @xmath342 , we have @xmath343 recall that @xmath344 . in particular , we have @xmath345 for @xmath346 . [ finiteness ] although a folding @xmath324 of @xmath320 is an infinite sequence if the root system is infinite , we are , in fact , always working with finite objects . indeed , we are examining @xmath324 by considering only one root at a time . given a folding @xmath324 of @xmath320 , we associate to each pair in @xmath324 ( or the corresponding index @xmath225 in @xmath22 ) an integer @xmath347 , which we call _ level _ ; the sequence @xmath348 will be called the _ level sequence _ of @xmath324 . the definition is as follows : @xmath349 where @xmath350 we make the convention that the sum is 0 if it contains no terms . the definition makes sense since the sum is always finite . in particular , we have the level sequence @xmath351 of @xmath320 . we will now consider certain affine reflections corresponding to foldings @xmath324 of @xmath320 . let @xmath352 ; recall that the latter is the reflection in the affine hyperplane @xmath353 . in particular , we have the affine reflections @xmath354 corresponding to @xmath320 . [ defweighted ] given @xmath355 and @xmath337 , we let @xmath356 and call @xmath152 the _ weight _ of @xmath324 ( respectively @xmath57 ) . we also let @xmath357 ( recall that @xmath358 ) . given a root @xmath27 , we will use the following notation : @xmath359 where @xmath360 . the following proposition is our main technical result , which relies heavily on the defining properties of @xmath0-chains . [ hyperplane ] let @xmath337 for some @xmath355 , and let @xmath361 ( the first or the second inequality is dropped if @xmath362 or @xmath363 , respectively ) . using the notation above , we have @xmath364 it suffices to consider @xmath363 . let us define the roots @xmath365 and @xmath366 for @xmath367 ; note that @xmath368 . throughout this proof , we use the maps @xmath369 and @xmath370 defined at the beginning of section [ sec : lambdachains ] to count roots in @xmath320 . we need to show that @xmath371 where , according to the definitions of folding operators and of levels , the integer @xmath372 can be expressed as @xmath373 indeed , by examining the segment of @xmath320 between @xmath374 and @xmath375 for @xmath367 ( we let @xmath376 ) , we note that only the root @xmath377 gets changed to @xmath378 in @xmath324 ; more precisely , the sign of the corresponding root in @xmath324 is @xmath379 . the definition of levels now leads us to the above formula for @xmath32 . we now use induction on @xmath380 , which starts at @xmath381 . let us assume that the statement to prove holds for @xmath382 . based on the discussion above , this means that @xmath383 where @xmath384 an easy linear algebra computation shows that @xmath385 where @xmath386 let us now substitute into this formula the expression for @xmath387 given by the induction hypothesis , and compare the result with the expression for @xmath32 given above . let us first assume that @xmath388 . it turns out that verifying @xmath389 amounts to proving the following identity : @xmath390 let us now multiply both sides by @xmath391 , then substitute @xmath392 with @xmath393 , and finally write @xmath27 for @xmath394 , @xmath38 for @xmath395 , and @xmath225 for @xmath396 . the identity to prove becomes precisely the one in lemma [ countroots ] . in the special case @xmath397 , we need to correct the identity above by adding @xmath391 to its right - hand side . this is proved in a similar way , based on the correction of lemma [ countroots ] in the case @xmath398 , which amounts to adding @xmath115 to the right - hand side of the corresponding formula . [ affinerefl ] let @xmath337 for some @xmath355 . using the notation above , we have @xmath399 in particular , we have @xmath400 for @xmath346 . furthermore , if @xmath401 , then @xmath402 . the first part follows from proposition [ hyperplane ] by applying the following basic result : if @xmath403 for some affine reflections @xmath404 , then @xmath405 . the rest of the corollary follows easily from the first part . the next proposition shows that all inner products of @xmath406 with positive roots can be easily read off from the level sequence @xmath407 . recall that , according to ( [ levelseq ] ) , the latter is computed by applying a simple counting procedure to the sequence of pairs in @xmath324 . [ lastlevel ] given a positive root @xmath27 , let @xmath408 , assuming that @xmath409 . then we have @xmath410 on the other hand , if @xmath411 , then we have @xmath412 we prove this result by induction on @xmath380 , which starts at @xmath381 . if @xmath413 , we assume that the result holds for @xmath414 , where @xmath415 . let @xmath416 we have @xmath417 recall that @xmath418 , by ( [ reflection ] ) . we have the following cases . we have @xmath419 . based on ( [ weightcomp ] ) and induction , we have @xmath420 where @xmath421 is the correction term in the formula for @xmath422 . on the other hand , an easy linear algebra computation shows that @xmath423 maps the affine hyperplane @xmath424 to @xmath425 but then , by proposition [ hyperplane ] , we have @xmath426 . hence , all we have to prove is that @xmath427 , where @xmath153 is the correction term in the formula for @xmath428 . this follows from the fact that , if @xmath429 is the pair indexed by @xmath283 in @xmath430 , then we have @xmath431 it is clear that @xmath30 . we have two subcases , corresponding to @xmath432 nonempty and empty . in the first subcase , letting @xmath433 , we have @xmath434 , and , thus , @xmath435 ; here the counting function @xmath369 refers to our fixed @xmath0-chain . based on ( [ weightcomp ] ) , induction , and lemma [ countroots ] , we have @xmath436 where @xmath437 is as above . on the other hand , note that @xmath438 , so we have to prove @xmath439 hence , it remains to show that @xmath440 , where @xmath153 is as above . this follows from the fact that @xmath441 the subcase @xmath442 is now immediate , since ( [ mucount ] ) still holds , with @xmath437 defined as in ( [ defeps ] ) ; indeed , in this case the induction is based on the second formula in the proposition to be proved . this case is easily reduced to case 1.2 above , since the formula to be proved can still be written as in ( [ toprove ] ) , with @xmath153 defined as in ( [ defeps ] ) . [ graphrep ] ( 1 ) since @xmath443 , we have @xmath444 . hence , proposition [ lastlevel ] can be restated in terms of @xmath445 . \(2 ) it is often useful to use the following graphical representation . let @xmath446 , and let us define the continuous piecewise - linear function @xmath447\rightarrow{\mathbb r}$ ] by @xmath448 then , according to the defining relation ( [ levelseq ] ) and proposition [ lastlevel ] , we have @xmath449 for instance , assume that the entries of @xmath324 indexed by the elements of @xmath450 are @xmath451 , @xmath452 , @xmath453 , @xmath453 , @xmath451 , @xmath452 , @xmath451 , @xmath453 , in this order ; also assume that @xmath454 . the graph of @xmath455 is shown in figure [ fig : walls ] . we will now define some special foldings @xmath340 of @xmath320 . the notation is the same as in section [ sec : chains - foldings ] . [ admsubset ] an _ admissible subset _ is a finite subset of @xmath22 ( possibly empty ) , that is , @xmath335 , such that we have the following saturated chain in the bruhat order on @xmath5 : @xmath456 if @xmath57 is an admissible subset , we will call @xmath337 an _ admissible folding _ ( of @xmath320 ) . we denote by @xmath457 the collection of all admissible subsets corresponding to our fixed @xmath0-chain . [ lskey ] the weyl group element @xmath458 is a generalization of the _ left - key _ of semistandard young tableaux , which was defined by lascoux - schtzenberger @xcite , and which is relevant to _ demazure characters_. indeed , for finite types , a demazure character formula in terms of @xmath458 appears in @xcite . according to ( [ reflection ] ) , we have the following intrinsic criterion for @xmath340 to be an admissible folding . [ intrinsic ] given arbitrary @xmath355 , @xmath340 is an admissible folding if and only if @xmath459 in this section , we will prove some results about admissible foldings . let us fix an admissible subset @xmath335 , and let @xmath324 be the associated admissible folding ; also , let @xmath407 , @xmath460 , and @xmath461 . recall the notation @xmath462 . we start with a basic result involving the bruhat order . [ bruhat ] let @xmath463 be a covering relation in the bruhat order , where @xmath38 is some positive root , and @xmath27 is a simple root . assume that we have @xmath464 and @xmath465 . then @xmath466 . for a simple root @xmath27 , the following conditions are equivalent : ( 1 ) @xmath467 ; ( 2 ) @xmath468 ; ( 3 ) there exists a reduced decomposition for @xmath92 that ends with @xmath40 . let us pick such a reduced decomposition @xmath469 , so @xmath470 . all elements @xmath96 that are covered by @xmath92 in the bruhat order are obtained by skipping one term in the reduced decomposition . we know that @xmath464 , so the element @xmath96 can not have a reduced decomposition that ends with @xmath471 . thus @xmath96 is obtained from @xmath92 by skipping the last term @xmath472 in the reduced decomposition . so @xmath473 , where @xmath474 . [ existroot ] assume that @xmath475 and @xmath476 for some simple root @xmath27 and @xmath477 ( if @xmath275 , then the first condition is void ) . then there exists @xmath225 with @xmath478 such that @xmath479 . find @xmath225 with @xmath478 such that @xmath480 and @xmath481 . by lemma [ bruhat ] , we have @xmath482 . this means that @xmath483 . [ folding ] if @xmath484 , then @xmath485 . we can assume that @xmath486 . since @xmath487 , where @xmath488 , we need to prove that the right - hand side is a positive root . this follows from the fact that @xmath489 , according to a well - known property of the bruhat order on a coxeter group , cf . * proposition 5.7 ) . [ signchange ] assume that @xmath27 is a simple root for which @xmath490 . let @xmath491 be either the minimum of @xmath450 , or an element for which its predecessor @xmath32 ( in @xmath450 ) satisfies @xmath492 . then we have @xmath493 . assume that @xmath494 . recall that the corresponding admissible subset is @xmath335 . note that @xmath495 , based on proposition [ folding ] ( indeed , we must have @xmath496 ) . let us define the index @xmath224 by the condition @xmath497 ( possibly , @xmath498 , in which case the second inequality is dropped ) . we also define the index @xmath499 by setting @xmath500 if @xmath501 , and by the condition @xmath502 , otherwise ( if @xmath275 in the second case , the corresponding first inequality is dropped ) . we clearly have @xmath503 , which implies @xmath476 . if @xmath504 , we also have @xmath505 , so @xmath475 . but then lemma [ existroot ] applies and leads to a contradiction . [ lastneg ] assume that , for some simple root @xmath27 , we have either @xmath506 , or @xmath507 for @xmath508 . then we have @xmath509 . assume that the conclusion fails , which means that @xmath510 ( cf . remark [ graphrep ] ( 2 ) ) . define the index @xmath499 by setting @xmath500 if @xmath506 , and by the condition @xmath511 , otherwise ( if @xmath275 or @xmath512 in the second case , one of the two inequalities is dropped ) . if @xmath504 , we have @xmath513 , so @xmath475 . it means that the hypotheses of lemma [ existroot ] are satisfied for @xmath514 . this lemma now leads to a contradiction . [ maximum ] assume that @xmath413 and let @xmath515 . we have @xmath516 in other words , @xmath517 is a positive multiple of a positive root . in particular , if @xmath518 then @xmath519 belongs to the positive cone in the root lattice . using the definitions and notation in section [ sec : chains - foldings ] , we have @xmath520 the fact that @xmath521 is a positive root is the content of proposition [ folding ] . [ extremalelement ] given @xmath522 and @xmath523 , the following hold . 1 . if @xmath381 , then @xmath524 for all simple roots @xmath27 with @xmath525 . 2 . if @xmath413 , then there is a simple root @xmath27 in @xmath324 such that @xmath525 and @xmath526 . the first part follows directly from the definitions . if @xmath413 , we can find a simple root @xmath27 such that @xmath510 . hence , proposition [ lastneg ] applies , and , letting @xmath527 , we have @xmath528 . but then @xmath529 , by proposition [ lastlevel ] , so the second statement is verified . let us now fix a simple root @xmath27 , and recall the description of the sets @xmath530 and @xmath531 based on the continuous piecewise - linear function @xmath455 , which was introduced in remark [ graphrep ] ( 2 ) . we will rephrase some of the above results in a simple way in terms of @xmath455 , and we will deduce some consequences ( to be used in the subsequent sections ) , which are formulated in the same language . assume that @xmath532 , so that @xmath455 is defined on @xmath533 $ ] , and let @xmath534 be the maximum of @xmath455 . note first that the function @xmath455 is determined by the sequence @xmath535 , where @xmath536 for @xmath537 , and @xmath538 . we have the following restrictions . 1 . @xmath539 for @xmath537 ( by proposition [ folding ] ) . 2 . @xmath540 or @xmath541 implies @xmath542 ( by propositions [ signchange ] and [ lastneg ] ) . [ cor1 ] we have @xmath543 . if @xmath544 , then @xmath545 for @xmath546 , and @xmath547 . the fact that @xmath543 follows from condition ( c2 ) for @xmath540 . thus , @xmath548 . if @xmath549 for @xmath550 , then @xmath551 by condition ( c1 ) . but then condition ( c2 ) leads to a contradiction . the last statement is obvious . [ cor2 ] assume that @xmath552 , and let @xmath283 be such that @xmath553 . we have @xmath554 , @xmath555 , and @xmath556 . moreover , we have @xmath557 for @xmath558 . if @xmath559 , then @xmath560 , so @xmath561 . it is also easy to see that both @xmath562 and @xmath563 contradict the definition of @xmath283 . now assume that the last statement fails . then there exists @xmath564 such that @xmath565 and @xmath566 . condition ( c2 ) implies @xmath567 and @xmath568 , which is a contradiction to the definition of @xmath283 . the next corollary can be proved in a similar way to corollary [ cor2 ] . [ cor3 ] assume that @xmath569 , and let @xmath32 be such that @xmath570 . we have @xmath571 , @xmath572 , and @xmath573 . moreover , we have @xmath557 for @xmath574 . we will now define _ root operators _ on the collection @xmath457 of admissible subsets corresponding to our fixed @xmath0-chain . let @xmath57 be such an admissible subset , let @xmath324 be the associated admissible folding , and @xmath407 its level sequence , denoted as in section [ sec : chains - foldings ] . also recall from section [ sec : chains - foldings ] the definitions of the finite sequences @xmath575 , @xmath530 , @xmath576 , and @xmath531 , where @xmath27 is a root . we will first define a partial operator @xmath155 on admissible subsets @xmath57 for each @xmath181 in @xmath88 , that is , for each simple root @xmath106 . let @xmath181 in @xmath88 be fixed throughout this section . let @xmath577 be the maximum of the finite set of integers @xmath578 . we know that @xmath543 from corollary [ cor1 ] . assume that @xmath552 . let @xmath579 be the minimum index @xmath225 in @xmath580 for which we have @xmath581 . if no such index exists , then @xmath582 ; in this case , we let @xmath583 . now let @xmath584 be the predecessor of @xmath283 in @xmath585 . by corollary [ cor2 ] , this always exists and we have @xmath586 . let us now define @xmath587 where @xmath588 is the identity map . note that the folding of @xmath320 associated to @xmath589 , which will be denoted by @xmath590 , is defined by a similar formula . more precisely , we have @xmath591 and @xmath592 in other words , based on proposition [ propf ] below , we can say that applying the root operator @xmath155 amounts to performing a `` folding '' in position @xmath32 , and , if @xmath593 , an `` unfolding '' in position @xmath283 . [ propf ] given the above setup , the following hold . 1 . if @xmath593 , then @xmath594 and @xmath595 . we have @xmath596 and @xmath597 . we have @xmath598 let @xmath599 . the first two statements follow immediately from corollaries [ cor1 ] and [ cor2 ] . for the third statement , note that , by corollary [ affinerefl ] , the weight of @xmath589 is @xmath600 if @xmath593 , and @xmath601 otherwise . using the formula @xmath602 , and the similar one for @xmath603 , we compute ( in both cases ) @xmath604 we now intend to define a partial inverse @xmath177 to @xmath155 . assume that @xmath605 . let @xmath606 be the maximum index @xmath225 in @xmath580 for which we have @xmath581 . note that such indices always exist , by corollary [ cor3 ] . now let @xmath607 be the successor of @xmath32 in @xmath585 . corollary [ cor3 ] implies that , if @xmath608 , then we have @xmath609 , while , otherwise , we have @xmath610 . finally , we define @xmath611 by the same formula as @xmath589 , namely ( [ defroot ] ) . hence , the folding of @xmath320 associated to @xmath611 is also defined in the same way as above . the following analog of proposition [ propf ] is proved in a similar way , by invoking corollaries [ cor1 ] and [ cor3 ] . [ prope ] given the above setup , the following hold . 1 . we have @xmath596 and @xmath612 . 2 . if @xmath593 , then @xmath613 and @xmath496 . we have @xmath614 [ closed ] if @xmath589 is defined , then it is also an admissible subset . similarly for @xmath611 . moreover , if @xmath589 and @xmath611 are both defined , then @xmath615 ; otherwise , we have @xmath616 . we consider @xmath155 first . the cases corresponding to @xmath593 and @xmath608 can be proved in similar ways , so we only consider the first case . let @xmath335 , as usual , and , based on proposition [ propf ] ( 1)-(2 ) , let @xmath617 be such that @xmath618 if @xmath275 or @xmath619 , then the corresponding indices @xmath620 , respectively @xmath621 , are missing . if @xmath622 , there is nothing to prove , so we assume @xmath623 . we use the criterion in corollary [ intrinsic ] . based on it , we have @xmath624 let @xmath625 ; if @xmath275 , then set @xmath626 . we need to prove that @xmath627 where @xmath628 for @xmath629 . indeed , the pair with index @xmath225 in @xmath630 is @xmath631 , for @xmath632 . note that ( [ chain2 ] ) also implies that @xmath615 . recall that @xmath633 , by ( [ reflection ] ) . on the other hand , based on proposition [ propf ] ( 2 ) , we have @xmath634 , which implies @xmath635 . hence , we have @xmath636 , which gives us the first covering relation in ( [ chain2 ] ) . but we also have @xmath637 , by ( [ chain1 ] ) , and @xmath638 , by the choice of @xmath32 . based on a well - known property of the bruhat order on a coxeter group ( * ? ? ? * theorem 1.1 ( iv ) ( iii ) ) , we deduce that @xmath639 and @xmath640 . the latter statement gives us the second covering relation in ( [ chain2 ] ) . we can proceed in this way until we get the whole chain ( [ chain2 ] ) . the proof of the result for @xmath611 is completely similar , based on proposition [ prope ] . [ bijection ] ( 1 ) assume that @xmath589 is defined . then we have @xmath641 hence , @xmath642 is defined and equal to @xmath57 . \(2 ) assume that @xmath611 is defined . then we have @xmath643 hence @xmath644 is defined and equal to @xmath57 . \(3 ) let @xmath499 be maximal such that @xmath645 is defined , and let @xmath224 be maximal such that @xmath646 is defined . then @xmath647 . the proof is straightforward based on the results already proved in this section , and is left to the interested reader . we note that that it is convenient to use the graphical representation described in remark [ graphrep ] ( 2 ) . in this section , we derive our main result . we start with the following definitions and lemma . given an admissible folding @xmath337 , denoted as in section [ sec : chains - foldings ] , we denote by @xmath648 the sequence indexed by @xmath649 given by @xmath650 and @xmath651 for @xmath652 . let us define @xmath653 where @xmath654 was defined at the beginning of section [ sec : rootoperators ] . [ lemtiming ] let @xmath655 and @xmath324 be two admissible foldings , with @xmath656 and @xmath657 . assume that @xmath658 , that @xmath659 is defined , and that @xmath660 for @xmath661 . then @xmath662 . moreover , equality holds if and only if @xmath663 . let @xmath664 and @xmath665 be the folding positions of @xmath655 and @xmath324 , respectively , where @xmath666 note that the pair in @xmath655 indexed by @xmath225 is @xmath667 , by proposition [ prope ] ( 1 ) . hence , we have @xmath668 , so @xmath669 . by proposition [ lastlevel ] , we have @xmath670 recall that @xmath457 is the collection of all admissible subsets corresponding to our fixed @xmath0-chain . the collection @xmath457 of admissible subsets together with the root operators form a semiperfect crystal . thus we have the following character formula : @xmath671 as usual , we denote by @xmath57 a generic element of @xmath672 , by @xmath152 its weight , and by @xmath337 the corresponding admissible folding . also , @xmath106 will be a generic simple root . axiom ( a1 ) follows from corollary [ cor1 ] . axiom ( a2 ) is the content of proposition [ closed ] and proposition [ bijection ] ( 1)-(2 ) . axiom ( a3 ) is the content of proposition [ propf ] ( 3 ) and proposition [ bijection ] ( 1 ) . according to proposition [ extremalelement ] , @xmath673 is the unique admissible subset which is maximal with respect to all partial orders @xmath197 , for @xmath104 . since the height of @xmath674 is strictly larger than the height of @xmath52 for any other admissible subset @xmath57 ( cf . proposition [ maximum ] ) , and since the height of @xmath675 is larger by 1 than the height of @xmath52 , we conclude that @xmath673 is the maximum object of @xmath457 . hence , axiom ( a5 ) is satisfied . the definition of a coherent timing pattern and the related verification of axiom ( a4 ) are analogous to those for ls chains in ( * ? ? ? * theorem 8.3 ) ; nevertheless , there are some features specifically related to our setup , such as the reversal of the total order on the set in which the timing pattern takes values , and the use of proposition [ lastlevel ] in lemma [ lemtiming ] ( used below ) . assume that @xmath676 , so that @xmath611 is defined . we define @xmath677 , where @xmath678 was defined in section [ sec : rootoperators ] in connection with the root operator @xmath177 . assuming that @xmath589 is defined , and applying proposition [ bijection ] ( 1 ) , we have @xmath679 by iteration , it follows that @xmath680 for @xmath681 . therefore , given @xmath682 for @xmath188 with @xmath683 and @xmath684 , we have @xmath685 we have @xmath686 , by proposition [ closed ] . hence , by lemma [ lemtiming ] , we have @xmath687 , so there exists @xmath688 such that @xmath689 . we claim that @xmath690 . let @xmath691 and @xmath692 . we have @xmath693 also note that @xmath694 , by proposition [ closed ] . assume that @xmath695 . then the fact that @xmath696 implies , based on lemma [ lemtiming ] , that the maximum of @xmath697 is attained at @xmath698 as well ; but this contradicts the definition of @xmath699 . the case @xmath700 is similar . on the other hand , similar reasoning proves conversely that given @xmath688 such that @xmath701 and @xmath702 , there is @xmath703 such that @xmath704 and @xmath705 . the formula for @xmath55 now follows from theorem [ thmst ] . ( littlewood - richardson rule ) . we have @xmath706 where the summation is over all @xmath57 in @xmath457 satisfying @xmath707 for all @xmath104 . this follows immediately from theorem [ thmst ] . there , the condition for @xmath57 to contribute to the summation was @xmath708 , which is easily seen to be equivalent to the condition stated above . ( branching rule ) . given @xmath205 , we have the following rule for decomposing @xmath55 as a sum of weyl characters relative to @xmath206 : @xmath709 where the summation is over all @xmath57 in @xmath457 satisfying @xmath710 for all @xmath212 . immediate , based on corollary [ branching ] . in this section , we explain the connection between our model and ls chains . we start with the relevant definitions , by closely following ( * ? ? ? * section 8) . the bruhat order on the orbit @xmath711 of a dominant or antidominant weight is the transitive closure of the relations @xmath712 the bruhat orders on @xmath711 and @xmath713 are dual isomorphic ; in fact , @xmath714 if and only if @xmath715 . as usual , we write @xmath716 to indicate that @xmath152 covers @xmath717 ; this happens only if @xmath718 for some @xmath16 , but not conversely . given @xmath719 and a fixed real number @xmath224 , one defines the @xmath224-_bruhat order _ @xmath720 as the transitive closure of the relations @xmath721 thus , @xmath152 covers @xmath717 in @xmath224-bruhat order if and only if @xmath152 covers @xmath717 in the usual bruhat order and @xmath722 is an integer multiple of a root . given @xmath719 , we say that a pair consisting of a chain @xmath723 in the @xmath5-orbit of @xmath0 and an increasing sequence of rational numbers @xmath724 is a _ lakshmibai - seshadri chain _ ( ls chain ) if @xmath725 following @xcite , we identify an ls chain ( denoted as above ) with the map @xmath726\rightarrow w\lambda$ ] given by @xmath727 for @xmath728 , where @xmath729 and @xmath730 , @xmath731 . note that the piecewise - constant left - continuous maps that correspond to ls chains can be characterized by the property @xmath732 where @xmath733 denotes the right - hand limit of @xmath37 at @xmath699 . to each ls chain @xmath37 , we associate the continuous piecewise - linear path @xmath734\rightarrow\hr$ ] given by @xmath735 in other words , we define @xmath736 for @xmath737 . the root operators @xmath155 and @xmath177 on ls chains ( for a simple root @xmath106 ) were defined by littelmann @xcite as follows . let @xmath738 be the minimum of the function @xmath739\rightarrow { \mathbb r}$ ] given by @xmath740 . if @xmath741 , then @xmath177 is undefined . otherwise , we let @xmath742 $ ] be minimal such that @xmath743 , and we let @xmath744 $ ] be maximal such that @xmath745 for @xmath746 $ ] . we define @xmath747 the definition of @xmath155 is similar . more precisely , if @xmath748 , then @xmath155 is undefined . otherwise , we let @xmath749 $ ] be maximal such that @xmath750 , and we let @xmath751 $ ] be minimal such that @xmath745 for @xmath752 $ ] . given the latter values for @xmath753 and @xmath754 , we define @xmath755 by the same formula as @xmath756 . [ defrootops ] the above definition of @xmath177 and @xmath155 applies to any continuous piecewise - linear path @xmath757 with @xmath758 , if we replace the map @xmath37 above with the left - hand derivative of @xmath757 ( which is a piecewise - constant left - continuous map defined on @xmath759 $ ] ) . with each ls chain @xmath37 ( denoted as above ) , one can associate the _ dual _ ls chain @xmath760 , which is defined as @xmath761 where @xmath762 . it is easy to see that @xmath763 and @xmath764 . the paths corresponding to ls chains are special cases of _ littelmann paths_. the latter were defined by littelmann @xcite , still in the setup of complex symmetrizable kac - moody algebras . more precisely , littelmann defined root operators @xmath177 and @xmath155 on continuous paths @xmath734\rightarrow \hr$ ] with @xmath758 . in fact , the operator @xmath177 ( respectively @xmath155 ) is defined as in remark [ defrootops ] if the function @xmath739\rightarrow { \mathbb r}$ ] given by @xmath740 is weakly decreasing ( respectively weakly increasing ) between @xmath753 and @xmath754 . in general , the definition is more involved ; for convenience , we stated it in section [ sec : finite ] . however , only the simpler version of the definition is needed for an ls chain , since it is known that the corresponding function @xmath765 satisfies the condition stated above . littelmann considered the collection @xmath766 of all paths obtained by applying the operators @xmath155 to a fixed continuous path from @xmath293 to @xmath0 which lies inside the dominant weyl chamber . he showed that these paths form a crystal , that the associated colored directed graph does not depend on the initial path , and that one can express @xmath767 moreover , there is a corresponding littlewood - richardson rule @xcite . stembridge @xcite reproved the special case of the above results corresponding to ls chains by showing that they form an admissible system . kashiwara @xcite , lakshmibai @xcite , and joseph @xcite proved independently that littelmann paths ( obtained from a fixed path via root operators ) have the structure of a perfect crystal . let us now return to ls chains , and fix @xmath0 in @xmath121 . recall the set @xmath22 in ( [ indset ] ) , and the @xmath0-chain @xmath19 given by proposition [ constr - lambdachain ] , which depends on a total order on the set of simple roots @xmath768 . we will now describe a bijection between the corresponding admissible subsets ( cf . definition [ admsubset ] ) and the ls chains corresponding to the antidominant weight @xmath769 . given an index @xmath230 , we let @xmath770 and @xmath771 . we have an order - preserving map from @xmath22 to @xmath772 given by @xmath773 . recall the notation @xmath358 and @xmath774 ; we also let @xmath775 . consider an admissible subset @xmath335 and let @xmath776 let @xmath777 be such that @xmath778 if and only if @xmath779 , for @xmath729 . define weyl group elements @xmath780 for @xmath781 and @xmath782 for @xmath729 by @xmath783 , @xmath784 , and @xmath785 . let also @xmath786 . for any @xmath787 , we have the following saturated chain in bruhat order of minimum ( left ) coset representatives modulo @xmath11 : @xmath788 indeed , none of the reflections @xmath789 lies in @xmath11 , since @xmath790 for all @xmath23 . the above chain gives rise to a saturated increasing chain from @xmath791 to @xmath792 in the bruhat order on @xmath713 . we will now show that this chain is , in fact , a chain in @xmath793-bruhat order . let @xmath794 , so that @xmath795 , for @xmath796 . we need to check that @xmath797 , for @xmath779 . but @xmath798 while , by definition , @xmath799 is a fraction with denominator @xmath800 . hence @xmath801 is an ls chain in the @xmath5-orbit of @xmath769 . we denote it by @xmath802 , and the associated continuous piecewise - linear path by @xmath803 . note that @xmath804 is the ls chain consisting only of @xmath769 , while @xmath805\rightarrow\hr$ ] is the path @xmath806 . the path @xmath807 intersects the affine hyperplane @xmath808 at @xmath809 for @xmath23 ; moreover , these and @xmath810 are the only intersections of @xmath807 with affine hyperplanes @xmath811 , for @xmath83 and @xmath812 . it is not hard to see that the path @xmath803 can be described using folding operators as follows : @xmath813 here , the folding operators @xmath329 are defined as follows on the relevant paths @xmath757 : @xmath814 [ weightls ] based on ( [ describefold ] ) , we can easily show that @xmath815 ( cf . definition [ defweighted ] ) . indeed , we have @xmath816 . [ rootopscorr ] we have @xmath817 for all admissible subsets @xmath57 ( here @xmath177 is the root operator for paths , while @xmath155 in the one defined in section [ sec : rootoperators ] ) . consider the point @xmath818 , where @xmath819 is a small positive real number . let @xmath820\to \h_\r^*$ ] be the path @xmath821 . given @xmath230 in @xmath22 and a sufficiently small @xmath819 , the path @xmath822 crosses the affine hyperplane @xmath823 at @xmath824 . fix a simple root @xmath106 and an admissible subset @xmath335 . let @xmath337 be the corresponding admissible folding , and @xmath407 the corresponding level sequence . we can find @xmath825 such that , for all @xmath243 in @xmath826 and @xmath827 , the points @xmath828 exist , and we have @xmath829 . let us now extend this path by the segments from 0 to @xmath830 and from @xmath831 to @xmath769 . more precisely , we consider the path @xmath832\to \h_\r^*$ ] given by @xmath833 since @xmath825 , the point @xmath834 does not lie on an affine hyperplane @xmath835 , for @xmath836 , whenever @xmath837\cup[1-\varepsilon,1)$ ] and @xmath827 . from now on , we assume that @xmath827 . let @xmath838 be the value of @xmath699 in @xmath839 $ ] for which @xmath840 crosses the affine hyperplane @xmath823 for @xmath230 in @xmath22 ( assuming that such a value exists ) . clearly , it is still true that , for all @xmath243 in @xmath826 , the points @xmath841 exist , and we have @xmath842 . now consider the path @xmath843 where the folding operators @xmath329 are defined as in ( [ simpledeffold ] ) , except that @xmath844 is replaced by @xmath838 . it is easy to see that the only intersections of @xmath845 with affine hyperplanes @xmath835 , for @xmath836 , occur at @xmath846 , @xmath810 , and @xmath847 for @xmath848 . moreover , note that the signs of the roots in the pairs @xmath849 for @xmath850 in @xmath324 with @xmath851 , as well as the sign of @xmath852 indicate the sides of the mentioned hyperplane on which the path @xmath840 lies before and/or after the intersection . corollary [ cor1 ] shows that the minimum of the function @xmath853 is @xmath854 ; furthermore , this minimum is attained at @xmath847 for @xmath848 with @xmath855 , as well as at @xmath810 if @xmath856 . let @xmath857 , @xmath858 , and assume @xmath593 . we define @xmath617 by @xmath502 and @xmath859 ( possibly @xmath275 , in which case the corresponding inequality is dropped ) . let us define @xmath860 as in remark [ defrootops ] . by corollary [ cor2 ] , the corresponding points @xmath753 and @xmath754 are as follows : @xmath861 consider @xmath862 , and let @xmath863 be its successor in @xmath864 . given a subset @xmath865 of @xmath22 , we will use the notation @xmath866 for @xmath867 ; we also set @xmath868 . let @xmath869 and @xmath870 . by ( [ defeppaths ] ) , the direction of @xmath860 for @xmath871 $ ] is @xmath872 . since @xmath873 , the direction of @xmath874 for the same values of @xmath699 is @xmath875 . but @xmath876 , so @xmath877 , and , therefore , the two directions above coincide . the directions of the two paths also coincide for @xmath878 , since @xmath615 ( cf . proposition [ closed ] ) . the case @xmath608 is similar . all this shows that @xmath879 now let us take the limit as @xmath880 . note that @xmath845 converges uniformly to @xmath803 , since @xmath881 converges to @xmath882 , for @xmath346 ; so the minimum of @xmath803 is also @xmath854 , and the points @xmath883 in the construction of @xmath884 are precisely @xmath885 ( which are the limits of @xmath886 ) . this implies that @xmath177 commutes with limits , so the proposition follows by taking the limit in ( [ foldepsilon ] ) . the map @xmath887 is a bijection between the admissible subsets considered above and the ls chains corresponding to the antidominant weight @xmath769 . surjectivity follows directly from theorem [ rootopscorr ] , based on the fact that all ls chains corresponding to @xmath769 can be obtained from the one consisting only of @xmath769 by applying the root operators @xmath155 . injectivity then follows from the character formula ( [ litchar ] ) , since @xmath815 , as noted in remark [ weightls ] . \(1 ) the proof of theorem [ rootopscorr ] contains the justification of the fact that the minima of the paths associated to ls chains are integers . this justification is based only on the combinatorics in section [ sec : folded ] . note that the same fact was proved by littelmann in @xcite using different methods . \(2 ) the proof of theorem [ rootopscorr ] shows that ls chains can be viewed as a limiting case of a special case of our construction . the special choices of @xmath0-chains that lead to ls chains represent a very small fraction of all possible choices . based on the independent results of kashiwara @xcite , lakshmibai @xcite , and joseph @xcite , which were discussed above , we deduce the following corollary . given a complex symmetrizable kac - moody algebra @xmath888 , consider the colored directed graph defined by the action of root operators ( cf . section [ sec : rootoperators ] ) on the admissible subsets corresponding to the special choice of a @xmath0-chain above . this graph is isomorphic to the crystal graph of the irreducible representation with highest weight @xmath0 of the associated quantum group @xmath218 . we make the following conjecture , which is the analog of a result due to littelmann , that was discussed above . the colored directed graph defined by the action of root operators on the admissible subsets corresponding to any @xmath0-chain does not depend on the choice of this chain . the conjecture was proved for finite types ( i.e. , complex semisimple lie algebras ) in @xcite . it implies that any choice of a @xmath0-chain leads to a perfect crystal . in this section , we discuss the way in which the model in this paper specializes to the one in @xcite in the case of finite irreducible root systems . let @xmath74 be the root system of a simple lie algebra . let @xmath6 be the _ affine weyl group _ for @xmath239 , that is , the group generated by the affine reflections @xmath48 ( defined in ( [ affref ] ) ) . the corresponding affine hyperplanes @xmath811 divide the real vector space @xmath9 into open regions , called _ alcoves . _ the _ fundamental alcove _ @xmath889 is given by @xmath890 we say that two alcoves are adjacent if they are distinct and have a common wall . for a pair of adjacent alcoves , let us write @xmath891 if the common wall of @xmath13 and @xmath892 is orthogonal to the root @xmath83 , and @xmath27 points in the direction from @xmath13 to @xmath892 . an _ alcove path _ is a sequence of alcoves @xmath893 such that @xmath894 and @xmath895 are adjacent , for @xmath896 . we say that an alcove path is _ reduced _ if it has minimal length among all alcove paths from @xmath897 to @xmath898 . let @xmath899 be the alcove obtained via the affine translation of the fundamental alcove @xmath889 by a weight @xmath0 . the reduced alcove paths from @xmath889 to @xmath900 are in bijection with the reduced decompositions of the element @xmath901 in @xmath6 defined by @xmath902 ; see @xcite . let us fix a dominant weight @xmath0 . this proposition is an analog of the fact that the normal ordering of roots can be described in terms of dihedral subgroups . each alcove @xmath13 is given by the inequalities @xmath905 where @xmath906 are some integers . we need the following characterization due to shi of the collection of integers @xmath907 associated with alcoves . @xcite an arbitrary collection of integers @xmath908 corresponds to some alcove @xmath13 , i.e. , @xmath909 for all @xmath16 , if and only if , for any triple of roots @xmath910 such that @xmath911 , we have @xmath912 . for a sequence of positive roots @xmath913 , define @xmath914 , for @xmath16 and @xmath915 . the axioms of a @xmath0-chain in the finite case ( condition ( 1 ) in definition [ lambdachain ] and condition ( @xmath916 ) in proposition [ equivalence ] ) can be rewritten in terms of the integers @xmath917 as follows : ( 1 ) @xmath918 , for @xmath16 ; and ( 2 ) for any triple @xmath910 such that @xmath919 and @xmath915 , we have @xmath920 ( interlacing condition ) . shi s result implies that these conditions are equivalent to the fact that @xmath904 is a reduced alcove path , where @xmath921 is the alcove associated with the collection of integers @xmath922 , i.e. , @xmath923 . \(2 ) reduced alcove paths from @xmath889 to @xmath924 ( where @xmath925 is the longest weyl group element ) correspond to _ reflection orderings _ @xcite . if @xmath0 is regular , then some reduced alcove paths from @xmath889 to @xmath926 start with an alcove path from @xmath889 to @xmath927 . hence , we can say that @xmath0-chains extend the notion of a reflection ordering . a _ gallery _ is a sequence @xmath928 such that @xmath929 are alcoves ; @xmath930 is a codimension one common face of the alcoves @xmath894 and @xmath895 , for @xmath896 ; @xmath931 is a vertex of the first alcove @xmath897 ; and @xmath932 is a vertex of the last alcove @xmath898 . furthermore , we require that @xmath933 , @xmath934 , and @xmath935 for some weight @xmath936 , which is called the _ weight _ of the gallery . the folding operator @xmath937 is the operator which acts on a gallery by leaving its initial segment from @xmath897 to @xmath894 intact and by reflecting the remaining tail in the affine hyperplane containing the face @xmath930 . in other words , we define @xmath938 where @xmath939 , @xmath940 , and @xmath941 , for @xmath942 . let us fix a reduced alcove path @xmath904 , which determines the @xmath0-chain @xmath19 with @xmath943 . the alcove path also determines an obvious gallery @xmath944 of weight @xmath769 . we use the same notation as in sections [ sec : lambdachains]-[sec : rootoperators ] . for instance , @xmath358 and @xmath774 . we also let @xmath945 be the affine reflection in the hyperplane containing @xmath946 . the weight of the gallery @xmath802 is @xmath949 ( cf . definition [ defweighted ] ) . indeed , we have @xmath816 . hence , the model in this paper specializes to the one in @xcite , whose construction was based on the geometry of the generalized flag variety . since we assumed that @xmath74 is irreducible , there is a unique _ highest coroot _ @xmath950 , i.e. , a unique coroot that has maximal height . we will also use the _ coxeter number _ , that can be defined as @xmath951 ( in the finite case , the dominant weight @xmath149 considered at the end of section [ sec : notation ] is unique , and is given by @xmath952 ) . let @xmath953 be the set of the elements of the lattice @xmath954 that do not belong to any affine hyperplane @xmath811 . each alcove @xmath13 contains precisely one element @xmath955 of the set @xmath953 ( cf . @xcite ) ; this will be called the _ central point _ of @xmath13 . in particular , @xmath956 . let us now associate to the gallery @xmath804 a continuous piecewise - linear path . consider the points @xmath958 , @xmath959 for @xmath960 , @xmath961 for @xmath962 , and @xmath963 . note that @xmath964 lies on @xmath946 for @xmath965 . let @xmath807 be the piecewise - linear path obtained by joining @xmath966 . given an admissible subset @xmath57 , let @xmath967 be the points on the faces of the gallery @xmath802 that are obtained ( in the obvious way ) from @xmath968 in the process of constructing @xmath802 from @xmath804 via folding operators . clearly , @xmath969 are the central points of the corresponding alcoves in @xmath802 , for @xmath960 . by joining @xmath970 , we obtain a piecewise - linear path that we call @xmath803 . note that @xmath803 can be described using folding operators , as in ( [ describefold ] ) , once these operators are appropriately defined . it turns out that , in general , the collection of paths @xmath803 , for @xmath57 ranging over admissible subsets , does _ not _ coincide with the collection of littelmann paths obtained from @xmath807 by applying the root operators @xmath177 . indeed , it is not true in general that @xmath977 , as was the case with the paths corresponding to ls chains ( cf . theorem [ rootopscorr ] ) . the reason is that , given @xmath978 , the function @xmath739\rightarrow { \mathbb r}$ ] given by @xmath740 is usually not weakly decreasing between the corresponding points @xmath753 and @xmath754 ( see section [ sec : ls ] for the definition of these points ) . this happens , for instance , when applying @xmath979 to the path @xmath807 in example [ exg2 ] below . such situations can also arise if we define @xmath807 by joining the centers of the faces @xmath946 , or the centers of both the alcoves @xmath921 and the faces @xmath946 ( in the order they appear in the gallery @xmath804 ) . in all these situations , we need the general definition of @xmath177 for littelmann paths , which we now recall from @xcite . as in section [ sec : ls ] , the definition is easier to state if we replace the continuous piecewise - linear paths @xmath757 ( satisfying @xmath758 ) with their left - hand derivatives @xmath37 ( which are piecewise - constant left - continuous maps defined on @xmath759 $ ] ) . recall that we denoted the minimum of the function @xmath765 by @xmath738 . if @xmath980 , then @xmath756 is undefined , as before , so assume that @xmath981 . choose @xmath982 such either our model can be based on _ any _ @xmath0-chain ( that is , not necessarily on the ones given by proposition [ constr - lambdachain ] ) , and we still have a definition of root operators for admissible subsets / admissible foldings that corresponds to the simpler version of their definition on paths , given in ( [ defeppaths ] ) . suppose that @xmath992 . proposition [ constr - lambdachain ] gives the following @xmath993-chain : @xmath994 thus , we have @xmath995 , @xmath996 , @xmath997 , @xmath998 , @xmath999 , @xmath1000 , @xmath1001 , @xmath1002 , @xmath1003 , @xmath1004 . there are six saturated chains in the bruhat order ( starting at the identity ) on the corresponding weyl group that can be retrieved as subchains of the @xmath993-chain . we indicate each such chain and the corresponding admissible subsets in @xmath1005 . 1 . 1 : @xmath1006 ; 2 . @xmath1007 : @xmath1008 ; 3 . @xmath1009 : @xmath1010 , @xmath1011 , @xmath1012 ; 4 . @xmath1013 : @xmath1014 , @xmath1015 , @xmath1016 ; 5 . @xmath1017 : @xmath1018 , @xmath1019 , @xmath1020 , @xmath1021 , @xmath1022 ; 6 . @xmath1023 : @xmath1024 . the weight of each admissible subset is now easy to compute ( by applying the corresponding affine reflections above to @xmath993 , cf . definition [ defweighted ] ) . this leads to the expression for the character @xmath1025 as the following sum over admissible subsets : @xmath1026 figure [ fig : g2 ] displays the galleries @xmath802 corresponding to the admissible subsets @xmath57 indicated above , the associated paths @xmath803 , as well as the action of the root operators @xmath155 on @xmath57 . for each path , we shade the fundamental alcove , mark the origin by a white dot `` @xmath1027 '' , and mark the endpoint of a black dot `` @xmath1028 '' . since some linear steps in @xmath803 might coincide , we display slight deformations of these paths , so that no information is lost in their graphical representations . as discussed above , the weights of the irreducible representation @xmath1029 are obtained by changing the signs of the endpoints of the paths @xmath803 ( marked by black dots ) . the roots in the corresponding admissible foldings @xmath340 can also be read off ; see proposition [ chain_path ] . at each step , a path @xmath803 either crosses a wall of the affine coxeter arrangement or bounces off a wall . the associated admissible subset @xmath57 is the set of indices of bouncing steps in the path . c. chevalley . sur les dcompositions cellulaires des espaces . in _ algebraic roups and eneralizations : lassical ethods _ , volume 56 part 1 of _ proceedings and symposia in pure mathematics _ , pages 123 . 1994 . a. lascoux and m .- keys and standard bases . in d. stanton , editor , _ invariant theory and tableaux _ , volume 19 of _ the i m a vol . in math . and its appl . _ , pages 125144 , berlin - heidelberg - new york , 1990 . springer - verlag .

[ sec : intro ] the collapsar model @xcite was originally proposed as an explanation for common , `` long - soft '' gamma - ray bursts ( grbs ) with durations near 20 s. this is a typical time scale for the gravitational collapse of the carbon - oxygen core of a massive star and the congruence of these time scales was an argument in favor of the model . it is well known , however , that forming a disk so quickly requires an unusually large amount of angular momentum to be stored in the core up to the point when the star dies ( e.g. , * ? ? ? this is presumably one reason why grbs are rare compared with supernovae . black holes are expected to form in a significant fraction of massive star deaths ( e.g. , * ? ? ? * ) and it seems possible that , in at least some of these , the black hole will form without generating an outgoing shock . what observable signal would accompany such an event ? would they all be `` un - novae '' @xcite , stars that simply suddenly disappear from view ? or will some sort of transient display announce the birth of any black hole ? here we explore the possibility that these black hole systems form an accretion disk , but only after almost all of the star has accreted into the hole . torques in differentially rotating contracting stars tend to concentrate angular momentum in the star s outer layers . the surface will also have a large angular momentum in a tidally locked system . on the other hand , stellar mass loss preferentially removes these same layers , so , just as in the usual collapsar model , reducing mass loss , e.g. , by reducing the metallicity , favors the retention of high angular momentum at the stellar surface . because the outer layers of the star take longer to fall in - months to years in the case of a red supergiant - the duration of the transient produced by accretion through a disk in these cases is much longer than for the common grbs studied so far . they might even be confused with other forms of gamma - ray transients like those arising from super - massive black hole accretion in active galaxies @xcite . here we shall consider four varieties of long duration transients categorized by the kind of system where they occur : 1 ) single supergiants with metallicity less than about @xmath4 and low mass loss ( [ sec : bsg ] ) ; 2 ) red or blue supergiants in tidally - locked , or nearly tidally - locked binaries ( [ sec : rsg ] ) ; 3 ) pair - instability induced collapse to a black hole ( [ sec : pair ] ) ; and 4 ) helium stars in tidally locked binaries ( [ sec : helium ] ) . in each case , the duration of the transient that is produced is correlated with the radius of the presupernova star , hence shortest for the helium stars , while the luminosity is inversely correlated with the radius , hence faintest for the red supergiants . as with common grbs , we are unable to predict _ ab initio _ the spectrum of these transients or the efficiency for converting black hole accretion rate into electromagnetic radiation . presumably , as in other forms of transients powered by very rapidly accreting , rotating black holes , accretion energy and black hole spin energy power relativistic bi - polar outflows . poynting flux , shocks within these outflows , and shocks with the circumstellar medium make x - rays and gamma - rays . it is common to take an efficiency of @xmath5 to @xmath6 times @xmath7 for the efficiency ( e.g. , * ? ? ? * ) thus suggesting an integrated power of @xmath8 per solar mass accreted . interestingly these `` failed supernovae '' @xcite are far more brilliant and powerful than their ( model - wise ) more successful cousins , ordinary supernovae . the idea that late time accretion in supernovae that make black holes might power long gamma - ray ( or hard x - ray ) transients was suggested by @xcite . the transients discussed there , referred to as `` type 2 collapsars '' , were produced because of fall - back in an explosion that launched a relatively strong outgoing shock in a red supergiant progenitor . two dimensional models explored jet propagation and break out for a variety of fall - back rates , but little attention was given to the angular momentum distribution necessary to form a disk . here we consider a wider variety of progenitors and pay closer attention to the angular momentum , especially in the outer layers . @xcite also discussed the possibility of very - long duration x - ray transients powered by fall back and accretion through a disk . he suggested these might look like `` ultra - luminous x - ray sources '' seen in nearby galaxies . these would be much longer in duration ( thousands of years ) and have much lower luminosity than the events considered here , but are an example of the same general idea . more recently @xcite and @xcite suggested that fallback or a failed explosion in a red supergiant could give a long duration gamma - ray transient that might be confused with a gamma - ray blazar . quataert and kasen , in particular , suggest this as an alternate interpretation of swift transient sw-1644 + 57 @xcite . woosley s paper was accepted prior to the discovery of this event . a variety of other models may lead to long duration bursts . for example a white dwarf @xcite or helium star merging with a black hole or neutron star @xcite . a neutron star or black hole merging with a a giant star might give a thorne - zytkow object @xcite or a black hole accreting at a rapid rate @xcite . helium star merger with a compact object also gives a long transient of some sort @xcite . here we consider only those transients coming from stars that are not actively merging with a companion . [ sec : black ] a black hole will form inside a massive star when a sufficient concentration of matter and energy accumulates to form an event horizon . the simplest possibility is that some portion of a massive star experiences an instability and falls freely , or almost freely , until it reaches such a high density that a trapped surface is formed . this rarely happens in nature though . only in the most massive stars experiencing a pair instability @xcite or a general relativistic instability @xcite can a trapped surface form before the center reaches nuclear density and bounces . that is , black hole formation in a massive star is almost always preceded by protoneutron star formation @xcite . for stars in the range @xmath9 , however , the protoneutron star that is formed experiences rapid accretion from the surrounding layers of heavy elements - silicon , oxygen , etc . if rotation , magnetic fields and neutrinos do not readily produce an outgoing shock , the hot protoneutron star grows beyond a critical mass and collapses to a black hole . the accretion rate is larger for more massive helium cores @xcite , and black hole formation without an outgoing shock may be a frequent occurrence in stars with main sequence masses over about @xmath10 @xcite . whether _ all _ stars above about @xmath10 produce black holes without outgoing shock waves is uncertain , but the possibility is not precluded by observations or theory . `` prompt '' black hole formation is probably more likely in very low metallicity stars that have experienced little mass loss and thus have especially large helium cores , but it could occur for solar metallicity stars , depending upon the uncertain mass loss rate @xcite . it is also possible that such a weak outgoing shock is generated that most of the star fails to explode and a black hole is made by fall back @xcite . these were the systems explored by @xcite . the ones treated here differ in assuming that any outgoing shock is insufficient to eject more than a small fraction of the surface layers of the star . we shall call such events , where only the surface layers have enough rotation to form a disk and a shock so weak that these layers are not ejected `` type 3 collapsars '' . this is to distinguish them from ordinary grbs where the disk forms promptly from material deep inside a highly differentially rotating helium core ( type 1 ; * ? ? ? * ; * ? ? ? * ) and from systems where longer duration transients result from the delayed fall back of other material also deep inside the star ( type 2 ; * ? ? ? * ; * ? ? ? [ sec : stars ] the preservation of angular momentum in the surface layers of the star can occur either because mass loss is inefficient in a single star or because of tidal interaction with a close companion . here we consider examples of both sorts . [ sec : bsg ] a paucity of metals in a star affects its evolution in several ways - reduced energy generation by the cno cycle in the hydrogen burning shell , reduced opacity , and reduced mass loss . of these effects , the reduction in mass loss is least well understood , but the lack of atomic lines in the atmosphere on the main sequence and of grains as a red giant seems certain to lead to some reduction . a common assumption is that the mass loss for massive main sequence stars depends upon some fractional power of the metallicity @xcite . more important and less certain is the dependence of mass lost as a red or blue supergiant on metallicity . it is thought that mass loss from cool giants is more dependent upon pulsations and grain formation @xcite . one might expect therefore a rapid fall off in mass loss below some value necessary for significant grain production . it has been estimated that red giant mass loss will be significantly less below @xmath4 @xcite . if stars do not lose mass then they conserve angular momentum . since the natural course of evolution leads to the contraction and spin up of the inner star , shear instabilities and magnetic torques will concentrate an increasing amount of angular momentum in the outer regions of the star . whereas the inner part does spin faster due to contraction , it actually loses most of its initial angular momentum by the time it reaches central carbon burning . @xcite surveyed the evolution of non - rotating zero - metal massive stars from @xmath11 to @xmath12 . in a parallel study currently underway @xcite , we are surveying the evolution of rotating stars in the same mass range with metallicity @xmath13 , @xmath14 and @xmath15 that of the sun . the two stars discussed here are extracted from that survey and are typical in the way that they accumulate large amounts of angular momentum in their outer layers before dying as supergiants . all calculations of stellar evolution presented here used the kepler code @xcite and included angular momentum transport by magnetic torques @xcite . models v24 and v36 are @xmath16 and @xmath17 main sequence stars with metallicity @xmath18 solar . both stars rotated rigidly on the main sequence with a moderate speed of about @xmath19 which is about @xmath20 of the keplerian value . both presupernova stars had hydrogenic envelopes that were , throughout most of their mass , radiative . model v36 , however , had a low mass surface convection zone that included @xmath21 . model v36 was consequently a yellow supergiant at death ( @xmath22 ; @xmath23 ; @xmath24 ) . model v24 was a rather large blue supergiant ( @xmath25 , @xmath26 ; @xmath27 ) . these two stars ended their lives with cores of helium and heavy elements of @xmath28 and @xmath29 respectively and angular momentum distributions as shown in fig . [ fig : bsg ] . whereas the lack of sufficient angular momentum within the helium core precludes making a disk around a black hole within that mass , there is ample rotation in the outer part of the hydrogen envelope to do so . this could be a very common occurrence . from the survey of @xcite , the total angular momentum of a massive star at birth that has an equatorial rotation speed of about @xmath20 keplerian on the main sequence is @xmath30 if the star does not lose mass , most of this angular momentum becomes concentrated , in the presupernova star , in a nearly rigidly rotating hydrogenic envelope with radius either @xmath31 ( blue supergiant ) or @xmath32 ( red supergiant ) . whereas the density declines with radius in the actual envelope , one can obtain some interesting scaling relations by assuming constant density . for a moment of inertia , @xmath33 , the angular velocity at the surface , @xmath34 , is @xmath35 the specific angular momentum of this rigidly rotating envelope is @xmath36 , or about @xmath37 at the surface of a red supergiant and @xmath38 for a blue supergiant , if no mass is lost . these can be compared with the angular momentum required to make a disk around a non - rotating black hole , @xmath39 or @xmath40 for a maximally rotating hole ( kerr parameter @xmath41 ) . all massive stars that do not lose mass have sufficient angular momentum in their outermost layers to make a disk around any black hole formed in their collapse . of course stars do lose mass and the fact that , for constant @xmath42 , @xmath43 means that the first mass to be lost contains most of the angular momentum . in practice , we find that a red supergiant that loses more than a few percent of its total mass before dying will probably not make a black hole - disk system . this sensitivity to mass loss is illustrated by the evolution of a @xmath44 model , o25 , with @xmath6 solar metallicity in a calculation that included mass loss at one - half the standard value for that metallicity . the star lost @xmath45 ( @xmath46 ) of its mass and died as a red supergiant with a radius of @xmath47 . the value of @xmath48 in its outermost layers was @xmath49 , too little to form a disk around a schwarzschild black hole , but possibly just enough to form a disk around a mildly rotating one . the point where @xmath48 declined to one - half its surface value in the presupernova star was located @xmath50 interior . this is borderline case . stars with less mass loss , i.e. , lower metallicity , and stars that never become red giants could make long gamma - ray transients ( table [ tab : models ] ) . [ sec : rsg ] based upon theoretical models for single massive stars , the red supergiant progenitors of common type iip supernovae are not expected to have rapidly rotating envelopes . expansion to a red giant slows the surface layers . torques transfer the angular momentum of the more rapidly rotating inner regions outwards concentrating it in the outermost layers . for stars of sufficient metallicity , stellar winds are efficient at removing these layers and , along with them , most of the star s angular momentum . for example , a @xmath51 solar metallicity star rotating rigidly with an equatorial speed of @xmath19 on the main sequence ends up a @xmath52 supernova progenitor if commonly used estimates for mass loss are employed @xcite . the surface rotation rate of that red supergiant ( r = @xmath53 ) is @xmath54 and its angular momentum at the surface is @xmath55 . nowhere in the star does any matter have enough angular momentum to form a disk around a black hole . if a black hole forms without making an outgoing shock , the star abruptly disappears . this sort of `` un - nova '' might be a common occurrence in the modern universe @xcite , though see @xcite . the outcome can be quite different in a binary system , however . a stellar merger could lead to a rapidly rotating envelope @xcite , but such mergers probably lead to ejection of a common envelope . an interesting alternative is detached binary systems with a small enough separation for tidal interaction , but too distant to merge or exchange mass . the circularization and synchronization time scales for binary stars , one of which is a red giant , have been considered by @xcite , @xcite and @xcite . for a primary with mass , @xmath56 , envelope mass @xmath57 , luminosity @xmath58 , and radius @xmath59 , in an orbit having radius @xmath60 , with a companion of mass @xmath61 , the _ circularization _ time , @xmath62 is @xcite @xmath63 where @xmath64 is a number close to unity . the lifetime of a @xmath51 main sequence star as a red supergiant is approximately @xmath65 , depending upon uncertain convection parameters . the mass of the envelope that is convective at any one time , however , is only about @xmath66 and the star loses more than this in its lifetime . a more appropriate time scale might be the time the star spends losing its last @xmath66 , which is about @xmath67 . the mass of the star when it becomes a supernova is @xmath52 and we assume a companion mass of , e.g. , @xmath68 . the luminosity of the star as a red supergiant is @xmath69 and its moment of inertia is approximately @xmath70 . equation ( [ eq : circ ] ) then implies that during its last @xmath67 , orbits with radius @xmath71 will circularize . the _ synchronization _ timescale for co - rotation is somewhat different @xcite and , in the present case , is easier to achieve . @xmath72 most of the angular momentum is in the hydrogen envelope and , from the same @xmath51 model , @xmath73 . with @xmath74 and @xmath75 @xcite , @xmath76 which only requires @xmath77 for synchronization in @xmath67 . in such a system , the period will be shorter than about @xmath78 years ( @xmath79 ) . systems much closer will merge or experience mass exchange , so we take this to be a typical rotation period for the envelope in a detached binary system that might achieve corotation . longer period systems will still transfer significant rotation to the envelope , however . this is about four orders of magnitude more angular momentum in the outer envelope than in the case of the single star . since the outer several solar masses of the presupernova star is convective , this rotation rate will persist to some depth . all of that mass would have sufficient angular momentum to form an accretion disk if its helium core collapsed to a black hole . even then though , the total angular momentum of the envelope , @xmath80 , would be a small fraction of the orbital angular momentum of the binary pair of stars . the addition of only @xmath5 of the angular momentum estimated here would still result in large quantities of material forming a disk , so the range of binary separations allowed is actually larger , perhaps by a factor @xmath81 , or about two . since the time scale for material to fall in from @xmath82 is longer than any realistic viscous time scale for the disk , the accretion rate will be given by the collapse timescale of the envelope . for matter at the base of the hydrogen shell where a disk would first form , the enclosed mass is @xmath83 and the radius about @xmath84 . at the outer edge of the star the mass is @xmath52 and the radius , @xmath85 . the corresponding free fall times ( @xmath86 ) are @xmath87 and @xmath88 suggesting that the event would last for months . accreting @xmath66 in @xmath89 with @xmath6 efficiency for converting accreted mass to outgoing energy would give a jet power of @xmath90 . conversion of only @xmath6 of the jet power ( @xmath5 of the accreted mass energy ) into gamma - rays would explain the recently discovered transient sw 1644 + 57 @xcite , especially if moderate beaming were involved . whereas the focus here is on red supergiants because they give the longest transients , tidal locking would also occur in blue supergiants with smaller radii . the key point is that forced co - rotation with a body in keplerian orbit gives , for a large range of orbital separations , envelope material that is rotating too fast to accrete , without hindrance , into a stellar mass black hole . since the time scale for accretion is set by the collapse time for the outer part of the star , not the viscous time of the disk , a transient from a tidally locked blue supergiant would be similar in duration within a factor of a few to those studied for single blue supergiants in [ sec : bsg ] . blue supergiants , however , will have less mass in the outer layers of the envelope and hence should have less powerful outbursts ( appendix [ sec : append ] and @xcite ) . [ sec : pair ] following helium burning , non - rotating helium cores with masses greater than @xmath91 will collapse directly into black holes with no outgoing shock @xcite . the collapse is caused by the pair - instability and , above this mass , nuclear burning is unable to reverse the implosion . rotation , unless it is very rapid and highly differential , does not change this mass limit greatly since the instability exists in the deep interior where the ratio of centrifugal force to gravity is small . the collapse of a rotating pair unstable star to a black hole was followed in two dimensions by @xcite . first , a @xmath92 main sequence star was evolved to the point of instability in the kepler ( 1d ) code . this star produced a helium core of @xmath93 . the initial star was assumed to rotate rigidly on the main sequence with a ratio of equatorial speed to keplerian of 20% . angular momentum transport , especially by shear mixing and eddington sweet circulation , was followed throughout the evolution , but magnetic torques were neglected . as a result , the collapsing helium core rotated sufficiently rapidly to produce a disk outside of a massive black hole once about @xmath94 of the @xmath93 core of helium and heavy elements had collapsed . @xcite speculated that the accretion of the remaining @xmath95 might produce a very bright , lengthy grb . we have recalculated the evolution of similar stars , but with magnetic torques @xcite included in the simulation . models z250a , b , and c are stars with zero metallicity and a mass at the time they formed of @xmath96 . they differ in the amount of angular momentum each had on the main sequence . z250a had @xmath97 corresponding to a rotation rate of @xmath98 , or @xmath99 keplerian . models z250b and z250c had @xmath100 and @xmath101 of angular momentum and rotational speeds of @xmath102 ( @xmath103 keplerian ) and @xmath104 ( @xmath105 keplerian ) . these made helium cores of @xmath106 , @xmath107 , and @xmath108 respectively , all of which collapsed directly to black holes following helium depletion . unlike in @xcite however , none of these helium cores had sufficient angular momentum to make a disk around a black hole anywhere in their interior . because mass loss was neglected in these models , the angular momentum lost from the helium core due to magnetic coupling instead ended up concentrated in the outer part of the hydrogen envelope . [ fig : pair ] shows the evolution of the angular momentum in model z250b . the evolution of the other two models was similar . inadequate angular momentum exists in the helium core to form a disk but , for model z250b , the outer @xmath109 can still do so . as in the tidally locked red supergiant case , the collapse time scale for this material is days to months , so a high energy transient of around @xmath110 , possibly boosted by beaming , could exist for that time . it should be noted , however , that mass loss is even more uncertain in these models than in the others . due to rotationally induced mixing , the envelopes of all three models were appreciably enhanced in c , n , and o and thus were red supergiants when they died . neglecting mass loss in such a situation is probably not justified . also there may be other mechanisms for losing mass in such massive stars besides line and grain driven mass loss @xcite . [ sec : helium ] helium stars in close tidally locked binaries have been frequently suggested as progenitors for common grbs @xcite . on the positive side , such events should occur frequently in nature and , if the core of a massive wolf - rayet star collapses to a black hole while the surface is forced to rotate at a fraction of its keplerian speed , disk formation around a black hole is assured . on the negative side , however , the production of a grb with typical duration @xmath010 s , requires that the _ inner _ part of the helium star rotate rapidly , not just its surface . the inclusion of magnetic torques in tidally locked stars leads , by itself , to too little rotation to make a disk in that part of the star which might accrete rapidly @xcite . on the other hand , leaving out magnetic torques makes it difficult to produce the spin rates of ordinary pulsars @xcite , and could lead to an _ overabundance _ of grb progenitors . this is not to say that close binary systems do not produce _ some _ form of gamma - ray transient though . the angular momentum in their cores may be adequate to produce a millisecond magnetar , and thus a common grb , just not a ( type 1 ) collapsar . if a black hole forms , the accretion of the _ outer _ layers of the star may lead to the formation of a disk and grb much longer than typical . it is this latter possibility that we explore here . following @xcite , we consider the evolution of helium cores of @xmath111 and @xmath109 whose surface s have become tidally locked with a closely orbiting companion . @xcite found the shortest possible periods were 2.05 hours ( @xmath112 ) and @xmath113 hours ( @xmath114 ) for @xmath111 and @xmath109 respectively , and they gave examples of systems containing a compact object and a helium core that might approach these limiting values . an @xmath111 he core in a tidally locked binary with a @xmath115 companion filling its roche lobe had a longer period of @xmath116 hours ( @xmath117 ) . here we calculate two models for each helium core mass . model 8a has an enforced rotation rate in its outer @xmath118 of @xmath119 throughout its evolution . models 8b , 16a , and 16b have rotation rates of @xmath120 , @xmath121 , and @xmath122 respectively . [ fig : binaryw ] shows the evolution of the angular momentum distribution as a function of mass for these four models . during helium burning , magnetic coupling maintains a state of nearly rigid rotation with an angular speed given by the surface boundary condition . for both model 8a and 16a , this corresponds to a ratio of rotational speed to keplerian of @xmath123 . for models 8b and 16b , the ratio is @xmath124 . roughly the inner @xmath125 ( @xmath111 ) or @xmath126 ( @xmath109 ) of the helium star is convective during helium burning and rigid rotation is enforced in that region . the outer part of the star is radiative , but magnetic torques there suppress differential rotation . during carbon burning and more advanced stages , the inner core contracts and rotates more rapidly . magnetic torques are too inefficient to enforce rigid rotation . later , during carbon burning , the stars ignite a convective helium burning shell which persists until the star dies . this convective shell does not extend all the way to the surface though , but does extend beyond the @xmath118 at the surface where corotation is enforced . thus all stars die with an angular velocity in their helium convective shell given by the surface boundary condition , but have larger rotation rates in their cores , due to contraction during advanced burning stages . [ fig : binaryj ] shows the corresponding angular momentum in the evolving models . shear instabilities and magnetic torques transport angular momentum out of the inner core and to the surface . in the absence of a binary companion , the surface would have actually spun up , but would have been removed by mass loss . in the present parametrized calculations though , there is no mass loss and the set rotation rate of the surface absorbs any excess angular momentum delivered to it . consequently the total angular momentum of the star decreases with time . for model 8a at helium ignition , it was @xmath127 , but for the presupernova star it was @xmath128 . [ fig : binaryjfin ] shows the final distribution of angular momentum in the stars compared with that required to make a disk around a black hole , once the mass interior to the given sample point has collapsed . four curves are given . three are for the minimum angular momentum required to form a disk at the last stable orbit of a ) a non - rotating ( schwarzschild ) black hole ; b ) a kerr black hole ; and c ) a black hole with rotation given by the actual angular momentum distribution of the presupernova star . the fourth curve is the angular momentum distribution in the presupernova star . because the inner parts of the star are not rotating very rapidly , curve a ) and c ) are similar except near the surface . only where the actual angular momentum lies above curve c ) can a disk form . material with less angular momentum plunges directly into the hole . none of the models has enough angular momentum to form a disk in its inner core . thus a type 1 collapsar will not occur . the angular momentum in the inner @xmath129 of the stars is still appreciable , however . models 8a and 16a would form @xmath130 ( gravitational mass ) pulsars with periods of @xmath131 and @xmath132 respectively . models 8b and 16b would form pulsars of @xmath133 and @xmath134 . model 16b thus qualifies as a `` millisecond magnetar '' candidate and might power a common grb if a pulsar is able to form before accretion turns it into a black hole . models 8b , 16a and 16b could also make powerful pulsar - powered supernovae . model 8a and other more slowly rotating systems probably would not . of greater interest for the present paper , however , is the large angular momentum in the _ outer _ layers of three of the models . these would be relevant if the center of the star collapsed to a black hole rather than a millisecond pulsar . model 8a has angular momentum sufficient to form a disk in its outer @xmath135 . two other models ( table [ tab : models ] ) have similar large rotation in their surface layers . this is consistent with the results of @xcite who found that their most rapidly rotating @xmath111 and @xmath109 helium cores ended their lives with @xmath48 in outer layers of @xmath136 and @xmath137 . the radius and gravitational potential of the high angular momentum surface layers allows an order of magnitude estimate of the free fall time scale ( table [ tab : models ] ) , about @xmath138 . the tidal interaction actually acts to brake the rotation of the surface layers compared to what they would have had in a star without mass loss and no companion star . to illustrate this , model 8c was calculated , starting from the rigidly rotating helium burning stage of model 8b ( total angular momentum @xmath139 ) , but with no surface boundary condition and no mass loss . that is the rotation rate of the surface layers was allowed to adjust to be consistent with whatever angular momentum was transported to them . without mass loss , the star was forced to conserve angular momentum overall and the very outer layers ended up rotating very rapidly ( fig . [ fig : binaryjfin ] ) . the angular speed , rather than being @xmath140 was @xmath141 . in fact the outer tenth of a solar mass rotated so rapidly that it would be centrifugally ejected . in a realistic calculation , this matter and more underlying matter would probably have been ejected as a disk or a centrifugally boosted wind . the deep interior still lacked sufficient angular momentum to form a disk around a black hole , but the amount of surface material that could form a disk ( if mass loss did not remove it ) was significant , about 2 @xmath142 , comparable to that in model 8b . two other models like 8c were calculated that included mass loss appropriate for a wolf - rayet star at solar metallicity and one - tenth that value . the first ended up with a final mass of @xmath143 and no surface layers with sufficient angular momentum to form a disk . the second ended up with a mass of @xmath144 and still had enough angular momentum in its outer solar mass to make a disk . so the major role of the close companion in the calculations done here is to preserve the angular momentum in the outer layers that might otherwise be removed by mass loss . a companion may also be essential so that the helium core rotates rapidly in the first place , following the loss of the star s hydrogen envelope . when magnetic torques are included in the calculation , the helium cores of supernova progenitors that are red and blue supergiants rotate too slowly for any portion of that core to make a disk . a close companion could either preserve angular momentum in the core by removing the envelope very early in the evolution or add angular momentum by tidal locking later on . [ sec : conclude ] table [ tab : models ] and table [ tab : models1 ] summarize the wide variety of transients that can be produced by type 3 collapsars . in table [ tab : models ] @xmath145 is the amount of mass capable of forming a centrifugally supported disk around a black hole with a kerr parameter given by accreting all the mass interior to @xmath145 and @xmath59 is the range of stellar radii where that mass exists . the accretion time scale has been estimated using the radius for the center of mass of @xmath145 . the luminosity has been estimated by multiplying the effective accretion rate ( mass / time scale ) by @xmath146 . this assumes @xmath6 efficiency for making a jet and @xmath6 efficiency for converting jet energy into hard radiation . most models produce extreme kerr black holes ( @xmath41 , table [ tab : models1 ] ) , so the radiative efficiencies may be higher than assumed here @xcite . the numbers in the table are only order of magnitude estimates . table [ tab : models1 ] gives the the estimated pulsar period formed in the stellar collapse assuming the angular momentum that exists in the inner @xmath129 of the core , a final gravitational mass of @xmath130 , and a moment of inertia of @xmath147 @xcite . if a black hole forms before the protoneutron star has radiated away most of its gravitational binding energy , the rotation rate will be slower as will the rotational kinetic energy available for explosion . @xmath148 and @xmath149 are the masses of the presupernova star and its helium core respectively and @xmath150 is the lagrangian mass coordinates of the stellar layers that can form a centrifugally supported disk . the kerr parameter , @xmath151 , is what would develop if the entire presupernova , including its rapidly rotating outer layers , collapsed to a black hole . in case where it exceeds unity it has been capped at @xmath152 . the kerr parameter for the helium core alone is @xmath153 . @xmath154 is the net binding energy , in units of @xmath155 , of the material that could form a disk . a range of masses can form disks and a large range of time scales characterize the accretion , hence the possible luminosities span many orders of magnitude . the shorter brighter transients from wolf - rayet stars in tidally locked binaries might comprise an interesting subset of long , but otherwise ordinary grbs , as others have noted . this channel could be especially important in regions where high metallicity and high mass loss preclude the production of grbs by solitary stars . the mass of @xmath156ni produced in the collapsar model is sensitive to the temperature , density , time scale , and ejection fraction of the disk and is difficult to estimate . the lower accretion rate , larger black hole mass , and smaller kerr parameter for the models presented here , however , will give lower temperatures and densities in the disk @xcite . below about @xmath157 , the nature of the accretion changes and the disk is no longer neutrino dominated . a thicker disk might qualitatively change the nature of the transient ( macfadyen , private communication ) . it is thus reasonable to expect that the @xmath156ni production , per solar mass of material accreted , will be different here and perhaps much smaller . these bursts might not be accompanied by supernovae ( e.g. , * ? ? ? on the longer end , blue and red supergiants will produce transients of order @xmath158 and @xmath159 respectively . these could result from either tidally locked systems or more distant systems with low metallicity and mass loss . these fainter and potentially more common systems may not have been detected yet and certainly should be searched for . they could easily be confused with flaring gamma - ray blazars @xcite , but would differ in that they only happen once and would be associated with regions of massive star formation , not particularly galactic nuclei . they too would probably not be accompanied by supernovae and might have a different beaming factor and spectrum from ordinary grbs . the black holes produced in these supergiant progenitors would frequently be rotating at near the maximum value ( fig . [ fig : bsg ] and fig . [ fig : rsg ] ) . an exciting prospect is that the first generations of stars , those with metallicities from zero to @xmath160 solar might be studied directly using the hard x - rays they emit when they die . whereas predictions are still clouded by the vagaries of mass loss and the explosion mechanism , our studies predict that the fraction of massive stars emitting such bursts could be substantial , essentially limited only by the fraction that make black holes promptly . if none make black holes promptly , then the death of every massive star must make a supernova , though perhaps a faint one . this too is an interesting possibility worth further exploration . because of their long duration and potentially high event rate , it is possible that several type 3 collapsars will be visible in distant galaxies at any one time . @xcite estimate @xmath161 core - collapse supernovae occur each year in a field of view @xmath162 in the redshift range @xmath163 . this is about @xmath164 supernovae per second spread over the whole sky . the observation of grbs beyond redshift @xmath165 shows that massive stars were dying at an even earlier time . black hole formation probably occurs in a significant fraction of these supernovae @xcite , perhaps all those with mass over @xmath10 and metallicity less than @xmath6 solar . in order to estimate , crudely , an event rate , assume that prompt black hole formation occurs in @xmath6 of core collapses and @xmath5 of these stars either have such low mass loss rates or close companions that their outer layers make a disk . then @xmath18 of all massive star deaths would make type 3 collapsars . for low metallicity stars , the latter could be an underestimate . the universal supernova rate of @xmath164 per second then implies a gamma - ray transient event rate of @xmath166 per second . if each event lasted @xmath158 , there would @xmath167 events in progress at all times . most of these would be at high redshift ( low metallicity ) and the spectrum would be softer and duration longer than in the lab frame , so the number of active sources could be higher . on the other hand , beaming could increase the brightness and decrease the event rate significantly . in an extreme limiting case where @xmath6 of all massive star deaths made type 3 collapsars with duration @xmath159 with negligible beaming , one would have @xmath168 active sources at any one time , or about @xmath152 per square degree . if each source radiated @xmath169 , this could potentially contribute to a `` diffuse '' cosmological background of soft gamma - rays . the greatest uncertainty in these results is whether the matter with high angular momentum will get ejected , by mass loss either before the explosion or during the explosion , without falling to the center of the star . as table [ tab : models1 ] shows , the binding energy of the critical material is very much less than the kinetic energy of a typical supernova ( @xmath170 ) , so a weak explosion or even strong pulsations during silicon burning could eject it . it seems unlikely that this would happen all of the time in all sorts of models , so perhaps this enters in chiefly as a major uncertainty in the event rates . we also note that the converse problem , the optical appearance of a `` supernova '' exploding with only @xmath171 of kinetic energy is an interesting one too . one way or another , it seems unlikely that these stars will all disappear without a trace . the authors appreciate valuable discussions with christian ott regarding core - collapse in massive stars and with ron taam on tidal interaction in close binaries . this research has been supported by at ucsc by the national science foundation ( ast 0909129 ) and the nasa theory program ( nnx09ak36 g ) . ah acknowledges support from the doe program for scientific discovery through advanced computing ( scidac ; de - fc02 - 09er41618 ) , by the us department of energy under grant de - fg02 - 87er40328 , by nsf grant ast-1109394 , and by the joint institute for nuclear astrophysics ( jina ; nsf grant phy02 - 16783 ) . [ sec : append ] starting from the usual definition of free - fall time , @xmath172 we obtain an order - of - magnitude estimate of the accretion rate from @xmath173 @xmath174 @xmath175 where we used @xmath176 and @xmath177 , and @xmath178 is the mass coordinate and @xmath179 is the radius coordinate of a shell . note that usually @xmath180 in most parts of the star ; density inversions can exist at the outer edge of convective regions in mixing - length theory especially in rsgs due to recombination and opacity drops . if @xmath181 , as in the case of the outer layers of a bsgs , we obtain @xmath182 , so bsgs may have a more tenuous accretion rate and fainter , softer grbs than rsgs based on the estimate from @xmath183 . on the other hand the average density of the star is higher as it is more compact , which may be a slight compensation . density inversions at the edge of rsg envelopes could lead to spikes on accretion rate in this simple approximation . of model v24 and the outer @xmath11 of model v36 will form a disk . the edge of the helium cores of the two models is apparent in the sharp inflection in the angular momentum at @xmath28 and @xmath29 . mass loss was included in the calculation , but due to the low metallicity , only @xmath184 and @xmath185 was lost in v24 and v36 , respectively . note that , if all the surface material accreted here , the black hole would rotate at nearly its maximum allowed value ( i.e. , the red line intersects the green one at the end ) . [ fig : bsg],title="fig:",scaledwidth=47.5% ] of model v24 and the outer @xmath11 of model v36 will form a disk . the edge of the helium cores of the two models is apparent in the sharp inflection in the angular momentum at @xmath28 and @xmath29 . mass loss was included in the calculation , but due to the low metallicity , only @xmath184 and @xmath185 was lost in v24 and v36 , respectively . note that , if all the surface material accreted here , the black hole would rotate at nearly its maximum allowed value ( i.e. , the red line intersects the green one at the end ) . [ fig : bsg],title="fig:",scaledwidth=47.5% ] , the solid dark line shows the actual distribution of angular momentum while the lighter lines show the values required to form a disk around a non - rotating black hole , a kerr black hole , or a black hole with the integrated angular momentum interior to the given mass . the angular velocity in the envelope is constant with @xmath79 . note that the black hole formed here would be an extreme kerr hole with @xmath186 . [ fig : rsg],scaledwidth=90.0% ] ccccccc bsg & v24 & 10.0 & 0.4 - 10 @xmath187 & 20,000 & 5 & 9 + & v36 & 10.2 & 1 - 50 @xmath187 & 60,000 & 1 & 2 + rsg - loz & o25 & 0 . & 1.2 @xmath188 & 10@xmath189 & ... & ... + rsg - bin & s25 & 2.8 & 0.1 - 9 @xmath190 & @xmath191 & 0.01 & 0.02 + pair sn & z250a & 19.6 & 0.9 - 8 @xmath190 & @xmath192 & 0.7 & 1 + & z250b & 20.6 & 0.4 - 6 @xmath190 & 10@xmath193 & 2 & 4 + & z250c & 12.1 & 0.2 - 2 @xmath190 & @xmath194 & 4 & 8 + wr - bin & 8a & 0.18 & 2.5 - 5 @xmath195 & 100 & 20 & 40 + & 8b & 1.19 & 1 - 5 @xmath195 & 100 & 100 & 200 + & 16a & 0.01 & 3 - 4 @xmath195 & 100 & 1 & 2 + & 16b & 0.83 & 2 - 4 @xmath195 & 40 & 200 & 300 + [ tab : models ] ccccccccc bsg & v24 & 5.4 & 23.96 & 14.0 - 24 & 2.1 & 1 & 8.32 & 0.13 + & v36 & 3.7 & 35.85 & 25.7 - 35.9 & 1.2 & 1 & 14.97 & 0.11 + rsg - loz & o25 & 5.1 & 23.76 & ... & ... & 0.52 & 10.07 & 0.12 + rsg - bin & r25 & 4.6 & 12.09 & 9.3 - 12.1 & 0.0025 & 1 & 9.19 & 0.13 + pair sn & 250a & ... & 250 & 230.4 - 250 & 2.9 & 1 & 142 & 0.05 + & 250b & ... & 250 & 229.4 - 250 & 8.2 & 1 & 166 & 0.09 + & 250c & ... & 250 & 237.9 - 250 & 6.8 & 1 & 222 & 0.13 + wr - bin & 8a & 7.4 & 8.043 & 7.86 - 8.04 & 1.0 & ... & 8.04 & 0.25 + & 8b & 3.2 & 8.043 & 6.85 - 8.04 & 7.8 & ... & 8.04 & 1 + & 16a & 2.5 & 16.087 & 16.08 - 16.09 & 0.08 & ... & 16.08 & 0.18 + & 16b & 1.5 & 16.087 & 15.26 - 16.09 & 12 & ... & 16.08 & 0.48 + [ tab : models1 ]

detecting planetary transits is highly valuable not only because it makes an independent confirmation of a planet from radial - velocity observations but also because the photometric transits provide unambiguous information on planet masses and radii , and thus mean density and interior structure ( e.g. * ? ? ? * and references therein ) . various high - precision follow - up studies for transiting planets can also uncover planetary atmospheres and the ( mis-)alignment between the stellar spin and planetary orbital axes via the rossiter - mclaughlin effect ( e.g. * ? ? ? * ; * ? ? ? these properties of transiting planets provide valuable hints for planet formation and evolution processes including dynamical interaction between planets . although transiting planets are as such important , it is difficult to detect them around evolved stars , especially giant stars , due to the large sizes of the host stars . targets for on - going radial - velocity surveys for planets around giants have typical radii of @xmath110 @xmath6 ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and then the relative flux variation of such a giant host star caused by a transit of a jupiter - sized planet is only @xmath7 , which is comparable to that for transit of an earth - sized planet across a solar - type star . it is impossible to detect such a transit from ground - based photometry and thus no transiting planets had been found around giant stars . our understanding of properties of planets around such stars are thus far behind from those around solar - like stars , although tens of planets have been found around giants by precise radial - velocity surveys ( e.g. * ? ? ? * and references therein ) . the _ kepler _ mission was successful in detecting transiting planets with very high photometric precision ( @xmath8 ) since 2009 . to date it has discovered more than 4000 transiting planet candidates from sub - earth - size to super - jupiter - size in @xmath10.32000 d orbits ( e.g. * ? ? ? although most of them are orbiting solar - type stars , _ kepler _ s photometric precision is high enough to detect a jupiter - sized planet transiting a giant star . as expected , _ kepler _ has identified several planet candidates around giant stars with radii larger than @xmath9 @xcite . the planet candidates have radii comparable to or larger than jupiter s , and it is interesting that many of them have short - period orbits . since such short - period planets have rarely been found by radial - velocity surveys around evolved stars ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , they will provide us unique opportunities to investigate planet formation and evolution processes around such stars . here we report the results of radial - velocity follow - up observations using subaru 8.2 m telescope for two of the _ kepler _ evolved stars , kepler-91 ( koi-2133 ) and koi-1894 . thanks to the high precision in their radial - velocity measurements , we independently confirmed a jovian planet ( kepler-91b , koi-2133.01 ) previously reported around kepler-91 @xcite and newly found a hint for the existence of additional companion in the system . we also detected a possible sub - saturn - mass planet around koi-1894 ( koi-1894.01 ) with 23@xmath4 confidence level . the rest of the paper is organized as follows . the adopted stellar parameters for the two stars are presented in section [ targets ] . the observations are described in section [ obs ] and the results of global analysis for the radial velocities and light curves are presented in section [ analysis ] . section [ summary ] is devoted to discussion and summary . kepler-91 ( kic 8219268 , koi-2133 ; @xmath10=12.495 ) was identified as a candidate star hosting a transiting short - period jupiter - sized planet ( koi-2133.01 ; @xmath11 , @xmath12 d ) by @xcite . after that , @xcite reported the confirmation of its planetary nature based on the detailed analysis of orbital brightness modulation seen in the light curve caused by ellipsoidal variation , doppler boosting , and reflected / emitted light from planet ( e.g. * ? ? ? * ; * ? ? ? * ) , and then the planet was named kepler-91b with the radius @xmath13 and mass @xmath14 . although @xcite and @xcite claimed a possible non - planertary nature for the system , @xcite and @xcite very recently reconfirmed the planetary nature via radial - velocity measurements with a precision of @xmath1100 m s@xmath2 and @xmath120 m s@xmath2 , respectively , and obtained the planetary mass of 1.09@xmath150.20@xmath16 and 0.73@xmath150.13@xmath16 , respectively . the stellar parameters ( effective temperature @xmath17 , surface gravity @xmath18 , mass @xmath19 , and radius @xmath20 ) of kepler-91 were reported by @xcite to be @xmath21 k , @xmath22 cgs , @xmath23 , and @xmath24 . after that , the parameters have been updated by the spectroscopic and asteroseismic analyses in @xcite and @xcite , which are consistent with each other . thus , we here adopted the values listed in @xcite : @xmath25 k , @xmath26 cgs , @xmath27 , and @xmath28 . koi-1894 ( kic 11673802 ; @xmath10=13.427 ) was also reported to be a planet - host candidate having a transiting short - period jupiter - sized planet ( koi-1894.01 ; @xmath29 , @xmath30 d ) by @xcite , though radial - velocity follow - up observations for the star have not been reported yet . the stellar parameters of koi-1894 were derived by @xcite to be @xmath31 k , @xmath32 cgs , @xmath33 , and @xmath34 , and they have been updated by the spectroscopic and asteroseismic analyses in @xcite as @xmath35 k , @xmath36 , and @xmath37 . here we adopted the updated values in this paper . recently @xcite reported a non - detection of blended stars for koi-1894 , which could have been physically associated companions and/or responsible for transit false positives if they were within @xmath10.15@xmath382.5@xmath38 separation and with magnitude difference up to @xmath39 , by high - angular - resolution ao imaging . the parameters for the two stars we adopt here are summarized in table [ tbl : stars ] . lrr [ tbl : stars ] @xmath17 ( k ) & 4550@xmath1575 & 4992@xmath1575 + @xmath18 ( cgs ) & 2.953@xmath150.007 & 2.87 + @xmath19 ( @xmath40 ) & 1.31@xmath150.10 & 1.410@xmath150.214 + @xmath20 ( @xmath41 ) & 6.30@xmath150.16 & 3.790@xmath150.190 we obtained high - precision radial - velocity data for the stars with the 8.2 m subaru telescope and the high dispersion spectrograph ( hds ; * ? ? ? * ) in 2013 and 2014 . we used the setups of stdi2b ( 2013 june 28 - 30 and 2014 july 8 , 10 , 14 - 16 ) and stdi2a ( 2013 december 6 and 2014 july 12 - 13 ) , which simultaneously cover a wavelength region of 35006200@xmath42 and 49007600@xmath42 respectively , the image slicer no.2 ( is#2 ; * ? ? ? * ) yielding a spectral resolution ( @xmath43 ) of 80000 , and an iodine absorption cell ( i@xmath44 cell ; * ? ? ? * ) for precise radial - velocity measurements . we obtained a total of 29 and 18 data points for kepler-91 and koi-1894 with typical signal - to - noise ratio of s / n@xmath453080 pix@xmath2 and 3045 pix@xmath2 , respectively , by an exposure time of 1200 sec depending on weather condition . the reduction of echelle data ( i.e. bias subtraction , flat - fielding , scattered - light subtraction , and spectrum extraction ) was performed using the iraf software package in the standard manner . we performed radial - velocity analysis for i@xmath44-superposed stellar spectra ( star+i@xmath44 ) by the method described in @xcite and @xcite , which is based on the method by @xcite and @xcite . a star+i@xmath44 spectrum is modeled as a product of a high resolution i@xmath44 and a stellar template spectrum convolved with a modeled instrumental profile ( ip ) of the spectrograph . we obtain the stellar template spectrum by deconvolving a pure stellar spectrum with the ip estimated from an i@xmath44-superposed flat spectrum . we achieved a radial - velocity precision of 517 m s@xmath2 for kepler-91 and 915 m s@xmath2 for koi-1894 . the derived radial velocities are listed in table [ tbl : rv - koi2133 ] and table [ tbl : rv - koi1894 ] together with the estimated uncertainties , and are plotted in figure [ fig : rv_koi2133_notrend ] and figure [ fig : rv_koi1894 ] . lrr [ tbl : rv - koi2133 ] 6473.00247 & 43.20 & 4.61 + 6473.01719 & 33.85 & 4.58 + 6473.03192 & 45.26 & 5.03 + 6473.96075 & 17.35 & 7.36 + 6473.97549 & 8.93 & 7.34 + 6473.99022 & 30.04 & 6.39 + 6474.95764 & @xmath4624.28 & 5.43 + 6474.97235 & @xmath4622.62 & 5.03 + 6474.98708 & @xmath4639.46 & 5.50 + 6633.69753 & @xmath4642.12 & 17.38 + 6847.93547 & 42.30 & 10.56 + 6847.95020 & 37.46 & 13.63 + 6847.96513 & 11.49 & 8.76 + 6847.98006 & 29.77 & 7.96 + 6847.99479 & 17.07 & 6.12 + 6848.00952 & 19.94 & 8.61 + 6848.08121 & 23.25 & 6.59 + 6849.75412 & @xmath4652.12 & 5.23 + 6849.76886 & @xmath4636.10 & 5.06 + 6849.78360 & @xmath4634.47 & 5.48 + 6853.04854 & @xmath4640.23 & 13.48 + 6853.97273 & 5.73 & 5.72 + 6853.98746 & 9.02 & 4.73 + 6854.00219 & 7.66 & 5.17 + 6854.84234 & 2.14 & 4.99 + 6854.85708 & 0.36 & 5.25 + 6854.87181 & @xmath466.69 & 6.03 + 6855.80523 & @xmath4643.37 & 6.63 + 6855.81828 & @xmath4641.91 & 5.28 lrr [ tbl : rv - koi1894 ] 6473.04710 & @xmath469.95 & 9.45 + 6473.06182 & @xmath460.82 & 11.01 + 6473.07656 & @xmath4611.81 & 9.94 + 6474.00523 & @xmath465.31 & 10.97 + 6474.01995 & 1.95 & 10.25 + 6474.03468 & 17.33 & 10.36 + 6474.04941 & 5.73 & 10.49 + 6475.00210 & @xmath463.24 & 10.03 + 6475.01683 & @xmath4621.07 & 9.68 + 6475.03159 & @xmath4613.24 & 8.79 + 6475.04632 & @xmath466.92 & 9.62 + 6848.09625 & @xmath464.14 & 15.02 + 6849.94269 & 0.52 & 10.34 + 6849.95743 & 13.69 & 9.86 + 6849.97216 & 10.09 & 9.04 + 6855.88939 & 0.65 & 9.63 + 6855.90411 & @xmath463.06 & 8.95 + 6855.91885 & 6.48 & 9.09 the variations in the observed radial velocities for both targets show a sign of planetary companions to kepler-91 and koi-1894 . in order to obtain accurate and precise estimates for system parameters of those systems , we here present a global analysis that makes use of all the available information from the _ kepler _ photometry and our spectroscopy . as is well known , a very precise light curve of a star orbited by planet(s ) shows a periodic modulation due to several astrophysical effects : the ellipsoidal variation , doppler boosting , and reflected / emission light from the planet ( e.g. , * ? ? ? * ; * ? ? ? the last two effects ( doppler boosting and planetary reflection / emission ) are synchronous with the planet s orbital period @xmath47 , with different peak locations along the planet phase ( @xmath48 for boosting and @xmath49 for the planetary reflection / emission , respectively ) . on the other hand , the ellipsoidal variation , which is caused by a tidal distortion by the planet s gravity , have two flux peaks ( @xmath48 and 0.75 ) within one orbit . thanks to the different phase dependence of these three effects , a very precise light curve and its modeling enable us to distinguish these three effects , and we can extract physical parameters such as the planet - to - star mass ratio @xmath50 and scaled semi - major axis @xmath51 , where @xmath52 is semi - major axis and @xmath20 is stellar radius . moreover , for transiting systems as in the present cases , incorporating the transit and/or secondary eclipse model into the above phase - curve variation lets us learn more about the planet properties ( e.g. , planet - to - star radius ratio @xmath53 and orbital inclination @xmath54 ) . to obtain the phase - folded light curves for kepler-91 and koi-1894 , and estimate system parameters , we downloaded all available public light curves ( q0q17 ) from _ kepler _ mast archive . while only long - cadence data were available for kepler-91 , koi-1894 s light curves involve some short - cadence data . adopting the pdc - sap flux data , for which unphysical artifacts are detrended , we reduced the light curves by the following procedure . first , after removing planetary transits , we further detrended and normalized the light curve for each quarter by fitting it with a fifth - order polynomial so as to remove the long - term trends that were not removed in the pdc - sap flux . this process was repeated implementing a @xmath55 clipping to remove outliers . we then combined the light curves for all quarters and phase - folded them with the ephemerides derived by the official _ kepler _ team using the q1q16 data ; orbital period @xmath56 d and transit center @xmath57 bjd for kepler-91 , and @xmath58 d and @xmath59 bjd for koi-1894 ( from nasa exoplanet archive ) . we later consider the impact of incorrect ephemeris . finally , the folded light curves were binned into 300 and 250 phase bins for kepler-91 and koi-1894 , respectively . these bin numbers were adopted so that each bin approximately covers a cycle span of long - cadence data ( @xmath60 minutes ) . the flux error for each bin @xmath61 , which is the standard deviation of the mean flux , was computed based on the dispersion of the flux values within the bin . the long - cadence and short - cadence flux data were separately folded and binned for koi-1894 , but the binned light curve of the short - cadence data was much noisier than that for long - cadence . we thus decided to ignore the short - cadence data in the following analysis . the analysis below is based on the method by @xcite , who employed the evil - mc model @xcite for the phase - curve variation ( i.e. , ellipsoidal variations , doppler boosting , and planetary reflection ) with some revisions ( e.g. , application to an eccentric orbit ) . this phase - curve model is multiplied by the analytic transit model by @xcite , and the relevant parameters ( e.g. , @xmath51 and @xmath54 ) are simultaneously determined . in addition , we also model and fit the observed radial velocities @xmath62 . thus , the @xmath63 statistics in the present case is @xmath64 where @xmath65 and @xmath66 are the @xmath67-th observed radial velocity and its error , and @xmath68 and @xmath69 are the @xmath67-th observed light curve flux and its error , which are the mean flux and its standard deviation in the @xmath67-th bin , respectively ( e.g. , * ? ? ? * ) . the radial velocity is modeled as @xmath70 where @xmath71 , @xmath72 , @xmath73 , @xmath74 , and @xmath75 are the radial velocity semi - amplitude , true anomaly , orbital eccentricity , argument of periastron , and radial velocity offset of our dataset , respectively . for the light curve model @xmath76 , we refer the readers to @xcite , for details . in the next subsections , we describe the fitting procedure for each of the targets . the binned light curve of kepler-91 in figure [ fig : phase_koi2133 ] shows a clear pattern of phase - curve variation ; the double peaks at @xmath48 and 0.75 are representative of the ellipsoidal variation . in addition , figure [ fig : phase_koi2133 ] suggests a possible detection of secondary eclipse at @xmath77 , which means the reflected / emitted light from the planet is visible in the folded light curve . therefore , following @xcite and @xcite , we model the planet s light @xmath78 by the following expression : @xmath79 where @xmath80 is the flux variation amplitude of planet s reflected / emitted light and @xmath81 is the planetary flux offset arising from the homogeneous surface emission . following @xcite , @xcite , and @xcite , we here introduce the phase - delay " @xmath82 of the brightest part on the planet surface from the substellar point . the planetary light @xmath78 is added to the flux model for the beaming and ellipsoidal variation , integrated over the stellar visible hemisphere and wavelength through the _ kepler _ band . in integrating the local flux calculated from the evil - mc model , we assume that the stellar effective temperature is 4550 k , and adopt the gravity darkening exponent of @xmath83 for equation ( 10 ) of @xcite based on the theoretical calculation by @xcite . the free parameters relevant to the light curve model are @xmath51 , the transit impact parameter @xmath84 , limb - darkening coefficients @xmath85 and @xmath86 for the quadratic limb - darkening law , @xmath50 ( @xmath87 ) , @xmath81 , @xmath80 , @xmath82 , the overall normalization factor @xmath88 for the folded light curve , @xmath53 , @xmath89 , @xmath90 , and @xmath91 , which represents the small time deviation of the transit center from the ephemeris reported by the _ kepler _ team . in our updated ephemeris , the initial transit center becomes @xmath92 . we fix the orbital period to be @xmath93 d that are derived by the _ kepler _ team based on the q1q16 data ( see section [ method ] ) . assuming that the likelihood is proportional to @xmath94 in equation ( [ eq : chisq ] ) , we simultaneously model the observed radial velocity and light curve , and compute the posterior distribution for each fitting parameter by implementing markov chain monte carlo ( mcmc ) simulation . in addition to the above mentioned twelve parameters , we add @xmath75 in equation ( [ eq : rvmodel ] ) to the fitting parameters . note that the radial velocity semi - amplitude @xmath71 in equation ( [ eq : rvmodel ] ) is related to the planet - to - star mass ratio @xmath50 by @xmath95 in computing the posteriors , we do not impose priors on the fitting parameters except for @xmath96 and @xmath97 ; due to the sparse sampling of the long - cadence data ( @xmath60 minutes ) and quality of the binned light curve , the limb - darkening coefficients are poorly constrained in the absence of priors , and thus we decide to put gaussian priors on the limb - darkening coefficients based on the theoretical table by @xcite as @xmath98 and @xmath99 . in our mcmc algorithm , originally developed in @xcite , the step size of each fitting parameter is iteratively scaled so that the overall acceptance ratio falls between 15% and 35% . after running 1,000,000 chains , the best - fit value and uncertainty for each fitting parameter are estimated from the median , and 15.87 and 84.13 percentiles of the marginalized posterior distribution of that parameter . shortly after we performed the first mcmc trial , we noticed a possible trend or drift in the observed radial velocities . figure [ fig : rv_koi2133_notrend ] plots the best - fit radial velocity curve as a function of the orbital phase @xmath100 . for clarity , the data taken in 2013 are shown in green and those by the 2014 campaign are plotted in red . while we have detected a clear modulation with an amplitude of @xmath101 m s@xmath2 , radial velocities take lower values for the 2014 data , which is evident in the bottom panel indicating the residual of the observed radial velocities from the best - fit model . in order to confirm the presence of the radial velocity drift , we try to fit the data with an additional parameter : the radial velocity drift @xmath102 . adding the drift term @xmath103 to the right side of equation ( [ eq : rvmodel ] ) , we performed again the mcmc simulation to fit both radial velocity and light curve . the derived best - fit parameters and their uncertainties are summarized in table [ tbl : best - fit - kepler91 ] . to compare between the radial velocity models with and without a drift term , we compute bayesian information criteria ( bic ) for the two cases , which is computed as @xmath104 , where @xmath105 is the number of free parameters and @xmath106 is the number of data points . in the absence of a trend , we obtain @xmath107 , and for the case of the radial velocity model with @xmath102 , bic becomes 383 ; @xmath108 is much larger than 10 , meaning that the model with @xmath102 is strongly favored . therefore , we conclude that a radial velocity drift ( or trend ) is present in our dataset , and report the best - fit parameters for the case with @xmath102 as the final result . we also tried several periods around @xmath47 and found that the @xmath109 gave slightly better results in the global fitting than @xmath47 did ( @xmath110 ) . however , the resultant parameters for the two cases are well consistent within 0.2@xmath4 level . differs by @xmath111 level . ] figures [ fig : phase_koi2133 ] and [ fig : rv_koi2133 ] plot the folded light curve and radial velocities ( red points ) along with their best - fit models ( blue line ) . the constant radial velocity offset and trend are both removed from the radial velocity data in figure [ fig : rv_koi2133 ] . the bottom panel in each figure shows the residuals from the best - fit model . from the posterior distribution of the fitting parameters , we also estimate the orbital and planetary parameters ( e.g. , the orbital inclination @xmath54 , planet mass @xmath112 and radius @xmath113 ) assuming the stellar properties reported by asteroseismology ( table [ tbl : stars ] ) to be @xmath114 deg , @xmath115 , and @xmath116 . these estimates are also shown in table [ tbl : best - fit - kepler91 ] . lrr [ tbl : best - fit - kepler91 ] @xmath51 & 2.253@xmath117 & 2.238@xmath118 + @xmath84 & 0.9041@xmath119 & 0.9062@xmath120 + @xmath85 & 0.737@xmath121 & 0.7370@xmath122 + @xmath86 & 0.493@xmath123 & 0.493@xmath124 + @xmath125 ( 10@xmath126 ) & 4.82@xmath127 & 4.72@xmath128 + @xmath81 ( 10@xmath129 ) & 5.65@xmath130 & 5.63@xmath131 + @xmath80 ( 10@xmath129 ) & 1.41@xmath132 & 1.42@xmath133 + @xmath82 & 0.394@xmath134 & 0.395@xmath135 + @xmath53 & 0.02286@xmath136 & 0.02294@xmath137 + @xmath89 & 0.0280@xmath138 & 0.0277@xmath139 + @xmath90 & @xmath460.043@xmath140 & @xmath460.046@xmath140 + @xmath91 ( 10@xmath141 day ) & @xmath461.09@xmath142 & @xmath461.11@xmath142 + @xmath75 ( m s@xmath2 ) & @xmath4642.5@xmath143 & @xmath4633.2@xmath144 + @xmath102 ( m s@xmath2 day@xmath2 ) & @xmath460.0612@xmath145 & + @xmath63 & 296 & 396 + bic & 383 & 477 + @xmath54 ( deg ) & 67.37@xmath146 & 67.20@xmath147 + @xmath73 & 0.0519@xmath148 & 0.0535@xmath149 + @xmath74 ( deg ) & @xmath4657.2@xmath150 & @xmath4658.8@xmath151 + @xmath112 ( @xmath16 ) & 0.66@xmath150.06 & 0.65@xmath150.06 + @xmath113 ( @xmath152 ) & 1.40@xmath150.04 & 1.41@xmath150.04 the radial - velocity variations are less visible for koi-1894 owing to the large radial - velocity error compared to the small semi - amplitude . the phase - folded light curve also shows a very tiny variation , if any , along the orbital phase . to extract possible planetary signals , we simultaneously model the radial velocities and light curve as in the case of kepler-91 . since the observed transit depth and koi-1894 s stellar radius suggest that the radius of the transiting companion ( koi-1894.01 ) is no more than @xmath153 , the reflected / emitted light from the planet is expected to be very small . according to equations ( 1 ) ( 3 ) in @xcite , the planetary reflection is estimated to be @xmath154 ppm at the location of koi-1894.01 , which is smaller than the expected amplitude of the ellipsoidal variation ( @xmath155 ppm ) . visual inspection of the binned light curve also suggests the absence of secondary eclipse . thus for simplicity , we here neglect the planet flux @xmath78 for koi-1894 , and model the folded light curve with the ellipsoidal variation ( including doppler boosting ) and transit only . neglecting @xmath78 also helps to avoid @xmath63 being stuck at a local minimum during the optimization . fixing the system parameters as @xmath156 k , @xmath157 , the gravity darkening exponent of @xmath158 @xcite , and @xmath159 derived with the q1q16 data ( see section [ method ] ) , we perform mcmc simulations for koi-1894 to infer the posterior distributions for the fitting parameters . in the present case , we have the eleven free parameters : @xmath51 , @xmath84 , @xmath85 , @xmath86 , @xmath50 , @xmath88 , @xmath53 , @xmath89 , @xmath90 , @xmath91 , and @xmath75 . again , we assume gaussian priors for the limb - darkening parameters as @xmath160 , and @xmath161 from the table by @xcite . due to the weak radial - velocity signal of koi-1894.01 , we found that the global fit to the radial velocities and light curve exhibits a degeneracy between the fitting parameters ( e.g. , @xmath162 and @xmath50 ) . thus , for koi-1894 , we also impose an additional prior on the host star s density from table [ tbl : stars ] ( @xmath163 ) , assuming a gaussian distribution @xcite . otherwise , the fitting algorithm is exactly the same as for kepler-91 . the result of the fit is summarized in table [ tbl : best - fit - koi1894 ] . the best - fit value for the mass ratio is @xmath164 , implying @xmath165 detection of the planet . the mass and radius ratios are translated as @xmath166 and @xmath167 assuming the stellar mass and radius in table [ tbl : stars ] . table [ tbl : best - fit - koi1894 ] also shows the result of our fit in the absence of the gaussian prior on @xmath168 . as expected , the mass constraint is slightly weaker for this case . the orbital eccentricity is poorly constrained due to the small planetary signal and small number of radial velocity data , but we can rule out a high eccentricity ( @xmath169 ) with our current datasets . we also try to fit the data with a circular orbit ( @xmath170 ) , and compare between the eccentric and circular cases . as shown in table [ tbl : best - fit - koi1894 ] , the circular orbital fit yields a slightly better constraint on the mass ratio ; @xmath171 with the stellar - density prior , leading to a @xmath172 detection . the mass and radius ratios are translated as @xmath173 and @xmath174 assuming the stellar mass and radius . based on the @xmath63 values for the best - fit parameters , we compare bic values for the two cases ; bic values of @xmath175 and @xmath176 are obtained for the eccentric and circular cases ( with the stellar - density prior ) , respectively , so that fitting with a circular orbit is favored . the observed radial velocities and whole light curve are plotted in figures [ fig : rv_koi1894 ] and [ fig : phase_koi1894 ] , with their best - fit models for @xmath170 . we also show zoomed - in versions of the transit and phase - curve variation in figures [ fig : transit_koi1894 ] and [ fig : ev_koi1894 ] , respectively . in figure [ fig : ev_koi1894 ] , the binned flux data used for the fit ( 250 bins ) are plotted by the red crosses , and black points with errorbars indicate the flux data binned into 30 bins . we also tested several periods around @xmath47 and found that about @xmath177 gave the minimum @xmath63 value in the global fitting ( @xmath178 ) , suggesting that the true period might exist around it . nonetheless , the resultant parameters are well consistent with those for the case of @xmath47 within 0.2@xmath4 level . differs by @xmath179 level . ] our global analysis indicates the detection of koi-1894.01 is still marginal , with only @xmath180 level , and the planet mass seems to be mainly constrained by the radial velocity data . in order to see if an independent estimate for the planet mass from the light curve alone gives a comparable result , we fit the folded light curve and estimate system parameters without radial velocity data . as a result , we obtain @xmath181 , consistent with the above result by the global fit . we reported the results of high - precision radial - velocity measurements with subaru / hds for the two _ kepler _ evolved stars kepler-91 and koi-1894 , which are candidates to host transiting planets . based on the simultaneous modeling of the radial - velocity data and _ kepler _ light curves , we independently confirmed the planetary nature of kepler-91b . the radial velocity semi - amplitude of @xmath182 m s@xmath2 we derived is consistent with what @xcite found ( @xmath183 m s@xmath2 ) with @xmath184 , but incompatible with the result by @xcite who reported @xmath185 m s@xmath2 with @xmath186 , although the reason for this discrepancy is unknown . we newly detected a drift of @xmath187 m s@xmath2 between the radial velocities taken at @xmath1 1-yr interval thanks to our better measurement precision and longer period of time of observations compared to the previous ones . the persistent radial velocity drift is suggestive of a possible presence of another companion ( planet ) outside of kepler-91b . due to the lack of data , however , we are not able to pin down the period nor mass of the companion . intensive high - precision radial - velocity monitoring of the star will uncover the unseen companion and provide hints for formation and evolution of the planetary system . the estimated parameters for kepler-91b in table [ tbl : best - fit - kepler91 ] are reasonably in good agreement with the previous study by @xcite except for the mass ratio @xmath50 and scaled semi - major axis @xmath51 . this is likely because our radial velocity data show a smaller amplitude and planet mass is estimated to be small . since the flux amplitude due to ellipsoidal variation is approximately proportional to @xmath188 @xcite , a small @xmath50 resulted in the smaller @xmath51 in order to explain the observed flux amplitude . the westward phase - shift ( @xmath189 ) of the flux maximum in planetary light ( from the substellar point ) is indicative of the inhomogeneous cloud coverage on the planetary surface . this is consistent with the trend that @xcite found , claiming that westward phase - shifts are preferentially seen for relatively cool close - in planets with equilibrium temperatures of @xmath190 k ; @xcite estimated the equilibrium temperature of kepler-91b to be @xmath191 k , depending on the assumed heat redistribution parameter . the magnitude of the phase - shift ( @xmath192 ) is , however , unusually large in our best - fit model compared with the previously reported values for other close - in planets @xcite . the reason is unknown , but the huge coverage of the planet surface illuminated by the host star of the planet surface is always illuminated by the host star due to its huge radius . ] might be responsible . note that we also detected a dip in the folded light curve ( figure [ fig : phase_koi2133 ] ) around @xmath193 reported in @xcite , and this dip should have more or less affected the fitting result . lrrrr [ tbl : best - fit - koi1894 ] @xmath51 & 3.79@xmath194 & 3.68@xmath195 & 4.25@xmath196 & 3.24@xmath197 + @xmath84 & 0.634@xmath198 & 0.56@xmath199 & 0.48@xmath200 & 0.68@xmath201 + @xmath85 & 0.724@xmath124 & 0.723@xmath202 & 0.724@xmath123 & 0.7271@xmath203 + @xmath86 & 0.391@xmath204 & 0.3908@xmath205 & 0.392@xmath206 & 0.391@xmath207 + @xmath125 ( 10@xmath126 ) & 1.19@xmath208 & 1.02@xmath209 & 1.29@xmath210 & 0.70@xmath211 + @xmath81 ( 10@xmath129 ) & 0 ( fixed ) & 0 ( fixed ) & 0 ( fixed ) & 0 ( fixed ) + @xmath80 ( 10@xmath129 ) & 0 ( fixed ) & 0 ( fixed ) & 0 ( fixed ) & 0 ( fixed ) + @xmath82 & 0 ( fixed ) & 0 ( fixed ) & 0 ( fixed ) & 0 ( fixed ) + @xmath53 & 0.01777@xmath212 & 0.01739@xmath213 & 0.01717@xmath214 & 0.01790@xmath215 + @xmath89 & 0 ( fixed ) & 0.052@xmath216 & 0 ( fixed ) & 0.047@xmath217 + @xmath90 & 0 ( fixed ) & 0.085@xmath218 & 0 ( fixed ) & 0.10@xmath219 + @xmath91 ( 10@xmath141 day ) & @xmath460.25@xmath220 & @xmath460.53@xmath221 & @xmath460.24@xmath222 & @xmath460.57@xmath223 + @xmath75 ( m s@xmath2 ) & @xmath4610.6@xmath224 & @xmath4610.0@xmath225 & @xmath4611.3@xmath226 & @xmath466.8@xmath227 + @xmath63 & 234 & 233 & 235 & 233 + bic & 285 & 295 & 286 & 294 + @xmath54 ( deg ) & 80.4@xmath228 & 80.2@xmath229 & 83.5@xmath230 & 76.2@xmath231 + @xmath73 & 0 ( fixed ) & 0.149@xmath232 & 0 ( fixed ) & 0.186@xmath233 + @xmath74 ( deg ) & 0 ( fixed ) & 61@xmath234 & 0 ( fixed ) & 65@xmath235 + @xmath112 ( @xmath16 ) & 0.18@xmath150.07 & 0.15@xmath236 & 0.19@xmath150.08 & 0.10@xmath237 + @xmath113 ( @xmath152 ) & 0.66@xmath150.04 & 0.64@xmath238 & 0.63@xmath239 & 0.66@xmath240 as for koi-1894 , we did not detect any statistically significant radial - velocity variations with our measurement precision of 915 m s@xmath2 and the number of data points . we excluded the possibility of a grazing transit by a binary companion for the star and set an upper limit on the mass of koi-1894.01 to be @xmath241 by our non - detection of radial - velocity variations with 3@xmath4 level . however , we detected possible radial - velocity variations with a semi - amplitude of @xmath115 m s@xmath2 in phase with ellipsoidal variations of the star with 23@xmath4 level . although we can not say the detection is statistically significant at this stage , it suggests that the koi-1894.01 could be a sub - saturn - mass planet . figure [ fig : plm - strad ] shows distribution of mass of exoplanets currently known plotted against their host star s radius . as seen in the figure , koi-1894.01 could be one of the lowest mass planets ever discovered around evolved stars together with kepler-56b , a super - neptune - mass planet ( @xmath242 ) detected via ttv ( transit timing variation ) method @xcite . actually the stellar parameters for koi-1891 are similar to those of kepler-56 ( * ? ? ? * @xmath243 , @xmath244 ) . confirmation of koi-1894.01 is highly encouraged in order to uncover such a new population of sub - saturn and super - neptune planets around relatively massive evolved stars , which have rarely been found so far either by radial - velocity surveys or transit ones . this research is based on data collected at subaru telescope , which is operated by national astronomical observatory of japan ( naoj ) . we are grateful to all the staff members of subaru for their support during the observations . this research has made use of the simbad database , operated at cds , strasbourg , france . is supported by japan society for promotion of science ( jsps ) fellowship for research ( pd:25 - 3183 ) . is supported by jsps research fellowships for young scientists ( no . 26 - 7182 ) and by the leading graduate course for frontiers of mathematical sciences and physics . barclay , t. et al . 2014 , , submitted ( arxiv:1408.3149 ) batalha , n.m . , et al . 2013 , , 204 , 24 butler , r. p. , marcy , g. w. , williams , e. , mccarthy , c. , dosanjh , p. , & vogt , s. s. 1996 , , 108 , 500 claret , a. 1998 , a&as , 131 , 395 claret , a. , & bloemen , s. 2011 , , 529 , a75 da silva , l. , et al . 2006 , , 458 , 609 demory , b .- o . , et al . 2013 , , 776 , l25 esteves , l.j . , de mooij , e.j.w . , & jayawardhana , r. 2013 , , 772 , 51 esteves , l. j. , de mooij , e. j. w. , & jayawardhana , r. 2014 , , submitted ( arxiv:1407.2245 ) faigler , s. , & mazeh , t. 2011 , mnras , 415 , 3921 faigler , s. , & mazeh , t. 2014 , , submitted ( arxiv:1407.2361 ) hirano , t. , narita , n. , shporer , a. , sato , b. , aoki , w. , & tamura , m. 2011 , , 63 , 531 hirano , t. , et al . 2012 , , 759 , l36 hirano , t. , et al . 2015 , , 799 , 9 huber , d. et al . 2013a , , 767 , 127 huber , d. et al . 2013b , science , 342 , 331 jackson , b. k. , lewis , n. k. , barnes , j. w. , drake deming , l. , showman , a. p. , & fortney , j. j. 2012 , , 751 , 112 johnson , j.a . , et al . 2007 , , 665 , 785 jones , m.i . , jenkins , j.s . , bluhm , p. , rojo , p. , & melo , c.h.f . 2014 , , 566 , 113 kambe , e. , et al . 2002 , , 54 , 865 law , n.m . 2014 , , 791 , 35 lillo - box , j. et al . 2014a , , 562 , 109 lillo - box , j. et al . 2014b , , 568 , 1 liu , y. , sato , b. , takeda , y. , ando , h. , & zhao , g. 2010 , , 62 , 1071 mazeh , t. , nachmani , g. , sokol , g. , faigler , s. , & zucker , s. 2012 , , 541 , a56 noguchi , k. et al . 2002 , , 54 , 855 ohta , y. , taruya , a. , & suto , y. 2009 , , 690 , 1 sato , b. , kambe , e. , takeda , y. , izumiura , h. , & ando , h. 2002 , , 54 , 873 sato , b. , et al . 2008 , , 60 , 539 sato , b. , et al . 2012 , , 64 , 97 sato , b. , et al . 2013 , , 65 , 85 seager , s. & malln - ornelas , g. 2003 , , 585 , 1038 seager , s. & deming , d. 2010 , , 48 , 631 shporer , a. , et al . 2011 , , 142 , 195 sliski , d. h. & kipping , d.m . 2014 , , 788 , 148 tajitsu , a. , aoki , w , & yamamuro , t , 2012 , , 64 , 77 takeda , y. , sato , b. , & murata , d. , 2008 , , 60 , 781 valenti , j. a. , butler , r. p. & marcy , g. w. 1995 , , 107 , 966 . wang , l. , liu , y. , zhao , g. , & sato , b. 2011 , , 63 , 1035 winn , j. n. 2011 , in exoplanets , ed . s. seager ( tucson , az : university of arizona press ) , p.5577 zielinski , p. , niedzielski , a. , wolszczan , a. , admow , m. , & nowak , g. 2012 , , 547 , 91

in the theory of critical phenomena it is usually assumed that a physical system near a continuous phase transition is characterized by a single length scale ( the `` correlation length '' ) in terms of which all other lengths should be measured . when combined with the experimental observation that the correlation length diverges at the phase transition , this simple but strong assumption , known as the scaling hypothesis , leads to the belief that at criticality the system has no characteristic length , and is therefore invariant under scale transformations . this suggests that all thermodynamic functions at criticality are homogeneous functions , and predicts the appearance of power laws . it also means that it should be possible to rescale a critical system appropriately and obtain a continuum model ( the `` continuum scaling limit '' ) which may have more symmetries and be easier to study than the original discrete model defined on a lattice . indeed , thanks to the work of polyakov @xcite and others @xcite , it was understood by physicists since the early seventies that critical statistical mechanical models should possess continuum scaling limits with a global conformal invariance that goes beyond simple scale invariance , as long as the discrete models have `` enough '' rotation invariance . this property gives important information , enabling the determination of two- and three - point functions at criticality , when they are nonvanishing . because the conformal group is in general a finite dimensional lie group , the resulting constraints are limited in number ; however , the situation becomes particularly interesting in two dimensions , since there every analytic function @xmath1 defines a conformal transformation , at least at points where @xmath2 . as a consequence , the conformal group in two dimensions is infinite - dimensional . after this observation was made , a large number of critical problems in two dimensions were analyzed using conformal methods , which were applied , among others , to ising and potts models , brownian motion , self - avoiding walk ( saw ) , percolation , and diffusion limited aggregation ( dla ) . the large body of knowledge and techniques that resulted , starting with the work of belavin , polyakov and zamolodchikov @xcite in the early eighties , goes under the name of conformal field theory ( cft ) . in two dimensions , one of the main goals of cft and its most important application to statistical mechanics is a complete classification of all universality classes via irreducible representations of the infinite - dimensional virasoro algebra . partly because of the success of cft , work in recent years on critical phenomena seemed to slow down somewhat , probably due to the feeling that most of the leading problems had been resolved nonetheless , however powerful and successful it may be , cft has some limitations and leaves various open problems . first of all , the theory deals primarily with correlation functions of _ local _ ( or quasi - local ) operators , and is therefore not always the best tool to investigate other quantities . secondly , given some critical lattice model , there is no way , within the theory itself , of deciding to which cft it corresponds . a third limitation , of a different nature , is due to the fact that the methods of cft , although very powerful , are generally speaking not completely rigorous from a mathematical point of view . in a somewhat surprising twist , the most recent developments in the area of two - dimensional critical phenomena have emerged in the mathematics literature and have followed a new direction , which has provided new tools and a way of coping with at least some of the limitations of cft . the new approach may even provide a reinterpretation of cft , and seems to be complementary to the traditional one in the sense that questions that are difficult to pose and/or answer within cft are easy and natural in this new approach and vice versa . these new developments came on the heels of interesting results on the scaling limits of discrete models ( see , e.g. , the work of aizenman @xcite , benjamini - schramm @xcite , aizenman - burchard @xcite , aizenman - burchard - newman - wilson @xcite , aizenman - duplantier - aharony @xcite and kenyon @xcite ) but they differ greatly from those because they are based on a radically new approach whose main tool is the stochastic loewner evolution ( @xmath3 ) , or schramm - loewner evolution , as it is also known , introduced by schramm @xcite . the new approach , which is probabilistic in nature , focuses directly on non - local structures that characterize a given system , such as cluster boundaries in ising , potts and percolation models , or loops in the @xmath4 model . at criticality , these non - local objects become , in the continuum limit , random curves whose distributions can be uniquely identified thanks to their conformal invariance and a certain markovian " property . there is a one - parameter family of @xmath3s , indexed by a positive real number @xmath5 , and they appear to be the only possible candidates for the scaling limits of interfaces of two - dimensional critical systems that are believed to be conformally invariant . in particular , substantial progress has been made in recent years , thanks to @xmath3 , in understanding the fractal and conformally invariant nature of ( the scaling limit of ) large percolation clusters , which has attracted much attention and is of interest both for intrinsic reasons , given the many applications of percolation , and as a paradigm for the behavior of other systems . the work of schramm @xcite and smirnov @xcite has identified the scaling limit of a certain percolation interface with @xmath0 , providing , along with the work of lawler - schramm - werner @xcite and smirnov - werner @xcite , a confirmation of many results in the physics literature , as well as some new results . however , @xmath0 describes a single interface , which can be obtained by imposing special boundary conditions , and is not in itself sufficient to immediately describe the scaling limit of the unconstrained model ( without boundary conditions ) in the whole plane . in particular , not only the nature and properties , but the very existence of the scaling limit of the collection of all interfaces remained an open question . this is true of all models , such as ising and potts models , that are represented in terms of clusters , and where the set of all interfaces forms a collection of loops . as already indicated by smirnov @xcite , such a collection of loops should have a continuum limit , that we will call the full " scaling limit of the model . the single interface limit is ideal for analyzing certain crossing / connectivity probabilities but not so good for others ; in section [ quickresults ] we give a few examples showing the use of the full scaling limit to represent such probabilities . in the context of percolation , in @xcite the authors used @xmath0 to construct a random process of continuous loops in the plane , which was identified with the full scaling limit of critical two - dimensional percolation , but without detailed proofs . ( for a discussion of whether this full scaling limit is a black noise , " see @xcite . for an analysis of random processes of loops related to @xmath6 for other values of @xmath5 , and _ conjectured _ to correspond to the full scaling limits of other statistical mechanics models , see @xcite . ) in this paper , we complete the analysis of @xcite , making rigorous the connection between the construction given there and the full scaling limit of percolation , and we prove some properties of the full scaling limit , the continuum nonsimple loop process , including ( one version of ) conformal invariance . the present work , as well as that of smirnov @xcite , builds on a collection of papers , including @xcite , which provided both inspiration and essential technical results . the proofs are based on the fact that the percolation exploration path converges in distribution to the trace of chordal @xmath0 , as argued by schramm and smirnov @xcite , and in particular on a specific version of this convergence that we will call statement ( s ) ( see section [ main - results ] ) . we note that no detailed proof of any version of convergence to @xmath0 has been available . nevertheless , at the request of the editor , we do not include a detailed proof of statement ( s ) in the present paper , as originally planned @xcite , due to length considerations . however , a detailed proof of statement ( s ) , based on smirnov s theorem about the convergence of crossing probabilities to cardy s formula @xcite , is now the topic of a separate paper @xcite . we note that statement ( s ) is restricted to jordan domains while no such restriction is indicated in @xcite . the rest of the paper is organized as follows . in section [ quickresults ] , we provide a quick presentation of theorems [ main - result1 ] , [ features ] , and [ main - result2 ] , which represent most of our main results , with some definitions postponed until section [ defs ] , including that of @xmath0 . section [ construct ] is devoted to the construction mentioned in theorem [ main - result2 ] of the continuum nonsimple loop process in a finite region @xmath7 of the plane . in section [ lapa ] , we introduce the discrete lattice model and a discrete construction analogous to the continuum one presented in section [ construct ] . most of the main technical results of this paper are stated in section [ main - results ] , while section [ proofs ] contains the proofs , using ( s ) , of those results and the results in section [ quickresults ] . at the percolation critical point , with probability one there is no infinite cluster ( in two dimensions ) , therefore the percolation cluster boundaries form loops ( see figure [ loops ] , where site percolation on the triangular lattice @xmath8 is depicted exploiting the duality between @xmath8 and the hexagonal lattice @xmath9 ) . we will refer to the continuum scaling limit ( as the mesh size @xmath10 of the rescaled hexagonal lattice @xmath11 goes to zero ) of the collection of all these loops as the continuum nonsimple loop process . its existence is the content of theorem [ main - result1 ] and some of its properties are described in theorem [ features ] below . . each hexagon of the hexagonal lattice @xmath9 represents a site of @xmath8 and is assigned one of two colors . in the critical percolation model , colors are assigned randomly with equal probability . the cluster boundaries are indicated by heavy lines ; some small loops appear , while other boundaries extend beyond the finite window . , width=302 ] the continuum nonsimple loop process can be described as a gas " of loops , or more precisely , a probability measure on countable collections of continuous , nonsimple , fractal loops in the plane . later in this paper , we will provide precise definitions of the objects involved in the next three theorems as well as detailed proofs . [ main - result1 ] in the continuum scaling limit , the probability distribution of the collection of all boundary contours of critical site percolation on the triangular lattice converges to a probability distribution on collections of continuous , nonsimple loops . [ features ] the continuum nonsimple loop process whose distribution is specified in theorem [ main - result1 ] has the following properties , which are valid with probability one : 1 . it is a random collection of countably many noncrossing continuous loops in the plane . the loops can and do touch themselves and each other many times , but there are no triple points ; i.e. no three or more loops can come together at the same point , and a single loop can not touch the same point more than twice , nor can a loop touch a point where another loop touches itself . any deterministic point @xmath12 in the plane ( i.e. , chosen independently of the loop process ) is surrounded by an infinite family of nested loops with diameters going to both zero and infinity ; any annulus about that point with inner radius @xmath13 and outer radius @xmath14 contains only a finite number @xmath15 of those loops . consequently , any two distinct deterministic points of the plane are separated by loops winding around each of them . any two loops are connected by a finite `` path '' of touching loops . the next theorem makes explicit the relation between the percolation full scaling limit and @xmath0 . its proof ( see section [ proofs ] ) relies on an inductive procedure that makes use of @xmath0 at each step and allows to obtain collections of loops with the correct distribution . [ main - result2 ] a continuum nonsimple loop process with the same distribution as in theorem [ main - result1 ] can be constructed by a procedure in which each loop is obtained as the concatenation of an @xmath0 path with ( a portion of ) another @xmath0 path ( see figure [ fig - sec3 ] ) . this procedure is carried out first in a finite disk @xmath16 of radius @xmath17 in the plane ( see section [ unit - disc ] ) , and then an infinite volume limit , @xmath18 , is taken . [ conf - inv ] there are various possible ways to formulate the conformal invariance properties of the continuum nonsimple loop process . one version is given in theorem [ thm - conf - inv ] . next we give some examples showing how the scaling limit of various connectivity / crossing probabilities can be expressed in terms of the loop process . although we can not say whether this fact may eventually lead to exact expressions going beyond cardy s formula , it at least shows that scaling limits of such probabilities exist and are conformally invariant ( early discussions of scaling limits of connectivity functions and of the consequences of conformal invariance for such quantities are given in @xcite ) . the examples will also highlight the natural nested structure of the collection of percolation cluster boundaries in the scaling limit . consider first an annulus centered at @xmath12 with inner radius @xmath19 and outer radius @xmath20 ( see figure [ annulus ] ) . the scaling limit @xmath21 of the probability of a crossing of the annulus ( by crossing here we refer to a monochromatic " crossing , i.e. , a crossing by either of the two colors , as discussed in section [ lapa ] see also figure [ loops ] ) can be expressed as follows in terms of the random variable @xmath22 defined in theorem [ features ] above : @xmath21 is the probability that @xmath22 equals zero . more generally , @xmath22 represents the scaling limit of the minimal number of cluster boundaries traversed by any path connecting the inner and outer circles of the annulus . an example with more geometric structure involves two disjoint discs @xmath23 and @xmath24 in the plane and the scaling limit @xmath25 of the probability that there is a crossing from @xmath23 to @xmath24 ( see figure [ discs ] ) . here we let @xmath26 denote the number of distinct loops in the plane that contain @xmath23 in their interior and @xmath24 in their exterior , and define @xmath27 in the complementary way . then @xmath25 is the probability that @xmath28 , and the scaling limit of the minimal number of cluster boundaries that must be crossed to connect @xmath23 to @xmath24 is @xmath29 . one can also consider , as in @xcite , the probability of a single monochromatic cluster in the exterior @xmath30 of the union of @xmath31 disjoint discs ( or other regions ) connecting all @xmath31 disc boundaries . in the scaling limit , this can be expressed as the probability of the event that there is a single continuous ( nonsimple ) curve in @xmath30 touching all @xmath31 disc boundaries that _ does not cross any of the loops of the continuum nonsimple loop process . _ we conclude this section by remarking that the continuum nonsimple loop process is just one example of a family of conformal loop ensembles " that are related to @xmath3 and to the gaussian free field ( see @xcite ) , and are conjectured to describe the full scaling limit of statistical mechanics models such as percolation , ising and potts models . the work of lawler , schramm , sheffield and werner has provided tools to define such loop ensembles in the continuum and to study some of their properties . but the key question concerning the proof of their connection to the discrete models via a continuum scaling limit remains an important open challenge , the only exception currently being percolation , as we show in this paper . we will find it convenient to identify the real plane @xmath32 and the complex plane @xmath33 . we will also refer to the riemann sphere @xmath34 and the open upper half - plane @xmath35 ( and its closure @xmath36 ) , where chordal @xmath3 will be defined ( see section [ sle1 ] ) . @xmath37 will denote the open unit disc @xmath38 . a domain @xmath7 of the complex plane @xmath33 is a nonempty , connected , open subset of @xmath33 ; a simply connected domain @xmath7 is said to be a jordan domain if its ( topological ) boundary @xmath39 is a jordan curve ( i.e. , a simple continuous loop ) . we will make repeated use of riemann s mapping theorem , which states that if @xmath7 is any simply connected domain other than the entire plane @xmath33 and @xmath40 , then there is a unique conformal map @xmath41 of @xmath7 onto @xmath37 such that @xmath42 and @xmath43 . when taking the scaling limit as the lattice mesh size @xmath44 one can focus on fixed finite regions , @xmath45 , or consider the whole @xmath32 at once . the second option avoids dealing with boundary conditions , but requires an appropriate choice of metric . a convenient way of dealing with the whole @xmath32 is to replace the euclidean metric with a distance function @xmath46 defined on @xmath47 by @xmath48 where the infimum is over all smooth curves @xmath49 joining @xmath50 with @xmath51 , parameterized by arclength @xmath52 , and where @xmath53 denotes the euclidean norm . this metric is equivalent to the euclidean metric in bounded regions , but it has the advantage of making @xmath32 precompact . adding a single point at infinity yields the compact space @xmath54 which is isometric , via stereographic projection , to the two - dimensional sphere . in dealing with the scaling limit we use the approach of aizenman - burchard @xcite . denote by @xmath55 the complete separable metric space of continuous curves in the closure @xmath56 of the disc @xmath16 of radius @xmath17 with the metric ( [ distance ] ) defined below . curves are regarded as equivalence classes of continuous functions from the unit interval to @xmath56 , modulo monotonic reparametrizations . @xmath57 will represent a particular curve and @xmath58 a parametrization of @xmath57 ; @xmath59 will represent a set of curves ( more precisely , a closed subset of @xmath55 ) . @xmath60 will denote the uniform metric on curves , defined by @xmath61 } |\gamma_1(t ) - \gamma_2(t)|,\ ] ] where the infimum is over all choices of parametrizations of @xmath62 and @xmath63 from the interval @xmath64 $ ] . the distance between two closed sets of curves is defined by the induced hausdorff metric as follows : @xmath65 the space @xmath66 of closed subsets of @xmath55 ( i.e. , collections of curves in @xmath56 ) with the metric ( [ hausdorff ] ) is also a complete separable metric space . we denote by @xmath67 its borel @xmath68-algebra . for each fixed @xmath69 , the random curves that we consider are polygonal paths on the edges of the hexagonal lattice @xmath70 , dual to the triangular lattice @xmath71 . a superscript @xmath10 is added to indicate that the curves correspond to a model with a `` short distance cutoff '' of magnitude @xmath10 . we will also consider the complete separable metric space @xmath72 of continuous curves in @xmath54 with the distance @xmath73 } \delta(\gamma_1(t),\gamma_2(t)),\ ] ] where the infimum is again over all choices of parametrizations of @xmath62 and @xmath63 from the interval @xmath64 $ ] . the distance between two closed sets of curves is again defined by the induced hausdorff metric as follows : @xmath74 the space @xmath75 of closed sets of @xmath76 ( i.e. , collections of curves in @xmath54 ) with the metric ( [ hausdorff - d ] ) is also a complete separable metric space . we denote by @xmath77 its borel @xmath68-algebra . when we talk about convergence in distribution of random curves , we always mean with respect to the uniform metric ( [ distance ] ) , while when we deal with closed collections of curves , we always refer to the metric ( [ hausdorff ] ) or ( [ hausdorff - d ] ) . [ ab ] in this paper , the space @xmath75 of closed sets of @xmath76 is used for collections of exploration paths ( see section [ explo ] ) and cluster boundary loops and their scaling limits , @xmath0 paths and continuum nonsimple loops . the stochastic loewner evolution ( @xmath3 ) was introduced by schramm @xcite as a tool for studying the scaling limit of two - dimensional discrete ( defined on a lattice ) probabilistic models whose scaling limits are expected to be conformally invariant . in this section we define the chordal version of @xmath3 ; for more on the subject , the interested reader can consult the original paper @xcite as well as the fine reviews by lawler @xcite , kager and nienhuis @xcite , and werner @xcite , and lawler s book @xcite . let @xmath78 denote the upper half - plane . for a given continuous real function @xmath79 with @xmath80 , define , for each @xmath81 , the function @xmath82 as the solution to the ode @xmath83 with @xmath84 . this is well defined as long as @xmath85 , i.e. , for all @xmath86 , where @xmath87 } | g_s(z ) - u_s| > 0 \}.\ ] ] let @xmath88 and let @xmath89 be the unbounded component of @xmath90 ; it can be shown that @xmath91 is bounded and that @xmath92 is a conformal map from @xmath89 onto @xmath78 . for each @xmath93 , it is possible to write @xmath82 as @xmath94 when @xmath95 . the family @xmath96 is called the * loewner chain * associated to the driving function @xmath97 . [ def - sle ] * chordal @xmath6 * is the loewner chain @xmath96 that is obtained when the driving function @xmath98 is @xmath99 times a standard real - valued brownian motion @xmath100 with @xmath101 . for all @xmath102 , chordal @xmath6 is almost surely generated by a continuous random curve @xmath57 in the sense that , for all @xmath103 , @xmath104 is the unbounded connected component of @xmath105 $ ] ; @xmath57 is called the * trace * of chordal @xmath6 . let @xmath106 be a jordan domain . by riemann s mapping theorem , there are ( many ) conformal maps from the upper half - plane @xmath78 onto @xmath7 . in particular , given two distinct points @xmath107 , there exists a conformal map @xmath41 from @xmath78 onto @xmath7 such that @xmath108 and @xmath109 . in fact , the choice of the points @xmath110 and @xmath111 on the boundary of @xmath7 only characterizes @xmath112 up to a multiplicative factor , since @xmath113 would also do . suppose that @xmath96 is a chordal @xmath6 in @xmath78 as defined above ; we define chordal @xmath6 @xmath114 in @xmath7 from @xmath110 to @xmath111 as the image of the loewner chain @xmath96 under @xmath41 . it is possible to show , using scaling properties of @xmath6 , that the law of @xmath114 is unchanged , up to a linear time - change , if we replace @xmath112 by @xmath113 . this makes it natural to consider @xmath114 as a process from @xmath110 to @xmath111 in @xmath7 , ignoring the role of @xmath41 . we are interested in the case @xmath115 , for which @xmath96 is generated by a continuous , nonsimple , non - self - crossing curve @xmath57 with hausdorff dimension @xmath116 . we will denote by @xmath117 the image of @xmath57 under @xmath41 and call it the trace of chordal @xmath0 in @xmath7 from @xmath110 to @xmath111 ; @xmath117 is a continuous nonsimple curve inside @xmath7 from @xmath110 to @xmath111 , and it can be given a parametrization @xmath118 such that @xmath119 and @xmath120 , so that we are in the metric framework described in section [ space ] . it will be convenient to think of @xmath117 as an oriented path , with orientation from @xmath110 to @xmath111 . we present here rad s theorem @xcite ( see also theorem 2.11 of @xcite ) , which deals with sequences of jordan domains and the corresponding conformal maps from the unit disc , and will be used in the proof of the key lemma [ strong - smirnov ] . since the theorem deals with jordan domains , the conformal maps from the unit disc to those domains have a continuous extension to @xmath121 . with a slight abuse of notation , we do not distinguish between the conformal maps and their continuous extensions . [ rado - thm ] for @xmath122 , let @xmath123 and @xmath124 be jordan curves parameterized respectively by @xmath125 and @xmath126 , @xmath127 $ ] , and let @xmath128 and @xmath41 be conformal maps from @xmath37 onto the inner domains of @xmath123 and @xmath124 such that @xmath129 and @xmath130 , @xmath131 for all @xmath132 . if @xmath133 as @xmath134 uniformly in @xmath64 $ ] then @xmath135 as @xmath134 uniformly in @xmath136 . the type of convergence of sequences of jordan domains @xmath137 to a jordan domain @xmath7 that will be encountered in lemma [ strong - smirnov ] is such that @xmath138 converges , as @xmath134 , to @xmath39 in the uniform metric ( [ distance ] ) on continuous curves , which is clearly sufficient to apply theorem [ rado - thm ] . as a preview to the full construction , we explain how to construct a single loop using two @xmath0 paths inside a domain @xmath7 whose boundary is assumed to have a given orientation ( clockwise or counterclockwise ) . this is done in three steps ( see figure [ fig - sec3 ] ) , of which the first consists in choosing two points @xmath110 and @xmath111 on the boundary @xmath39 of @xmath7 and `` running '' a chordal @xmath0 , @xmath139 , from @xmath110 to @xmath111 inside @xmath7 . as explained in section [ sle2 ] , we consider @xmath57 as an oriented path , with orientation from @xmath110 to @xmath111 . the set @xmath140 $ ] is a countable union of its connected components , which are open and simply connected . if @xmath12 is a deterministic point in @xmath7 , then with probability one , @xmath12 is not touched by @xmath57 @xcite and so it belongs to a unique domain in @xmath140 $ ] that we denote @xmath141 . in three steps . a domain @xmath7 is formed by the solid curve . the dashed curve is an excursion @xmath142 ( from a to b ) of an @xmath0 in @xmath7 that creates a subdomain @xmath143 containing @xmath12 . the dotted curve @xmath144 is an @xmath0 in @xmath143 from b to a. a loop is formed by @xmath142 followed by @xmath144.,width=302 ] the elements of @xmath140 $ ] can be conveniently characterized in terms of how a point @xmath12 in the interior of the component was first `` trapped '' at some time @xmath145 by @xmath146 $ ] , perhaps together with either @xmath147 or @xmath148 ( the portions of the boundary @xmath39 from @xmath110 to @xmath111 counterclockwise or clockwise respectively ) see figure [ four - types ] : ( 1 ) those components whose boundary contains a segment of @xmath148 between two successive visits at @xmath149 and @xmath150 to @xmath148 ( where here and below @xmath151 ) , ( 2 ) the analogous components with @xmath148 replaced by the other part of the boundary @xmath147 , ( 3 ) those components formed when @xmath152 with @xmath57 winding about @xmath12 in a counterclockwise direction between @xmath153 and @xmath145 , and finally ( 4 ) the analogous clockwise components . from @xmath110 to @xmath111 inside a domain whose boundary is the solid curve.,width=226 ] we give to the boundary of a domain of type 3 or 4 the orientation induced by how the curve @xmath57 winds around the points inside that domain . for a domain @xmath154 of type 1 or 2 which is produced by an excursion " @xmath142 from @xmath155 to @xmath156 , the part of the boundary that corresponds to the inner perimeter of the excursion @xmath142 ( i.e. , the perimeter of @xmath57 seen from @xmath12 ) is oriented according to the direction of @xmath57 , i.e. , from @xmath157 to @xmath158 . if we assume that @xmath39 is oriented from @xmath110 to @xmath111 clockwise , then the boundaries of domains of type 2 have a well defined orientation , while the boundaries of domains of type 1 do not , since they are composed of two parts which are both oriented from the beginning to the end of the excursion that produced the domain . now , let @xmath143 be a domain of type 1 and let @xmath159 and @xmath160 be respectively the starting and ending point of the excursion that generated @xmath143 . the second step to construct a loop is to run a chordal @xmath0 , @xmath161 , inside @xmath143 from @xmath160 to @xmath159 ; the third and final step consists in pasting together @xmath142 and @xmath144 . running @xmath144 inside @xmath143 from @xmath160 to @xmath159 partitions @xmath162 into new domains . notice that if we assign an orientation to the boundaries of these domains according to the same rules used above , all of those boundaries have a well defined orientation , so that the construction of loops just presented can be iterated inside each one of these domains ( as well as inside each of the domains of type 2 , 3 and 4 generated by @xmath117 in the first step ) . this will be done in the next section . in this section we define the continuum nonsimple loop process inside the unit disc @xmath163 via an inductive procedure . later , in order to define the continuum nonsimple loops in the whole plane , the unit disc will be replaced by a growing sequence of large discs , @xmath16 , with @xmath164 ( see theorem [ thm - therm - lim ] ) . the basic ingredient in the algorithmic construction , given in the previous section , consists of a chordal @xmath0 path @xmath117 between two points @xmath110 and @xmath111 of the boundary @xmath39 of a given simply connected domain @xmath106 . we will organize the inductive procedure in steps , each one corresponding to one @xmath0 inside a certain domain generated by the previous steps . to do that , we need to order the domains present at the end of each step , so as to choose the one to use in the next step . for this purpose , we introduce a deterministic countable set of points @xmath165 that are dense in @xmath33 and are endowed with a deterministic order ( here and below by deterministic we mean that they are assigned before the beginning of the construction and are independent of the @xmath0 s ) . the first step consists of an @xmath0 path , @xmath166 , inside @xmath37 from @xmath167 to @xmath168 , which produces many domains that are the connected components of the set @xmath169 $ ] . these domains can be priority - ordered according to the maximal @xmath170- or @xmath171- coordinate distances between points on their boundaries and using the rank of the points in @xmath165 ( contained in the domains ) to break ties , as follows . for a domain @xmath7 , let @xmath172 be the maximal @xmath170- or @xmath171-distance between points on its boundary , whichever is greater . domains with larger @xmath173 have higher priority , and if two domains have the same @xmath173 , the one containing the highest ranking point of @xmath165 from those two domains has higher priority . the priority order of domains of course changes as the construction proceeds and new domains are formed . the second step of the construction consists of an @xmath0 path , @xmath63 , that is produced in the domain with highest priority ( after the first step ) . since all the domains that are produced in the construction are jordan domains , as explained in the discussion following corollary [ jordan ] , for all steps we can use the definition of chordal @xmath3 given in section [ sle2 ] . as a result of the construction , the @xmath0 paths are naturally ordered : @xmath174 . it will be shown ( see especially the proof of theorem [ thm - convergence ] below ) that every domain that is formed during the construction is eventually used ( this is in fact one important requirement in deciding how to order the domains and therefore how to organize the construction ) . so far we have not explained how to choose the starting and ending points of the @xmath0 paths on the boundaries of the domains . in order to do this , we give an orientation to the boundaries of the domains produced by the construction according to the rules explained in section [ single - loop ] . we call * monochromatic * a boundary which gets , as a consequence of those rules , a well defined ( clockwise or counterclockwise ) orientation ; the choice of this term will be clarified when we discuss the lattice version of the loop construction below . we will generally take our initial domain @xmath175 ( or @xmath16 ) to have a monochromatic boundary ( either clockwise or counterclockwise orientation ) . it is easy to see by induction that the boundaries that are not monochromatic are composed of two `` pieces '' joined at two special points ( call them a and b , as in the example of section [ single - loop ] ) , such that one piece is a portion of the boundary of a previous domain , and the other is the inner perimeter of an excursion ( see again section [ single - loop ] ) . both pieces are oriented in the same direction , say from a to b ( see figure [ fig - sec3 ] ) . for a domain whose boundary is not monochromatic , we make the `` natural '' choice of starting and ending points , corresponding to the end and beginning of the excursion that produced the domain ( the points b and a respectively , in the example above ) . as explained in section [ single - loop ] , when such a domain is used with this choice of points on the boundary , a loop is produced , together with other domains , whose boundaries are all monochromatic . for a domain whose boundary is monochromatic , and therefore has a well defined orientation , there are various procedures which would yield the `` correct '' distribution for the resulting continuum nonsimple loop process ; one possibility is as follows . given a domain @xmath7 , @xmath110 and @xmath111 are chosen so that , of all pairs @xmath176 of points in @xmath39 , they maximize @xmath177 if @xmath178 , or else they maximize @xmath179 . if the choice is not unique , to restrict the number of pairs one looks at those pairs , among the ones already obtained , that maximize the other of @xmath180 . notice that this leaves at most two pairs of points ; if that s the case , the pair that contains the point with minimal real ( and , if necessary , imaginary ) part is chosen . the iterative procedure produces a loop every time a domain whose boundary is not monochromatic is used . our basic loop process consists of the collection of all loops generated by this inductive procedure ( i.e. , the limiting object obtained from the construction by letting the number of steps @xmath134 ) , to which we add a trivial " loop for each @xmath12 in @xmath37 , so that the collection of loops is closed in the appropriate sense @xcite . the continuum nonsimple loop process in the whole plane is introduced in theorem [ thm - therm - lim ] , section [ main - results ] . there , a trivial " loop for each @xmath181 has to be added to make the space of loops closed . we will denote by @xmath8 the two - dimensional triangular lattice , whose sites we think of as the elementary cells of a regular hexagonal lattice @xmath9 embedded in the plane as in figure [ loops ] . two hexagons are * neighbors * if they are adjacent , i.e. , if they have a common edge . a sequence @xmath182 of hexagons such that @xmath183 and @xmath184 are neighbors for all @xmath185 and @xmath186 whenever @xmath187 will be called a * @xmath8-path * and denoted by @xmath188 . if the first and last sites of the path are neighbors , the path will be called a * @xmath8-loop*. a finite set @xmath7 of hexagons is * connected * if any two hexagons in @xmath7 can be joined by a @xmath8-path contained in @xmath7 . we say that a finite set @xmath7 of hexagons is * simply connected * if both @xmath7 and its complement are connected . for a simply connected set @xmath7 of hexagons , we denote by @xmath189 its * external site boundary * , or * s - boundary * ( i.e. , the set of hexagons that do not belong to @xmath7 but are adjacent to hexagons in @xmath7 ) , and by @xmath39 the topological boundary of @xmath7 when @xmath7 is considered as a domain of @xmath33 . we will call a bounded , simply connected subset @xmath7 of @xmath8 a * jordan set * if its s - boundary @xmath189 is a @xmath8-loop . for a jordan set @xmath190 , a vertex @xmath191 that belongs to @xmath39 can be either of two types , according to whether the edge incident on @xmath170 that is not in @xmath39 belongs to a hexagon in @xmath7 or not . we call a vertex of the second type an * e - vertex * ( e for `` external '' or `` exposed '' ) . given a jordan set @xmath7 and two e - vertices @xmath192 in @xmath39 , we denote by @xmath193 the portion of @xmath39 traversed counterclockwise from @xmath170 to @xmath171 , and call it the * right boundary * ; the remaining part of the boundary is denote by @xmath194 and is called the * left boundary*. analogously , the portion of @xmath195 of @xmath189 whose hexagons are adjacent to @xmath193 is called the * right s - boundary * and the remaining part the * left s - boundary*. a * percolation configuration * @xmath196 on @xmath8 is an assignment of @xmath197 ( equivalently , yellow ) or @xmath198 ( blue ) to each site of @xmath8 ( i.e. , to each hexagon of @xmath9 see figure [ loops ] ) . for a domain @xmath7 of the plane , the restriction to the subset @xmath199 of @xmath8 of the percolation configuration @xmath68 is denoted by @xmath200 . on the space of configurations @xmath201 , we consider the usual product topology and denote by @xmath202 the uniform measure , corresponding to bernoulli percolation with equal density of yellow ( minus ) and blue ( plus ) hexagons , which is critical percolation in the case of the triangular lattice . a ( percolation ) * cluster * is a maximal , connected , monochromatic subset of @xmath8 ; we will distinguish between blue ( plus ) and yellow ( minus ) clusters . the * boundary * of a cluster @xmath7 is the set of edges of @xmath9 that surround the cluster ( i.e. , its peierls contour ) ; it coincides with the topological boundary of @xmath7 considered as a domain of @xmath33 . the set of all boundaries is a collection of `` nested '' simple loops along the edges of @xmath9 . given a percolation configuration @xmath68 , we associate an arrow to each edge of @xmath9 belonging to the boundary of a cluster in such a way that the hexagon to the right of the edge with respect to the direction of the arrow is blue ( plus ) . the set of all boundaries then becomes a collection of nested , oriented , simple loops . a * boundary path * ( or * b - path * ) @xmath57 is a sequence @xmath203 of distinct edges of @xmath9 belonging to the boundary of a cluster and such that @xmath204 and @xmath205 meet at a vertex of @xmath9 for all @xmath185 . to each b - path , we can associate a direction according to the direction of the edges in the path . given a b - path @xmath57 , we denote by @xmath206 ( respectively , @xmath207 ) the set of blue ( resp . , yellow ) hexagons ( i.e. , sites of @xmath8 ) adjacent to @xmath57 ; we also let @xmath208 . for a jordan set @xmath190 and two e - vertices @xmath192 in @xmath39 , imagine coloring blue all the hexagons in @xmath195 and yellow all those in @xmath209 . then , for any percolation configuration @xmath200 inside @xmath7 , there is a unique b - path @xmath57 from @xmath170 to @xmath171 which separates the blue cluster adjacent to @xmath195 from the yellow cluster adjacent to @xmath209 . we call @xmath210 a * percolation exploration path * ( see figure [ fig2-sec4 ] ) . an exploration path @xmath57 can be decomposed into * left excursions * @xmath142 , i.e. , maximal b - subpaths of @xmath57 that do not use edges of the left boundary @xmath194 . successive left excursions are separated by portions of @xmath57 that contain only edges of the left boundary @xmath194 . analogously , @xmath57 can be decomposed into * right excursions * , i.e. , maximal b - subpaths of @xmath57 that do not use edges of the right boundary @xmath193 . successive right excursions are separated by portions of @xmath57 that contain only edges of the right boundary @xmath193 . notice that the exploration path @xmath211 only depends on the percolation configuration @xmath200 inside @xmath7 and the positions of the e - vertices @xmath170 and @xmath171 ; in particular , it does not depend on the color of the hexagons in @xmath189 , since it is defined by imposing fictitious @xmath212 boundary conditions on @xmath7 . to see this more clearly , we next show how to construct the percolation exploration path dynamically , via the * percolation exploration process * defined below . given a jordan set @xmath190 and two e - vertices @xmath192 in @xmath39 , assign to @xmath193 a counterclockwise orientation ( i.e. , from @xmath170 to @xmath171 ) and to @xmath194 a clockwise orientation . call @xmath213 the edge incident on @xmath170 that does not belong to @xmath39 and orient it in the direction of @xmath170 ; this is the `` starting edge '' of an exploration procedure that will produce an oriented path inside @xmath7 along the edges of @xmath9 , together with two _ nonsimple _ monochromatic paths on @xmath8 . > from @xmath213 , the process moves along the edges of hexagons in @xmath7 according to the rules below . at each step there are two possible edges ( left or right edge with respect to the current direction of exploration ) to choose from , both belonging to the same hexagon @xmath214 contained in @xmath7 or @xmath189 . * if @xmath214 belongs to @xmath7 and has not been previously explored , " its color is determined by flipping a fair coin and then the edge to the left ( with respect to the direction in which the exploration is moving ) is chosen if @xmath214 is blue ( plus ) , or the edge to the right is chosen if @xmath214 is yellow ( minus ) . * if @xmath214 belongs to @xmath7 and has been previously explored , the color already assigned to it is used to choose an edge according to the rule above . * if @xmath214 belongs to the right external boundary @xmath195 , the left edge is chosen . * if @xmath214 belongs to the left external boundary @xmath209 , the right edge is chosen . * the exploration process stops when it reaches @xmath111 . we can assign an arrow to each edge in the path in such a way that the hexagon to the right of the edge with respect to the arrow is blue ; for edges in @xmath39 , we assign the arrows according to the direction assigned to the boundary . in this way , we get an oriented path , whose shape and orientation depend solely on the color of the hexagons explored during the construction of the path . boundary conditions on the first column , corresponding to the boundary of the region where the exploration is carried out . the colored hexagons that do not belong to the first column have been `` explored '' during the exploration process . the heavy line between yellow ( light ) and blue ( dark ) hexagons is the exploration path produced by the exploration process.,width=226 ] when we present the discrete construction , we will encounter jordan sets @xmath7 with two e - vertices @xmath215 assigned in some way to be discussed later . such domains will have either monochromatic ( plus or minus ) boundaries or @xmath212 boundary conditions , corresponding to having both @xmath195 and @xmath209 monochromatic , but of different colors . as explained , the exploration path @xmath216 does not depend on the color of @xmath189 , but the interpretation of @xmath216 does . for domains with @xmath212 boundary conditions , the exploration path represents the interface between the yellow cluster containing the yellow portion of the s - boundary of @xmath7 and the blue cluster containing its blue portion . for domains with monochromatic blue ( resp . , yellow ) boundary conditions , the exploration path represents portions of the boundaries of yellow ( resp . , blue ) clusters touching @xmath194 and adjacent to blue ( resp . , yellow ) hexagons that are the starting point of a blue ( resp . , yellow ) path ( possibly an empty path ) that reaches @xmath193 , pasted together using portions of @xmath194 . in order to study the continuum scaling limit of an exploration path , we introduce the following definitions . [ approx ] given a jordan domain @xmath7 of the plane , we denote by @xmath217 the largest jordan set of hexagons of the scaled hexagonal lattice @xmath70 that is contained in @xmath7 , and call it the * @xmath10-approximation * of @xmath7 . it is clear that @xmath218 converges to @xmath39 in the metric ( [ distance ] ) . [ exp - path ] let @xmath7 be a jordan domain of the plane and @xmath217 its @xmath10-approximation . for @xmath107 , choose the pair @xmath219 of e - vertices in @xmath218 closest to , respectively , @xmath110 and @xmath111 ( if there are two such vertices closest to @xmath110 , we choose , say , the first one encountered going clockwise along @xmath218 , and analogously for @xmath111 ) . given a percolation configuration @xmath68 , we define the * exploration path * @xmath220 . for a fixed @xmath69 , the measure @xmath202 on percolation configurations @xmath68 induces a measure @xmath221 on exploration paths @xmath222 . in the continuum scaling limit , @xmath44 , one is interested in the weak convergence of @xmath221 to a measure @xmath223 supported on continuous curves , with respect to the uniform metric ( [ distance ] ) on continuous curves . one of the main tools in this paper is the result on convergence to @xmath0 announced by smirnov @xcite ( see also @xcite ) , whose detailed proof is to appear @xcite : _ the distribution of @xmath224 converges , as @xmath44 , to that of the trace of chordal @xmath0 inside @xmath7 from @xmath110 to @xmath111 , with respect to the uniform metric ( [ distance ] ) on continuous curves_. actually , we will rather use a slightly stronger conclusion , given as statement ( s ) at the beginning of section [ main - results ] below , a version of which , according to @xcite ( see p. 734 there ) , and @xcite , will be contained in @xcite . this stronger statement is that the convergence of the percolation process to @xmath0 takes place _ locally uniformly _ with respect to the shape of the domain @xmath7 and the positions of the starting and ending points @xmath110 and @xmath111 on its boundary @xmath39 . we will use this version of convergence to @xmath0 to identify the continuum nonsimple loop process with the scaling limit of _ all _ critical percolation clusters . a detailed proof of statement ( s ) can be found in @xcite . although the convergence statement in ( s ) is stronger than those in @xcite , we note that it is restricted to jordan domains , a restriction not present in @xcite . before concluding this section , we give one more definition . consider the exploration path @xmath225 and the set @xmath226 . the set @xmath227 is the union of its connected components ( in the lattice sense ) , which are simply connected . if the domain @xmath7 is large and the e - vertices @xmath228 are not too close to each other , then with high probability the exploration process inside @xmath217 will make large excursions into @xmath217 , so that @xmath227 will have more than one component . given a point @xmath229 contained in @xmath227 , we will denote by @xmath230 the domain corresponding to the unique element of @xmath227 that contains @xmath12 ( notice that for a deterministic @xmath231 , @xmath230 is well defined with high probability for @xmath10 small , i.e. , when @xmath232 and @xmath233 ) . next , we show how to construct , by twice using the exploration process described in section [ explo ] , a loop @xmath234 along the edges of @xmath235 corresponding to the external boundary of a monochromatic cluster contained in a large , simply connected , jordan set @xmath7 with monochromatic blue ( say ) boundary conditions ( see figures [ fig3-sec4 ] and [ fig4-sec4 ] ) . consider the exploration path @xmath236 and the sets @xmath207 and @xmath206 ( see figure [ fig3-sec4 ] ) . the set @xmath237 is the union of its connected components ( in the lattice sense ) , which are simply connected . if the domain @xmath7 is large and the e - vertices @xmath238 are chosen not too close to each other , with large probability the exploration process inside @xmath7 will make large excursions into @xmath7 , so that @xmath237 will have many components . there are four types of components which may be usefully thought of in terms of their external site boundaries : ( 1 ) those components whose site boundary contains both sites in @xmath207 and @xmath209 , ( 2 ) the analogous components with @xmath209 replaced by @xmath195 and @xmath207 by @xmath206 , ( 3 ) those components whose site boundary only contains sites in @xmath239 , and finally ( 4 ) the analogous components with @xmath207 replaced by @xmath206 . notice that the components of type 1 are the only ones with @xmath212 boundary conditions , while all other components have monochromatic s - boundaries . for a given component @xmath143 of type 1 , we can identify the two edges that separate the yellow and blue portions of its s - boundary . the vertices @xmath240 and @xmath241 of @xmath9 where those two edges intersect @xmath242 are e - vertices and are chosen to be the starting and ending points of the exploration path @xmath243 inside @xmath143 . if @xmath244 are respectively the ending and starting points of the left excursion @xmath142 of @xmath216 that created " @xmath143 , by pasting together @xmath142 and @xmath243 with the help of the edges of @xmath39 contained between @xmath240 and @xmath245 and between @xmath241 and @xmath246 , we get a loop @xmath234 which corresponds to the boundary of a yellow cluster adjacent to @xmath194 ( see figure [ fig4-sec4 ] ) . notice that the path @xmath243 in general splits @xmath143 into various other domains , all of which have monochromatic boundary conditions . to the e - vertex @xmath171 . the starting edge " and ending edge " of the exploration path are indicated by dotted segments next to @xmath170 and @xmath171 . the outer layer of hexagons does not belong to the domain where the explorations are carried out , but represents its monochromatic blue / plus external site boundary . @xmath245 and @xmath246 are the ending and starting points of a left excursion that determines a new domain @xmath143 , and @xmath240 and @xmath241 are the vertices where the edges that separate the yellow and blue portions of the s - boundary of @xmath143 intersect @xmath242 . @xmath240 and @xmath241 will be respectively the beginning and end of a new exploration path whose starting edge " and ending edge " are indicated by dotted segments next to those points.,width=302 ] to @xmath241 whose resulting path ( heavy broken line ) is pasted to the left excursion generated by the previous exploration with the help of edges ( indicated again by a heavy broken line ) of @xmath39 contained between @xmath240 and @xmath245 and between @xmath241 and @xmath246.,width=302 ] we now give the algorithmic construction for discrete percolation which is the analogue of the continuum one . each step of the construction is a single percolation exploration process ; the order of successive steps is organized as in the continuum construction detailed in section [ unit - disc ] . we start with the largest jordan set @xmath247 of hexagons that is contained in the unit disc @xmath37 . we will also make use of the countable set @xmath165 of points dense in @xmath33 that was introduced earlier . the first step consists of an exploration process inside @xmath248 . for this , we need to select two points @xmath170 and @xmath171 in @xmath249 ( which identify the starting and ending edges ) . we choose for @xmath170 the e - vertex closest to @xmath167 , and for @xmath171 the e - vertex closest to @xmath168 ( if there are two such vertices closest to @xmath167 , we can choose , say , the one with smallest real part , and analogously for @xmath168 ) . the first exploration produces a path @xmath250 and , for @xmath10 small , many new domains of all four types . these domains are ordered according to the maximal @xmath170- or @xmath171- distance @xmath173 between points on their boundaries and , if necessary , with the help of points in @xmath165 , as in the continuum case , and that order is used , at each step of the construction , to determine the next exploration process . with this choice , the exploration processes and paths are naturally ordered : @xmath251 . each exploration process of course requires choosing a starting and ending vertex and edge . for domains of type 1 , with a @xmath212 or @xmath252 boundary condition , the choice is the natural one , explained before . for a domain @xmath253 ( used at the @xmath132th step ) of type other than 1 , and therefore with a monochromatic boundary , the starting and ending edges are chosen with a procedure that mimics what is done in the continuum case . once again , the exact procedure used to choose the pair of points is not important , as long as they are not chosen too close to each other . this is clear in the discrete case because the procedure that we are presenting is only `` discovering '' the cluster boundaries . in more precise terms , it is clear that one could couple the processes obtained with different rules by means of the same percolation configuration , thus obtaining exactly the same cluster boundaries . as in the continuum case , we can choose the following procedure . ( in theorem [ thm - convergence ] we will slightly reorganize the procedure by using a coupling to the continuum construction to guarantee that the order of exploration of domains of the discrete and continuum procedures match despite the rules for breaking ties . ) given a domain @xmath7 , @xmath170 and @xmath171 are chosen so that , of all pairs @xmath176 of points in @xmath39 , they maximize @xmath177 if @xmath178 , or else they maximize @xmath179 . if the choice is not unique , to restrict the number of pairs one looks at those pairs , among the ones already obtained , that maximize the other of @xmath180 . notice that this leaves at most two pairs of points ; if that s the case , the pair that contains the point with minimal real ( and , if necessary , imaginary ) part is chosen . the procedure continues iteratively , with regions that have monochromatic boundaries playing the role played in the first step by the unit disc . every time a region with @xmath212 boundary conditions is used , a new loop , corresponding to the outer boundary contour of a cluster , is formed by pasting together , as explained in section [ single - loop ] , the new exploration path and the excursion containing the region where the last exploration was carried out . all the new regions created at a step when a loop is formed have monochromatic boundary conditions . in this section we collect our main results about the continuum nonsimple loop process . before doing that , we state a precise version , called statement ( s ) , of convergence of exploration paths to @xmath0 that we will use in the proofs of these results , presented in section [ proofs ] . statement ( s ) is an immediate consequence of theorem 5 of @xcite . the proof given in @xcite , which relies among other things on the result of smirnov @xcite concerning convergence of crossing probabilities to cardy s formula @xcite , is an expanded and corrected version of appendix a of @xcite . we note that ( s ) is both more general and more special than the convergence statements in @xcite more general in that the domain can vary with @xmath10 as @xmath44 , but more special in the restriction to jordan domains . given a jordan domain @xmath7 with two distinct points @xmath107 on its boundary , let @xmath223 denote the law of @xmath117 , the trace of chordal @xmath0 , and let @xmath221 denote the law of the percolation exploration path @xmath224 . let @xmath254 be the space of continuous curves inside @xmath7 from @xmath110 to @xmath111 . we define @xmath255 ( where @xmath256 denotes the open ball of radius @xmath257 centered at @xmath170 in the metric ( [ distance ] ) ) and denote by @xmath258 the prohorov distance ; weak convergence is equivalent to convergence in the prohorov metric . statement ( s ) is the following ; it is used in the proofs of all the results of this section _ except _ for lemmas [ sub - conv]-[boundaries ] . * for jordan domains , there is convergence in distribution of the percolation exploration path to the trace of chordal @xmath0 that is _ locally uniform _ in the shape of the boundary with respect to the uniform metric on continuous curves ( [ distance ] ) , and in the location of the starting and ending points with respect to the euclidean metric ; i.e. , for @xmath259 a jordan domain with distinct @xmath107 , @xmath260 , @xmath261 and @xmath262 such that for all @xmath263 with @xmath143 jordan and with @xmath264 and @xmath265 , @xmath266 . we first give some important results which are needed in the proofs of the main theorems . we start with two lemmas which are consequences of @xcite , of standard bounds on the probability of events corresponding to having a certain number of monochromatic crossings of an annulus ( see lemma 5 of @xcite , appendix a of @xcite , and also @xcite ) , but which do _ not _ depend on statement ( s ) . [ sub - conv ] let @xmath267 be the percolation exploration path on the edges of @xmath70 inside ( the @xmath10-approximation of ) @xmath37 between ( the e - vertices closest to ) @xmath167 and @xmath168 . for any fixed point @xmath268 , chosen independently of @xmath267 , as @xmath44 , @xmath267 and the boundary @xmath269 of the domain @xmath270 that contains @xmath12 jointly have limits in distribution along subsequences of @xmath10 with respect to the uniform metric ( [ distance ] ) on continuous curves . moreover , any subsequence limit of @xmath269 is almost surely a simple loop @xcite . [ boundaries ] using the notation of lemma [ sub - conv ] , let @xmath271 be the limit in distribution of @xmath267 as @xmath44 along some convergent subsequence @xmath272 and @xmath273 the boundary of the domain @xmath274 of @xmath275 $ ] that contains @xmath12 . then , as @xmath134 , @xmath276 converges in distribution to @xmath277 . the two lemmas above are important ingredients in the proof of theorem [ thm - convergence ] below . the second one says that , for every subsequence limit , the discrete boundaries converge to the boundaries of the domains generated by the limiting continuous curve . if we use statement ( s ) , then the limit @xmath271 of @xmath278 is the trace of chordal @xmath0 for every subsequence @xmath279 , and we can use lemmas [ boundaries ] and [ sub - conv ] to deduce that all the domains produced in the continuum construction are jordan domains . the key step in that direction is represented by the following result , our proof of which relies on ( s ) . [ jordan ] for any deterministic @xmath268 , the boundary @xmath273 of a domain @xmath274 of the continuum construction is almost surely a jordan curve . the corollary says that the domains that appear after the first step of the continuum construction are jordan domains . the steps in the second stage of the continuum construction consist of @xmath0 paths inside jordan domains , and therefore corollary [ jordan ] , combined with riemann s mapping theorem and the conformal invariance of @xmath0 , implies that the domains produced during the second stage are also jordan . by induction , we deduce that all the domains produced in the continuum construction are jordan domains . we end this section with one more lemma which is another key ingredient in the proof of theorem [ thm - convergence ] ; we remark that its proof requires ( s ) in a fundamental way . [ strong - smirnov ] let @xmath259 denote a _ jordan domain , with @xmath280 two points on @xmath39 . let @xmath281 , be a sequence of _ random _ jordan domains with points on their boundaries such that , as @xmath134 , @xmath282 converges in distribution to @xmath283 with respect to the uniform metric ( [ distance ] ) on continuous curves , and the euclidean metric on @xmath284 . for any sequence @xmath285 with @xmath279 as @xmath134 , @xmath286 converges in distribution to @xmath117 with respect to the uniform metric ( [ distance ] ) on continuous curves . in this section we state the main technical theorems of this paper . our main results , presented in section [ quickresults ] , are consequences of these theorems . the proofs of these theorems rely on statement ( s ) . as noted before , a detailed proof of statement ( s ) can be found in @xcite . [ thm - convergence ] for any @xmath287 , the first @xmath132 steps of ( a suitably reorganized version of ) the full discrete construction inside the unit disc ( of section [ full ] ) converge , jointly in distribution , to the first @xmath132 steps of the full continuum construction inside the unit disc ( of section [ unit - disc ] ) . furthermore , the scaling limit of the full ( original or reorganized ) discrete construction is the full continuum construction . moreover , if for any fixed @xmath288 we let @xmath289 denote the number of steps needed to find all the cluster boundaries of euclidean diameter larger than @xmath257 in the discrete construction , then @xmath289 is bounded in probability as @xmath44 ; i.e. , @xmath290 . this is so in both the original and reorganized versions of the discrete construction . the second part of theorem [ thm - convergence ] means that both versions of the discrete construction used in the theorem find all large contours in a number of steps which does not diverge as @xmath44 . this , together with the first part of the same theorem , implies that the continuum construction does indeed describe all macroscopic contours contained inside the unit disc ( with blue boundary conditions ) as @xmath44 . the construction presented in section [ unit - disc ] can of course be repeated for the disc @xmath16 of radius @xmath17 , for any @xmath17 , so we should take a `` thermodynamic limit '' by letting @xmath291 . in this way , we would eliminate the boundary ( and the boundary conditions ) and obtain a process on the whole plane . such an extension from the unit disc to the plane is contained in the next theorem . let @xmath292 be the ( limiting ) distribution of the set of curves ( all continuum nonsimple loops ) generated by the continuum construction inside @xmath16 ( i.e. , the limiting measure , defined by the inductive construction , on the complete separable metric space @xmath293 of collections of continuous curves in @xmath16 ) . for a domain @xmath7 , we denote by @xmath294 the mapping ( on @xmath75 or @xmath66 ) in which all portions of curves that exit @xmath7 are removed . when applied to a configuration of loops in the plane , @xmath294 gives a set of curves which either start and end at points on @xmath39 or form closed loops completely contained in @xmath7 . let @xmath295 be the same mapping lifted to the space of probability measures on @xmath75 or @xmath66 . [ thm - therm - lim ] there exists a unique probability measure @xmath296 on the space @xmath75 of collections of continuous curves in @xmath54 such that @xmath297 as @xmath291 in the sense that for every bounded domain @xmath7 , as @xmath291 , @xmath298 . we we will generally take monochromatic blue boundary conditions on the disc @xmath16 of radius @xmath17 , but this arbitrary choice does not affect the results . the next theorem states a conformal invariance property of the continuum nonsimple loop processes of theorem [ main - result1 ] . [ thm - conf - inv ] given two disjoint discs , @xmath23 and @xmath24 , let @xmath299 ( respectively , @xmath300 ) be the smallest loop from the continuum nonsimple loop process x that surrounds @xmath23 ( resp . , @xmath24 ) and let @xmath301 ( resp . , @xmath302 ) be the connected component of @xmath303 ( resp . , @xmath304 ) that contains @xmath23 ( resp . , @xmath24 ) . assume that @xmath301 and @xmath302 are disjoint and let @xmath305 , @xmath306 , denote the distribution of the loops inside @xmath307 . then , conditioned on @xmath301 and @xmath302 , the configurations inside @xmath301 and @xmath302 are independent and moreover @xmath308 ( here @xmath309 denotes the probability distribution of the loop process @xmath310 when @xmath311 is distributed by @xmath312 ) , where @xmath313 is a conformal homeomorphism from @xmath301 onto @xmath302 . we remark that the result is still valid ( without the independence ) even if @xmath301 and @xmath302 are not disjoint , but for simplicity we do not consider that case . to conclude this section , we show how to recover chordal @xmath0 from the continuum nonsimple loop process , i.e. , given a ( deterministic ) jordan domain @xmath7 with two boundary points @xmath110 and @xmath111 , we give a construction that uses the continuum nonsimple loops of @xmath296 to generate a process distributed like chordal @xmath0 inside @xmath7 from @xmath110 to @xmath111 . remember , first of all , that each continuum nonsimple loop has either a clockwise or counterclockwise direction , with the set of all loops surrounding any deterministic point alternating in direction . for convenience , let us suppose that @xmath110 is at the bottom " and @xmath111 is at the top " of @xmath7 so that the boundary is divided into a left and right part by these two points . fix @xmath288 and call @xmath314 the set of all the directed segments of loops that connect from the left to the right part of the boundary touching @xmath39 at a distance larger than @xmath257 from both @xmath110 and @xmath111 , and @xmath315 the analogous set of directed segments from the right to the left portion of @xmath39 . for a fixed @xmath288 , there is only a finite number of such segments , and , if they are ordered moving along the left boundary of @xmath7 from @xmath110 to @xmath111 , they alternate in direction ( i.e. , a segment in @xmath314 is followed by one in @xmath315 and so on ) . between a segment in @xmath315 and the next segment in @xmath314 , there are countably many portions of loops intersecting @xmath7 which start and end on @xmath39 and are maximal in the sense that they are not contained inside any other portion of loop of the same type ; they all have counterclockwise direction and can be used to make a `` bridge '' between the right - to - left segment and the next one ( in @xmath314 ) . this is done by pasting the portions of loops together with the help of points in @xmath39 and a limit procedure to produce a connected ( nonsimple ) path . if we do this for each pair of successive segments on both sides of the boundary of @xmath7 , we get a path that connects two points on @xmath39 . by letting @xmath316 and taking the limit of this procedure , since almost surely @xmath110 and @xmath111 are surrounded by an infinite family of nested loops with diameters going to zero , we obtain a path that connects @xmath110 with @xmath111 ; this path is distributed as chordal @xmath0 inside @xmath7 from @xmath110 to @xmath111 . the last claim follows from considering the analogous procedure for percolation on the discrete lattice @xmath70 , using segments of boundaries . it is easy to see that in the discrete case this procedure produces exactly the same path as the percolation exploration process . by theorems [ main - result1 ] and [ main - result2 ] , the scaling limit of this discrete procedure is the continuum one described above , therefore the claim follows from ( s ) . in this section we present the proofs of the results stated in sections [ quickresults ] and [ main - results ] . in order to do that , we will use the following lemma . [ arms ] let @xmath317 be the event that the annulus @xmath318 centered at @xmath319 contains six disjoint monochromatic crossings , not all of the same color , and let @xmath320 be the event , for some @xmath321 , that @xmath322 contains three disjoint monochromatic crossings , not all of the same color . then , for any @xmath288 , @xmath323 and @xmath324 * proof . * we know from @xcite that there exist @xmath325 and @xmath326 so that for @xmath327 , and @xmath10 small enough ( in particular , @xmath328 ) , @xmath329 for any @xmath330 . if we cover @xmath37 with @xmath331 balls of radius @xmath332 centered at @xmath333 , we have that , for @xmath334 and @xmath10 small enough , @xmath335 where the first inequality follows from the observations that for any @xmath336 , @xmath337 and @xmath338 for some @xmath339 , and the second inequality uses ( [ ksz ] ) . using the fact that @xmath331 is @xmath340 , we can let first @xmath44 and then @xmath341 to obtain ( [ six ] ) . we also know , as a consequence of lemma 5 of @xcite or as proved in appendix a of @xcite , that for any @xmath342 , the probability that the semi - annulus @xmath343 contains three disjoint monochromatic crossings , not all of the same color , is bounded above by @xmath344 for some @xmath345 and @xmath346 . ( we remark that the result still applies when @xmath347 is replaced by any other half - plane . ) since the unit disc is a convex subset of the half - plane @xmath348 and therefore the intersection of an annulus centered at @xmath167 with the unit disc @xmath37 is a subset of the intersection of the same annulus with the half - plane @xmath348 , we can use that bound to conclude that for @xmath349 , and in fact for any @xmath321 , there exists a constant @xmath350 such that @xmath351 for some @xmath346 . we can then use similar arguments to those above , together with ( [ lsw ] ) , to obtain ( [ three ] ) and conclude the proof . + * proof of lemma [ sub - conv ] . * the first part of the lemma is a direct consequence of @xcite ; it is enough to notice that the ( random ) polygonal curves @xmath267 and @xmath269 satisfy the conditions in @xcite and thus have a scaling limit in terms of continuous curves , at least along subsequences of @xmath10 . let us assume that this is not the case and that the limit @xmath352 of @xmath353 along some subsequence @xmath285 touches itself , i.e. , @xmath354 for @xmath355 with positive probability . if that happens , we can take @xmath356 small enough so that the annulus @xmath357 is crossed at least four times by @xmath352 ( here @xmath358 is the ball of radius @xmath359 centered at @xmath50 ) . because of the choice of topology , the convergence in distribution of @xmath353 to @xmath352 implies that we can find coupled versions of @xmath353 and @xmath352 on some probability space @xmath360 such that @xmath361 , for all @xmath362 as @xmath134 ( see , for example , corollary 1 of @xcite ) . using this coupling , we can choose @xmath132 large enough ( depending on @xmath363 ) so that @xmath353 stays in an @xmath364-neighborhood @xmath365 of @xmath352 . this event however would correspond to ( at least ) four paths of one color ( corresponding to the four crossings by @xmath366 , which shadows @xmath353 ) and two of the other color ( belonging to percolation clusters adjacent to the cluster of @xmath366 , and of the opposite color ) , of the annulus @xmath367 ( see , for example , @xcite see also figure [ fig - lemma5 - 1 ] ) . as @xmath368 , we can let @xmath341 , in which case the probability of seeing the event just described somewhere inside @xmath37 goes to zero by an application of lemma [ arms ] , leading to a contradiction . + * proof . * let @xmath285 be a convergent subsequence for @xmath267 and let @xmath372 be the limit in distribution of @xmath278 as @xmath134 . for simplicity of notation , in the rest of the proof we will drop the @xmath132 and write @xmath10 instead of @xmath373 . because of the choice of topology , the convergence in distribution of @xmath374 to @xmath57 implies that we can find coupled versions of @xmath375 and @xmath57 on some probability space @xmath360 such that @xmath376 , for all @xmath363 as @xmath134 ( see , for example , corollary 1 of @xcite ) . using this coupling , we first consider the case of @xmath377 such that @xmath370 but @xmath371 . since @xmath378 is an open subset of @xmath33 , there exists a continuous curve @xmath379 joining @xmath50 and @xmath51 and a constant @xmath288 such that the @xmath257-neighborhood @xmath380 of the curve is contained in @xmath378 , which implies that @xmath57 does not intersect @xmath380 . now , if @xmath375 does not intersect @xmath381 , for @xmath10 small enough , then there is a @xmath8-path @xmath188 of unexplored hexagons connecting the hexagon that contains @xmath50 with the hexagon that contains @xmath51 , and we conclude that @xmath382 . this shows that the event that @xmath370 but @xmath371 implies the existence of a curve @xmath379 whose @xmath257-neighborhood @xmath380 is not intersected by @xmath57 but whose @xmath383-neighborhood @xmath381 is intersected by @xmath375 . this implies that @xmath384 , @xmath385 such that @xmath386 . but the right hand side goes to zero for every @xmath288 as @xmath44 , which concludes the proof of one direction of the claim . to prove the other direction , we consider two points @xmath387 such that @xmath388 but @xmath389 . assume that @xmath50 is trapped before @xmath51 by @xmath57 and suppose for the moment that @xmath378 is a domain of type 3 or 4 ; the case of a domain of type 1 or 2 is analogous and will be treated later . let @xmath145 be the first time @xmath50 is trapped by @xmath57 with @xmath390 the double point of @xmath57 where the domain @xmath378 containing @xmath50 is `` sealed off . '' at time @xmath145 , a new domain containing @xmath50 is created and @xmath51 is disconnected from @xmath50 . choose @xmath288 small enough so that neither @xmath50 nor @xmath51 is contained in the ball @xmath391 of radius @xmath257 centered at @xmath392 , nor in the @xmath257-neighborhood @xmath393,\varepsilon)$ ] of the portion of @xmath57 which surrounds @xmath50 . then it follows from the coupling that , for @xmath10 small enough , there are appropriate parameterizations of @xmath57 and @xmath375 such that the portion @xmath394 $ ] of @xmath395 is inside @xmath393,\varepsilon)$ ] , and @xmath396 and @xmath397 are contained in @xmath391 . for @xmath50 and @xmath51 to be contained in the same domain in the discrete construction , there must be a @xmath8-path @xmath188 of unexplored hexagons connecting the hexagon that contains @xmath50 to the hexagon that contains @xmath51 . > from what we said in the previous paragraph , any such @xmath8-path connecting @xmath50 and @xmath51 would have to go though a `` bottleneck '' in @xmath391 ( see figure [ fig - lemma6 - 2 ] ) . assume now , for concreteness but without loss of generality , that @xmath378 is a domain of type 3 , which means that @xmath57 winds around @xmath50 counterclockwise , and consider the hexagons to the left " of @xmath394 $ ] ( these are all lightly shaded in figure [ fig - lemma6 - 2 ] ) . those hexagons form a `` quasi - loop '' around @xmath50 since they wind around it ( counterclockwise ) and the first and last hexagons are both contained in @xmath391 . the hexagons to the left of @xmath394 $ ] belong to the set @xmath398 , which can be seen as a ( nonsimple ) path by connecting the centers of the hexagons in @xmath398 by straight segments . such a path shadows @xmath375 , with the difference that it can have double ( or even triple ) points , since the same hexagon can be visited more than once . consider @xmath398 as a path @xmath399 with a given parametrization @xmath400 , chosen so that @xmath400 is inside @xmath391 when @xmath395 is , and it winds around @xmath50 together with @xmath395 . -path @xmath188 of unexplored hexagons from @xmath50 to @xmath51 having to go through a bottleneck " due to the fact that the exploration path ( heavy line ) comes close to itself . an approximate location of the continuum double point at @xmath401 is indicated by the small disc in one of the hexagons in the bottleneck area.,width=302 ] now suppose that there were two times , @xmath402 and @xmath403 , such that @xmath404 and @xmath405 $ ] winds around @xmath50 . this would imply that the `` quasi - loop '' of explored yellow hexagons around @xmath50 is actually completed , and that @xmath406 . thus , for @xmath50 and @xmath51 to belong to the same discrete domain , this can not happen . for any @xmath407 , if we take @xmath10 small enough , @xmath399 will be contained inside @xmath408 , due to the coupling . following the considerations above , the fact that @xmath50 and @xmath51 belong to the same domain in the discrete construction but to different domains in the continuum construction implies , for @xmath10 small enough , that there are four disjoint yellow @xmath8-paths crossing the annulus @xmath409 ( the paths have to be disjoint because , as we said , @xmath399 can not , when coming back to @xmath391 after winding around @xmath50 , touch itself inside @xmath391 ) . since @xmath410 is also crossed by at least two blue @xmath8-paths from @xmath411 , there is a total of at least six @xmath8-paths , not all of the same color , crossing the annulus @xmath412 . we can then use lemma [ arms ] to conclude that , if we keep @xmath257 fixed and let @xmath44 and @xmath341 , the probability to see such an event anywhere in @xmath37 goes to zero . in the case in which @xmath50 belongs to a domain of type 1 or 2 , let @xmath142 be the excursion that traps @xmath50 and @xmath413 be the point on the boundary of @xmath37 where @xmath142 starts and @xmath414 the point where it ends . choose @xmath288 small enough so that neither @xmath50 nor @xmath51 is contained in the balls @xmath415 and @xmath391 of radius @xmath257 centered at @xmath416 and @xmath392 , nor in the @xmath257-neighborhood @xmath417 of the excursion @xmath142 . because of the coupling , for @xmath10 small enough ( depending on @xmath257 ) , @xmath375 shadows @xmath57 along @xmath142 , staying within @xmath417 . if this is the case , any @xmath8-path of unexplored hexagons connecting the hexagon that contains @xmath50 with the hexagon that contains @xmath51 would have to go through one of two `` bottlenecks , '' one contained in @xmath415 and the other in @xmath391 . assume for concreteness ( but without loss of generality ) that @xmath50 is in a domain of type 1 , which means that @xmath57 winds around @xmath50 counterclockwise . if we parameterize @xmath57 and @xmath375 so that @xmath418 and @xmath419 , @xmath394 $ ] forms a `` quasi - excursion '' around @xmath50 since it winds around it ( counterclockwise ) and it starts inside @xmath420 and ends inside @xmath421 . notice that if @xmath375 touched @xmath422 , inside both @xmath420 and @xmath421 , this would imply that the `` quasi - excursion '' is a real excursion and that @xmath406 . for any @xmath407 , if we take @xmath10 small enough , @xmath375 will be contained inside @xmath408 , due to the coupling . therefore , the fact that @xmath423 implies , with probability going to one as @xmath44 , that for @xmath288 fixed and any @xmath407 , @xmath375 enters the ball @xmath424 and does not touch @xmath422 inside the larger ball @xmath425 , for @xmath426 or @xmath427 . this is equivalent to having at least two yellow and one blue @xmath8-paths ( contained in @xmath428 ) crossing the annulus @xmath429 . as @xmath44 , we can let @xmath332 go to zero ( keeping @xmath257 fixed ) and use lemma [ arms ] to conclude that the probability that such an event occurs anywhere on the boundary of the unit disc goes to zero . we have shown that , for two fixed points @xmath369 , having @xmath430 but @xmath382 or vice versa implies the occurrence of an event whose probability goes to zero as @xmath44 , and the proof of the lemma is concluded . + * proof of lemma [ boundaries ] . * as in the proof of lemma [ claim - c ] , we let @xmath285 be a convergent subsequence for @xmath267 and let @xmath372 be the limit in distribution of @xmath278 as @xmath134 , and in the rest of the proof consider coupled versions of @xmath431 and @xmath57 . let us introduce the hausdorff distance @xmath432 between two closed nonempty subsets of @xmath136 : @xmath433 with this metric , the collection of closed subsets of @xmath136 is a compact space . we will next prove that @xmath353 converges in distribution to @xmath273 as @xmath368 , in the topology induced by ( [ hausdorff - dist ] ) . ( notice that the coupling between @xmath434 and @xmath57 provides a coupling between @xmath353 and @xmath273 , seen as boundaries of domains produced by the two paths . ) we will now use lemma [ sub - conv ] and take a further subsequence @xmath435 of the @xmath373 s that for simplicity of notation we denote by @xmath436 such that , as @xmath437 , @xmath438 converge jointly in distribution to @xmath439 , where @xmath352 is a simple loop . for any @xmath288 , since @xmath352 is a compact set , we can find a covering of @xmath440 by a finite number of balls of radius @xmath383 centered at points on @xmath352 . each ball contains both points in the interior @xmath441 of @xmath352 and in the exterior @xmath442 of @xmath440 , and we can choose ( independently of @xmath443 ) one point from @xmath441 and one from @xmath442 inside each ball . once again , the convergence in distribution of @xmath444 to @xmath352 implies the existence of a coupling such that , for @xmath443 large enough , the selected points that are in @xmath441 are contained in @xmath445 , and those that are in @xmath442 are contained in the complement of @xmath446 . but by lemma [ claim - c ] , each one of the selected points that is contained in @xmath445 is also contained in @xmath274 with probability going to @xmath427 as @xmath437 ; analogously , each one of the selected points contained in the complement of @xmath446 is also contained in the complement of @xmath447 with probability going to @xmath427 as @xmath437 . this implies that @xmath273 crosses each one of the balls in the covering of @xmath352 , and therefore @xmath448 . > from this and the coupling between @xmath444 and @xmath352 , it follows immediately that , for @xmath443 large enough , @xmath449 with probability close to one . a similar argument ( analogous to the previous one but simpler , since it does not require the use of @xmath352 ) , with the roles of @xmath445 and @xmath274 inverted , shows that @xmath450 with probability going to @xmath427 as @xmath437 . therefore , for all @xmath288 , @xmath451 as @xmath437 , which implies convergence in distribution of @xmath444 to @xmath273 , as @xmath452 , in the topology induced by ( [ hausdorff - dist ] ) . but lemma [ sub - conv ] implies that @xmath444 converges in distribution ( using ( [ distance ] ) ) to a simple loop , therefore @xmath273 must also be a simple loop ; and we have convergence in the topology induced by ( [ distance ] ) . it is also clear that the argument above is independent of the subsequence @xmath453 ( and of the original subsequence @xmath272 ) , so the limit of @xmath269 is unique and coincides with @xmath273 . hence , we have convergence in distribution of @xmath269 to @xmath273 , as @xmath44 , in the topology induced by ( [ distance ] ) , and indeed joint convergence of @xmath454 to @xmath455 . + * proof of corollary [ jordan ] . * the corollary follows immediately from lemma [ sub - conv ] and lemma [ boundaries ] , as already seen in the proof of lemma [ boundaries ] . + * proof of lemma [ strong - smirnov ] . * first of all recall that the convergence of @xmath282 to @xmath283 in distribution implies the existence of coupled versions of @xmath282 and @xmath283 on some probability space @xmath360 such that @xmath456 , @xmath457 , @xmath458 for all @xmath363 as @xmath134 ( see , for example , corollary 1 of @xcite ) . this immediately implies that the conditions to apply rad s theorem ( see theorem [ rado - thm ] of section [ rado ] ) are satisfied . let @xmath128 be the conformal map that takes the unit disc @xmath37 onto @xmath459 with @xmath460 and @xmath461 , and let @xmath41 be the conformal map from @xmath37 onto @xmath7 with @xmath462 and @xmath131 . then , by theorem [ rado - thm ] , @xmath128 converges to @xmath41 uniformly in @xmath136 , as @xmath134 . let @xmath57 ( resp . , @xmath463 ) be the chordal @xmath0 inside @xmath7 ( resp . , @xmath459 ) from @xmath110 to @xmath111 ( resp . , from @xmath464 to @xmath465 ) , @xmath466 , @xmath467 , @xmath468 , and @xmath469 , @xmath470 , @xmath471 . we note that , because of the conformal invariance of chordal @xmath0 , @xmath352 ( resp . , @xmath472 ) is distributed as chordal @xmath0 in @xmath37 from @xmath473 to @xmath474 ( resp . , from @xmath475 to @xmath476 ) . since @xmath477 and @xmath478 for all @xmath363 , and @xmath135 uniformly in @xmath136 , we conclude that @xmath479 and @xmath480 for all @xmath363 . later we will prove a `` continuity '' property of @xmath0 ( lemma [ continuity ] ) that allows us to conclude that , under these conditions , @xmath472 converges in distribution to @xmath352 in the uniform metric ( [ distance ] ) on continuous curves . once again , this implies the existence of coupled versions of @xmath472 and @xmath352 on some probability space @xmath481 such that @xmath482 , for all @xmath363 as @xmath134 . therefore , thanks to the convergence of @xmath128 to @xmath41 uniformly in @xmath136 , @xmath483 , for all @xmath363 as @xmath134 . but since @xmath484 is distributed as @xmath485 and @xmath486 is distributed as @xmath117 , we conclude that , as @xmath134 , @xmath485 converges in distribution to @xmath117 in the uniform metric ( [ distance ] ) on continuous curves . we now note that ( s ) implies that , as @xmath44 , @xmath487 converges in distribution to @xmath485 _ uniformly _ in @xmath132 , for @xmath132 large enough . therefore , as @xmath134 , @xmath286 converges in distribution to @xmath117 , and the proof is concluded . [ continuity ] let @xmath488 be the unit disc , @xmath110 and @xmath111 two distinct points on its boundary , and @xmath57 the trace of chordal @xmath0 inside @xmath37 from @xmath110 to @xmath111 . let @xmath489 and @xmath490 be two sequences of points in @xmath491 such that @xmath492 and @xmath493 . then , as @xmath134 , the trace @xmath463 of chordal @xmath0 inside @xmath37 from @xmath464 to @xmath465 converges in distribution to @xmath57 in the uniform topology ( [ distance ] ) on continuous curves . * let @xmath494 be the ( unique ) linear fractional transformation that takes the unit disc @xmath37 onto itself , mapping @xmath110 to @xmath464 , @xmath111 to @xmath465 , and a third point @xmath495 distinct from @xmath110 and @xmath111 to itself . @xmath496 and @xmath497 depend continuously on @xmath464 and @xmath465 . as @xmath134 , since @xmath492 and @xmath493 , @xmath128 converges uniformly to the identity in @xmath136 . using the conformal invariance of chordal @xmath0 , we couple @xmath463 and @xmath57 by writing @xmath498 . the uniform convergence of @xmath128 to the identity implies that @xmath499 as @xmath134 , which is enough to conclude that @xmath463 converges to @xmath57 in distribution . + * proof of theorem [ thm - convergence ] . * let us prove the second part of the theorem first . we will do this for the original version of the discrete construction , but essentially the same proof works for the reorganized version we will describe below , as we will explain later . suppose that at step @xmath132 of this discrete construction an exploration process @xmath500 is run inside a domain @xmath501 , and write @xmath502 , where @xmath503 are the maximal connected domains of unexplored hexagons into which @xmath501 is split by removing the set @xmath504 of hexagons explored by @xmath500 . let @xmath505 and @xmath506 be respectively the maximal @xmath170- and @xmath171-distances between pairs of points in @xmath507 . suppose , without loss of generality , that @xmath508 , and consider the rectangle @xmath509 ( see figure [ fig3-thm1 ] ) whose vertical sides are aligned to the @xmath171-axis , have length @xmath505 , and are each placed at @xmath170-distance @xmath510 from points of @xmath507 with minimal or maximal @xmath170-coordinate in such a way that the horizontal sides of @xmath509 have length @xmath510 ; the bottom and top sides of @xmath509 are placed in such a way that they are at equal @xmath171-distance from the points of @xmath507 with minimal or maximal @xmath171-coordinate , respectively . it follows from the russo - seymour - welsh lemma @xcite ( see also @xcite ) that the probability to have two vertical @xmath8-crossings of @xmath509 of different colors is bounded away from zero by a positive constant @xmath512 that does not depend on @xmath10 ( for @xmath10 small enough ) . if that happens , then @xmath513 . the same argument of course applies to the maximal @xmath171-distance when @xmath514 . we can summarize the above observation in the following lemma . [ smaller ] suppose that at step @xmath132 of the full discrete construction an exploration process @xmath500 is run inside a domain @xmath501 . if @xmath515 , then for @xmath10 small enough ( i.e. , @xmath516 for some constant @xmath517 ) , @xmath518 with probability at least @xmath512 independent of @xmath10 . the same holds for the maximal @xmath171-distances when @xmath519 . here is another lemma that will be useful later on . ( for an example of the phenomenon described in the lemma , see figure [ fig2-sec4 ] , and assume that the unexplored hexagons there are all blue ; then the s - boundary of the small domain made of a single blue hexagon and that of the blue domain to the northeast share exactly two adjacent yellow hexagons . ) [ daughters ] two daughter " subdomains , @xmath520 and @xmath521 , either have disjoint s - boundaries , or else their common s - boundary consists of exactly two adjacent hexagons ( of the same color ) where the exploration path @xmath500 came within @xmath522 hexagons of touching itself just when completing the s - boundary of one of the two subdomains . * * suppose that the two daughter subdomains have s - boundaries @xmath523 and @xmath524 that are not disjoint and let @xmath525 be the set of ( sites of @xmath8 that are the centers of the ) hexagons that belong to both s - boundaries . @xmath526 can be partitioned into subsets consisting of single hexagons that are not adjacent to any another hexagon in @xmath526 and groups of hexagons that form simple @xmath8-paths ( because the s - boundaries of the two subdomains are simple @xmath8-loops ) . let @xmath527 be such a subset of hexagons of @xmath526 that form a simple @xmath8-path @xmath528 . then there is a @xmath8-path @xmath529 of hexagons in @xmath523 that goes from @xmath530 to @xmath531 without using any other hexagon of @xmath532 and a different @xmath8-path @xmath533 in @xmath524 that goes from @xmath531 to @xmath530 without using any other hexagon of @xmath532 . but then , all the hexagons in @xmath532 other than @xmath530 and @xmath531 are surrounded " by @xmath534 and therefore can not have been explored by the exploration process that produced @xmath520 and @xmath521 , and can not belong to @xmath523 or @xmath524 , leading to a contradiction , unless @xmath535 . similar arguments lead to a contradiction if @xmath526 is partitioned into more than one subset . if @xmath536 is not adjacent to any other hexagon in @xmath526 , then it is adjacent to two other hexagons of @xmath523 and two hexagons of @xmath524 . since @xmath184 has only six neighbors and neither the two hexagons of @xmath523 adjacent to @xmath184 nor those of @xmath524 can be adjacent to each other , each hexagon of @xmath523 is adjacent to one of @xmath524 . but then , as before , @xmath184 is surrounded " by @xmath537 and therefore can not have been explored by the exploration process that produced @xmath520 and @xmath521 , and can not belong to @xmath523 or @xmath524 , leading once again to a contradiction . the proof is now complete , since the only case remaining is the one where @xmath526 consists of a single pair of adjacent hexagons as stated in the lemma . + with these lemmas , we can now proceed with the proof of the second part of the theorem . lemma [ smaller ] tells us that large domains are chopped " with bounded away from zero probability ( @xmath538 ) , but we need to keep track of domains of diameter larger than @xmath257 in such a way as to avoid double counting " as the lattice construction proceeds . more accurately , we will keep track of domains @xmath539 having @xmath540 , since only these can have diameter larger than @xmath257 . to do so , we will associate with each domain @xmath539 having @xmath540 that we encounter as we do the lattice construction a non - negative integer label . the first domain is @xmath541 ( see the beginning of section [ full ] ) and this gets label @xmath427 . after each exploration process in a domain @xmath539 with @xmath540 , if the number @xmath542 of daughter " subdomains @xmath543 with @xmath544 is @xmath545 , then the label of @xmath539 is no longer used , if instead @xmath546 , then one of these @xmath542 subdomains ( chosen by any procedure e.g. , the one with the highest priority for further exploration ) is assigned the _ same _ label as @xmath539 and the rest are assigned the next @xmath547 integers that have never before been used as labels . note that once all domains have @xmath548 , there are no more labelled domains . * proof . * except for @xmath248 , every domain comes with ( at least ) a physically correct " monochromatic half - boundary " ( notice that we are considering s - boundaries and that a half - boundary coming from the artificially colored " boundary of @xmath248 is not considered a physically correct monochromatic half - boundary ) . let us assume , without loss of generality , that @xmath551 . if we associate with each label the last " ( in terms of steps of the discrete construction ) domain which used that label ( its daughter subdomains all had @xmath548 ) , then we claim that it follows from lemma [ daughters ] that ( with high probability ) any two such last domains that are labelled have disjoint s - boundaries . this is a consequence of the fact that the two domains are subdomains of two ancestors " that are distinct daughter subdomains of the _ same _ domain ( possibly @xmath248 ) and whose s - boundaries are therefore ( by lemma [ daughters ] ) either disjoint or else overlap at a pair of hexagons where an exploration path had a close encounter of distance two hexagons with itself . but since we are dealing only with macroscopic domains ( of diameter at least order @xmath257 ) , such a close encounter would imply , like in lemmas [ sub - conv ] and [ boundaries ] , the existence of six crossings , not all of the same color , of an annulus whose outer radius can be kept fixed while the inner radius is sent to zero together with @xmath10 . the probability of such an event goes to zero as @xmath44 and hence the unit disc @xmath37 contains , with high probability , at least @xmath549 disjoint monochromatic @xmath8-paths of diameter at least @xmath552 , corresponding to the physically correct half - boundaries of the @xmath549 labelled domains . now take the collection of squares @xmath553 of side length @xmath554 centered at the sites @xmath555 of a scaled square lattice @xmath556 of mesh size @xmath332 , and let @xmath557 be the number of squares of side @xmath332 needed to cover the unit disc . let @xmath558 and consider the event @xmath559 , which implies that , with high probability , the unit disc contains at least @xmath560 disjoint monochromatic @xmath8-paths of diameter at least @xmath552 and that , for at least one @xmath561 , the square @xmath562 intersects at least six disjoint monochromatic @xmath8-paths of diameter larger that @xmath552 , so that the annulus " @xmath563 is crossed by at least six disjoint monochromatic @xmath8-paths contained inside the unit disc . if all these @xmath8-paths crossing @xmath563 have the same color , say blue , then since they are portions of boundaries of domains discovered by exploration processes , they are shadowed " by exploration paths and therefore between at least one pair of blue @xmath8-paths , there is at least one yellow @xmath8-path crossing @xmath563 . therefore , whether the original monochromatic @xmath8-paths are all of the same color or not , @xmath563 is crossed by at least six disjoint monochromatic @xmath8-paths not all of the same color contained in the unit disc . let @xmath564 denote the @xmath565 as @xmath44 of the probability that such an event happens anywhere inside the unit disc . we have shown that the event @xmath559 implies a six - arms " event unless not all labelled domains have disjoint s - boundaries . but the latter also implies a six - arms " event , as discussed before ; therefore @xmath566 since @xmath567 , bounds in @xcite imply that , for @xmath257 fixed , @xmath568 as @xmath341 , which shows that @xmath569 and concludes the proof of the lemma . + now , let @xmath570 denote the number of distinct domains that had label @xmath168 ( this is equal to the number of steps that label @xmath168 survived ) . let us also define @xmath571 to be the smallest integer @xmath572 such that @xmath573 and @xmath574 to be the random variable corresponding to how many bernoulli trials ( with probability @xmath512 of success ) it takes to have @xmath575 successes . then , we may apply ( sequentially ) lemma [ smaller ] to conclude that for any @xmath168 @xmath576 where @xmath577 is an independent copy of @xmath574 . now let @xmath578 be i.i.d . random variables equidistributed with @xmath579 . let @xmath580 be the number of steps needed so that all domains left to explore have @xmath581 . then , for any positive integer @xmath582 , @xmath583 notice that , for fixed @xmath582 , @xmath584 as @xmath585 . moreover , for any @xmath586 , by lemma [ labels ] , we can choose @xmath587 large enough so that @xmath588 . so , for any @xmath586 , it follows that @xmath589 which implies that @xmath590 to conclude this part of the proof , notice that the discrete construction can not `` skip '' a contour and move on to explore its interior , so that all the contours with diameter larger than @xmath257 must have been found by step @xmath132 if all the domains present at that step have diameter smaller than @xmath257 . therefore , @xmath591 , which shows that @xmath289 is bounded in probability as @xmath44 . + for the first part of the theorem , we need to prove , for any fixed @xmath287 , joint convergence in distribution of the first @xmath132 steps of a suitably reorganized discrete construction to the first @xmath132 steps of the continuum one . later we will explain why this reorganized construction has the same scaling limit as the one defined in section [ full ] . for each @xmath132 , the first @xmath132 steps of the reorganized discrete construction will be coupled to the first @xmath132 steps of the continuum one with suitable couplings in order to obtain the convergence in distribution of those steps of the discrete construction to the analogous steps of the continuum one ; the proof will proceed by induction in @xmath132 . we will explain how to reorganize the discrete construction as we go along ; in order to explain the idea of the proof , we will consider first the cases @xmath592 , @xmath522 and @xmath593 , and then extend to all @xmath594 . + @xmath592 . the first step of the continuum construction consists of an @xmath0 @xmath62 from @xmath167 to @xmath168 inside @xmath37 . correspondingly , the first step of the discrete construction consists of an exploration path @xmath250 inside @xmath428 from the e - vertex closest to @xmath167 to the e - vertex closest to @xmath168 . the convergence in distribution of @xmath250 to @xmath62 is covered by statement ( s ) . + @xmath595 . the convergence in distribution of the percolation exploration path to chordal @xmath0 implies that we can couple @xmath250 and @xmath62 generating them as random variables on some probability space @xmath360 such that @xmath596 for all @xmath363 as @xmath134 ( see , for example , corollary 1 of @xcite ) . now , let @xmath23 be the domain generated by @xmath62 that is chosen for the second step of the continuum construction , and let @xmath597 be the highest ranking point of @xmath165 contained in @xmath23 . for @xmath10 small enough , @xmath598 is also contained in @xmath428 ; let @xmath599 be the unique connected component of the set @xmath600 containing @xmath598 ( this is well - defined with probability close to @xmath427 for small @xmath10 ) ; @xmath601 is the domain where the second exploration process is to be carried out . > from the proof of lemma [ boundaries ] , we know that the boundaries @xmath602 and @xmath603 of the domains @xmath601 and @xmath23 produced respectively by the path @xmath604 and @xmath62 are close with probability close to one for @xmath10 small enough . for the next step of the discrete construction , we choose the two e - vertices @xmath605 and @xmath606 in @xmath602 that are closest to the points @xmath607 and @xmath608 of @xmath603 selected for the coupled continuum construction ( if the choice is not unique , we can select the e - vertices with any rule to break the tie ) and call @xmath609 the percolation exploration path inside @xmath610 from @xmath605 to @xmath606 . it follows from @xcite that @xmath611 converge jointly in distribution along some subsequence to some limit @xmath612 . we already know that @xmath613 is distributed like @xmath62 and we can deduce from the joint convergence in distribution of @xmath614 to @xmath615 ( lemma [ boundaries ] ) , that @xmath616 is distributed like @xmath603 . therefore , if we call @xmath63 the @xmath0 path inside @xmath23 from @xmath607 to @xmath608 , lemma [ strong - smirnov ] implies that @xmath617 is distributed like @xmath63 and indeed that , as @xmath44 , @xmath618 converge jointly in distribution to @xmath619 . + @xmath620 . so far , we have proved the convergence in distribution of the ( paths and boundaries produced in the ) first two steps of the discrete construction to the ( paths and boundaries produced in the ) first two steps of the discrete construction . the third step of the continuum construction consists of an @xmath0 path @xmath621 from @xmath622 to @xmath623 , inside the domain @xmath24 with highest priority after the second step has been completed . let @xmath624 be the highest ranking point of @xmath165 contained in @xmath24 , @xmath625 the domain of the discrete construction containing @xmath626 after the second step of the discrete construction has been completed ( this is well defined with probability close to @xmath427 for small @xmath10 ) , and choose the two e - vertices @xmath627 and @xmath628 in @xmath629 that are closest to the points @xmath630 and @xmath631 of @xmath632 selected for the coupled continuum construction ( if the choice is not unique , we can select the e - vertices with any rule to break the tie ) . the third step of the discrete construction consists of an exploration path @xmath633 from @xmath627 to @xmath628 inside @xmath625 . it follows from @xcite that @xmath634 converge jointly in distribution along some subsequence to some limit @xmath635 . we already know that @xmath613 is distributed like @xmath62 , @xmath636 like @xmath603 and @xmath617 like @xmath63 , and we would like to apply lemma [ strong - smirnov ] to conclude that @xmath637 is distributed like @xmath621 and indeed that , as @xmath44 , @xmath638 converges in distribution to @xmath639 . in order to do so , we have to first show that @xmath640 is distributed like @xmath632 . if @xmath641 is a subset of @xmath642 , this follows from lemma [ boundaries ] , as in the previous case , but if the s - boundary of @xmath641 contains hexagons of @xmath643 , then we can not use lemma [ boundaries ] directly , although the proof of the lemma can be easily adapted to the present case , as we now explain . indeed , the only difference is in the proof of claim ( c ) and is due to the fact that , when dealing with a domain of type 1 or 2 , we can not use the bound on the probability of three disjoint crossings of a semi - annulus because the domains we are dealing with may not be convex ( like the unit disc ) . on the other hand , the discrete domains like @xmath610 and @xmath641 where we have to run exploration processes at various steps of the discrete construction are themselves generated by previous exploration processes , so that any hexagon of the s - boundary of such a domain has three adjacent hexagons which are the starting points of three disjoint @xmath8-paths ( two of one color and one of the other ) . two of these @xmath8-paths belong to the s - boundary of the domain , while the third belongs to the adjacent percolation cluster ( see figure [ fig2-thm1 ] ) . this allows us to use the bound on the probability of _ six _ disjoint crossings of an annulus . to the left of the exploration path indicated by a heavy line , has three neighbors that are the starting points of two disjoint yellow @xmath8-paths ( denoted 1 and 2 ) belonging to the s - boundary of @xmath644 and one blue @xmath8-path ( denoted 3 ) belonging to the adjacent percolation cluster.,width=302 ] to see this , let @xmath645 be the @xmath8-paths contained in the s - boundary of the discrete domain ( i.e. , @xmath610 in the present context ) and @xmath646 the @xmath8-path belonging to the adjacent cluster , all starting from hexagons adjacent to some hexagon @xmath214 ( centered at @xmath50 ) in the s - boundary of @xmath610 . for @xmath647 and @xmath10 small enough , let @xmath648 be the event that the exploration path @xmath649 enters the ball @xmath650 without touching @xmath651 inside the larger ball @xmath652 . @xmath648 implies having ( at least ) three disjoint @xmath8-paths ( two of one color and one of the other ) , @xmath653 and @xmath654 , contained in @xmath610 and crossing the annulus @xmath655 , with @xmath653 and @xmath654 disjoint from @xmath645 and @xmath646 . hence , @xmath648 implies the event that there are ( at least ) six disjoint crossings ( not all of the same color ) of the annulus @xmath655 . once claim ( c ) is proved , the rest of the proof of lemma [ boundaries ] applies to the present case . therefore , we have convergence in distribution of @xmath629 to @xmath632 , which allows us to use lemma [ strong - smirnov ] and conclude that @xmath656 converges in distribution to @xmath657 . + @xmath658 . we proceed by induction in @xmath132 , iterating the steps explained above ; there are no new difficulties ; all steps for @xmath659 are analogous to the case @xmath620 . + to conclude the proof of the theorem , we need to show that the scaling limit of the original full discrete construction defined in section [ full ] is the same as that of the reorganized one just used in the proof of the first part of the theorem . in order to do so , we can couple the two constructions by using the same percolation configuration for both , so that the two constructions have at their disposal the same set of loops to discover . we proved above that the original discrete construction finds _ all _ the macroscopic " loops , so we have to show that this is true also for the reorganized version of the discrete construction . this is what we will do next , using essentially the same arguments as those employed for the original discrete construction ; we present these arguments for the sake of completeness since there are some changes . consider the reorganized discrete construction described above , where the starting and ending points of the exploration processes at each step are chosen to be close to those of the corresponding ( coupled ) continuum construction . suppose that at step @xmath132 of this discrete construction an exploration process @xmath500 is run inside a domain @xmath501 , and write @xmath502 , where @xmath503 are the connected domains into which @xmath501 is split by the set @xmath504 of hexagons explored by @xmath500 . let @xmath660 ( resp . , @xmath505 ) and @xmath661 ( resp . , @xmath506 ) be respectively the maximal @xmath170- and @xmath171-distance between pairs of points in @xmath662 ( resp . , @xmath507 ) . if @xmath508 and the e - vertices on @xmath507 are chosen to be closest to two points of @xmath662 with maximal @xmath170-distance , then the same construction and argument spelled out earlier in the first part of the proof ( corresponding to the second part of the theorem ) show that @xmath513 with bounded away from zero probability . if the e - vertices on @xmath507 are chosen to be closest to two points of @xmath662 with maximal @xmath170-distance but @xmath663 , then consider the rectangle @xmath664 whose vertical sides are aligned to the @xmath171-axis , have length @xmath506 , and are each placed at the same @xmath170-distance from the points of @xmath507 with minimal or maximal @xmath170-coordinate in such a way that the horizontal sides of @xmath664 have length @xmath665 ; the bottom and top sides of @xmath664 are placed in such a way that they touch the points of @xmath507 with minimal or maximal @xmath171-coordinate , respectively . notice that , because of the coupling between the continuum and discrete constructions , for any @xmath666 , for @xmath132 large enough , @xmath667 and @xmath668 . since in the case under consideration we have @xmath669 and @xmath670 , for @xmath10 large enough , we must also have @xmath671 . once again , it follows from the russo - seymour - welsh lemma that the probability to have two vertical @xmath8-crossings of @xmath664 of different colors is bounded away from zero by a positive constant that does not depend on @xmath10 ( for @xmath10 small enough ) . if that happens , then @xmath672 . all other cases are handled in the same way , implying that the maximal @xmath170- and @xmath171-distances of domains that appear in the discrete construction have a positive probability ( bounded away from zero ) to decrease by ( approximately ) a factor @xmath673 at each step of the discrete construction in which an exploration process is run in that domain . with this result at our disposal , the rest of the proof , that for any @xmath288 the number of steps needed to find all the loops of diameter larger than @xmath257 is bounded in probability as @xmath44 ( which implies that _ all _ the macroscopic " loops are discovered ) , proceeds exactly like for the original discrete construction . + * proof of theorem [ thm - therm - lim ] . * first of all , we want to show that @xmath674 does not depend on @xmath17 , provided @xmath7 is strictly contained in @xmath16 and @xmath675 . in order to do this , we assume that the above conditions are satisfied for the pair @xmath676 and show that @xmath677 for all @xmath678 . take two copies of the scaled hexagonal lattice , @xmath11 and @xmath679 , their dual lattices @xmath680 and @xmath681 , and two percolation configurations , @xmath682 and @xmath683 , both with blue boundary conditions and coupled in such a way that @xmath684 . the laws of the boundaries of @xmath68 and @xmath685 are also coupled , in such a way that the boundaries or portions of boundaries contained inside @xmath7 are identical for all small enough @xmath10 . therefore , letting @xmath44 and using the convergence of the percolation boundaries inside @xmath16 and @xmath686 to the continuum nonsimple loop processes @xmath292 and @xmath687 respectively , we conclude that @xmath677 . > from what we have just proved , it follows that the probability measures @xmath688 on @xmath689 , for @xmath690 , satisfy the consistency conditions @xmath691 for all @xmath692 . since @xmath66 , @xmath75 are complete separable metric spaces , the measurable spaces @xmath689 , @xmath693 are standard borel spaces and so we can apply kolmogorov s extension theorem ( see , for example , @xcite ) and conclude that there exists a unique probability measure on @xmath693 with @xmath694 for all @xmath690 . it follows that , for @xmath695 and all @xmath7 strictly contained in @xmath16 and such that @xmath675 , @xmath696 , which concludes the proof . + * proof of theorems [ main - result1 ] and [ main - result2 ] . * these are immediate consequences of theorems [ thm - convergence ] and [ thm - therm - lim ] , where the full scaling limit is intended in the topology induced by ( [ hausdorff - d ] ) . + * proof of theorem [ features ] . * _ 1 . _ the fact that the continuum nonsimple loop process is a random collection of noncrossing continuous loops is a direct consequence of its definition . the fact that the loops touch themselves is a consequence of their being constructed out of @xmath0 , while the fact that they touch each other follows from the observation that a chordal @xmath0 path @xmath117 touches @xmath39 with probability one . therefore , each new loop in the continuum construction touches one or more previous ones ( many times ) . the nonexistence of triple points follows directly from lemma 5 of @xcite on the number of crossings of an annulus , combined with theorem [ main - result1 ] , which allows to transport discrete results to the continuum case . in fact , a triple point would imply , for discrete percolation , at least six crossings ( not all of the same color ) of an annulus whose ratio of inner to outer radius goes to zero in the scaling limit , leading to a contradiction . _ 3 . _ combining russo - seymour - welsh type arguments for percolation ( see , for example , lemma 3 of @xcite ) with theorem [ main - result1 ] , we know that @xmath296-a.s . there exists a ( random ) @xmath697 , with @xmath698 , such that @xmath16 is surrounded by a continuum nonsimple loop contained in @xmath699 . > from ( the proof of ) theorem [ thm - therm - lim ] , we also know that @xmath700 for all @xmath701 . this implies that by taking @xmath702 large enough and performing the continuum construction inside @xmath686 , we have a positive probability of generating a loop @xmath703 contained in the annulus @xmath704 , with @xmath705 . if that is the case , all the loops contained inside @xmath16 are connected , by construction , to the loop @xmath703 surrounding @xmath16 by a finite sequence ( a `` path '' ) of loops ( remember that in the continuum construction each loop is generated by pasting together portions of @xmath0 paths inside domains whose boundaries are determined by previously formed loops or excursions ) . therefore , any two loops contained inside @xmath16 are connected to each other by a path " of loops . using again the fact that @xmath700 for all @xmath701 , and letting first @xmath702 and then @xmath706 go to @xmath707 , we see from the discussion above ( with @xmath291 as well ) that any two loops are connected by a finite path " of intermediate loops , @xmath296-a.s . + * proof of theorem [ thm - conf - inv ] . * combining russo - seymour - welsh type arguments for percolation ( see , for example , lemma 3 of @xcite ) with theorem [ main - result1 ] , we know that @xmath296-a.s . there exists a bounded continuum nonsimple loop that surrounds both @xmath299 and @xmath300 , so that @xmath301 and @xmath302 are both bounded . we can then take @xmath708 such that @xmath299 and @xmath300 ( and therefore @xmath301 and @xmath302 ) are both contained in the disc @xmath16 with probability tending to @xmath427 as @xmath291 . consider now the continuum construction inside the disc @xmath16 for some large @xmath17 . let @xmath709 ( resp . , @xmath710 ) be the smallest loop surrounding @xmath23 ( resp . , @xmath24 ) produced by the construction and let @xmath711 ( resp . , @xmath712 ) be the connected component of @xmath713 ( resp . , @xmath714 ) that contains @xmath23 ( resp . , @xmath24 ) . it follows from the previous observation and from ( the proof of ) theorem [ thm - therm - lim ] that as @xmath291 , @xmath711 ( resp . , @xmath712 ) is ( with probability tending to @xmath427 ) distributed like @xmath301 ( resp . , @xmath302 ) and moreover the loop configuration inside @xmath711 ( resp . , @xmath712 ) is distributed by @xmath312 ( resp . , @xmath715 ) . this already proves the first claim of the theorem , since it is clear from the continuum construction inside @xmath16 that the loop configurations inside @xmath711 and @xmath712 are independent . it also means that in order to complete the proof of theorem , it suffices to prove the second claim for the case of the continuum construction inside @xmath16 , for all large @xmath17 . in order to do that , we consider a modified discrete construction inside @xmath16 , as explained below . in view of the above observations , we take @xmath17 large and condition on the existence inside @xmath16 of two disjoint loops , @xmath716 and @xmath717 , surrounding @xmath23 and @xmath24 respectively , and let @xmath718 ( resp . , @xmath719 ) be the domain of @xmath720 ( resp . , @xmath721 ) containing @xmath23 ( resp . , @xmath24 ) . the modified discrete construction inside @xmath16 is analogous to the ordinary " one except inside the domains @xmath718 and @xmath719 , where the the exploration paths are coupled to a continuum construction inside the unit disc in the following way . roughly speaking , the discrete construction inside @xmath718 is one in which the @xmath722 pairs ( the starting and ending points of the exploration paths ) at each step are chosen to be closest to the @xmath723 points in @xmath718 mapped from the unit disc @xmath724 via @xmath725 , where the pairs @xmath284 are those that appear at the corresponding steps of the continuum construction inside @xmath724 and @xmath725 is a certain conformal map from @xmath724 onto @xmath718 , as specified below . the discrete construction inside @xmath719 is coupled in the same way to the _ same _ continuum construction inside @xmath37 via a certain conformal map @xmath726 from @xmath724 onto @xmath719 . the conformal map @xmath725 will be defined for @xmath10 sufficiently small and is specified in the following way . we fix a point @xmath727 in @xmath711 and denote by @xmath728 the unique conformal map from @xmath37 onto @xmath711 such that @xmath729 and @xmath730 . for @xmath10 sufficiently small , so that @xmath727 is contained in @xmath718 , we let @xmath725 be the unique conformal map from @xmath37 onto @xmath718 such that @xmath731 and @xmath732 . we also denote by @xmath733 the unique conformal map from @xmath37 onto @xmath712 such that @xmath734 and @xmath735 , where @xmath736 is any fixed conformal map from @xmath711 to @xmath712 . note that , by the uniqueness part of riemann s mapping theorem , we can conclude that @xmath737 . for @xmath10 sufficiently small , so that @xmath738 is contained in @xmath719 , we let @xmath726 be the unique conformal map from @xmath37 onto @xmath719 such that @xmath739 and @xmath740 . as @xmath44 , @xmath741 and @xmath742 , and by an application of rad s theorem ( theorem [ rado - thm ] ) , ( the continuous extensions of ) @xmath725 and @xmath726 converge uniformly in @xmath136 to the ( continuous extensions of ) @xmath728 and @xmath733 respectively . we now describe more precisely the modified construction inside @xmath718 . let @xmath62 be the first @xmath0 path in @xmath37 from @xmath607 to @xmath608 ; because of the conformal invariance of @xmath0 , the image @xmath743 of @xmath62 under @xmath725 is a path distributed as the trace of chordal @xmath0 in @xmath718 from @xmath744 to @xmath745 . the uniform convergence of @xmath725 to @xmath728 and statement ( s ) imply that the exploration path @xmath250 inside @xmath718 from @xmath605 to @xmath606 , chosen to be closest to @xmath744 and @xmath745 respectively , converges in distribution to @xmath746 , as @xmath44 , which means that there exists a coupling so that the paths @xmath250 and @xmath743 stay close for @xmath10 small . one can use the same strategy as in the proof of the first part of theorem [ thm - convergence ] , and obtain a discrete construction whose exploration paths are coupled to the @xmath0 paths @xmath747 that are the images of the paths @xmath463 in @xmath37 . then , for this discrete construction inside @xmath718 , the scaling limit of the exploration paths will be paths inside @xmath711 distributed as the images of the @xmath0 paths in @xmath37 under the conformal map @xmath728 . analogously , for the discrete construction inside @xmath719 , the scaling limit of the exploration paths will be paths inside @xmath712 distributed as the images of the @xmath0 paths in @xmath37 under the ( continuous extension of the ) conformal map @xmath748 which is the uniform limit of @xmath726 as @xmath44 . therefore , the path inside @xmath712 obtained as the scaling limit of an exploration path at a given step of the construction inside @xmath719 is the image under the conformal map @xmath749 of the path inside @xmath711 obtained as the scaling limit of the exploration path at corresponding step of the construction inside @xmath718 . in order to conclude the proof , we have to show that the discrete constructions inside @xmath750 and @xmath751 defined above find all the boundaries ( with diameter greater than @xmath257 ) in a number of steps that is bounded in probability as @xmath44 . this is as in the second part of theorem [ thm - convergence ] , but since we are now dealing with a modified construction inside general jordan domains , we need to show that we can reach the same conclusion . in order to do so , we will use the fact that the modified construction is coupled to an ordinary " continuum construction in the unit disc . we work out the details only for @xmath718 , since the proof is the same for @xmath719 . > from the second part of theorem [ thm - convergence ] it follows that , for any fixed @xmath752 and @xmath753 , the probability that the number of steps of the continuum construction in @xmath37 that are necessary to ensure that only domains with diameter less than @xmath754 are present is larger than @xmath517 , goes to zero as @xmath585 . since @xmath725 ( can be extended to a function that ) is continuous in the compact set @xmath136 , @xmath725 is uniformly continuous and so we can choose @xmath755 such that any subdomain of @xmath37 of diameter at most @xmath756 is mapped by @xmath725 to a subdomain of @xmath718 of diameter at most @xmath257 . since @xmath757 , as @xmath44 , where @xmath728 ( can be extended to a function that ) is continuous in the compact set @xmath136 and is therefore uniformly continuous , we can choose @xmath758 such that any subdomain of @xmath37 of diameter at most @xmath759 is mapped by @xmath725 to a subdomain of @xmath711 of diameter at most @xmath257 , and moreover such that @xmath760 . this , combined with the coupling between @xmath0 paths and exploration paths inside @xmath539 , assures that the number of steps necessary for the new discrete construction inside @xmath718 to find all the loops of diameter at least @xmath257 is bounded in probability as @xmath44 . therefore , the scaling limit , as @xmath44 , of the modified discrete constructions for @xmath718 and @xmath719 give the measures @xmath761 and @xmath762 , and it follows by construction that @xmath763 for _ any _ conformal map @xmath736 from @xmath711 onto @xmath712 . since this is true for all large @xmath17 , by letting @xmath291 we can conclude that @xmath308 . * acknowledgements . * we are grateful to greg lawler , oded schramm and wendelin werner for various interesting and useful conversations and to stas smirnov for communications about a paper in preparation . f. c. thanks wendelin werner for an invitation to universit paris - sud 11 , and vincent beffara and luiz renato fontes for many helpful discussions . we thank michael aizenman , lai - sang young and an anonymous referee for comments about the presentation of our results . f. c. and c. m. n. acknowledge respectively the kind hospitality of the courant institute and of the vrije universiteit amsterdam where part of this and related work was done . m. aizenman , scaling limit for the incipient spanning clusters , in _ mathematics of multiscale materials ; the i m a volumes in mathematics and its applications _ ( k. golden , g. grimmett , r. james , g. milton and p. sen , eds . ) , springer ( 1998 ) . p. d. seymour , d. j. a. welsh , percolation probabilities on the square lattice , in _ advances in graph theory _ ( b. bollobs ed . ) , annals of discrete mathematics * 3 * , north - holland , amsterdam , pp . 227 - 245 ( 1978 ) . s. smirnov , critical percolation in the plane . i. conformal invariance and cardy s formula . ii . continuum scaling limit . ( long version of @xcite , dated nov . 15 , 2001 ) , available at http://www.math.kth.se/@xmath765stas/papers/index.html .

the currently best - studied theories of gamma - ray bursts ( grbs ) and their afterglows ( ags ) are the _ fireball _ models ( see , e.g. , zhang & meszaros 2003 for a recent review ) and the _ cannonball _ ( cb ) model ( see , e.g. , dar & de rjula 2003a ; dado , dar & de rjula , 2002a ; 2003a and references therein ) . the first set of models is often considered to be _ the standard model _ of grbs . in spite of their similarly - sounding names , these two models are ( or were initially ) completely different in their basic hypothesis , in their description of the data , and in their predictions . in this note we concentrate on a cb - model prediction which is not ( to date ) a standard - model one , the apparently superluminal motion of the source of grbs and their afterglows : the cannonballs " . in quite exceptional cases relatively close - by grbs with sufficiently bright ( radio ) ags it may be possible to observe this superluminal motion of cbs relative to `` fixed stars '' . prior to grb 030329 , the case in which a possible superluminal motion ( dar & de rjula 2000 ) came closest to being observable was that of grb 980425 , which could `` almost '' be classified ( dar & de rjula 2003a ) as an x - ray flash ( xrf ) . for xrfs , the observation of a superluminal motion may be simpler than for grbs , for the source s apparent displacement in the sky is proportional to the ( small ) observer s viewing angle , and we interpret xrfs as jetted grbs viewed at larger angles ( dar & de rjula 2003a ; dado , dar & de rjula 2003f ) . in a different model , this interpretation of xrfs has also been advocated by yamazaki , yonetoku & nakamura ( 2003 ) . the cb - model is a very explicit elaboration of the original proposal by shaviv and dar ( 1995 ) : that the @xmath0-rays of a grb would be generated by inverse compton scattering ( ics ) of stellar light by the electron constituents of transient narrow jets , emitted in stellar processes leading to gravitational collapse . in the cb model , long - duration grbs , as well as xrfs ( dar & de rjula 2003a ; dado , dar & de rjula 2003f ) are produced in the explosions of _ ordinary _ core - collapse sne , akin to sn1998bw ( dar & de rjula 2000 ) , the first sn to be observed in `` association '' with a grb ( grb 980425 , galama et al . two opposite jets of cbs are emitted in the process , travelling with initially large lorentz factors : @xmath1 . the cbs initially expand ( in their rest system ) at a velocity comparable to , or smaller than , the speed of sound in a relativistic plasma ( @xmath2 ) , so that the jet opening angle ( subtended by a cb s radius as observed from its emission point ) is @xmath3 . the cbs highly relativistic motion collimates their emitted radiation the grb and its ag within a forward beam of characteristic opening angle @xmath4 . an observer sees the `` doppler - favoured '' jet , travelling at a small angle @xmath5 relative to the line of sight . typically @xmath6 , so that the jet s opening angle can be neglected and the observer s angle is the _ only _ relevant one . in the cb model , during the ag , the jet opening angle _ diminishes _ with time , while the beaming angle @xmath7 increases , both evolutions being due to the cb s interaction with the matter of the interstellar medium ( dado , dar & de rjula , 2002a ) . as a consequence , the ag s source becomes increasingly `` pointlike '' , and its motion in the sky can in principle be followed . since a cb s motion is relativistic for days or months of ( observer s ) time , its apparent displacement in the sky is _ superluminal _ ( courdec 1939 , rees 1967 ) , as we argued in dar & de rjula ( 2000 ) . the closest - by grb observed so far grb 980425 , at a redshift @xmath8 came very close to having an observable superluminal motion ( see dado , dar & de rjula 2003a for a detailed discussion ) . in a gcn note ( dar & de rjula 2003b ) we argued that this motion may be observable in the next - closest grb ( 030329 , at @xmath9 ) . in this note we sharpen the predictions for this putative motion , given the current availability of precise optical data at early times and times later than the first @xmath10 days ( lipkin et al . 2003 and references therein ) , as well as sparse x - ray data ( marshall & swank 2003 ; marshall , markwardt & swank 2003 ; tiengo et al . 2003 ) and abundant radio data ( sheth et al . 2003 ; berger et al . 2003 ; pooley 2003 ; kuno et al . 2004 ) . a very relevant new information ( lipkin et al . 2003 and references therein ) on the ag of grb 030329 is the abundance of multiple deviations of the optical light curves relative to a smoothly declining behaviour . we shall refer to these deviations as _ features_. in dado et al . ( 2003c ) we attributed the most obvious optical - ag feature a shoulder " starting at @xmath11 1 day ( after burst) to a transition between a first to a second dominant cb , a choice supported by the fact that the @xmath0-ray light - curve of this grb has a very clear two - pulse structure ( vanderspek et al . 2003 ; http:// space.mit.edu/ hete/ bursts/ grb030329/ ; see also vanderspeck et al . 2004 ) , as shown in fig . ( [ grb ] ) . with the emergence of a handful of other similarly - significant features in the first week fast ups and downs of the optical fluences , @xmath12 , by some 20 to 40% a more elaborate interpretation is required . in the ag model @xmath12 is a direct and _ quasi - local _ tracer of the density of the ism through which a cb travels , a fact to which we have attributed previous similar observations , e.g. the humps " in the optical ags of grbs 000301c and 970508 ( dado et al . it is therefore necessary to investigate the effect of local ism density - inhomogeneities on the expected superluminal motion . this is what we do in this note for the case of grb 030329 . we conclude that the sources motion results in a displacement of @xmath13 ( 0.6 ) mas from day 3 to day 30 ( 100 ) after burst . detecting such a motion may not be out of the question . for the benefit of readers not familiar with the cb model , we present in an appendix a brief overview of the model and its current confrontation with data . we also offer there some commentary on the evolution of the standard model . the lorentz factor @xmath14 of a cb diminishes with the distance @xmath15 from the parent sn , as its motion is decelerated by collisions with the constituents of the interstellar medium ( ism ) . in an approximately hydrogenic ism of local number density @xmath16 , the evolution of @xmath17 is determined by energy - momentum conservation to be : @xmath18 with @xmath19 the cb s baryon number and @xmath20 its calculable asymptotic radius , reached within minutes of observer s time . for typical parameters , by the time @xmath21 , the function @xmath17 is still extremely well approximated by its initial value @xmath22 ( dado et al . 2002a ) . for an object travelling at nearly the speed of light the relation between @xmath23 and the element of observer s time , @xmath24 , is : @xmath25 where @xmath26 is the cosmological redshift and @xmath27 is the doppler factor by which the energy of photons emitted by the cb is locally boosted by its relativistic motion . for the lorentz factors and viewing angles relevant to our discussion , @xmath28 and @xmath29 , and @xmath30 to an excellent approximation . we may integrate eqs . ( [ dgamma1],[dgamma2 ] ) to obtain : @xmath31 which , for any given @xmath16 , yields the explicit @xmath15-dependent lorentz factor @xmath17 . substitute this result , with use of eq . ( [ doppler ] ) , into eq . ( [ dxdt ] ) and integrate : @xmath32 to obtain the explicit relation between observer s time @xmath33 and travelled distance @xmath34 the last two equations are a parametric description of the function of actual interest , @xmath35 . for some simple density profiles the above formal exercise can be carried explicitly to the end . for a constant - density ism , for instance , we may define : @xmath36 and conclude that @xmath35 is the real root of the cubic : @xmath37= { 2\,c\ , t\over 3\ , ( 1+z)\ , x_\infty}\ ; , \label{cubic0}\ ] ] that is : @xmath38\ , , \nonumber\\ c&\equiv & \left[{2/ \left(b^2 + 2\,\theta^6+b\,\sqrt{b^2 + 4\,\theta^6}\right)}\right]^{1/3}\ , , \nonumber\\ b&\equiv & { 1/ \gamma_0 ^ 3}+{3\,\theta^2/\gamma_0}+ { 6\,c\ , t/ [ ( 1+z)\ , x_\infty]}\ , , \label{cubic}\end{aligned}\ ] ] while the distance travelled by a cb at a given observer s time is : @xmath39\ ; , \label{xoft}\ ] ] so that it takes a distance @xmath40 for the cb to half its original lorentz factor . as we shall see in detail in the case of grb 030329 , by the time the distance @xmath41 is reached , the ag fluence , proportional to a high power of @xmath35 , has decreased by more than one order of magnitude . we have performed cb - model fits to the ag measurements of all the grbs of known redshift , and found that the values of @xmath41 are spread over more than an order of magnitude , centering at @xmath42 pc ( dado et al 2002a ) . these are comparable to the radii of the _ superbubbles _ wherein most sn explosions take place , created by the stellar winds and explosions of many previous sne in dense star - formation regions . the approximation of a constant density @xmath43 , employed in the quoted fits , may be a fair first try for the density within the superbubble and in the galactic - halo regions close - to but without the superbubble . in the case of grb 030329 we shall see that the density profile of the superbubble s surface , as traced by the cb s x - ray and optical fluences , is quite rich and interesting . the transverse projected velocity in the sky of a cb relative to its parent sn can be obtained from eq . ( [ doppler ] ) : @xmath44 which , for typical parameters , is extremely superluminal . the resulting angular separation at time @xmath33 is : @xmath45\ ; , \label{deltaalpha}\ ] ] where @xmath46 is the angular distance to the sn / cb system and @xmath47 is the luminosity distance ( we use throughout a cosmology with @xmath48 and @xmath49 ) . the last expression in the r.h.s . of eq . ( [ deltaalpha ] ) is valid for a constant - density ism . at late time , when @xmath50 ^ 2\ll 1 $ ] , eq . ( [ doppler ] ) implies that @xmath51 , while eq . ( [ cubic ] ) implies that when @xmath52 ^ 2\ll 1 $ ] , @xmath35 approaches its asymptotic behaviour , @xmath53^{-1/3}$ ] . the corresponding asymptotic behaviour of @xmath54 is , for a constant @xmath43 : @xmath55^{1/3}\ , , \label{latealpha}\ ] ] which , in practice , is a fair approximation for @xmath33 larger than a few days . since xrfs are grbs seen at a relatively large @xmath56 , eq . ( [ latealpha ] ) implies that the superluminal motion in the former systems is relatively easier to observe . moreover , the larger the value of @xmath56 , the smaller the values of @xmath57 favoured by selection effects ( detection thresholds ) . this also privileges xrfs over grbs as potential targets for a search of a superluminal cb motion . in the cb model , the ags of grbs and xrfs consist of three contributions , from the cbs themselves , the concomitant sn , and the host galaxy : @xmath58 the latter contribution is usually determined by late - time observations , when the cb and sn contributions become negligible , or from measurements with sufficient angular resolution to tell apart @xmath59 from @xmath60 . let the unattenuated energy flux density of sn1998bw at redshift @xmath61 ( galama et al . 1998 ) be @xmath62 $ ] . for a similar sn placed at a redshift @xmath26 ( dar 1999 , dar & de rjula 2000 , dado et al . 2002a ) : @xmath63 = { 1+z \over 1+z_{bw}}\ ; { d_l^2(z_{bw})\over d_l^2(z)}\ , a(\nu , z)\ , f_{bw}[\nu',t']\ , , \label{bw}\ ] ] where @xmath64 is the attenuation along the line of sight , @xmath65 , and @xmath66 . the simple ansatz that _ all _ long - duration grbs would be associated with sn1998bw - like sne ( dar & plaga 1999 ; dar & de rjula 2000 ; dado et al . 2002a ) has proven to be unexpectedly precise and successful . for the most precise test so far , that of the grb 030329/sn2003dh pair , see , e.g. , dado et al . ( 2003c ) , stanek et al . ( 2003 ) , matheson et al . ( 2003 ) and hjorth et al . ( 2003 ) . the ag of the cbs is mainly due to synchrotron radiation from accelerated electrons in the cb s chaotic magnetic field . at optical and higher frequencies , the ag has the approximate form ( dado et al . 2003a , c ) : @xmath67\propto n_p^{(1+\hat\alpha)/2}\ , r_\infty^2\ , \gamma^{3\hat\alpha-1}\ , \delta^{3+\hat\alpha}\ , a(\nu , t)\ , \nu^{-\hat\alpha}\ , , \label{afterglow}\ ] ] with @xmath68 changing from @xmath69 to @xmath70 as each given observed frequency exceeds the time - dependent _ `` injection bend '' : _ @xmath71 ^ 3\ , \left[{n_p\over 10^{-3}\;cm^3}\right]^{1/2}\ , { \rm hz}\ , , \label{nubend}\ ] ] where @xmath43 is the baryon density of the interstellar medium changes from @xmath72 to @xmath73 , is interpolated by @xmath74^{1.1}}$ ] ( dado et al . 2003a ) . ] . in the same approximation in which eq . ( [ latealpha ] ) was derived , the ags of grbs and xrfs have the asymptotic behaviour @xmath75 ( dado et al . 2002a ) . at radio frequencies , the ag spectrum is affected by self - absorption in the cbs themselves , characterizable by a single parameter per cb : a `` free - free '' absorption frequency , @xmath76 ( dado et al . the attenuation @xmath77 is a product of the attenuation in the host galaxy , in the intergalactic medium , and in our galaxy . the attenuation in our galaxy in the direction of the grb or xrf is usually estimated from the galactic maps of selective extinction , @xmath78 , of schlegel , finkbeiner & davis ( 1998 ) , using the extinction functions of cardelli et al . the extinction in the host galaxy and the intergalactic medium , @xmath79 in eq . ( [ bw ] ) , can be estimated from the difference between the observed spectral index _ at very early time when the cbs are still near the sn _ and that expected in the absence of extinction . indeed , the cb model predicts and the data confirm with precision the gradual evolution of the effective optical spectral index towards the constant value @xmath80 observed in all `` late '' ags ( dado et al . 2002a ; 2003a ) . the `` late '' index is independent of the attenuation in the host galaxy , since at @xmath81 ( observer s ) days after the explosion , the cbs are typically already moving in the low - column - density , optically - transparent halo of the host galaxy . the cb - model description of the r - band ags of six representative grbs of interest to the analysis of grb 030329 are shown in fig . ( [ tiriri ] ) , and discussed anon . in dado et al . ( 2002a ) we showed that , in the cb model , the ags of the grbs of known redshift measured at the time could be successfully explained by a single dominant cb launched in the explosion of a sn akin to sn1998bw and moving , after a couple of observer hours , in an approximately constant - density ism . a typical example is that of grb 990510 . two exceptions are grb 990123 and grb 021211 , whose ags were measured very early , at a time when their cbs may be piercing the expected @xmath82 density profile generated by the pre - sn `` wind '' of the parent star . in the cb model , the early fluence of optical ags is proportional to @xmath83 , in excellent agreement with the observations ( dado et al . 2003b ) . prior to grb 030329 , there were three cases ( grbs 970508 , 000301c and 021004 ) for which the data showed clear evidence for deviations from a smooth ag behaviour . in the case of grb 000301c , we attributed the `` residua '' of the observational data relative to its smooth cb - model fit to moderate deviations of the ism density ( far from the parent sn ) from a constant value . in the case of grb 970508 , we have actually fit the ag to a jump from one to another constant - density value ( dado et al . 2002a ) . for grb 021004 we found it more natural to fit the ag to the contribution of two cbs , since in this case the @xmath0-ray count - rate as a function of time has a very clear two - pulse ( i.e. two - cb ) structure ( dado et al . 2003d ) . moreover , unlike for grb 970508 , our attempts to describe the ag of grb 021004 in terms of density variations failed , while the two - cb description fits the general trend of the ag very well . grb 030329 also has a two - cb @xmath0-ray structure , but an unprecedentedly non - smooth ag , presumably for the simple reason that the precision and continuity of the data are also unprecedented . in the case of this grb , the question will arise whether or not we should interpret its peculiar ag shape to two cbs , to density fluctuations , or to a combination of both effects . in dado et al . ( 2003c ) we made a first - round " cb - model fit to the nir - optical observations of the ag of this grb , then extending up to day @xmath10 after burst . since experience with all previously - measured ags had given us confidence in the quality of the model , we extrapolated these fits to later times , and we predicted the presence of a sn akin to sn1998bw , luminous enough to compete with the ag and be discovered ten days after burst they have also reached the predicted asymptotic behaviour @xmath84 , and in the cb model there is no need to hypothesize whether `` breaks '' in the ag fluence have occurred by a given time , there being no breaks . ] . sn2003dh was discovered 9.6 days after burst and turned out to be surprisingly similar to sn1998bw ( stanek el al . 2003 ; matheson et al . 2003 ; hjorth et al . 2003 ) . in a second round , we have now extended our original fit to include the x - ray data of rxte ( marshall & swank , 2003 ; marshall , markwardt & swank , 2003 ) and xmm - newton ( tiengo et al . 2003 ) , as well as the radio data of sheth et al . ( 2003 ) and berger et al . ( 2003 ) and many more nir - optical measurements ( lipkin et al . 2003 and references therein ) . the optical and x - ray data and the broad - band cb - model fit are shown in fig . ( [ figone ] ) . the inclusion of the new data modifies the parameters of the original data - poor fit only at the 10% level , a satisfactory result . the photon - number light - curve of the @xmath0 rays of grb 030329 consisted of two clear pulses , corresponding in the cb model to two dominant cbs , see fig . ( [ grb ] ) . consequently , the fit to the ag was again performed with the additive contribution of two separate cbs . in the optical domain , the two contributions correspond to the two shoulders , as illustrated in fig . ( [ figtwo ] ) , wherein we show the r - band results and the separate contributions of the two cbs . the data variations relative to the predicted smooth ag light - curve , ups and downs of @xmath85 magnitude , are , as for grbs 970508 , 000301c ( dado et al . 2002a ) and 021004 ( dado et al . 2003b ) , to be expected : they trace moderate deviations from a constant - density interstellar medium , as implied by eqs . ( [ afterglow],[nubend ] ) . these density variations are discussed in detail in the next section . the radio data at different frequencies and their comparison with the cb - model fit are shown in fig . ( [ figthree ] ) . the fit to the data at @xmath86 ghz is shown fig . ( [ figfour ] ) , wherein the contributions of the two cbs are separately shown . as for the optical data , fig . ( [ figtwo ] ) , we conclude that the late " ( @xmath87 d ) radio data are also dominated by one of the cbs . the rich structure of the ag of this grb is shown in fig . ( [ figseven ] ) , in which lipkin et al . ( 2003 ) have compared the r - band data to a reference " ( a double power - law of indices @xmath88 and @xmath89 , with the break at day @xmath90 ) by plotting the magnitude difference between the data and the reference . the cb - model description of the optical and radio ags in figs . ( [ figone ] ) and ( [ figthree ] ) satisfactorily reproduces the observed general trends . but the description is unsatisfactory in that its residua are also significant , as we show in fig . ( [ figten ] ) . these residua are not unlike those in fig . ( [ figseven ] ) , but for the absence of a prominent @xmath91 " feature , which we view as the result of the comparison of a power law unmotivated at early times with the smoothly - varying data . the remaining features , particularly the more prominent ones occurring after @xmath92 day , require an explanation . the cb - model s fluence of eq . ( [ afterglow ] ) is proportional to a power of the ism s electron density @xmath93 and , in that sense , it traces " its local , instantaneous value . this tracing is not perfectly _ local _ , because @xmath94 depends on @xmath35 , and this function reflects the integral effect of the ism density that a cb has swept through its prior voyage . to reproduce the features still prominent in fig . ( [ figten ] ) , we must assume a given non - trivial density profile and solve eqs . ( [ dgamma2 ] ) to ( [ tint ] ) explicitly to obtain the function @xmath35 to be used as input in the expression for the fluence , eq . ( [ afterglow ] ) . we have in the past done this for the ag of grb 970508 , as reproduced in fig . ( [ tiriri ] ) . the cbs of grbs 970508 and 030329 , as we shall see explicitly for the second one , are at a distance @xmath95 pc from the sne that emitted them , at an observer s time @xmath92 day . at a distance of that order , we expect the cbs to be exiting the superbubble where the sn explosion is very likely to have taken place . as they exit it , they may encounter successive density inhomogeneities produced by a succession of past sn explosions and stellar mass ejections , which created the bubble in the first place . we have considered the _ structured density profile _ shown ( as a function of observer s time ) in fig . ( [ figeight ] ) , in which successive onion peals " appear as abrupt density increases , followed by a decline proportional to @xmath96 ( or , approximately , @xmath97 ) , analogous to the closer - by profile of a wind - fed " circumburst material . with such a profile of overdensities we obtain a description of the r - band ag whose residua ( relative to the reference broken power - law introduced by the observers ) are shown in fig . ( [ figseven ] ) as the ( red ) line . the parameters of this fit are only slightly different from the ones for a constant @xmath98 , but for a somewhat larger initial value of @xmath99 for the cb dominating the ag at late times . the @xmath91 feature is fairly well reproduced with the density profile of fig . ( [ figeight ] ) , which is constant up to day @xmath100 . the feature is , in this sense , a fake . the description of the later - time features , however , requires the structured density profile of fig . ( [ figeight ] ) . clearly the many - parameter exercise we just described is not a fit " : it could easily be improved to obtain an even better description , and it is certainly not as predictive " as our prior fits to the early part of the ags of grbs 990123 and 021211 , from which we claimed to have successfully traced the magnitude and @xmath96 profile of the close - by circumburst density ( dado et al . yet , the description is phenomenologically satisfactory , in the sense that we can once again claim that , in the cb model , the shapes of ags provide interesting and consistent _ quasi - local _ tracers of the ism density through which the cbs travel . the actual distance @xmath101 from the sn at which the late - ag - dominating cb in grb 030329 was , as seen at a given observer s time @xmath33 , is shown in fig . ( [ figkpc ] ) , constructed with use of eq . ( [ tint ] ) . indeed , at the time at which the density inhomogeneities are apparent ( @xmath33 between @xmath102 and @xmath90 days ) the cb is 100 to 200 pc away from the sn , a reasonable radius for the overdensities surrounding a superbubble . given the need to introduce density variations to describe the detailed behaviour of the ag , the question arises whether or not two cbs as opposed to just one are needed . our attempts to describe the ag with just one cb and a variety of density profiles failed . the reason is simple and unavoidable : the density increase required to produce the ag s shoulder " at @xmath87 day inescapably increases the rate at which @xmath35 decreases with time . this entails a fast fall - down of the fluence , following soon after the initial increase produced by the local density increase . the resulting light - curves rise and fall fast like the one shown in fig . ( [ tiriri ] ) does in the case of grb 970508 and they fail in the description of the ag of grb 030329 . thus , two cbs appear to be necessary to understand the ag of grb 030329 , as they are for the description of its two - pulse grb structure , shown in fig . ( [ grb ] ) . we are primarily concerned in this note with the putative superluminal motion of the cbs of grb 030329 , and not with a complete description of its wide - band ag . thus , we have limited our detailed analysis with a structured density profile to the r - band ag , as in figs . ( [ figseven ] ) , the complete broad - bend analysis with density variations being very laborious . the outcome of the discussion that follows is that the structure seen in the r - band ag should be _ almost _ achromatic for data ranging from x - ray to nir frequencies , while the structures should progressively disappear at lower and lower radio frequencies . the x - ray and optical ags are not perfectly achromatic because the injection - bend frequency of eq . ( [ nubend ] ) `` crosses '' a given frequency at different times . but the crossing of the optical frequencies typically occurs at @xmath92 day and the light curves become increasingly achromatic thereafter . the radio ags , on the other hand , are predicted to be strongly chromatic ( dado et al . 2003a ) . to be more specific , we quote from the mentioned reference : _ electrons that enter a cb with an injection lorentz factor @xmath35 are rapidly fermi accelerated ... on a longer time scale , they lose energy by synchrotron radiation , and their energy distribution evolves ... electrons with a large @xmath103 $ ] emit synchrotron radiation , with no significant time - delay , at the observer s optical and x - ray wavelengths . but the emission of radio is delayed by the time it takes the electrons to `` descend '' to an energy at which their characteristic emission is in the observer s radio band ... thus , the optical and x - ray ag emission starts ... a few observer s seconds after the corresponding @xmath0-ray pulse . the radio signal , on the other hand , must await a time @xmath104 _ _ for the cumulated electrons to cool down . " @xmath105 was explicitly estimated in dado et al . ( 2003a ) to be of @xmath106 day . applied to the case at hand of a varying ism density , what this means is that the features observed in the optical ags should be smoothed in the radio over intervals of @xmath106 day . these intervals are frequency - dependent : longer for the lower radio frequencies . the expected trend of features that progressively disappear as the frequency diminishes is precisely the trend observed in the radio data reproduced in fig . ( [ figthree ] ) . the `` second - round '' fit shown in the figure systematically overestimates the late results , also a consequence of having ignored the density enhancements . the location of the source of the ag of this grb has been followed in detail by vlba observers from day @xmath107 3 to day @xmath107 84 after burst ( taylor et al . , to be published ) . the associated sn2003dh dominates the optical ag after day @xmath108 , but is negligible in the radio ag at all measured times . thus , the location of the radio source(s ) is the location of the cb(s ) , expected to be extremely superluminal in the cb model . the values of the fit observer s angles are very similar for the two cbs : @xmath109=2.2 $ ] mrad , @xmath110=2.3 $ ] mrad . the values of the initial lorentz factors are not so different : @xmath111=1037 $ ] , @xmath112=1606 $ ] . but the values of the deceleration parameter are very different : @xmath113=0.033 $ ] mpc , @xmath114=0.37 $ ] mpc ( these numbers refer to the initial @xmath99 as in eq . ( [ xinf ] ) , still unaffected by the density variations occurring at @xmath87 day ) . this is the main reason why the contribution of cb1 to the ag decays much faster with time than that of cb2 , as implied by eq . ( [ cubic0 ] ) and illustrated in figs . ( [ figtwo],[figfour ] ) . it is also the reason why the superluminal motion of cb1 is much slower than that of cb2 , as implied by eq . ( [ deltaalpha ] ) . we are therefore interested in the fastest - moving and late - ag - dominating cannonball : cb2 . in the absence of the density fluctuations of fig . ( [ figeight ] ) , the predicted angular displacement @xmath115 of cb2 , as given by the integrated form of eq . ( [ deltaalpha ] ) , would be that of the upper panel of fig . ( [ tororo ] ) . it is on the basis of our first - round " constant - density fit that we predicted such a displacement in dar & de rjula ( 2003b ) . in the presence of the observed density fluctuations , the predicted @xmath115 is shown in the lower panel of fig . ( [ tororo ] ) , the result of the integration of the velocity @xmath116 in eqs . ( [ supervelocity],[deltaalpha ] ) , obtained from the full - fledged determination of @xmath35 from eqs . ( [ dgamma1 ] ) to ( [ tint ] ) . the density fluctuations are dominantly density enhancements , implying a faster deceleration of the cb . thus the significantly reduction in the predicted apparent superluminal motion . alas , this prediction of the cb - model is , in the case of grb 030329 , harder to test than we originally thought . in its current form , the theory of ags in the cb model is based on an analogy with high - resolution observations of relativistic jets emitted by systems such as the microquasars ss 433 ( http://chandra.harvard.edu/ photo/2002/ 0214/ index.html ; http:// chandra . edu/ press/04@xmath117releases/ press@xmath117010504.html ) , grs 1915 + 105 ( mirabel & rodriguez 1999 ; dhawan et al . 2000 ) and xte j1550 - 564 ( corbel et al . 2002 ) ; and active galactic nuclei , e.g. the quasar pictor a , whose emitted cbs , as seen at x - ray wavelengths , seem to stop expanding laterally shortly after ejection , and to travel for hundreds of kpc , before they finally stop and blow up ( wilson , young & shopbell 2001 ; grandi et al . the cb - model s simple explanation for this surprising fact is the following ( dado et al . the ambient protons intercepted by a cb in its voyage encounter its inner chaotic magnetic field and are reemitted roughly isotropically ( in the cb s rest system ) . this implies an inwards pressure that stops the cb s expansion and that is asymptotically equal to the pressure of the cb s inner magnetic field , which is thereby calculable . the corresponding deceleration law ( at constant radius @xmath20 and ism density @xmath43 ) is that of eq . ( [ cubic0 ] ) . the magnitude and time - dependence of the magnetic field determined as in the previous paragraph plays a crucial role in the prediction of the broad - band ag spectra , which in the cb model is extremely simple , parameter - thrifty , and successful ( dado et al . 2003a , 2003e ) , particularly in comparison with its standard counterpart ( see , e.g. granot & sari 2002 ) . in spite of their phenomenological success , our assumptions leading to calculable cb radii and magnetic fields are very bold , and are no doubt oversimplifications . to investigate their solidity , in dado et al . ( 2002a ) we also studied a different ansatz : that the cb s radius @xmath118 would keep on increasing constantly with @xmath23 , with the ism density still kept approximately constant . this ansatz , which corresponds to a much faster cb deceleration than that of eq . ( [ cubic0 ] ) , failed miserably . but in between the two extremes ( a non - increasing and a steadily increasing cb s radius ) , it is clear that there should be choices of @xmath119 and @xmath16 that are phenomenologically satisfactory and yet , correspond to a _ faster _ cb deceleration than the simple one dictated by eq . ( [ cubic0 ] ) at constant @xmath120 . for such choices the simplest predictions of superluminal velocities that we have discussed would result in overestimates . but it would be premature to modify the simple predictions of the cb model , which so far are succesful . in the internal / external shock model of grbs , the variability of ag light curves the set of features " that we have discussed is attributed to patchy shells and refreshed shocks ( see , e.g. piran , nakar & granot 2003 ; granot , nakar & piran 2003 and references therein ) . in the case of grb 030329 , the features are attributed to delayed collisions between late , relatively low-@xmath0 shells and the earlier higher-@xmath0 shells having decelerated in their interactions with the ism . the timing and the lorentz factors are chosen so that the early fast shells collide in a matter of ( observer ) seconds to produce the grb peaks , while the slow shells catch up with the decelerated earlier ones at times of order days : some five order of magnitudes longer . the required amount of fine - tuning is considerable . the late shells bring in an energy injection an order of magnitude larger than that of the original blast wave . it would be interesting to know whether their collisions with the slowed - down shells are expected to give rise to grb pulses as well , in which case grbs should be repeaters " . in the cb model , as we have seen , all that is required to explain the ag features " are ism density inhomogeneities occurring at a natural distance from the parent sn : that corresponding to the radius of a superbubble . no new phenomena must be invoked and no grb parameters must be fine - tuned with special care . the data on grb 030329 are now sufficiently complete to allow for a detailed prediction of the motion of the source of its ag allegedly superluminal in the cb model . the @xmath0-ray light - curve and the optical ag require the presence of two cbs , one of which dominates the ag at late times . the parameters needed to predict the motion of the two cbs in the sky are determined by the optical data , so that the individual motion of each cb is predicted . the main result of this paper is the prediction of the sky - motion of the fastest - moving cb , which dominates the late ag , and is shown in the lower panel of fig . ( [ tororo ] ) . the main caveat concerning a putative superluminal signature ( dar & de rjula 2000 , 2003b ) concerns the input to the estimate of its magnitude . indeed , we have shown in this paper that the predictions are very sensitive to the details of the density profile of the ism . but we are encouraged by the fact that in the only case in which the superluminal jets of cbs made by a core - collapse sn could be seen , they were seen . indeed , observations of sn 1987a ( ninenson and papaliolios , 1999 ) showed two sources , moving in opposite directions along the sn s axis at ( real ) velocities compatible with the speed of light and at an apparent superluminal velocity for the approaching source . mercifully , the jet of that sn was not pointing in our direction ( dar , laor & shaviv 1998 ; dar & de rjula 2001b ) . regarding the search for a superluminal motion , we learned by reading the e - version of nyt 030529 ( the new york times of that date , in grb s parlance ) that , according to dale frail _ [ our observations ] are sufficient to rule out predictions of the cannonball model"_. we have shown that , indeed , the observations of complicated features in the optical ag of grb 030329 imply that our earlier results ( dar & de rjula 2003b ) which ignored the presence of these features constituted an overestimate of the predicted superluminal displacement . in this sense , frail was right in stating that the observations ruled out the _ predictions _ , as opposed to the model itself . in a setting more scientific than the nyt , bloom et al . ( 2003 ) state : _ owing to the proximity and bright radio emission , high - resolution ( @xmath121 pc ) _ very long baseline array _ imaging of the compact afterglow was used by frail ( 2003 ) to _ * unequivocally disprove * _ the cannonball model for the origin of grbs . " _ the emphasis is ours . we have seen that these news of the death of the cb model may have been premature . even though mark twain eventually died for sure , the cb model though probably not immortal is still in an excellent shape . yet , trying to disprove the best available model(s ) , or even the proof of a difficult theorem , is the acceptable standard attitude in many a realm of the exact sciences . in this sense , the apparently strong motivation of the quoted observers to disprove the cb model is in our opinion , and regarding this particular model the healthiest of all possible attitudes . * acknowledgements . * one of us ( a. de r. ) is indebted to the physics department and space research institute of technion for its hospitality . this research was supported in part by the helen asher space research fund for research at the technion . in this model , the _ engines _ generating long - duration grbs and xrfs are ordinary core - collapse sne . following the collapse of the stellar core into a neutron star or a black hole , and given the characteristically large specific angular momentum of stars , an accretion disk or torus is hypothesized to be produced around the newly - formed compact object , either by stellar material originally close to the surface of the imploding core and left behind by the explosion - generating outgoing shock , or by more distant stellar matter falling back after its passage ( de rjula 1987 ) . a cb is emitted , as observed in microquasars , when part of the accretion disk falls abruptly onto the compact object ( e.g. mirabel & rodrigez 1999 ; rodriguez & mirabel 1999 and references therein ) . the cbs of a grb are assumed to me made of ordinary hydrogenic matter . this is also in analogy to microquasar cbs , observed to be made of ordinary matter , as opposed to the standard - model contrived mixture of baryons and @xmath122 pairs . indeed , blueshifted and redshifted h , he , metal and heavy - element optical and uv lines were detected from the approaching and receding cbs emitted by the microquasar ss433 ; see , e.g. eikenberry et al . ( 2001 ) , gies et al . ( 2002 ) . in the cb model , sn1998bw , associated with grb 980425 , is an ordinary core - collapse sn ( dar and plaga 1999 ; dar and de rjula 2000 ) : its `` peculiar '' x - ray and radio emissions were not emitted by the sn , but were part of the grb s ag ( dado et al . 2002a , 2003a ) . the grb appeared to be underluminous because it was viewed further off axis than others ( e.g. , dar and de rjula 2000 ) . thus , it makes sense in this model to expect an association between ( long ) grbs and sn1998bw - like engines , transported to the grb location ( dar 1999 ) . the a - posteriori analysis of dado et al . ( 2002a ) of all the available data supported the conclusion that _ all _ ( long ) grbs are indeed associated with such sne . in several more recent cases we used the cb model to _ predict _ how the associated sn would compare with the rest of the ag ( grb 011121 , 020405 , 021211 ; dado et al . 2002b , 2002c , 2003d ) . the most interesting case is that of grb 030329/sn2003dh , for which we also foretold the date when the sn would be dominant enough to be discovered ( dado et al . 2003c ) . the fraction of visible grbs , relative to the sne that produce them , is @xmath123 , where @xmath124 is some effective value of the maximum observer s angle for which grbs have been detectable . in practice @xmath124 is a few milliradians ( dado , dar & de rjula 2002a ) . this brings the _ total _ rate of ( observed or unobserved ) grbs to a value close to that of core collapse sne . this implies that a good fraction of such sne may produce long - duration grbs ( dar & plaga 1999 ; dar & de rjula 2000 ) . in dar & de rjula ( 2000 ) the initial lorentz factor of the cbs was argued to be @xmath1 , a choice corroborated by all our subsequent analyses . the high - energy photons of a single pulse in a grb or an xrf are produced as a cb coasts through the `` ambient light '' permeating the surroundings of the parent sn . the electrons enclosed in the cb compton up - scatter photons to energies that , close to the cbs direction of motion , correspond to the @xmath0-rays of a grb and less close to it , to the x - rays of an xrf . each pulse of a grb or an xrf corresponds to one cb . the timing sequence of emission of the successive individual pulses ( or cbs ) reflects the chaotic accretion process and its properties are not predictable , but those of the single pulses are ( dar & de rjula 2003a and references therein ) . in practice grbs are observable for @xmath125 , xrfs are the same phenomenon , observed from somewhat larger angles ( dar & de rjula 2003a ; dado , dar & de rjula 2003f ) . grbs are notorious for their variety . yet , they have some two dozen common properties . the characteristic @xmath0-ray energy is surprisingly narrowly distributed around @xmath126 kev . the energy spectra are well fit by a specific band " function . the total `` equivalent spherical energies '' are distributed over a wide range , that becomes very narrow for the `` true '' ( beaming - corrected ) energy values . the photons arrive in successive pulses with fairly similar time - profiles which , in long " grbs , have a median width of 0.5 seconds . the @xmath0-ray polarization is nearly maximal . a long list of phenomenologically - observed correlations between many different grb observables appear to be well satisfied . _ all _ of these properties are very simply and successfully explained by the cb model , in a manner that does not require the introduction of _ any _ ad - hoc parameters ( dar & de rjula 2003a and references therein ) . in the ag phase , and within a few observer minutes from the end of a grb s intense @xmath0 emission , the emissivity of a cb is dominated by synchrotron emission from the electrons that penetrate in it as it propagates in the ism . integrated over frequency , this synchrotron emissivity is proportional to the energy - deposition rate of the ism electrons in the cb . these electrons are fermi - accelerated " in the cb s tangled magnetic maze to a broken power - law energy distribution with a `` bend '' energy equal to their incident energy in the cbs rest frame , @xmath127 . their synchrotron radiation the afterglow is collimated and doppler - boosted by the relativistic motion of the cbs . the radiation is also redshifted by the cosmological expansion and attenuated on its way to an earthly observer , during its passage through the cb itself , the host galaxy , the intergalactic space and our own galaxy . the approximate cb - model analysis of the fluences of grb ags , as functions of time and frequency , is fairly simple . besides @xmath26 and @xmath56 ( which are not parameters specific to the model ) it involves only three quantities : @xmath128 , @xmath99 , an overall normalization , and @xmath76 ( a _ single _ free - parameter frequency pertinent to the radio - absorption within a cb ) . in spite of its economy of parameters ( notorious in the case of the wide - band spectral shapes ) the cb model satisfactorily describes in a unified manner all measured grb ags , including that of grb 980425 ( e.g. , dado et al . 2002a , 2003a , 2003e ) . our impression is that no such claims can be made in the realm of standard grb models . berger , e. , et al . 2003 , nature , 426 , 154 blake , c. & bloom , j. s. 2003 , gcn circ . 2011 bloom , j. s. , et al . 2003 , astro - ph/0308034 cardelli , j. a. , clayton , g. c. & mathis , j. s. 1989 , apj , 345 , 245 corbel , s. , et al . 2002 , science , 298 , 196 courdec , p. 1939 , annales dastrophysique 2 , 271 dado , s. , dar a. & de rjula , a. 2002a , a&a , 388 , 1079 dado , s. , dar a. & de rjula a. 2002b , apj , 572 , l143 dado , s. , dar a. & de rjula a. 2002c , a&a , 393 , l25 dado , s. , dar a. & de rjula a. 2003a , a&a , 401 , 243 dado , s. , dar a. & de rjula a. 2003b , apj , 585 , l15 dado , s. , dar a. & de rjula a. 2003c , apj , 594 , l89 dado , s. , dar a. & de rjula a. 2003d , apj , 593 , 961 dado , s. , dar a. & de rjula a. 2003e , phys . lett . b562 , 161 dado s. , dar a. & de rjula a. 2003f , astro - ph/0309294 dahawan , v. , mirabel , f. & rodriguez , l. f. 2000 , apj , 543 , 373 dar , a. & de rjula , a. 2000 , astro - ph/0008474 dar , a. & de rjula , a. 2001 , in _ astrophysics and gamma ray physics in space _ ; eds . a. morselli and p. picozza ; frascatti physics series vol . xxiv ( 2002 ) , pp . 513 - 523 ; astro - ph/0110162 dar , a. & de rjula , a. 2003a , astro - ph/0308248 dar , a. & de rjula , a. 2003b , gcn circ 2133 dar , a. , laor , a. & shaviv , n. j. 1998 , phys . 80 , 5813 dar , a. , & plaga , r. 1999 , a&a , 349 , 259 dar , a. 1999 , a&as , 138 , 505 dar , a. 2003 , astro - ph/0301389 de rjula , a. 1987 , phys . 193 , 514 de rjula , a. 2002 , astro - ph/0207033 donaghy , t. q. , lamb , d. q. & graziani . c , 2003 , astro - ph/031496 eikenberry , s. s. , et al . 2001 , apj , 561 , 1027 esin , a. a. & blandford , r. 2000 , apj , 534 , l151 frail , d. a. , et al . , 2001 , apj , 562 , l55 frail , d. a. 2003 , american astronomical society meeting , 2003 frederiksen , j. t. , hedelal , c. b. , haugbolle , t. & nordlund , a. astro - ph/0303360 galama , t. j. , et al . 1998 , nature 395 , 670 ghisellini , g. & lazzati , d. 1999 , astro - ph/9906471 gies , d. r. , et al . 2002 , apj , 566 , 1069 grandi , p. , et al . 2003 , 2003 , apj , 586 , 123 granot , j. , nakar , e. , piran , t. 2003 , nature 426 , 138 granot , j. & sari , r. 2002 , apj . 568 , 820 granot , j. , panaitescu , a. , kumar , p. & woosley , s. e. 2002 , apj , 570 , l61 greiner , j. , et al . 2003a , gcn circ . 2020 greiner , j. , et al . 2003b , astro - ph/0311282 hjorth , j. , et al . nature 2003 , 423 , 847 hurley , k. , sari , r. & djorgovski , s. g. 2002 , astro - ph/0211620 henden , a. a. , et al . 2003 , gcn circ . 2082 kuno , n. , et al . 2004 , astro - ph/0401258 lamb , d. q. , donaghy , t. q. & graziani , c. 2003a , astro - ph/0309456 lamb , d. q. , donaghy , t. q. & graziani , c. 2003b , astro - ph/0309463 lamb , d. q. , donaghy , t. q. & graziani , c. 2003c , astro - ph/030504 lazzati , d. , rossi , e. , ghisellini , g. & rees , m. 2003 , astro - ph/0309038 , lazzati , d. , ghisellini , g. , celotti , a. & rees , m. 2000 , apj , 529 , l17 lazzati , d. 2003 , astro - ph/0312331 li , w. , filippenko , a.v . , chornock , r. & jha , s. 2003 , apj , 586 , l9 lipkin , h. y. , et al . 2003 , astro - ph/0312594 margon , b. a. 1984 , ara&a , 22 , 507 marshall , f. e. & swank , j. h. 2003 , gcn circ . 1996 marshall , f. e. , markwardt , c. & swank , j. h. 2003 , gcn circ . 2052 matheson , t. , et al . apj , 2003 , 599 , 394 meszaros , p. & rees , m. j. 1999 , mnras , 306 , l39 mirabel , f. & rodriguez , l. f. 1999 , ara&a , 37 , 409 nisenson , p. & papaliolios , c. 1999 , apj , 518 , l29 patat , f. , et al . 2001 , apj , 555 , 900 peterson , b. a. & price , p. a. 2003 , gcn circ . 1985 piran , t. , nakar , e. & granot , j. 2003 , astro - ph/0312138 pooley , g. 2003 , gcn circs . 2043 , 2072 ramirez - ruiz , e. , dray , l. m. , madau , p. & tout , c. a. 2001 , mnras , 327 , 829 rees , m. j. 1967 , mnras , 135 , 345 rees , m. j. & meszaros , p. 1992 , mnras , 258 , l41 reichart , d. e. , 2001 , apj , 554 , 643 rodriguez , l. f. & mirabel , i. f. 1999 , apj , 511 , 398 rykoff , e. s. , et al . 2003 , gcn circ . 1995 sari , r. & piran , t. 1999 , a&as , 138 , 537 sari , r. 1999 , apj , 524 , l43 shaviv , n. j. & dar , a. 1995 , apj , 447 , 863 sheth , k. , et al . 2003 , apj , 595 , l33 schlegel , d. j. , finkbeiner d. p. & davis m. 1998 , apj , 500 , 525 stanek , k. z. , et al . 2003 , apj , 591l , 17 tiengo , a. , mereghetti , s. , ghisellini , g. , rossi , e. , ghirlanda , g. & schartel , n. 2003 , astro - ph/0305564 torii , k. 2003 , gcn circ . 1986 vanderspek , v. , et al . 2003 , gcn circ . 1997 vanderspeck , r. , et al . 2004 , astro - ph/0401311 waxman , e. 2003 , astro - ph/030351 waxman , e. & draine , b. t. 2000 , apj , 537 , 796 wilson , a. s. , young , a. j. & shopbell , p. l. 2001 , apj , 547 , 740 yamazaki , r. , yonetoku , d. & nakamura , t. apj , 2003 , 594 , l79 zeh , a. , klose , s. & hartmann , d. h. astro - ph/0311610 zhang , b. & meszaros , p. 2003 , astro - ph/0311321

recent advances in the construction of low cost , low power , and mass produced micro sensors and micro - electro - mechanical ( mem ) systems have ushered in a new era in system design using distributed sensor networks @xcite . the advent of sensor network technology provides a variety of applications that have been considered unrealistic in the past . one such application is in the area of space communications : with ad hoc sensor networks , a number of sensor nodes randomly placed on a planet can collaboratively collect information and then , also collaboratively , send the information back to earth . in this scenario , the sensors must have an ability to transmit information over very long distances with high energy efficiency . in this kind of point - to - point communication , directional antennas are a preferred means to avoid interference . in general , this can be achieved by adaptive beamforming . given a number of well - designed antenna elements at the transmitting / receiving sensor nodes , each node could in principle autonomously transmit / receive the information to / from any desired direction . the advantages and applications of beamforming with antenna arrays are well known ; in wireless communications , this enables space - division multiplex access ( sdma ) , a technology which has the potential to significantly increase the capacity of multiple access channels . one of the most important constraints on wireless sensors is energy efficiency . since the sensor nodes are often distributed in places where manual maintenance is costly , such as remote locations , on top of buildings and so on , it should be possible to operate these for several months without battery replacement . considering the fact that each antenna element requires analog circuitry ( and thus leads to costly hardware ) , in practice each distributed sensor is likely to be equipped with only a single antenna and this precludes the use of autonomous beamforming in scenarios of very energy efficient communication . nevertheless , if the sensors in the cluster share the information _ a priori _ and synchronously transmit the data collaboratively as sketched in fig.[fig : adhoc_network_model ] , it may be possible to beamform when transmitting ( and also receiving ) the data in a distributed manner . the resultant overhead due to intra - cluster information sharing may be relatively small as this can be done by low - cost short distance broadcasting - type communication among nodes . thus , with distributed collaborative beamforming , the nodes can send the collected information to the far - end receiver over long distances with high efficiency . also , only the sensor cluster in the specified target direction receives the data with high signal power and no significant interference occurs for clusters in other directions . overall there is thus a potential to increase the capacity of the multiple access channel significantly despite the additional overhead for information sharing . the obvious question is whether one can form a nice beampattern with a narrow mainbeam , or achieve a reasonable directional gain ( directivity ) . the sensor nodes in ad hoc networks are located randomly by nature , and the resultant beampattern depends on the particular realization of the sensor node locations . therefore , it may be quite natural to treat the beampattern with probabilistic arguments . in this paper , assuming idealized channel model conditions and antenna properties , we analyze the achievable performance of collaborative beamforming based on distributed sensor nodes in a probabilistic sense . specifically , the statistical properties of the achievable beampattern of the random sensor arrays are studied based on the following assumptions . the sensors form an ad hoc network and the geometry of the cluster is given by a two - dimensional disk of a given radius over which all sensor nodes are distributed uniformly as illustrated in fig.[fig : adhoc_network_model ] . since the corresponding far - field beampattern depends on the particular realization of the random array of nodes , the probability distribution of the far - field beampattern is of particular interest . to the best of the authors knowledge , the beampattern aspects of collaborative beamforming using random arrays have not been analyzed before in the framework of wireless ad hoc sensor networks . nevertheless , in the antenna design literature , probabilistic analysis of random arrays is not new . in the framework of linear array design with a large number of sensors , lo @xcite has developed a comprehensive theory of random arrays in the late 1960s . it has been shown that randomly generated linear arrays with a large number of nodes can in fact form a good beampattern with high probability and that with linear random arrays of @xmath0 sensors , the directivity approaches @xmath0 asymptotically . although our scenario is quite different in that our main goal is not to design array geometry but to _ exploit _ the randomness of the distributed sensor network , it turns out that the results we shall develop in this paper can be seen as an extension of the theory of linear random arrays , @xcite , to random arrays on a disk . thus , the same conclusion will be reached : with @xmath0 collaborative sensor nodes , one can achieve a directivity of order @xmath0 asymptotically . the major difference between classical beamforming by antenna arrays and distributed beamforming is that whereas the geometry of the former is usually known _ a priori _ , the exact location of the sensor nodes in ad hoc network is not , and it should be acquired dynamically . even if their relative location is estimated by some adaptive algorithm ( e.g. , @xcite for receiver beamforming ) , considering the low snr operation of the sensor nodes , it is almost certain that the acquired geometric information has some inaccuracy . also , since all nodes are operated with physically different local oscillators , each node may suffer from statistically independent phase offsets . in order to model and elucidate the effect of these impairments , we consider the following two scenarios : _ closed - loop _ and _ open - loop_. the closed - loop scenario may be described as follows . each node independently synchronizes itself to the beacon sent from the destination node ( such as a base station ) and adjusts its initial phase to it . thus , the beam will be formed in the direction of arrival of the beacon . this kind of system is often referred to as a _ self - phasing _ array in the literature , and may be effective for systems operating in time - division duplex ( tdd ) mode . the residual phase jitter due to synchronization and phase offset estimation among sensor nodes is then often the dominant impairment . on the other hand , in the open - loop scenario we assume that all nodes within the cluster acquire their relative locations from the beacon of a nearby reference point or cluster head . the beam will then be steered toward an arbitrary direction . thus , the destination need not transmit a beacon , but each node requires knowledge of its relative position from a predetermined reference point within the cluster . this case may occur in ad hoc sensor networks where sensor nodes do not have sufficient knowledge of the destination direction _ a priori_. in this scenario , since the acquisition of precise knowledge is not realistic , the effects of location estimation ambiguity among sensors upon the beampattern may be of particular interest . throughout the paper , the nodes and channel are assumed to be static over the communication period , and for simplicity the information rate is assumed to be sufficiently low that inter - symbol interference ( isi ) , due to residual timing offset , is negligible . it will also be assumed that all nodes share the same transmitting information _ a priori _ , as the main focus of the paper is on the beampattern , rather than the front - end communication performance . the paper is organized as follows . section [ sec : model ] describes the assumptions , model , and main parameters that describe beam characteristics associated with the framework of wireless ad hoc sensor networks . in section [ sec : average_property ] , the average properties of the beampattern are derived . the average beampattern of linear random arrays has been derived in @xcite , and our results can be seen as its extension to our sensor network model . for analytical purpose , we also introduce the concept of 3 db sidelobe region . in section [ sec : distribution_beam ] , the statistical distribution of the beampattern in a specific direction is derived . lo @xcite has derived the distribution of the beampattern in linear arrays based on a gaussian approximation of the array factor , which is a common assumption in the random array literature . in contrast , we shall develop a numerical method to calculate the exact distribution of the beampattern and also examine the accuracy of the gaussian approximation in detail . the distribution of the maximum of the sidelobe region is discussed in section [ sec : distribution_maximum ] . this aspect of beampattern was analyzed by steinberg @xcite , agrawal and lo @xcite , and donvito and kassam @xcite in the framework of linear random arrays . in this paper , we derive an upper bound on the distribution of the maximum sidelobe in our framework of collaborative beamforming based on the approach of @xcite . the effect of phase jitter or location estimation errors on the resultant beampattern , associated with the closed - loop and open - loop scenarios , is analyzed in section [ sec : imperfect ] . steinberg @xcite has analyzed the effects of phase estimation errors in linear arrays , and based on a similar approach we analyze the effects of the average beampattern with phase estimation errors . finally , section [ sec : conclusion ] concludes the paper . [ bl][bl][.8]@xmath1 [ bl][bl][.8]@xmath2 [ bl][bl][.8]@xmath3 [ bl][bl][.8]@xmath4 [ bl][bl][.8]@xmath5 [ bl][bl][.8]@xmath6 [ bl][bl][.8]@xmath7 [ bl][bl][.8]@xmath8 [ bl][bl][.8]@xmath9 the geometrical configuration of the distributed nodes and destination ( or target ) is illustrated in fig.[fig : notaion ] where , without loss of generality , all the collaborative sensor nodes are assumed to be located on the @xmath2-@xmath3 plane . the @xmath10th node location is thus denoted in polar coordinates by @xmath6 . the location of the destination is given in spherical coordinates by @xmath7 . following the standard notation in antenna theory @xcite , the angle @xmath11 $ ] denotes the elevation direction , whereas the angle @xmath12 $ ] represents the azimuth direction . in order to simplify the analysis , the following assumptions are made : 1 . the location of each node is chosen randomly , following a uniform distribution within a disk of radius @xmath4 . each node is equipped with a single ideal isotropic antenna . all sensor nodes transmit identical energies , and the path losses of all nodes are also identical . thus the underlying model falls within the framework of phased arrays . there is no reflection or scattering of the signal . thus , there is no multipath fading or shadowing . 5 . the nodes are sufficiently separated that any mutual coupling effects @xcite among the antennas of different sensor nodes are negligible . furthermore , we also assume that all the nodes are perfectly synchronized so that no frequency offset or phase jitter occurs . the effects of phase ambiguities on the beampattern will be discussed in section[sec : imperfect ] . let @xmath13 denote the euclidean distance between the @xmath10th node and the reference location @xmath14 , which is written as @xmath15 if the initial phase of node @xmath16 is set to @xmath17 the corresponding array factor , given the realization of node locations @xmath18 \in [ 0 , r]^n$ ] and @xmath19 \in [ -\pi , \pi]^n$ ] , is written as @xmath20 } , \label{far_field_f0}\end{aligned}\ ] ] where @xmath0 is the number of sensor nodes and @xmath21 is the wavelength of the radio frequency ( rf ) carrier . in this paper , we are interested in the radiation pattern in the far - field region , and we assume that the far - field condition @xmath22 holds . the far - field distance @xmath13 in can then be approximated as @xmath23 the far - field beampattern is thus approximated by @xmath24 } \triangleq \tilde{f}(\phi , \theta | { \boldsymbol r},{\boldsymbol \psi } ) . \label{far_field_f1}\end{aligned}\ ] ] alternatively , instead of applying @xmath25 as in , if we choose @xmath26 then we obtain the array factor as @xmath27 } \nonumber \\ & = e^{j \frac{2 \pi}{\lambda } a } \frac{1}{n } \sum_{k=1}^{n } e^{j \frac{2 \pi}{\lambda } r_k \left [ \sin \theta_0 \cos(\phi_0 - \psi_k ) - \sin \theta \cos(\phi - \psi_k ) \right ] } \triangleq \tilde{f}^{\dag } ( \phi , \theta \label{far_field_f2}\end{aligned}\ ] ] the only difference between @xmath28 in and @xmath29 . the far - field beampattern is thus identical for both systems , and the received signal exhibits no difference as long as the base station compensates for this phase rotation . therefore , there are two ways of forming a beam . one way is to use , but this approach requires accurate knowledge of the distance , relative to the wavelength @xmath21 , between each node and the destination . thus , this applies to the closed - loop case such as self - phasing arrays . alternatively , the use of requires knowledge of the node positions relative to some common reference ( such as the origin in this example ) , and thus corresponds to the open - loop case . knowledge of the elevation direction @xmath8 is also required , but this may be assumed to be known _ a priori _ in many applications . in both cases , the synchronization among sensors is critical , which may be achieved by the use of reference signals such as those of the global positioning system ( gps ) . of particular interest in practice is the case where @xmath30 , i.e. , the destination node is in the same plane as the collaborative sensor nodes . therefore , we will consider the beampattern in this plane and thus assume that @xmath31 for the rest of the paper . we then denote @xmath32 in by @xmath33 and @xmath34 for simplicity . by assumption , the node locations @xmath6 follow the uniform distribution over the disk of radius @xmath4 . thus , the probability density functions ( pdfs ) of @xmath35 and @xmath36 are given by @xmath37 from , we have ( with @xmath38 ) @xmath39 where @xmath40 and @xmath41 . the compound random variable @xmath42 has the following pdf : @xmath43 note that since the above model is symmetric with respect to the azimuth direction @xmath44 , any particular choice of @xmath9 does not change the results in the following . therefore , without loss of generality , we assume that @xmath45 , and the parameter @xmath44 simply corresponds to the difference angle between the target direction and the reference . also , note that @xmath46 . the array factor of can then be rewritten as @xmath47 where @xmath48 is the radius of the disk normalized by the wavelength . finally , the far - field beampattern can be defined as @xmath49 where @xmath50 we start by investigating the average beampattern of the random array resulting from the distributed sensor network model in the previous section . here , the average is taken over all realizations of @xmath51 , and from the average beampattern is expressed as @xmath52 where @xmath53 denotes expectation with respect to the random variables @xmath54 . from and , it can be readily shown that @xmath55 where @xmath56 is the @xmath57th order bessel function of the first kind . although the function @xmath58 is oscillatory , the local maxima of oscillation tend to decrease with increasing @xmath2 . in , the first term represents the average power level of the sidelobe , which does not depend on the node location , whereas the second term is the contribution of the mainlobe factor . since , conditioned on @xmath44 , the array factor of the form is an average of bounded independent and identically distributed ( i.i.d ) complex random variables , by the weak law of large numbers the beampattern converges to the ensemble average in probability as @xmath59 . the average beampattern of is plotted in fig.[fig : avrage_beam ] for several values of @xmath60 with @xmath61 and 256 . as can be observed , the sidelobe approaches @xmath62 as the beam angle moves away from the target direction . [ bl][bl][.8]average power [ db ] [ bl][bl][.8]angle [ degree ] [ bl][bl][.8]@xmath63 [ bl][bl][.8]@xmath64 [ bl][bl][.8]@xmath65 and @xmath66.,title="fig:",width=472 ] to gain further insight , consider the asymptotic expansion of the bessel function @xmath67 for @xmath68 as @xmath69 we then have @xmath70 and becomes , for @xmath71 , @xmath72 the @xmath57th peak of the average sidelobe will appear around @xmath73 , @xmath74 , and its corresponding value becomes @xmath75 ^ 3 , \label{peak_nth}\ ] ] which does not depend on @xmath60 . the @xmath57th peak and @xmath57th zero positions ( in the sense of the second term in ) can then be expressed asymptotically as @xmath76 since the peak sidelobe value does not depend on @xmath60 and is less sensitive to the value of @xmath0 , it is apparent that the only way one can avoid high peaks in the sidelobe region is to increase @xmath60 such that most of the major peaks are relatively concentrated around the mainlobe . this phenomenon will be further examined in the following subsections . one of the important figures of merit in directional antenna design is the 3db beamwidth . in the deterministic antenna , the 3db beamwidth is the threshold angle at which the power of the beampattern drops 3db below that in the target direction @xmath9 . in our scenario , the 3db beamwidth itself is a random variable and it is not easy to characterize analytically . thus , as an alternative measure , we may define the 3db beamwidth of the _ average _ beampattern denoted by @xmath77 as the angle @xmath44 that satisfies @xmath78 in the limit as @xmath79 , one may obtain @xmath80 by numerically solving . for @xmath81 , can be approximated as @xmath82 . therefore , the beamwidth is asymptotically independent of @xmath0 and is mainly determined by the inverse of the disk radius of the cluster . consequently , sparsely distributed sensors form a narrow beam on average provided that the cluster radius is sufficiently large . this sharp mainbeam property may be desirable , but if the far - field destination node has mobility , it should be designed carefully ; the calibration should take place before the mobile node moves out of the mainbeam , but the mainbeam width is inversely proportional to the normalized radius @xmath60 as observed in fig.[fig : avrage_beam ] . therefore , calibration should be performed more frequently if the destination node moves rapidly or when @xmath60 is increased . similar to the 3db beamwidth concept , it may also be convenient for our subsequent analysis to define the region within which the average of the sidelobe beampattern falls below some threshold level . as we have seen , for large @xmath60 , the sidelobe of the average beampattern is given by @xmath62 asymptotically . therefore , we shall define the _ 3db sidelobe region _ as the region in which neither neighboring sidelobe peak in the average beampattern exceeds 3db above @xmath62 . let @xmath83 denote the minimum index of the peak position such that the corresponding peak value satisfies this 3db condition . specifically , from , @xmath83 is the minimum integer @xmath57 that satisfies @xmath84 ^ 3 & \leq 2 , \label{3db_region_def}\end{aligned}\ ] ] and it can be bounded by @xmath85 let @xmath86 denote the angle corresponding to the zero point next to the @xmath83th peak sidelobe which can be obtained from with @xmath87 . consequently , in this paper , the 3db sidelobe region is defined as @xmath88 fig.[fig:3dbdefs ] illustrates the definitions of the 3db sidelobe region together with that of the 3db beamwidth . [ r][r][1.0]1 [ r][r][1.0]@xmath89 [ r][r][1.0]@xmath90 [ b][b][.8]1st peak [ b][b][.8]@xmath83th peak [ b][b][.8]1st zero [ b][b][.8]@xmath83th zero [ t][t][.8]3db beamwidth [ b][b][.8]mainbeam [ b][b][.8]3db sidelobe region as will be shown in section [ sec : zero_mean ] , the idea behind the introduction of 3db sidelobes is that in this region one may assume that the mean value of the random array factor of sampled at @xmath91 becomes a random variable with approximately zero mean . thus we may reasonably assume that the array factor has zero mean in this region , and this assumption significantly simplifies the analysis of the statistical distribution in the following sections . fig.[fig : three_side ] shows the threshold angle above which the 3db sidelobe region begins . the asymptotic 3db beamwidth is also shown for reference . as can be observed , whereas the 3db beamwidth is less sensitive to the number of nodes @xmath0 , the 3db sidelobe region will be considerably reduced as @xmath0 increases . this means that as @xmath0 increases the dominant non - negligible sidelobe peak may occur with high probability unless @xmath60 is also increased . this trade - off will be clarified by the study of directivity in the following subsection . [ bl][bl][.8]@xmath60 [ bl][bl][.8]threshold angle [ degree ] and number of nodes @xmath0.,title="fig:",width=472 ] the directivity or directional gain is the parameter that characterizes how much radiated energy is concentrated in the desired direction , relative to a single isotropic antenna . specifically , it may be defined as @xmath92 where @xmath93 is the radiated power density in the direction of @xmath44 . in the scenario of this paper , since @xmath93 depends on the particular realization of @xmath51 , the corresponding gain may be expressed , by substituting @xmath94 of into the above , as @xmath95^{-1 } . \label{directivity}\ ] ] the mean value of is given by @xmath96 unfortunately , direct calculation of does not result in a closed - form or insightful expression . thus , we shall consider the following as an alternative measure : @xmath97 note that by jensen s inequality , we have @xmath98 which means that @xmath99 in is a lower bound on @xmath100 . however , since by the law of large numbers the denominator of @xmath101 may approach its average value with high probability as @xmath0 increases , the above bound is expected to become tight as @xmath0 increases . this will be verified in the numerical results below . substituting into , we obtain @xmath102 where @xmath103 is a generalized hypergeometric function which monotonically decreases with increasing @xmath2 and converges to 0 as @xmath104 . therefore , unlike well - designed deterministic linear arrays , the gain of a given realization is very likely to be less than @xmath0 , and the limit @xmath0 can be approached only by increasing @xmath60 . this agrees with the previous observation that the average mainbeam becomes narrow as @xmath60 increases and thus improves the directivity . it should be noted that although has a simple form and offers some insight into the asymptotic behavior of directivity , the calculation of the generalized hypergeometric function involved in becomes numerically unstable as @xmath60 increases , and it is much easier to numerically integrate the denominator of directly . [ bl][bl][.8]normalized directivity @xmath105 [ bl][bl][.8]normalized radius @xmath106 and normalized radius @xmath60.,title="fig:",width=472 ] fig.[fig : directivity ] shows the relationship between the normalized directivity bound @xmath107 and @xmath60 . also shown in the figure as diamond - shaped points are the corresponding exact average directivities @xmath108 obtained by the simulation of 1000 realizations . as can be observed , the bound is very tight compared to the exact performance . thus , it follows that in order to achieve high normalized directivity ( i.e. directivity close to @xmath0 ) with @xmath0 nodes , the distribution of the nodes should be as sparse as possible . in fact , we have the following theorem : [ theorem_directivity ] for large @xmath60 and @xmath0 , @xmath109 is lower bounded by @xmath110 where @xmath111 is a positive constant independent of @xmath0 and @xmath60 ( @xmath112 ) . see appendix [ app : directive ] . note that the factor @xmath113 which appears in can be seen as a _ one - dimensional node density_. to verify the above theorem , fig.[fig : directivity_density ] shows the relationship between @xmath107 and the node density @xmath113 , as well as the lower bound in . [ bl][bl][.8]normalized directivity @xmath114 [ bl][bl][.8]exact ( n=16 ) [ bl][bl][.8]exact ( n=256 ) [ bl][bl][.8]lower bound ( theorem 1 ) [ bl][bl][.8]node density @xmath115 and node density @xmath113.,title="fig:",width=472 ] the above theorem indicates that there is a simple relationship between directivity and node density . it can be seen that the node density almost uniquely determines the normalized directivity @xmath109 . it is important to note that in order to achieve a certain normalized directivity with large numbers of nodes @xmath0 , the node density should be maintained to the desired value by spreading the nodes as sparsely as possible . alternatively , if the normalized region @xmath60 is fixed , it is not efficient in terms of normalized directivity to increase the number of sensor nodes . the above theorem also indicates that if the sensor nodes are uniformly distributed and if we are to choose @xmath0 nodes out of them , in terms of normalized directivity it may be better to choose them as sparsely as possible . in the previous section , we have seen that random arrays have nice average beampatterns with low sidelobes . however , the average behavior does not necessarily approximate a beampattern of any given realization unless @xmath116 . in fact , even though the average beampattern has a sharp mainbeam and sidelobes always close to @xmath62 , there is a large dynamic range of sidelobes among randomly generated beampatterns . as an example , the average beampattern and one particular realization of randomly generated sensor locations is shown in fig.[fig : av_vs_instant_beam ] . the mainbeam of the realization closely matches the average , but sidelobes may fluctuate with a large dynamic range and easily exceed the average level . and @xmath61.,width=472 ] therefore , in practice , the statistical _ distribution _ of beampatterns and sidelobes in particular , is of interest . by approximating the beampattern sidelobes as a complex gaussian process , lo @xcite has derived the distribution of the beampattern in the case of linear random arrays . in the following , we first derive a numerical method that allows calculation of the exact distribution of the beampattern without applying gaussian approximations . we then derive a convenient asymptotic form of the sidelobe distribution using a gaussian approximation similar to @xcite , and evaluate its validity in our framework . since the array factor is a sum of i.i.d . random variables , its distribution can be computed numerically by the characteristic function method . to this end , from let @xmath117 where @xmath118 and @xmath119 is defined in . the joint characteristic function of @xmath120 and @xmath121 is written as @xmath122 } \right\ } . \label{joint_chf1}\end{aligned}\ ] ] for a given pair of @xmath123 and @xmath124 , the above expectation is a single integral of a well - behaved function and can be calculated numerically . since @xmath125 is a sum of @xmath0 i.i.d . complex random variables , the joint probability density of @xmath126 and @xmath127 in is given by @xmath128^n e^{-j(\omega x + \nu y ) } \,d\omega \ , d\nu .\ ] ] the above integral can be computed efficiently using the two - dimensional fast fourier transform ( fft ) . finally , the complementary cumulative distribution function ( ccdf ) of the beampattern , i.e. , the probability that the instantaneous power of a given realization in the direction @xmath44 exceeds a threshold power , @xmath129 , is given by @xmath130 & = \pr \left [ \frac{\tilde{x}^2 + \tilde{y}^2}{n^2 } > p_0 \right ] = \int \!\!\ ! \int_{x^2 + y^2 > n^2 p_0 } f_{\tilde{x } , \tilde{y } } ( x , y ) \,dx\ , dy . \label{exact_dist_num}\end{aligned}\ ] ] the exact evaluation of the ccdf outlined above is computationally demanding , especially if the desired numerical precision is high . considering that the array factor consists of a sum of @xmath0 statistically independent random variables , as @xmath0 increases , by the central limit theorem we may expect that the array factor with any given direction , except at the deterministic angle @xmath131 , approaches a complex gaussian distribution . this approximation may typically result in a simpler distribution formula . to this end , we write as @xmath132 where @xmath133 since the @xmath134 s are i.i.d . random variables , as @xmath0 increases the distribution of @xmath135 and @xmath136 at the direction @xmath137 may approach that of a gaussian random variable with @xmath138 ^ 2 \triangleq \sigma^2_x \label{gauss_var_x } \\ e \left\ { y \right\ } & = 0 \label{gauss_mean_y } \\ \text{var } \left ( y \right ) & = \frac{1}{2}\left ( 1 - \frac{j_1\left ( 2 \alpha ( \phi ) \right ) } { \alpha ( \phi ) } \right ) \triangleq \sigma^2_y . \label{gauss_var_y}\end{aligned}\ ] ] note that @xmath139 , i.e. , @xmath135 and @xmath136 are orthogonal and thus statistically uncorrelated . the joint pdf of @xmath135 and @xmath136 is then given by @xmath140 the ccdf of @xmath141 can be expressed as @xmath142 & = \pr \left [ \frac{{x}^2 + { y}^2}{n } > { p}_0 \right ] = \pr \left [ \sqrt { { { x}^2 + { y}^2 } } > \sqrt{n { p}_0 } \right ] \nonumber \\ & = \int_{\sqrt{n{p}_0}}^{\infty } \int_{-\pi}^{\pi } \frac{r}{2 \pi \sigma_x \sigma_y } \exp \left ( -\frac{| r \cos\omega - m_x |^2 } { 2\sigma^2_x } -\frac { r^2 \sin^2 \omega } { 2 \sigma^2_y } \right ) d\omega \ , dr \nonumber \\ & = \int_{-\pi}^{\pi } \frac{1}{4 \pi \sigma_x \sigma_y u^2_{\omega } } e^{v^2_{\omega } - \frac{m_x^2}{2 \sigma_x^2 } } \left [ \sqrt{\pi } v_{\omega } \text { erfc}\left ( w_{\omega } - v_{\omega } \right ) + e^{-(w_{\omega } -v_{\omega } ) ^2 } \right ] d\omega , \label{exact_ccdf_gauss}\end{aligned}\ ] ] where @xmath143 for @xmath144 , the terms @xmath145 and @xmath146 in the variance expressions and rapidly decrease and their contribution to the resulting variances becomes minor . therefore , it is very likely that both variances are approximately equal in the sidelobe region . when this is the case , i.e. , if @xmath147 , the distribution of the complex envelope becomes a nakagami - rice distribution . consequently , the resulting integral can be expressed in terms of the first - order marcum - q function @xmath148 = q \left ( \frac{m_x}{\sigma_x } , \frac{\sqrt{n { p}_0}}{\sigma_x } \right ) = q \left ( \sqrt{2 } { m_x } , \sqrt{2 n { p}_0 } \right ) . \label{marcum_formula}\ ] ] furthermore , if the mean @xmath149 is zero , the envelope follows a rayleigh distribution and we simply have @xmath150 & = e^{- n { p}_0 } . \label{full_gaussian}\end{aligned}\ ] ] as we have seen , if the mean value of the array factor can be assumed to be zero , the distribution can be significantly simplified and thus analysis becomes readily tractable . from it is apparent that under the constant variance constraint the mean value increases as @xmath0 increases . therefore , when @xmath0 is large , the zero mean assumption may not be guaranteed in general . in section [ sec:3dbsidelobe_region ] , we have introduced the 3db sidelobe region , and in the following we derive properties of the mean value of the array factor in this region . from the definition of , the sidelobe in the 3db region satisfies @xmath151 it follows that @xmath152 from , , and , we have @xmath153 and thus we get @xmath154 therefore , in the 3db sidelobe region , the square of the mean is bounded by unity in the large-@xmath0 asymptote and thus the mean does not grow unbounded with the number of nodes @xmath0 . , and @xmath155.,width=472 ] in fig.[fig : distribution_r2 ] , the ccdf s computed with various formulae are shown with @xmath156 and @xmath157 , which corresponds to the sidelobe region . the exact formula of , the equal variance gaussian approximation of , and the zero - mean gaussian approximation of are shown in the figure . note that the precise gaussian case of was also calculated by numerical integration , but it is almost identical to the marcum - q function approximation in for this case and thus is not shown . as observed from fig.[fig : distribution_r2 ] , even the zero - mean gaussian approximation may be valid for this sidelobe region , but for @xmath158 the gaussian approximation will have some noticeable discrepancy with the exact value . this is due to the fact that the zero - mean approximation does not hold for this case . in fact , fig.[fig : three_side ] indicates that for this value of @xmath0 , the angle falls between the 3db sidelobe region and the mainlobe region and thus the zero mean assumption may not be accurate . fig.[fig : three_main ] shows the distribution at 3db beamwidth of the average beampattern defined by . in this case , the exact form , the precise gaussian , and the marcum - q approximation show different results even for relatively large @xmath0 , since at this angle the variance of the array factor is small and a large number of random variables must be summed in for a non - zero mean gaussian random variable to adequately approximate the sum . as observed , as the number of nodes increases , the mainbeam variance becomes small and approaches the mean value of -3db , as expected . therefore , for large @xmath0 the mainbeam can be made stable . this observation agrees with a similar result for linear arrays in @xcite . [ l][l][.8]exact [ l][l][.8]precise gaussian [ l][l][.8]marcum - q with exact , precise gaussian , and marcum - q formulas.,title="fig:",width=472 ] it is well known that unlike periodic or equally - spaced antenna arrays , many arrays with unequal spacing will yield no grating lobes for a large number of elements . this property is also preserved for random arrays @xcite , but in order to verify this , one may need to find the distribution of the maximum power of the sidelobes . in this section , we develop an approximate upper bound on the distribution of the peak sidelobes for random sensor networks . in the previous section , we have seen that the distribution of the beampattern samples within the sidelobe region can be characterized by a zero mean gaussian random variable if the zero - mean condition is satisfied . in the following , we further assume that the beampattern is a gaussian random _ process_. in this case , any two samples taken from the beampattern should be characterized by jointly gaussian random variables . in the linear random array framework , the distribution of peak sidelobes has been studied in @xcite , assuming the array factor is a gaussian process . for simplicity , only the 3db sidelobe region is considered and it will be assumed that the process is stationary with zero mean . the extension to the non - stationary case is studied in @xcite . in the following , the ccdf of the maximum peak sidelobe , which is the probability that the maximum peak sidelobe exceeds a given power level , will also be referred to as _ outage probability _ and denoted by @xmath159 . let @xmath160 denote the random variable representing the number of upward crossings at a given level @xmath161 per interval in the 3db sidelobe region @xmath162 . as shown in appendix [ app : level_crossing ] , assuming that the array factor in this region can be approximated as a zero - mean gaussian process , the mean of @xmath160 is given by @xmath163 note that the above function monotonically decreases with increasing @xmath161 only for @xmath164 , and thus is meaningful only in this region . finally , noticing that the outage probability is the probability that at least one peak exceeds level @xmath161 and is given by @xcite @xmath165 & = \sum_{k=1}^{\infty } \pr [ \nu(a ) = k ] \leq \sum_{k=1}^{\infty } k \pr [ \nu(a ) = k ] = e \left\ { \nu(a ) \right\},\end{aligned}\ ] ] then serves as an upper bound for the outage probability for the maximum sidelobe peak for @xmath164 . thus , we obtain the ccdf upper bound as @xmath166 & \leq 4 \left ( 1 - \sin \frac{\phi^{\text{zero}}_{n_0}}{2 } \right ) \sqrt{\pi } \tilde{r } \sqrt{n p_0 } e^{-n p_0 } , \qquad \text { for } \quad n p_0 > \frac{1}{2 } . \label{upperbound_sidelobe}\end{aligned}\ ] ] fig.[fig : ccdf_sidelobepeak ] shows a comparison between simulation results and the upper bound . for the simulation , the outage probability is calculated based on 10000 randomly generated realizations with the node density @xmath167 and only the peaks within the 3db sidelobe region are examined . also , in order to capture peak values accurately , the entire 3db sidelobe region of @xmath44 is sampled at a rate as large as @xmath168 . as can be observed , the bound is in good agreement with simulation for large @xmath0 . [ bl][bl][.8]@xmath167 .,title="fig:",width=472 ] let @xmath169 denote the threshold of the maximum peak value ( @xmath129 ) normalized by the average sidelobe level ( @xmath62 ) . since from @xmath170 approaches zero as @xmath60 increases , reduces to @xmath171 the above inequality illuminates the relationship between the outage probability and @xmath60 ( assuming that @xmath170 is negligibly small ) . fig.[fig : outage_sidelobepeak ] shows the maximum possible value of @xmath172 for a given outage probability and @xmath60 . as can be observed , the maximum sidelobe may grow as @xmath60 increases , but the amount is below 12db for many cases of interest . consequently , the maximum sidelobe level ( in the 3db region ) of randomly generated arrays may be written as @xmath173 , where the required margin @xmath174 depends on the parameters @xmath60 and @xmath175 , but not on @xmath0 . thus , increasing @xmath0 always results in a reduction of maximum sidelobe level . [ bl][bl][.8]@xmath176 [ bl][bl][.8]@xmath177 [ bl][bl][.8]@xmath178 [ bl][bl][.8]normalized radius@xmath60 [ b][b][.8]normalized sidelobe level @xmath172 .,title="fig:",width=472 ] so far , we have evaluated the beampattern assuming perfect knowledge of the initial phase for each node . in this section , we analyze the effect of the phase ambiguities in the closed - loop scenario as well as location estimation errors in the open - loop scenario . for each of the two scenarios , we derive the average beampattern and calculate the amount of mainbeam degradation . in the closed - loop case , the effects of imperfect phase may be easily derived , following the approach developed by steinberg @xcite . the initial phase of node @xmath10 in will now be given by @xmath179 where @xmath180 corresponds to the phase offset due to the phase ambiguity caused by carrier phase jitter or offset between the transmitter and receiver nodes . in the following , the phase offset @xmath180 s are assumed to be i.i.d . random variables . then , from , , , and , the far - field array factor ( with @xmath181 ) will be given by @xmath182 the average beampattern of will be replaced by @xmath183 similar to , direct calculation of results in @xmath184 where @xmath185 thus , as @xmath79 , the average beampattern will simply become a version of the original scaled by a factor of @xmath186 . let us now assume that the phase offset follows a tikhonov distribution , a typical phase jitter model for phase - locked loop ( pll ) circuits given in @xcite , @xmath187 where @xmath188 is the variance of the phase noise and @xmath189 is the @xmath57th order modified bessel function of the first kind . the corresponding attenuation factor is given by @xmath190 the variance of the phase noise @xmath188 is related to the loop snr of the pll by @xmath191 fig . [ fig : degradation_close ] shows the degradation factor @xmath192 with respect to the loop snr . as observed from the figure , a loop snr of at least 3db may be necessary for each node in order to reduce the overall beampattern degradation to less than 3db . in the open - loop case , our model of the initial phase is given in with @xmath30 , and if there are estimation errors in the location parameters @xmath35 and @xmath36 , the initial phase will be replaced by @xmath193 where @xmath194 and @xmath195 are the corresponding error random variables , each set assumed to be i.i.d . and also independent of @xmath35 and @xmath36 for simplicity . with the far - field approximation , we have @xmath196 + \delta { r}_k \cos(\phi_0 - ( { \psi}_k + \delta \psi_k ) ) \right\ } \nonumber \\ & = \frac{2\pi}{\lambda } a + \frac{4 \pi}{\lambda } { r}_k \left [ \sin\left ( \psi_k - \frac{\phi_0 + \phi - \delta \psi_k}{2 } \right ) \sin\left ( \frac{\phi_0 - \phi - \delta \psi_k}{2 } \right ) \right ] \nonumber \\ & \quad + \frac{2\pi}{\lambda } \delta { r}_k \cos ( { \psi}_k - ( \phi_0 - \delta { \psi}_k ) ) . \label{dist_phase2}\end{aligned}\ ] ] let @xmath197 . then , the right - hand side ( rhs ) of is given by @xmath198 the resulting far - field array factor of will then be given by @xmath199 and the beampattern is expressed as @xmath200 where @xmath201 conditioned on @xmath202 and @xmath195 , the angle @xmath203 can be seen as a uniformly distributed random variable , and thus the pdf of @xmath134 is given by . considering the fact that @xmath35 and @xmath194 are assumed to be statistically independent , we further assume for analytical purposes that @xmath134 and @xmath204 are statistically independent . then , again , the beampattern does not depend on the particular choice of @xmath9 . furthermore , on modeling @xmath205 as being uniformly distributed over @xmath206 $ ] and assuming the phase term of @xmath204 to be uniformly distributed over @xmath207 $ ] , the probability density function of @xmath204 will be given by @xmath208 , \qquad |v| \leq r_{\text{max } } .\ ] ] consequently , the average beampattern can be written as @xmath209 where @xmath210 and without loss of generality @xmath211 was assumed . in , @xmath212 denotes a generalized hypergeometric function which has an oscillatory tail but converges to zero as @xmath2 increases . also , assuming that the @xmath195 are uniformly distributed over @xmath213 $ ] and using the approximation @xmath214 which is valid for the beampattern around the mainbeam , we obtain @xmath215 since the hypergeometric function @xmath216 has a maximum peak value of 1 at @xmath217 , the above expression indicates that regardless of the value of @xmath60 , there may be two symmetric peaks around the mainbeam at @xmath218 resulting in a _ pointing error_. therefore , the mainbeam may spread over by a factor of @xmath219 . at the center of the mainbeam , we have @xmath220 fig . [ fig : degradation_open ] shows the degradation factor @xmath221 and @xmath222 for a given @xmath223 and @xmath224 . as observed from the figure and discussion above , the angle estimation error has two effects , i.e. , pointing error and mainbeam degradation . in particular , if we wish to suppress the mainbeam degradation below 3db , from the figure , we should choose @xmath225 . this means that the maximum angle estimation error should satisfy @xmath226 and as @xmath60 becomes large , the requirement of minimum angle ambiguity from becomes severe . [ bl][bl][.8]@xmath227 @xmath228 in this paper , we have analyzed the stochastic performance of random arrays for distributed collaborative beamforming , in the framework of wireless ad hoc sensor networks . it has been shown that under ideal channel and system assumptions , directivity of order @xmath0 can be achieved asymptotically with @xmath0 sensor nodes , as long as the sensor nodes are located sparsely enough . we have studied the average and the distribution of the beampattern as well as the distribution of the sidelobe peaks . several forms of the ccdf of the beampattern have been derived and compared , with particular emphasis on the gaussian approximation of the array factor . we have considered two scenarios of distributed beamforming and investigated the effects of phase ambiguity and location estimation error upon the resultant average beampatterns . our main conclusion is that , given a number of nodes randomly distributed over a large disk , one may form a nice beampattern with narrow mainlobe and sidelobes as low as @xmath62 plus some margin for maximum sidelobe peaks . also , the directivity approaches @xmath0 if the nodes are located as sparsely as possible however , our analysis is based on a number of ideal assumptions on the system and channel model . in practice , a number of open issues remain , such as applicability of beamforming when the destination or nodes in the cluster are in rapid motion or the channel suffers severe multipath fading . also , specific algorithms should be developed for frequency offset correction of each node as well as methods for initial phase or location estimation . finally , efficient protocols for sharing the transmit as well as calibration information among nodes are required . we first prove the following lemma : [ lemma_hyper ] a generalized hypergeometric function @xmath103 with @xmath68 can be bounded as @xmath229 where @xmath230 is a constant ( @xmath231 ) . we start with the integral form @xmath232 since the asymptotic form of @xmath233 given by is valid for @xmath68 , we have the following inequalities @xmath234 @xmath235 for some threshold value @xmath236 which should be determined numerically , and for some interval @xmath237 . the parameter @xmath238 is chosen to be the smallest non - negative value such that @xmath239 should hold , and this guarantees a continuity of the function at the threshold @xmath240 . fig.[fig : function_approximation ] illustrates the relationship of and with @xmath236 chosen as a cross point of @xmath241 between the functions @xmath242 and @xmath243 , yielding @xmath244 . the corresponding value of @xmath238 is 0.4664 . [ bl][bl][1.0]function of @xmath241 [ bl][bl][1.0]@xmath241 [ bl][bl][.8]@xmath245 [ bl][bl][.8]@xmath246 [ bl][bl][.8]@xmath247 and its upper bound with @xmath248.,title="fig:",width=472 ] substituting and into , we get for @xmath249 @xmath250 the first term on the rhs of is given by @xmath251 the second term on the rhs of is given by @xmath252 \nonumber \\ & = \frac{8}{\pi^2 x^2_0 } \frac{1}{x } \left ( 1 - \left ( \frac{x_0}{x } \right)^2 + o\left(1/x^4 \right ) \right ) + o\left ( \ln(x)/x^3 \right ) \nonumber \\ & = \frac{1}{x } \frac{8}{\pi^2 x^2_0 } + o\left ( \ln(x)/x^3 \right ) . \label{fx3_2}\end{aligned}\ ] ] consequently , we may write @xmath253 and the second term on the rhs of drops as @xmath2 becomes large . with @xmath244 and @xmath254 , the coefficient of @xmath255 can be calculated to be @xmath256 . from , , and lemma [ lemma_hyper ] , we have @xmath257 for large @xmath0 , the rhs of converges to with @xmath258 . in this appendix , we obtain the mean number of upward crossings of a given level of the zero mean gaussian process based on the approach of rice @xcite . assume that @xmath135 and @xmath136 are uncorrelated zero - mean gaussian processes with variance @xmath259 . let @xmath260 and @xmath261 and @xmath262 denote the corresponding processes differentiated by @xmath263 . by assumption , @xmath261 and @xmath262 become zero mean gaussian processes . in order to calculate the variance , first consider the autocorrelation function of @xmath135 at instants @xmath264 and @xmath265 given by @xmath266 where the other terms become zero by the zero mean assumption . also for the same reason , the first term of the rhs of may be approximated by zero . therefore , letting @xmath267 , we obtain @xmath268 differentiating the above with respect to @xmath269 twice , setting @xmath270 , and carrying out the statistical average with respect to @xmath1 , the variance of @xmath261 is given by @xcite @xmath271 likewise , one may obtain @xmath272 , and the joint pdf of @xmath273 is given by @xmath274 on changing the random variables in the polar coordinates via @xmath275 , @xmath276 and integrating out @xmath277 and @xmath278 , we obtain @xmath279 the number of positive ( upward ) crossings of the process @xmath123 at level @xmath161 per interval @xmath280 is given by @xcite @xmath281 consequently , the mean number of upward crossings for the interval @xmath162 is given by @xmath282 which results in . j. c. chen , l. yip , j. elson , h. wang , d. maniezzo , r. e. hudson , k. yao , and d. estrin , `` coherent acoustic array processing and localization on wireless sensor networks , '' _ proc . ieee _ , vol . 91 , pp . 11541162 , aug .

neutron star cooling is an important tool for the study of dense matter . by comparing cooling models with observations of thermal emission from these objects , we can gain insight into the equation of state ( eos ) of dense matter , the signatures of exotic particles , superfluid energy gaps , and magnetic field properties [ for a recent review , see @xcite ] . a neutron star loses the thermal energy with which it was born initially through neutrino emission , and later through photon emission , the change between these two stages occurring about @xmath0 yr after birth . several heating mechanisms may become important during late stages of the evolution of a neutron star , most of them related to the delayed adjustment to the progressively changing equilibrium state as the rotation slows down . these include the dissipation of rotational energy due to interactions between superfluid and normal components of the star @xcite and release of strain energy stored by the solid crust due to spin - down deformation @xcite . another of these mechanisms is _ rotochemical heating _ @xcite , which has its origin in deviations from beta equilibrium due to spin - down compression . as a neutron star spins down , the centrifugal force acting on each fluid element diminishes , changing the local value of the pressure . since the composition of neutron star matter in beta equilibrium is a one - parameter function , this compression results in a displacement of the equilibrium concentration of each particle species . reactions which change the chemical composition are in charge of driving the system to the new equilibrium configuration . but if the rate at which reactions do this task is slower than the change of the equilibrium concentrations due to spin - down compression , the system is permanently out of chemical equilibrium . this implies an excess of energy , which is dissipated by enhanced neutrino emission and heat generation . after its introduction , this heating mechanism was studied by several authors : @xcite applied it to quark stars , @xcite made an order - of - magnitude estimate of the effects of superfluidity , and @xcite studied the heating due to compositional transitions in the crust due to spin - down compression . all studies agree that rotochemical heating is particularly important for old neutron stars with fast rotation and low magnetic fields , features that are characteristic of millisecond pulsars ( msps ) . the most striking prediction associated with rotochemical heating is that , if the spin - down timescale is substantially longer than any other timescale involved ( with the likely exception of magnetic field decay ) , the star arrives at a _ quasi - equilibrium state _ , in which the temperature depends only on the current , slowly changing value of @xmath1 ( the product of the angular velocity and its time derivative ) , proportional to the spin - down power , and not on the star s previous history @xcite . this provides a simple way to constrain the physics involved in theoretical models , once the spin parameters and observed surface temperature of a msp are known . although the qualitative behavior of the thermal evolution of neutron stars with rotochemical heating is known , all previous studies made only order - of - magnitude estimates , ignoring the spatial structure of the star and thus not offering reliable predictions to be compared with observations . also unknown is the dependence of the thermal evolution on eos and stellar mass . in this work , we calculate the thermal evolution of msps with rotochemical heating , taking the structure of the star fully into account in the frame of general relativity , and using realistic eoss of dense matter . in order to do so , we develop a general formalism to treat the evolution of the temperature and departures of chemical equilibrium . as a first approach , we have chosen the simplest possible core composition , namely neutrons , protons , electrons , and muons ( @xmath2 matter ) , to see if we can explain observations with it , before invoking exotic particles . the main difficulty of treating the spatial structure of the star is the lack of an expression for the spin - down compression in the frame of general relativity . we develop such an expression with the aid of the slow - rotation approximation of @xcite . the astrophysical situation to be modelled is the stage after the accretion - driven spin - up of the pulsar has finished . although it is generally believed that crusts and cores of neutron stars have superfluid components ( e.g. , @xcite ) , in this work we ignore superfluidity . our plan is to make an extensive study of rotochemical heating in non - superfluid stars and compare results with observations , in order to assess whether invoking superfluidity ( whose parameters are currently very uncertain ) is necessary to explain observations . the structure of this paper is the following : in 2 , we develop our general formalism , as well as the expression for the spin - down compression rate . in 3 , we detail the input necessary for numerical calculations ( eos , reactions , heat capacities , etc . ) . in 4 , we describe our results , comparing our predictions with the recent detection of likely thermal ultraviolet emission from the nearest msp , psr j0437 - 4715 @xcite , and the lack of optical emission from the even closer `` classical '' pulsar psr j0108 - 1413 @xcite . we also list some other msps for which the rotochemical heating luminosity might become detectable . a short summary of our main conclusions is given in 5 . it is conventional in neutron star cooling calculations to divide the star into two regions : a nearly isothermal _ interior _ , which ranges from the center out to density @xmath3 g @xmath4 , and a thin _ envelope _ , which ranges from @xmath5 to the surface and where a strong temperature gradient exists @xcite . since we are modelling the thermal evolution of a msp long after accretion has stopped , it is safe to assume that thermal relaxation from an initial non - uniform internal temperature profile has already occurred , so that the redshifted internal temperature , @xmath6 is uniform ( see , e.g. , @xcite ) . here , @xmath7 is the time component of the metric of a non - rotating reference star , of which @xmath8 is the radial spherical coordinate . although spherical symmetry is broken for a rotating star , we describe the latter as a perturbation to the corresponding non - rotating star ( with the same total baryon number ) . we establish a lagrangian correspondence between the surfaces of constant @xmath8 in the non - rotating star with the constant - pressure surfaces of its rotating counterpart , on which all local thermodynamic quantities will be shown to be constant ( see [ sec : perturbation ] ) . the evolution of the internal temperature is given by the thermal balance equation @xcite , which for an isothermal interior is given by @xmath9,\ ] ] where @xmath10 is the total heat capacity of the star , @xmath11 is the total power released by heating mechanisms , @xmath12 is the total neutrino luminosity , and @xmath13 is the photon luminosity . these quantities are calculated as @xmath14 respectively , where @xmath15 is the proper volume element , @xmath16 is the specific heat ( heat capacity per unit volume ) of each particle species , @xmath17 is the total neutrino emissivity contributed by reactions , @xmath18 is the total heating rate per unit volume , @xmath19 is the stefan - boltzmann constant , @xmath20 is the stellar coordinate radius , @xmath21 , @xmath22 is the effective radius as measured from infinity , and @xmath23 the redshifted effective temperature . the surface temperature @xmath24 is obtained from the internal temperature by assuming an envelope model @xcite . since the neutrino emissivity and heating rate are modified when the neutron star is out of chemical equilibrium @xcite , the evolution of the temperature depends on how strongly it departs from the beta equilibrium state . for @xmath2 matter , this departure can be quantified by the chemical imbalances @xcite : @xmath25 where @xmath26 is the deviation from the equilibrium chemical potential of species @xmath27 , at a given pressure . since diffusion timescales are short compared with the evolutionary timescales to be considered @xcite , we assume uniform redshifted chemical potential deviations throughout the core : @xmath28 to obtain the time evolution of the chemical imbalances , we start by writing down the chemical potential of each particle species as a function of the number density of all particles : @xmath29 . we assume small departures from chemical equilibrium , which can be expressed by requiring that @xmath30 . in this approximation , the departures from the equilibrium particle number densities @xmath31 are related to the @xmath32 by @xmath33 where the partial derivatives are evaluated at the beta equilibrium state . to eliminate the effect of particle diffusion between different regions of the star , we integrate equation ( [ eq : dni_dmui ] ) over regions where free particles exist . after integration , we obtain the deviation from the equilibrium number of particles @xmath34 as a function of the redshifted chemical potential lags : @xmath35 with @xmath36 where we have used equation ( [ eq : delta_mu_infty ] ) . since the @xmath37 do not depend on time , we invert and take the time derivative of equation ( [ eq : dni_dmui ] ) , obtaining the evolution of the @xmath38 : @xmath39 the rate of change of @xmath34 is given by @xmath40 where @xmath41 is the change in the total number of particles of species @xmath27 due to reactions @xcite . here , @xmath42 is the net creation rate of particles of species @xmath27 per unit volume due to reaction @xmath43 . it can be checked that the @xmath44 satisfy baryon number and charge conservation . the quantity @xmath45 is the change in the equilibrium value of the total number of particles of species @xmath27 , which depends on the spin - down compression rate . we defer its calculation to the next subsection . the evolution of the redshifted chemical imbalances @xmath46 follows from equations ( [ eq : eta1 ] ) , ( [ eq : eta2 ] ) , and ( [ eq : delta_mu_infty ] ) : @xmath47 equations ( [ eq : dot_t ] ) , ( [ eq : dot_eta1 ] ) , and ( [ eq : dot_eta2 ] ) give a complete description of the thermal evolution of a neutron star with rotochemical heating and @xmath2 composition , given an expression for @xmath45 . for slow enough rotation frequencies , the deviation of the star from the non - rotating configuration can be treated as a small perturbation , which we describe in terms of a lagrangian formalism . using the fact that the total baryon number @xmath48 of a star is conserved , we can describe its interior in terms of surfaces of constant pressure @xmath49 enclosing a fixed number of baryons @xmath50 ( e.g. , the surface @xmath51 encloses @xmath48 baryons inside it ) . as the stellar rotation rate changes , these surfaces will readjust their shapes , with a corresponding change in @xmath49 , but keeping @xmath50 constant . since there is a one - to - one relation between the enclosed number of baryons @xmath50 and the radial coordinate of the non - rotating configuration @xmath8 ( for fixed @xmath48 ) , we use them interchangeably to identify a given surface , relating them by @xmath52 . the underlying assumption is that , on a surface of constant pressure , the other thermodynamical quantities are also constant . this is straightforward for neutron - star matter in beta equilibrium ( described by a barotropic equation of state ) , but not so obvious for departures from this state ( which are not necessarily barotropic , as the pressure - density relation will also depend on the local particle abundances ) . we show in appendix [ sec : gradients ] that , in a uniformly rotating , perfect - fluid star in hydrostatic equilibrium , surfaces of constant pressure coincide with those of constant energy density and gravitational redshift , regardless of whether the equation of state is barotropic .. ] this , together with diffusive equilibrium ( uniform redshifted chemical potentials everywhere in the stellar interior ) , ensures that the chemical composition on each surface is also constant . to quantify the spin - down compression rate and thus @xmath53 , we need an expression for the change in pressure with rotation frequency on each surface enclosing a constant @xmath50 , keeping @xmath48 constant . @xcite developed a perturbative approach in the frame of general relativity , relating quantities on rotating and non - rotating stars with the same central energy density @xmath54 ( but , therefore , different @xmath48 ) , by expanding the metric of an axially symmetric non - rotating star in even powers of the angular velocity @xmath55 . this perturbative approach is valid for stars rotating at frequencies much smaller than the kepler frequency @xmath56 , at which mass shedding from the equator occurs . for realistic equations of state , this limiting frequency can be well approximated by means of empirical formulae @xcite . the corresponding kepler periods , @xmath57 , are lower than the shortest msp periods measured to date [ see , e.g. , the atnf pulsar catalogue @xcite ] . our task is to express our spin - down compression rate at constant @xmath48 in terms of hartle s results , which assume constant @xmath54 . if the baryon number enclosed by a surface of constant pressure @xmath49 in a rotating star is @xmath58 , some partial - derivative manipulation yields @xmath59 where the subscripts denote quantities which are kept constant in carrying out the differentiation . we note that @xmath60 and @xmath61 are just @xmath62 and @xmath63 , respectively , evaluated at the surface of the star ( @xmath51 ) . also , since we are perturbing around the non - rotating configuration , we can evaluate all the derivatives at @xmath64 . @xcite found that the lowest - order change in a given stellar property with rotation is linear in the squared angular frequency @xmath65 , for a star with fixed central density . we can thus write @xmath66 and consider it independent of @xmath65 . to find the number of baryons enclosed by a surface of constant pressure , we recall the baryon number conservation law @xcite , and write @xmath67 where @xmath68 is the determinant of the metric , @xmath69 is the time component of the fluid s four - velocity , @xmath70 is the rest frame baryon number density , and @xmath71 is the spatial volume enclosed by the constant pressure surface ( throughout this section , we use @xmath72 ) . for the non - rotating case , we have @xmath73 where @xmath74 is the radial component of the non - rotating metric . for the rotating case , one has to express the perturbed metric of @xcite in the non - rotating coordinate system . keeping terms to order @xmath65 , we get @xmath75\\ \label{eq : ut_rot } u^t & = & e^{-\phi}\left [ 1 - \xi\frac{d\phi}{dr } - h + \frac{1}{2}r^2\sin^2{\theta}{\bar{\omega}}^2e^{-2\phi}\right],\end{aligned}\ ] ] where @xmath76 , @xmath77 , and @xmath78 are metric - perturbing functions which are proportional to @xmath65 and can be separated into spherical and quadrupolar parts , @xmath79 is the difference between the star s rotation frequency and the frame - dragging frequency , and @xmath80 is a lagrangian displacement that relates surfaces of constant @xmath81 in rotating and non - rotating stars of the same @xmath54 @xcite . the spherical polar angle is @xmath82 , and the mass enclosed inside the radius @xmath8 is @xmath83 . replacing equations ( [ eq : g_rot ] ) and ( [ eq : ut_rot ] ) in ( [ eq : baryon_law ] ) , keeping terms to order @xmath65 , taking the angular integral , and integrating by parts in @xmath8 , we get ) is not present , which is correct when integrating up to the surface , but not at intermediate layers . ] @xmath84dy + 4\pi r^2 e^{\lambda(\rho_c , r ) } \xi_0(\rho_c , r ) n(\rho_c , r).\end{aligned}\ ] ] the functions @xmath79 , @xmath85 , @xmath86 are obtained by solving first and second - order differential equations in the coordinate @xmath8 , as explained in @xcite , and which need a value for the rotation rate of the star @xmath55 and a non - rotating relativistic stellar structure as input . the zero subscript on functions indicates their spherical part ( the quadrupolar part vanishes when performing the angular integral , and the function @xmath78 has no spherical part ) . the remaining derivatives in equation ( [ eq : dpdomega2_final ] ) can be obtained from the non - rotating configuration . we first have @xmath87 where @xmath88 can be calculated analytically from the relativistic stellar structure equations @xcite . the other derivative has to be obtained by differentiating equation ( [ eq : n_norot ] ) : @xmath89 dy.\end{aligned}\ ] ] for numerical calculations , it is convenient to write the derivative in the first term of the right - hand side as @xmath90 in order to calculate the three derivatives with constant @xmath8 in the same way . the resulting spin - down compression rate , equation ( [ eq : dpdomega2_final ] ) , is plotted in figure [ fig : dpdomega2 ] , normalized by pressure and kepler frequency . the left panel displays results obtained for different eoss , with fixed central pressure , while the right panel shows results for different stellar models and fixed eos . as expected , the expression is negative , since spin - down compression means increase in @xmath49 with decrease of @xmath65 . compression is strongest at the center of the star . it can be shown that the central value is given by @xmath91 where @xmath92 is the adiabatic index , and the subscripts @xmath93 mean central values . since @xmath94 for the maximum mass non - rotating configuration , the divergence of the result for @xmath95 , as shown in the right panel of figure [ fig : dpdomega2 ] , is easy to understand . the graphs confirm the order - of - magnitude expression @xmath96 which implies that the perturbation is small as long as @xmath97 . with our expression for the spin - down compression rate at hand , we can write the change in the equilibrium number of particles of species @xmath27 with time as @xmath98 where the derivative @xmath99 is calculated in beta equilibrium . it is straightforward to check that the @xmath45 satisfy baryon number and charge conservation . the variables @xmath50 and @xmath49 are clearly continuous functions of @xmath8 in the non - rotating star , regardless of the presence of a phase transition . in appendix [ sec : continuity ] , we show explicitly that the same is true for @xmath100 . on the other hand , the equilibrium particle concentrations may have a discontinuous jump @xmath101 at the transition , which would contribute a dirac delta function to @xmath102 . therefore , there would be a finite contribution to @xmath45 from the transition , @xmath103_{trans}.\ ] ] to order of magnitude , the fractional contribution of a phase transition to the total value of @xmath45 ( and to the total heating of the star , if the reactions are of the same kind everywhere ) is the ratio of the jump @xmath101 to the total variation in @xmath104 across the stellar core . in the present work , as described in the following sections , phase transitions are not important , as the only transition in one of our eoss is fairly weak in each integration step as the ratio of the ( not necessarily small ) increment in @xmath105 to the small increment in @xmath49 . ] . in other cases , such as a star with a core of deconfined quarks of fairly uniform relative abundances , the phase transition may dominate the heating . in order to maintain diffusive equilibrium , a finite number of particles per unit time , @xmath106 , must be brought to ( or , if negative , away from ) the phase transition much faster than the latter progresses due to the spin - down of the star . therefore , a phase transition is the place where our assumption of diffusive equilibrium is most strongly put into question . however , in the case of interest to us , the low temperatures ( implying long mean - free paths and therefore fast diffusion ) and the extremely slow evolution should still validate this approximation . @xcite has shown that , due to the presence of two separate conserved charges ( baryon number and electric charge ) , phase transitions in neutron stars may not occur as a single discontinuity , but instead through a mixed phase , in which both phases coexist over a finite range in pressure ( or stellar radius ) . in this case , any discontinuities would be considerably weakened , if not fully eliminated , and we do not expect any special effects ( such as discussed above ) associated with the phase transitions . to implement the formalism of 2 numerically for the non - superfluid case , we need as input an eos , neutrino emissivities , a specific heat , and an envelope model . we then have to calculate the stellar structure and the corresponding integrals over the star . since our formalism requires a detailed knowledge of the structure of the star , we need an eos to determine the spatial dependence of the thermodynamical quantities by solving the relativistic stellar structure equations @xcite . in addition to the usual variables given in most published eoss ( pressure , energy density , chemical composition , and adiabatic index , all evaluated in chemical equilibrium ) , we need to know the partial derivatives of equation ( [ eq : dni_dmui ] ) ( which are easy to compute if we know the energy density or energy per baryon of interacting particles as a function of baryon number density and proton fraction ) and the baryon effective masses for computing reaction rates and heat capacities . we also want to cover a broad range of eoss , in order to assess the dependence of the quasi - equilibrium temperature on their properties . given our requirements , we have used three sets of eoss : 1 . the two most realistic eoss from @xcite ( apr ) for the core , supplemented with that of @xcite and @xcite , for the inner and outer crust , respectively ; 2 . five representative eoss from @xcite ( bpal ) for the core , with the same eoss for the inner and outer crust as for the previous set ; and 3 . a non - interacting fermi gas ( see , e.g. , @xcite ) for the whole star . in table [ tab : eos_par ] we list , for each eos , the mass @xmath83 , central energy density @xmath54 , coordinate radius @xmath20 , and effective radius as seen from infinity @xmath107 of the maximum - mass non - rotating configuration . since in the apr set the thermodynamical quantities are obtained from an analytical fit to tabulated many - body calculations @xcite , care has to be taken when interpreting results obtained for densities higher than the highest tabulated value , since extrapolation can lead to significant errors . this is the reason why we do not list the maximum - mass configuration for the a18 + @xmath108 eos . another feature of this set is that the a18 + @xmath108 + uix * eos becomes non - causal at densities greater than @xmath109 , which is the central density of a star of @xmath110 , slightly below the maximum mass formally allowed for this eos , @xmath111 . however , this eos is considered to be the state of the art among those derived from non - relativistic potentials , since it incorporates three - nucleon interactions and relativistic boost corrections @xcite . for this reason , we will in what follows take it as our reference eos . the eoss of the bpal set are labelled according to @xcite , who characterize them by the different magnitude of their nuclear incompressibility and density dependence of their symmetry energy . the values for the fermi gas eos listed in table [ tab : eos_par ] correspond to a central density slightly below the appearance of @xmath112 hyperons . the a18 + @xmath113 + uix * eos has an additional feature : a phase transition associated with the appearance of a neutral pion condensate at a density @xmath114 g @xmath4 . in order to connect the normal phase with the pion - condensed one , we assumed a maxwell transition , which implies a discontinuity in all thermodynamical quantities but the temperature , pressure and neutron chemical potential , resulting in an energy - density jump of 6.6% . @xcite also considered the possibility of a mixed - phase transition ( e.g. , @xcite ) , concluding that ( for a standard , @xmath115 neutron star ) the shell containing the mixed phase is only @xmath116 m thick . they also estimated the debye screening length of electrons to be smaller than the characteristic size of structures ( by a factor 6 - 12 ) , which would favor the maxwell transition as a better approximation to the real situation @xcite . also listed in table [ tab : eos_par ] are the kepler periods @xmath117 corresponding to most of the eoss , which are necessary in order to determine the extent to which the slow - rotation approximation holds for treating the spin - down compression . the apr and pal values were computed with the empirical formula of @xcite . this gives , for each eos , the maximum allowed rotation frequency in terms of the mass and radius of the maximum - mass non - rotating configuration . the accuracy for non - causal eoss is somewhat lowered from its optimal value of 1.5% @xcite , which applies for a18 + @xmath118 + uix*. the reason why the entry corresponding to a18 + @xmath108 is empty is that its maximum mass configurations occur at a central density which requires substantial extrapolation , making the empirical formula an unreliable estimator of @xmath57 . for the fermi gas eos , we adopt the exact value calculated for a pure neutron gas by @xcite , since the empirical formula of @xcite does not work for this case . so far , the fastest known msp is psr b1937 + 21 , with a period of 1.56 ms @xcite , which is about three times longer than kepler periods for all realistic eoss . since we are not taking superfluidity into account , direct and modified urca reactions are the dominant neutrino emission mechanisms . we have therefore neglected additional neutrino emission processes in the core and the crust of the star , and ignored possible modifications in urca rates due to the neutral pion condensate of the apr set ( see , e.g. , @xcite ) . away from beta equilibrium , neutrino emissivities and net reaction rates per unit volume of urca - type reactions in non - superfluid matter are given by @xcite @xmath119 respectively , where @xmath120 is the neutrino emissivity in equilibrium due to reaction @xmath43 , @xmath121 is the chemical imbalance affected by reaction @xmath43 , @xmath122 is the local temperature , @xmath70 is the baryon number density , @xmath78 is boltzmann s constant , and @xmath123 and @xmath124 are dimensionless control functions which depend on whether the reaction is of direct of modified urca type . these functions are given by @xcite @xmath125 where the subscripts @xmath126 and @xmath83 mean direct and modified urca , respectively . a finite @xmath127 of either sign enhances neutrino emission due to the even nature of the functions @xmath123 . the amount of energy released by each reaction of type @xmath43 is @xmath127 @xcite , thus the total energy dissipation rate per unit volume is @xmath128 with the signs defined by @xmath129 where @xmath130 is the rate per unit volume of the reaction that transforms the particle set @xmath48 to the set @xmath131 . this sign convention ensures that reaction rates are enhanced in the direction which restores equilibrium , since the functions @xmath124 are odd . the equilibrium emissivities of both direct and modified urca reactions can be written as @xmath132 where @xmath133 is a slowly varying function of @xmath70 , @xmath134 for direct urca , and @xmath135 for modified urca ( e.g. , @xcite ) . this decoupling of the temperature from the spatial part introduces a great simplification . to obtain the non - equilibrium neutrino luminosities and net reaction rates , we first define @xmath136 where we have used equations ( [ eq : t_int ] ) , ( [ eq : l_nu ] ) and ( [ eq : delta_mu_infty ] ) , and where @xmath137 is the equilibrium neutrino luminosity due to reaction @xmath43 . combining equations ( [ eq : q_noneq ] ) , ( [ eq : dgamma_noneq ] ) , ( [ eq : xi_def ] ) and ( [ eq : int_qeq ] ) , we obtain @xmath138 we denote modified urca reactions with electrons and muons by @xmath43 = @xmath139 and @xmath140 , respectively , in each case adding the contributions of the neutron and proton branches @xcite . direct urca with electrons or muons is denoted by @xmath43 = @xmath141 and @xmath142 , respectively . from the apr set , the only eos which allows direct urca is a18 + @xmath108 + uix * , exceeding the threshold for direct urca with electrons at @xmath143 g @xmath4 , which is the central density of a star of 2 @xmath144 . direct urca with muons lies in the non - causal regime . with the exception of bpal31 , all the eoss of the pal set allow electron direct urca . muon direct urca is forbidden for both bpal21 and bpal31 . combining equations ( [ eq : l_h ] ) , ( [ eq : delta_mu_infty ] ) , ( [ eq : dgamma_noneq ] ) , ( [ eq : q_h ] ) , ( [ eq : xi_def ] ) , and ( [ eq : int_dgamma ] ) , we obtain the heating luminosity corresponding to the reaction @xmath43 @xmath145 we can combine the heating and cooling contributions of each reaction into a single expression , which will prove to be useful for physical insight . if we define @xmath146 we can write the difference of equations ( [ eq : l_nu_j ] ) and ( [ eq : l_h_j ] ) as @xmath147 thus , the relative size of @xmath148 and @xmath149 will determine , through the functions @xmath150 , whether there is net cooling or heating due to the neutrino emitting reactions . since the functions @xmath150 are even , they do not depend on the sign of the @xmath151 . we plot the functions @xmath152 for the direct and modified urca case in figure [ fig : m_j ] . with @xmath153 , the constant term in @xmath123 gives the conventional cooling case . as @xmath154 grows , the cooling is enhanced by the additional neutrino emission due to the chemical imbalance , reaching a maximum cooling at @xmath155 . for larger values of @xmath154 , the heating becomes important , completely balancing the cooling for @xmath156 . for very large values of @xmath154 , the heating dominates , growing as @xmath157 and @xmath158 in the direct and modified urca case , respectively . in this limit , a fixed fraction of the energy released is emitted as neutrinos , 3/8 and 1/2 for modified and direct urca , respectively . the remainder stays behind , heating the star . since we want to simulate the evolution of a msp which has finished accretion , we use the fully accreted envelope model of @xcite to calculate the effective surface temperature @xmath159 where @xmath160 is the surface gravity in units of @xmath161 cm s@xmath162 , @xmath163 is the internal temperature in units of @xmath164 k , and @xmath165 . the photon luminosity then follows from equation ( [ eq : l_gamma ] ) . an accreted envelope is more heat transparent than a pure iron one for @xmath166 k , leading to faster cooling , whereas for @xmath167 k the opposite happens @xcite . to calculate the integral in equation ( [ eq : capcal ] ) , we use the specific heat capacities at constant volume @xmath16 for degenerate , non - superfluid fermions ( see , e.g. , @xcite ) . in analogy with the neutrino emissivity , we can separate the temperature dependence from the spatial one . from equation ( [ eq : capcal ] ) , we can define @xmath168 where @xmath169 and @xmath170 are the fermi momentum and effective mass of species @xmath27 , respectively . the latter can be obtained analytically for the apr and pal eoss ( see , e.g. , @xcite ) , while for the noninteracting fermi gas we use @xmath171 , where @xmath172 is the chemical potential of species @xmath27 . the sum is performed over all particle species and the integral is done over the region where each particle species exists . we take only free particles into account , neglecting the heat capacity due to the lattice of ions in the crust . in this work , we are neglecting processes in the crust which might also modify the total number of neutrons , protons , or electrons . we may ask here if this choice is a good approximation for computing the evolution of the chemical imbalances , since the short diffusion timescale limit means that the chemical equilibrium everywhere in the star can be restored by reactions occurring wherever these are fastest . @xcite calculated the heating of a neutron star due to spin - down compression in the crust . they found that neutron absorptions in the inner crust are the dominant non - equilibrium process , producing a very small heating rate @xmath173 , where @xmath174 is the spin - down power . although they did not consider neutron diffusion between different layers in the star , and modelled the spin - down compression as a one - dimensional process , we take their result to suggest a very small contribution to the total heating . since the inner crust accounts for about 2% of the total baryons in the star and the neutrino emissivities of the core overwhelm crustal emission processes , we integrated equations ( [ eq : bij_def ] ) only over the core and neglected processes in the crust , under the assumption that the error introduced would be small . our results confirm our assumption , since for the more realistic eoss we obtained heating rates due to core processes much bigger than iida & sato s results ( see equation [ eq : ratio_edot ] ) . we have to warn that neglecting the crust may not be a good approximation if superfluidity is included , since non - equilibrium processes in the crust can even overtake suppressed urca reactions in the core . it has also been suggested by @xcite that direct urca reactions could occur at the crust - core interface . we have not explored this possibility . however , we are aware that even a small amount of direct urca reactions in the inner crust would have a significant effect on the evolution of chemical imbalances . finally , we write the coupled equations for the time evolution of the temperature and the chemical imbalances explicitly for the non - superfluid case with @xmath2 composition , taking direct or modified urca reactions into account . for the internal temperature , from equations ( [ eq : dot_t ] ) , ( [ eq : capcal ] ) , ( [ eq : l_h ] ) , ( [ eq : l_nu ] ) , ( [ eq : l_gamma ] ) , ( [ eq : net_cooling ] ) , and ( [ eq : c_int ] ) , we get @xmath175.\end{aligned}\ ] ] for the chemical disequilibria , using equations ( [ eq : dni_dmui ] ) , ( [ eq : dni_def ] ) , ( [ eq : dot_ni ] ) , ( [ eq : dot_eta1 ] ) , ( [ eq : dot_eta2 ] ) , and ( [ eq : nieq ] ) , we obtain @xmath176 where we have defined @xmath177 and @xmath178 typical values for the different constants are plotted in figure [ fig : ctes ] . since @xmath179 , equations ( [ eq : evolucion_eta1 ] ) and ( [ eq : evolucion_eta2 ] ) have a positive term , proportional to the spin - down power , which arises from the change in the equilibrium concentrations of each particle species with spin - down and makes the chemical imbalances grow . the remaining terms , opposite in sign to @xmath180 and @xmath181 , account for the effect of reactions trying to restore beta equilibrium . equation ( [ eq : evolucion_ti ] ) has the photon luminosity as a negative term , and the net contribution of neutrino reactions , equation ( [ eq : net_cooling ] ) , which may be positive or negative depending on the absolute value of @xmath182 and @xmath183 . for numerical calculations , the evolution of @xmath184 was computed assuming magnetic dipole braking with no field decay , relating magnetic field , rotation period , and period derivative by the conventional formula @xmath185 g , where @xmath49 is measured in seconds . a typical solution of equations ( [ eq : evolucion_ti ] ) , ( [ eq : evolucion_eta1 ] ) , and ( [ eq : evolucion_eta2 ] ) is shown in the left panel of figure [ fig : evol_xi ] . first , the temperature starts falling as in the conventional cooling case , while the chemical imbalances grow due to spin - down compression . at some point , the ratios @xmath182 and @xmath183 ( right panel of figure [ fig : evol_xi ] ) are big enough , so that the net contributions of neutrino reactions , equation ( [ eq : net_cooling ] ) , are positive , and their sum exceeds @xmath186 . the temperature then starts rising . the chemical imbalances continue to grow , more slowly than the temperature and hence reducing @xmath182 and @xmath183 . finally , the star can arrive at a _ quasi - equilibrium _ state , where the rate at which spin - down modifies equilibrium concentrations is the same as the rate at which reactions drive the system towards the equilibrium configuration , with heating and cooling balancing each other @xcite . the subsequent thermal evolution can then be approximated by the simultaneous solution of the equations @xmath187 and @xmath188 . we discuss this _ quasi - equilibrium solution _ in detail in 4.2 . the pure modified urca case ( @xmath189 ) is the simplest to analyze . figure [ fig : evol_tdmuvarios ] shows the thermal evolution of the same star as in figure [ fig : evol_xi ] , this time with different initial conditions of internal temperature ( left ) and chemical imbalances ( right ) . the spin - down parameters were chosen so that the arrival at the quasi - equilibrium state is clearly visible , the short - dashed lines being the quasi - equilibrium solutions . we note that , in both cases , the value of the temperature at the quasi - equilibrium state and the time required to arrive do not depend on initial conditions . if direct urca reactions are open , solutions change noticeably . in figure [ fig : evol_urca ] we show the thermal evolution of three different stellar models built with the same eos , with identical initial conditions and spin - down parameters . the 1.4@xmath144 star is the same as that of figure [ fig : evol_tdmuvarios ] , with only modified urca reactions in operation . the 2@xmath144 star has electron direct urca operating , with the corresponding muon reaction forbidden . we see that the minimum temperature is higher and occurs at a much earlier time , after which the system reaches a metastable " quasi - equilibrium state , corresponding to partial equilibration between @xmath190 and @xmath191 . however , @xmath192 continues to grow , leading the system to a final quasi - equilibrium state , determined by muon modified urca . the 2.17@xmath144 star has direct urca with electrons and muons . this time the final quasi - equilibrium state occurs even earlier than the metastable " quasi - equilibrium of the 2@xmath144 star , with the temperature at about the same level . when only modified urca reactions operate , it is possible to solve analytically for the quasi - equilibrium values of the photon luminosity @xmath193 and chemical imbalances @xmath194 and @xmath195 , as function of stellar model and current value of @xmath184 . what makes these approximations possible is the high value of @xmath182 and @xmath183 at quasi - equilibrium . from the right panel of figure [ fig : evol_xi ] , we see that , at quasi - equilibrium ( @xmath196 yr ) , @xmath197 . this enables us to ignore all but the greatest power in the functions @xmath198 and @xmath199 ( equations [ eq : f_m_def ] , [ eq : h_m_def ] , and [ eq : m_j_def ] ) : @xmath200 the error in this approximation , for @xmath201 , is less than one part in 100 . to solve for @xmath193 , we set @xmath202 and use equations ( [ eq : c_h_def ] ) and ( [ eq : c_m_def ] ) to rewrite ( [ eq : evolucion_ti ] ) , ( [ eq : evolucion_eta1 ] ) , and ( [ eq : evolucion_eta2 ] ) as @xmath203 solving for @xmath194 and @xmath195 in equations ( [ eq : dot_eta1_qe ] ) and ( [ eq : dot_eta2_qe ] ) , we get @xmath204 where we have used equations ( [ eq : w_npe_def ] ) and ( [ eq : w_npm_def ] ) and the fact that @xmath205 . both chemical imbalances at quasi - equilibrium are positive , since @xmath184 and @xmath206 are negative . we finally replace equations ( [ eq : eta1_qe ] ) and ( [ eq : eta2_qe ] ) in ( [ eq : dot_ti_qe ] ) , to obtain the photon luminosity : @xmath207|\omega \dot{\omega}|^{8/7}.\ ] ] an interesting feature of equation ( [ eq : l_gamma_qe ] ) is that it does not depend on envelope model . the reason is that the limit @xmath208 implies that reactions are completely determined by chemical imbalances , eliminating any dependence on internal temperature . the energy released is divided in fixed fractions between neutrino emission ( @xmath209 = 3/8 $ ] ) and heating ( @xmath210 ) , the latter being radiated entirely as photons . note also that the ratio @xmath211 depends only on eos and stellar model , and is of order unity , thus we will make reference to either of them as @xmath212 . for the entire range of eoss and stellar models , we can write @xmath213 where @xmath214 is the period derivative measured in units of @xmath215 , and @xmath216 the period in ms . values for different stellar models built with the a18 + @xmath108 + uix * eos are shown in figure [ fig : ctes ] . using equation ( [ eq : l_gamma ] ) , we can rewrite equation ( [ eq : l_gamma_plnum ] ) in terms of the effective temperature : @xmath217 this time , the uncertainty due to eos and stellar model is smaller than 25% . we can also write the ratio of quasi - equilibrium photon luminosity ( and thus total heating rate ) to the spin - down power @xmath174 @xmath218 it is also possible to calculate @xmath219 for the quasi - equilibrium state . its spin - down power dependence , @xmath220 where @xmath221 is the exponent of the envelope model of @xcite ( see [ sec : envelope ] ) , shows that increasing @xmath222 reduces @xmath223 , decreasing the accuracy of the analytical approximation , although very slowly . although we did not include hyperons in our calculations , we may ask how their presence could modify the thermal evolution . the first particles to appear after muons are probably the @xmath224 and @xmath225 hyperons , which require the introduction of two additional chemical imbalances : @xmath226 once hyperons are present , the following non - leptonic reactions are open ( e.g. , @xcite ) : @xmath227 reaction ( [ eq : nnsp ] ) proceeds via weak interactions , since it does not conserve strangeness , while ( [ eq : nnln ] ) proceeds via strong interactions . both reactions have timescales several orders of magnitude shorter than beta processes @xcite . therefore , imbalances ( [ eq : eta_nnsp ] ) and ( [ eq : eta_nl ] ) remain small compared to @xmath228 and @xmath229 , and reactions ( [ eq : nnsp ] ) and ( [ eq : nnln ] ) contribute negligibly to the total heat generation . moreover , since chemical imbalances associated to direct or modified urca reactions with hyperons are linear combinations of @xmath230 , @xmath231 , @xmath228 , and @xmath229 , the latter two will determine the heat generation through both nucleon and hyperon reactions . in order to assess the importance of including the urca processes involving hyperons in addition to the nucleon processes , consider , as an example , the @xmath112 direct urca reactions @xmath232 whose associated chemical imbalance is @xmath233 . their net effect on equations ( [ eq : evolucion_eta1 ] ) and ( [ eq : evolucion_eta2 ] ) is to enhance the electron direct urca rate according to @xmath234 , reducing the chemical imbalance and thus the stellar surface temperature . this correction is small , since direct urca reactions with hyperons are at least a factor 5 weaker than their nucleon analogs @xcite . for the pure modified urca case , it can be checked from equation ( [ eq : l_gamma_qe ] ) that the correction to the surface temperature introduced by adding several reactions involving hyperons is a factor @xmath235^{1/28}$ ] , where @xmath236 and @xmath237 are the nucleon and hyperon urca luminosities , respectively . thus , the corrections due to hyperons in a purely modified urca or purely direct urca scenario are fairly negligible . the only important effect of hyperons in the context of rotochemical heating is that , in their presence , conditions for direct urca processes may become more easily satisfied , as is the case with reactions involving @xmath238 @xcite . the steep increase of the proton concentration with the appearance of @xmath112 hyperons ( e.g. , @xcite ) can also cause the condition for nucleon direct urca to be satisfied at lower densities than if hyperons are excluded from the models . we may now ask which of the known msps are likely to be in the quasi - equilibrium state . to answer this question , we need to know how long it takes to arrive at the quasi - equilibrium state , and if real pulsars are older than this time . since the negative equilibration " terms due to reactions in equations ( [ eq : evolucion_eta1 ] ) and ( [ eq : evolucion_eta2 ] ) are important only for imbalances near their equilibrium values ( recall the limit @xmath239 ) , we can assume that the initial evolution of @xmath212 is due only to @xmath1 : @xmath240 integrating over time , we get @xmath241.\ ] ] for the true solution to have reached the quasi - equilibrium state , this approximate solution must exceed @xmath242 . using equations ( [ eq : eta1_qe ] ) or ( [ eq : eta2_qe ] ) , together with equation ( [ eq : eta_ell_aprox ] ) , we may express this condition as an upper limit on the initial spin period @xmath243 with @xmath244 ( here and in the rest of this subsection , we assume that only modified urca processes are active . ) given the current spin parameters of a pulsar , condition ( [ eq : p0_qe ] ) gives the highest initial period it can have had , so that enough time has elapsed for it to have reached the quasi - equilibrium state . for msps , the constant @xmath48 is generally small . for a 1.4@xmath144 star built with the a18 + @xmath108 + uix * eos , we get @xmath245 values for other stellar masses are shown in figure [ fig : ctes ] . from equation ( [ eq : etapto_aprox ] ) , we can estimate a characteristic timescale for equilibration as @xmath246 @xcite calculated the thermal evolution of neutron stars with rotochemical heating for different values of the magnetic field , using the magnetic dipole braking model with no field decay . his conclusion was that a necessary condition for the quasi - equilibrium solution to be a good approximation to the exact one is that the spin - down timescale has to be longer than that for the growth of chemical imbalances . to quantify this , we estimate the spin - down timescale as the characteristic age , @xmath247 and take the ratio of @xmath248 to it , getting @xmath249 figure [ fig : evol_chiteqvarios ] shows the internal temperature of our conventional star after the rise from the minimum temperature , for different initial values of @xmath48 , with fixed initial @xmath248 . dotted lines are the quasi - equilibrium solutions at the instantaneous value of @xmath250 . the figure suggests that , for high values of @xmath48 , exact and quasi - equilibrium solutions depart from each other , whereas for small @xmath48 , solutions remain `` together '' . in fact , all solutions depart from each other , the departure being slower with time for smaller @xmath48 . for more general spin - down laws with arbitrary braking index @xmath251 , it can be shown analytically that this is true for @xmath252 . the increase in @xmath253 , where @xmath254 is the exact solution and @xmath255 the quasi - equilibrium one , is roughly linear with time . for @xmath256 at @xmath257 , and assuming a standard braking index @xmath258 , @xmath259 is @xmath260 lower than @xmath254 at @xmath261 , the difference growing 2% every @xmath262 yr . rewriting equation ( [ eq : a_pppto ] ) in terms of the magnetic field , @xmath263 where @xmath264 is the surface magnetic field in units of @xmath262 g , we can easily understand the results of @xcite , since increasing @xmath131 with fixed @xmath265 increases the initial value of @xmath48 . relating @xmath48 at the present time and at @xmath257 for an arbitrary , but constant , braking index , @xmath266 shows that the former is greater than the latter for all reasonable values of @xmath70 . thus , a small current value of @xmath48 is a good indicator that the star remains close to the quasi - equilibrium state , as long as it satisfies @xmath267 . since constraints on the initial periods @xmath268 of a few msps have been obtained from the cooling ages of their white dwarf companions @xcite , we can check which objects are likely at the quasi - equilibrium state . in table [ tab : predictions ] , further discussed in [ sec : othermsps ] , we show the range in @xmath268 for some msps obtained by @xcite . we made no distinction between the @xmath269 and @xmath258 case , and adopted the widest possible range in @xmath268 . in table [ tab : predictions ] we also show the value of @xmath48 and the upper limit on initial spin period @xmath270 required for the msp to be currently in the quasi - equilibrium state . among msps with constraints on initial periods , psr j1012 + 5307 is the only one not satisfying the conditions . the small values of @xmath48 imply that any initial periods substantially shorter than the current ones will leave the msps in equilibrium by the present time . recently , @xcite measured ultraviolet emission from psr j0437 - 4715 , the most probable explanation being that it corresponds to thermal radiation . since this object satisfies @xmath271 with small @xmath48 , we assume that it is already at the quasi - equilibrium state . in figure [ fig : ts_rinf ] , we plot the blackbody fit of @xcite as dashed lines , which correspond to 68% and 90% confidence contours . we overplot our values of @xmath272 for the spin parameters of psr j0437 - 4715 @xcite , as function of @xmath107 , for the eoss listed in table [ tab : eos_par ] . the bold lines show the range in @xmath107 corresponding to the mass constraint of @xcite , @xmath273 . the complete mass range for each eos goes from 1@xmath144 to 0.95@xmath274 . the upper limit was chosen to avoid the divergence in the spin - down compression rate for masses near @xmath274 ( see figure [ fig : dpdomega2 ] ) , which makes @xmath272 also diverge due to @xmath206 ( see equation [ eq : l_gamma_qe ] ) . the best agreement with observations is reached if only modified urca reactions are present . when direct urca reactions open , there are two abrupt drops in @xmath272 with increasing stellar mass , the first one ( @xmath275 ) due to electron direct urca , and the second ( @xmath276 ) due to muon direct urca . this can be seen in figure [ fig : ts_rinf ] for the curves calculated with bpal11 and a18 + @xmath108 + uix * , as examples of only electron direct urca , and in the curves corresponding to bpal32 and bpal33 , which have electron and muon direct urca . perfect agreement could be obtained by allowing masses within @xmath277 of the maximum - mass non - rotating configuration , for the eoss with only modified urca . however , we consider this scenario very unlikely . using the eoss which have only modified urca reactions within the allowed mass for psr j0437 - 4715 , we can constrain the predicted effective temperature to the narrow range @xmath278 k , about 20% lower than the blackbody fit of @xcite . there are three possible reasons why we are not matching their results : 1 . we are not taking superfluidity into account . this would reduce urca reaction rates , lengthening the equilibration timescale and raising the quasi - equilibrium temperature @xcite . we are neglecting other heating mechanisms ( some of them directly related to superfluidity ) , which could further raise the temperature at any stage in the thermal evolution . nonetheless , in msps , all proposed mechanisms are less important than rotochemical heating @xcite . 3 . the thermal spectrum could deviate from a blackbody , in the same way as the isolated neutron star rx j1856 - 3754 , which has a well - determined blackbody x - ray spectrum that underpredicts the optical flux @xcite , indicating a more complex spectral shape of its thermal emission . kargaltsev and collaborators stress the fact that psr j0437 - 4715 , despite its much larger spin - down age , has a higher surface temperature than the upper limit for the younger , `` classical '' pulsar j0108 - 1431 , @xmath279 k , inferred from the optical non - detection by @xcite . in this regard , we note that the spin - down power ( @xmath280 ) of the latter pulsar is 680 times lower than that of j0437 , making rotochemical heating substantially less important . its equilibration timescale , according to equation ( [ eq : tau_eq_def ] ) , is @xmath281 yr , longer than the age of the universe and certainly much longer than the spin - down age of the pulsar . thus , it is not expected to be even close to reaching its quasi - equilibrium state ( although it is old enough to have lost its initial heat content ) . its actual temperature should therefore be substantially smaller than the predicted quasi - equilibrium surface temperature , which is already @xmath282 times lower than that of j0437 , well below the observational upper limit . in table [ tab : predictions ] , we provide predictions of @xmath272 ( assuming modified urca reactions and no superfluidity ) for several msps . since the high - energy part of their expected thermal spectrum is highly absorbed by neutral hydrogen in the interstellar medium , we chose to order them by decreasing predicted quasi - equilibrium flux in the low - energy ( rayleigh - jeans ) regime , @xmath283 where @xmath284 is the distance to the object . each @xmath285 entry in table [ tab : predictions ] is scaled by the value for psr j0437 - 4715 . the first group of msps after psr j0437 - 4715 are nearby , single msps , for which @xmath285 is expected to be larger than for the binary msps , for which the initial period has been estimated from their white dwarf companion ( see [ sec : conditions ] ) . we consider the former as the primary targets for observations analogous to those of @xcite . however , even for psr j0437 - 4715 , the nearest and brightest msp known to date , the detection of thermal emission is rather difficult , so the observation of other objects of table [ tab : predictions ] will be a real challenge . the second group of msps are those with estimates for @xmath268 . however , their low fluxes make the detection of their thermal emission impossible for current instruments . optical detections are even more daunting than those in the ultraviolet . the extrapolated rayleigh - jeans spectrum of psr j0437 - 4715 gives magnitudes @xmath286 , @xmath287 , and @xmath288 , which are completely overwhelmed by the white dwarf companion . the most promising single msps have about three times lower expected fluxes at earth , i. e. , are another 1.2 magnitudes fainter , beyond the reach of current telescopes . in this work , we have made an extensive study of the effect of rotochemical heating in non - superfluid millisecond pulsars . we have set up a general formalism for the thermal evolution in the framework of general relativity , which takes the stellar structure fully into account and can be extended to include both superfluidity and exotic particles . a key ingredient in this formalism is the spin - down compression rate based on the slow - rotation approximation of @xcite . the main consequence of rotochemical heating is the arrival at a quasi - equilibrium state , in which the effective temperature of a neutron star depends only on the current value of the spin - down power . we argue that most of the known msps are very likely in the quasi - equilibrium state . if only modified urca reactions are allowed , the quasi - equilibrium bolometric photon luminosity in the non - superfluid case can be well approximated by @xmath289 independent of the neutron star envelope model . the influence of eos and stellar model on the quasi - equilibrium state is very weak , the only significant factor being the occurrence of direct urca reactions . if they are open for both muons and electrons , quasi - equilibrium temperatures are low , as in the conventional cooling case . if they are open only for electrons , the system arrives at a metastable " quasi - equilibrium , after which it proceeds as if only modified urca reactions with muons were present . the highest temperatures are reached when all direct urca reactions are forbidden . even our highest predicted quasi - equilibrium effective temperatures are lower than the blackbody fit of @xcite to the uv emission of psr j0437 - 4715 by about 20% . the inclusion of superfluidity will likely raise our predicted temperatures , being the subject of future work . we thank g. pavlov and o. kargaltsev for letting us know about their work in advance of publication and for kindly providing the data for figure [ fig : ts_rinf ] . the authors are also grateful to m. taghizadeh , m. van kerkwijk , r. mignani , d. page , m. catelan , m. a. daz , c. dib , and p. jofr for discussions that benefited the present paper , and an anonymous referee for thoughtful comments that improved its final version . this work made extensive use of nasa s astrophysics data system service , and received financial support from fondecyt through grant # 1020840 . in this appendix , we show that , in uniformly rotating , relativistic stars in hydrostatic equilibrium , composed of a perfect fluid , the surfaces of constant pressure and those of constant energy density coincide , as in the newtonian case , even if the equation of state is not barotropic . in addition , the gravitational redshift is constant on the same surfaces . this result , together with the diffusive equilibrium condition , allows us to describe the spin - down compression in terms of lagrangian changes of thermodynamic variables on isobaric surfaces enclosing a fixed baryon number , since all thermodynamical quantities are constant on each isobar . as a by - product , we also provide an alternative derivation of the equation of hydrostatic equilibrium for uniformly rotating , relativistic stars . the assumption a perfect ( i.e. , non - dissipative ) fluid , whose stress - energy tensor depends only on the energy density @xmath81 and ( isotropic ) pressure @xmath49 , is of course not strictly true in the astrophysical situation of interest in this paper , in which entropy is generated by non - equilibrium weak interactions , particle diffusion , and heat conduction , and lost ( from the star ) through the emission of neutrinos and photons . however , the total energy dissipated ( and eventually lost ) along the star s lifetime is much smaller than that associated with the mass , random motions , and interactions of its particles , as well as its rotation , and the time scale of the release of this energy is many orders of magnitude longer than the dynamical time and the rotation period of the star . therefore , the contribution of these dissipative processes to the energy and momentum fluxes is extremely small and can be neglected in the stress - energy tensor . this is in line with the usual approach ( also used in neutron star cooling calculations and in the rest of this paper ) of neglecting thermal effects when calculating the structure of the star , which is regarded as a fixed background on which thermal processes take place . the metric of a stationary , axially symmetric system can be written in the form ( see , e.g. , @xcite ) : @xmath290 where the metric coefficients are functions of @xmath8 and @xmath82 only . the components of the 4-velocity of any fluid element of the uniformly rotating star are @xmath291 , and @xmath292 , with @xmath293\right)^{-1/2}.\ ] ] we note that @xmath294 is the gravitational redshift factor , which corrects the energy of a freely moving particle , as it is measured by observers co - moving with fluid elements on different surfaces inside the star . , where @xmath295 are the covariant components of the particle s 4-momentum . due to the symmetries of the metric , the expression within parenthesis is conserved along the particle s world line , and is its redshifted energy . ] the energy - momentum conservation equation can then be written as @xmath296 where @xmath297 are the components of the stress - energy tensor of a perfect fluid . applying to this equation a projection operator orthogonal to the four - velocity , @xmath298 @xcite , we get @xmath299 since the only non - vanishing terms are those with @xmath300 and @xmath301 taking the values @xmath302 , the relevant connection coefficients are @xmath303 hence , equation ( [ eq : cons_enmom ] ) yields @xmath304 using equation ( [ eq : ut_def ] ) , we arrive at the equation of hydrostatic equilibrium @xmath305 which is in agreement with the result of @xcite , for the special case of rigid rotation . this shows that the redshift factor @xmath294 is also constant on the surfaces of constant @xmath49 . thus , we can write @xmath306 which implies that @xmath81 is also constant on the same surfaces . in this appendix , we show that each term of the spin - down compression rate , equation ( [ eq : dpdomega2_final ] ) , is continuous across a phase transition , even if the energy density and baryon number density change discontinuously . we use units with @xmath307 . the discontinuous quantities in the derivative @xmath308 , equation ( [ eq : dndp_def ] ) , are @xmath70 and @xmath88 . the gibbs free energy per unit volume for @xmath2 matter in beta equilibrium at @xmath309 is @xmath310 since , at a phase transition , the pressure @xmath49 and neutron chemical potential @xmath311 are continuous , the fractional size of the number density jump can be written as @xmath312 where @xmath313 is the positive jump in number density . from the relativistic stellar structure equations , we have @xmath314 the fractional size of the corresponding jump across the phase transition is @xmath315 thus , from equations ( [ eq : dn_n ] ) and ( [ eq : ddpdr_dpdr ] ) , we see that the quantity @xmath316 is continuous across a phase transition , and so is @xmath317 . regarding the derivative @xmath318 , equation ( [ eq : dndomega2_n0 ] ) , since the jump in @xmath70 is integrated in equation ( [ eq : n_norot ] ) , @xmath319 is continuous . for @xmath320 , the jump in @xmath70 due to the last term of equation ( [ eq : n_rot ] ) is cancelled by the contribution from @xmath321 in the integral . using @xmath322 where @xmath301 is dirac s delta function and @xmath323 , with @xmath324 the pressure at which the phase transition occurs , we can write @xmath325 where @xmath326 is the step function . equation ( [ eq : disc_dndomega2 ] ) is equal to minus the jump ( increasing @xmath8 ) of the last term in equation ( [ eq : n_rot ] ) . finally , for the derivative @xmath327 , the contribution from @xmath328 to the integral in equation ( [ eq : dndn0_p ] ) cancels out the discontinuity due to @xmath70 in the first term . using @xmath329 where @xmath330 , and @xmath331 the radial coordinate at which @xmath332 , we can write @xmath333 which is the negative of the jump ( increasing @xmath8 ) of the first term in equation ( [ eq : dndn0_p ] ) , since the sign of @xmath334 is also @xmath335 . the right panel of figure [ fig : dpdomega2 ] shows the spin - down compression rate inside the core of several stellar models built with the a18 + @xmath108 + uix * eos , which has a phase transition . the occurrence of the transition can be seen as a ( fairly harmless ) discontinuity in the derivative of each curve . lccccc a18 + @xmath108 & 1.55 & 1.86 & 9.81 & 13.42 & - + a18 + @xmath108 + uix@xmath336 & 2.19 & 2.78 & 9.97 & 16.79 & 0.51 + bpal 11 & 1.42 & 4.45 & 8.42 & 11.86 & 0.54 + bpal 21 & 1.70 & 3.46 & 9.33 & 13.69 & 0.56 + bpal 31 & 1.91 & 2.86 & 10.10 & 15.18 & 0.59 + bpal 32 & 1.95 & 2.66 & 10.56 & 15.64 & 0.63 + bpal 33 & 1.97 & 2.53 & 10.92 & 15.94 & 0.66 + fermi gas & 0.62 & 1.10 & 12.77 & 13.80 & 0.98 lccccccccc j0437 - 4715 & 5.76 & 1.86 & 0.14 & 0.72 & 1 & 0.17 & 5.33 & 2.45.3 & 1 , 2 + j1024 - 0719 & 5.16 & @xmath3371.85 & @xmath3370.25 & @xmath3370.79 & @xmath338 0.35 & @xmath3370.14 & @xmath3394.82 & & 3 , 4 + j2124 - 3358 & 4.93 & 1.23 & 0.27 & 0.74 & 0.28 & 0.13 & 4.64 & & 3 , 4 + j0030 + 0451 & 4.87 & @xmath3371.00 & 0.32 & @xmath3370.70 & @xmath3370.19 & @xmath3370.12 & @xmath3394.60 & & 4 , 5 + j1744 - 1134 & 4.07 & 0.71 & 0.36 & 0.74 & 0.15 & 0.09 & 3.90 & & 6 + b1257 + 12 & 6.22 & 4.26 & 0.45 & 0.86 & 0.12 & 0.22 & 5.63 & & 3 , 4 + j0034 - 0534 & 1.88 & @xmath3370.67 & 0.53 & @xmath3371.41 & @xmath3370.14 & @xmath3370.03 & @xmath3391.85 & @xmath3371.4 & 2 , 4 , 7 + j1012 + 5307 & 5.26 & 1.34 & 0.41 & 0.71 & 0.11 & 0.14 & 4.93 & @xmath3395.1 & 2 , 4 , 8 + b1855 + 09 & 5.36 & 1.74 & 0.90 & 0.76 & 0.03 & 0.15 & 5.00 & @xmath3374.6 & 2 , 3 , 9 + j1713 + 0747 & 4.57 & 0.81 & 1.10 & 0.70 & 0.02 & 0.11 & 4.35 & 2.23.1 & 2 , 10 + j1640 + 2224 & 3.16 & 0.16 & 1.15 & 0.60 & 0.01 & 0.05 & 3.08 & @xmath3391.6 & 2 , 4 , 11 + j2019 + 2425 & 3.93 & @xmath3370.70 & 1.49 & @xmath3370.76 & @xmath3370.01 & @xmath3370.08 & @xmath3393.87 & 0.93.9 & 2 , 4 , 12

inflation is a successful paradigm for the physics of the early universe @xcite . besides solving the classical problems of modern cosmology ( e.g. the flatness , entropy , and horizon problems ) , it provides primordial perturbations in perfect agreement with the observations @xcite . a challenge for inflationary models is to protect the required flatness of the inflaton potential against radiative corrections . this protection can be provided by an approximate shift symmetry as in models of natural inflation @xcite . the symmetry can be broken by a controllably small amount , the most known example of this is the case of an axion field acquiring a potential from instantons . the application of this to inflation , however , requires a greater than planckian axion decay constant @xmath1 @xcite , which may not be stable against gravitational corrections @xcite and which may be impossible to realize in string theory @xcite . proposed solutions to this problem include using two @xcite or more @xcite axions , which provide an effective large scale evolution even if the decay constants of the original axions are sub - planckian , requiring nontrivial compactifications in string theory @xcite , suitably coupling the axion to a @xmath7-form @xcite , modifying the axion kinetic term @xcite , and slowing down the axion evolution through particle production @xcite as in warm inflation @xcite . in particular , in the mechanism of @xcite the dissipation occurs through the production of a u(1 ) field coupled to the inflaton @xmath0 through the interaction @xmath8 ( where @xmath9 is the u(1 ) field strength , and @xmath10 its dual ) . @xcite showed that this coupling can also affect the background evolution if the u(1 ) field is replaced by a su(2 ) field with a nonvanishing vacuum expectation value ( vev ) . specifically , due to the interaction with the vev of the vector multiplet , the inflaton can be in slow roll even if its potential would otherwise ( i.e. , in absence of this interaction ) be too steep to give inflation . such a model has been dubbed _ chromo - natural inflation _ in @xcite . term besides the usual @xmath11 yang - mills term - in the limit in which the axion is close to the bottom of its potential @xcite . perturbations of gauge - flation were studied in the last of @xcite . ] the model is characterized by the action @xmath12 , \label{action}\end{aligned}\ ] ] where @xmath0 is the axion inflaton and @xmath13 . we use the convention @xmath14 for the levi - civita tensor . the vector field has the vev for space indices , and @xmath15 for internal su(2 ) indices . the index @xmath16 should not be confused with the scale factor @xmath17 , that enters in the line element as @xmath18 $ ] . we denote by dot a derivative with respect to physical time @xmath19 , and by prime a derivative with respect to conformal time @xmath20 . ] @xmath21 which is chosen to give isotropic expansion . we note that , for a generic theory with vector fields , an isotropic solution may be unstable against anisotropic perturbations . for instance , this can be the case when a dilaton - like coupling @xmath22 of a scalar inflaton @xmath23 to a su(2 ) field with the vev ( [ a - vev ] ) or to an orthogonal u(1 ) triplet is arranged so to produce a scale invariant spectrum for the vectors . isotropy is instead preserved if the vector fields are massive @xcite , as in the current model ( if the mass arises from an explicit breaking , one should also check that the theory has no ghosts @xcite ) . ref . @xcite performs a thorough analysis of the background evolution of the model . the study of the chromo - natural inflation theory at the perturbative level was recently performed in @xcite , for @xmath24 where @xmath3 is the mass of the vector field fluctuations in this limit @xcite ( as we also discuss below ) . when this condition is realized , the vector field can be integrated out while , at the same time , leaving its mark on the inflationary dynamics ( this is an explicit realization of the so called _ gelaton _ @xcite mechanism ) . in this limit , chromo - natural inflation is equivalent to a single scalar field @xmath25 theory ( where @xmath26 and @xmath27 ) . the non - canonical kinetic term precisely encodes the effect of the gauge fields . in @xcite it is also shown that an effective field theory equivalence of chromo - natural inflation to a non - canonical @xmath28 lagrangian holds as long as @xmath29 . beyond this limit , a general perturbative study of the dynamics of the gauge field and the axion becomes necessary in order to test the stability of the theory and formulate its predictions . this is precisely the scope of the present work . we perform a full linear order study of scalar , tensor and vector perturbations and we show that the inflationary background solution of the model is stable for @xmath30 , and it is otherwise unstable . a full phenomenological study of the model is beyond our purposes . nonetheless , we explore the scalar and tensor modes production for a given choice of @xmath31 and @xmath32 for which the coupling with the vector fields is crucial to ensure slow - roll . specifically , we choose @xmath33 , and @xmath34 , while @xmath35 is a free parameter controlled by the value of @xmath36 . the stability condition @xmath37 provides a lower bound for this ratio . too small values are also excluded because they lead to a too red spectrum of the scalar modes ( we obtained this numerically ; this behavior is also seen in the analytic study of @xcite in the regime of validity of their analysis ) . on the other hand , the amount of gravity wave signal increases with @xmath35 , and a level which can be observed in the current or forthcoming experiments @xcite can be obtained even if the inflaton in the model spans a range of some orders of magnitude below @xmath38 , evading the so called lyth bound @xcite ( in contrast to what was expected in @xcite ) . therefore , requiring a stable solution , with a sufficiently flat scalar power spectrum , and sufficiently small tensor modes provides constraints on the parameters of the model that go in opposite directions . for our choice @xmath33 , and @xmath34 we could not find any acceptable solution . we expect that the situation should improve at larger @xmath1 , where the inflaton potential becomes flatter ( and the model approaches conventional slow roll inflation ) . the paper is organized as follows : in sec . [ sec : model ] we review the background evolution of the model for different ( stable or unstable ) ranges of the theory ; in sec . [ sec : eqs - perturbations ] we start the perturbation analysis by introducing the most general decomposition of the metric , the inflaton , and the gauge field fluctuations , identifying the physical degrees of freedom and verifying that scalar , vector and tensor perturbations decouple at linear order ; we also quantize the system and define the cosmological correlators ; in secs . [ sec : tensors ] , [ sec : vectors ] and [ sec : scalars ] we study the fluctuations for , respectively , tensor , vector and scalar modes ; finally in sec . [ sec : conclusions ] we draw our conclusions . + we supplement our perturbation analysis in appendix [ app - dg - scalar ] , where we show that neglecting scalar metric fluctuations does not affect the stability analysis of the model ( while it considerably simplifies the computations ) . chromo - natural inflation is described by the action ( [ action ] ) of a pseudo - scalar field @xmath0 coupled to an @xmath39 gauge multiplet . the vector multiplet has the vev ( [ a - vev ] ) . the _ 00 _ component of einstein equations reads @xmath40 inflation can only occur if the potential @xmath41 is the major contribution to the total energy density , i.e. @xmath42 ; the parameters @xmath43 , @xmath44 , @xmath45 and @xmath46 are all much smaller than unity . + the equations of motion for the inflaton and the gauge field are : @xmath47 if we neglect @xmath48 , @xmath49 and @xmath50 , one can solve eqs . ( [ eqone]-[eqtwo ] ) for @xmath51 and @xmath52 @xcite @xmath53 in these equations we then assume @xmath54 and we obtain @xmath55 eq . ( [ eqthree ] ) can be rewritten as an equation of motion for the gauge field in terms of an effective potential @xmath56 where @xmath57 is minimized by @xmath58 in the minimum , the two terms of @xmath57 are parametrically equal to each other . therefore , the same is true for the last two terms in ( [ eqfour ] ) . as a consequence , the first term on the right hand side of ( [ eqfour ] ) dominates over the other two in the regime @xmath59 studied in @xcite . this implies that @xcite @xmath60 for general values of the parameters , inserting @xmath61 , eq . ( [ eqfour ] ) becomes @xmath62 by looking at eq . ( [ eqone ] ) , one realizes that , when the gauge field settles in its minimum , it originates an actual damping term for the motion of the axion . notice also that , when @xmath63 , the kinetic energy of the fields can be disregarded ( see e.g. eq . ( [ eqthree ] ) ) ; therefore the slow - roll parameter ( which would normally receive contributions also from @xmath64 and @xmath65 ) will be , instead , mostly due to @xmath66 and @xmath67 @xmath68 finally we have @xmath69 from eq . ( [ chid ] ) we can compute the number of e - foldings @xmath70 where we defined @xmath71 . + let us introduce the parameter @xmath72 @xmath73 in terms of which the expression ( [ efolds ] ) rewrites @xmath74 an upper bound on @xmath75 can be obtained by `` pretending '' that the slow roll approximation holds at all values of @xmath0 , corresponding to @xmath76 . the resulting upper limit is maximized for @xmath77 , where it evaluates to @xmath78 . in the @xmath79 limit , the upper limit acquires the simple form @xmath80 . in terms of @xmath72 , the slow roll solutions read @xmath81 where we have defined @xmath82 figure [ fig : bck ] allows to appreciate the accuracy of these slow roll approximations . in the figure we show the evolution of @xmath51 for four choices of @xmath72 , comparing the exact evolution with the slow roll approximation ( [ slow ] ) . ; we then employ ( [ slow ] ) to have the initial value for @xmath51 and @xmath83 . we set @xmath84 , and we then obtain the initial value for @xmath85 by solving the @xmath86 einstein equations exactly at the initial time . we make sure that the initial value of @xmath0 leads to more than @xmath87 e - folds of inflation ( so that the slow - roll solution can be achieved ; we note that the system starts slightly displaced from the slow roll solution , since we set @xmath84 ; the displacement is however very small , since @xmath88 in the slow roll solution , and the background evolution quickly approaches the slow roll solution ) . the evolutions shown in figure [ fig : bck ] are restricted to the final @xmath87 e - folds of inflation . ] in all cases , we take @xmath89 , as the main purpose of this model is to allow slow roll for @xmath31 . this is `` compensated '' by @xmath32 , and in all cases we fixed for definiteness @xmath34 . we note that fixing the value of @xmath72 does specify a relation between @xmath90 and @xmath36 , but does not fix these two values . as typical in inflationary models , the parameters can be specified only from fixing the normalization of the power spectrum of the scalar modes to the observed value @xmath91 @xcite . the evolutions shown in the figure cover @xmath87 e - folds of inflation . we note from the figure that @xmath51 performs very small oscillations around the slow roll solution . we believe that they are due to the fact that @xmath92 is tracking a time dependent minimum ( [ qmin ] ) ( we do not show this here , but also this tracking is extremely accurate during the slow roll phase ) . the oscillations appear somewhat large in the logarithmic scale chosen , but we see that they do not lead to any net departure from the slow roll solution . moreover , in section [ sec : scalars ] we compare full numerical solutions of the scalar perturbations , obtained using the full numerical background solutions , with analytical solutions , for which the slow roll approximations are used , and we also find excellent agreement . as we discussed after eq . ( [ efolds2 ] ) , decreasing @xmath72 in the @xmath93 region , while keeping the other parameters ( including the initial value of @xmath0 ) fixed , decreases the amount of inflation . as a consequence , @xmath94 can result in sufficient inflation only if the inflaton is initially close to the top of the potential . another way to express this is to note that the inflaton rolls faster as @xmath72 decreases ( with the other parameters kept fixed ) , as it is clear both from the slow roll expression ( [ slow ] ) and from the figure . perhaps the most surprising feature of the model is that , although all the solutions shown in figure [ fig : bck ] appear to be acceptable inflationary solutions ( in all cases the slow roll solution appears to be an attractor ; we note however that the background dynamics only probes homogeneous departures from the slow roll solution ) , and although they only differ from each other by the value of @xmath72 , the background solution with @xmath95 is unstable , while the other ones are stable . this emerges from the study of the scalar perturbations around these solutions that we perform below . in this section we discuss at the formal level how we compute , we quantize , and we solve the linearized theory for the perturbations of the model around the background solution discussed in the previous section . the discussion is divided in two subsections . in the first subsection we discuss how the perturbations can be divided into three groups , decoupled from each other at the linearized level . in the second subsection we give the form of the quadratic action and we discuss how we compute the corresponding linearized theory for the perturbations . there are @xmath96 perturbations in the system , one of the inflaton , @xmath97 of the su(2 ) vector field , and @xmath98 of the metric , that we decompose as : @xmath99 \nonumber\\ g_{00 } & = & - a^2 \left ( 1 - 2 \phi \right ) \nonumber\\ g_{0i } & = & a^2 \left ( b_i + \partial_i b \right ) \nonumber\\ g_{ij } & = & a^2 \left [ \left ( 1 + 2 \psi \right ) \delta_{ij } + 2 \partial_i \partial_j e + \partial_i e_j + \partial_j e_i + h_{ij } \right ] \nonumber\\ \label{deco1}\end{aligned}\ ] ] in this expression , @xmath100 is the su(2 ) index ( we also denote by @xmath16 the scale factor , as there is no ambiguity between it and an su(2 ) index ) , while @xmath101 ranges over the spatial coordinates . we denote as `` tensor modes '' the perturbations @xmath102 and @xmath103 , which we impose to be transverse ( @xmath104 ) and traceless ( @xmath105 ) ; due to these properties , the tensor sector contains @xmath7 perturbations . we denote as `` vector modes '' the perturbations @xmath106 , which we impose to be transverse ( @xmath107 ) ; due to this , the vector sector contains @xmath98 perturbations . we denote as `` scalar modes '' the remaining @xmath108 perturbations . we point out that the terms tensor / vector / scalar " are appropriate for the perturbations of the metric and of the inflaton , as they indicate how these modes transform under a spatial rotation . we extend this terminology also to the perturbations of the vector field , following the notation of @xcite , even if , strictly speaking , these terms are inappropriate ( given that the su(2 ) index has been used in the decomposition ) . we nonetheless adopt it since the fact that the vector vev is diagonal ( @xmath109 ) plus the transversality properties that we have imposed guarantee that the tensor / vector / sectors that we have defined above remain decoupled from each other at the linearized level @xcite . we fourier transform these modes @xmath110 where @xmath111 denotes any of the perturbations , and we study the theory in fourier space . in our stability study , we solve for the perturbations at the linearized level , and therefore we study a mode with a given momentum @xmath112 ( modes of different momenta are coupled to each other at the nonlinear level ) . we can actually fix the orientation of @xmath112 along the @xmath113axis without loss of generality . starting from a general direction for @xmath112 , we rotate the coordinates so that @xmath114 , where @xmath115 is the unit vector along the @xmath113axis , and @xmath116 . after a general rotation , @xmath117 is no longer proportional to @xmath118 ; however we can re - obtain @xmath119 through a global su(2 ) rotation . therefore , we can set @xmath120 without loss of generality . this choice simplifies our algebra . we need to remove the redundancy associated to general coordinate and su(2 ) transformations . under an infinitesimal coordinate transformation with parameter @xmath121 , @xmath122 and we remove the freedom of infinitesimal coordinate transformations by setting @xmath123 . consider instead an su(2 ) transformation with infinitesimal parameter @xmath124 ( with @xmath125 transverse ) . under this transformation , @xmath126 and we can fix the su(2 ) freedom by setting @xmath127 . clearly , also other modes of the metric and of the gauge field change under these transformations , and different gauge choices can be made . our choices are motivated by the fact that ( i ) they completely fix the freedom , and ( ii ) they preserve all the @xmath128 and @xmath129 modes . these perturbations are nondynamical , as they enter in the quadratic action of the perturbations without time derivative , and can be immediately integrated out . with our gauge choices , and with @xmath130 , the decomposition ( [ deco1 ] ) acquires the explicit form @xmath131 and @xmath132 we verified explicitly that the scalar modes ( @xmath133 ) , the vector modes ( @xmath134 ) and the tensor modes ( @xmath135 ) are decoupled from each other at the linearized level . namely , the quadratic action for the perturbations splits into three decoupled parts @xmath136 we have seen that the total action for the perturbations splits in a sum of three decoupled quadratic actions . let us denote by @xmath137 the vector formed by the perturbations in one of these three systems . we can perform a transformation @xmath138 so that the action for the array @xmath139 is of the type @xmath140 \label{formal}\ ] ] hermitianity of the action implies that @xmath141 is an anti - hermitian matrix , and @xmath142 a hermitian matrix ; the matrices obtained in the current model are actually real . in the following we quantize the system ( [ formal ] ) , so to obtain the initial conditions for the modes , and we give an expression for the correlators between the modes . this discussion summarizes the one of @xcite . we first `` rotate '' @xmath143 where @xmath144 is a unitary matrix ( so that the kinetic term in ( [ formal ] ) remains canonical ) , satisfying @xmath145 as @xmath141 is real , @xmath144 can also be taken real , and , therefore , orthogonal . we note that @xmath144 is not uniquely determined by this condition , and we fix it by setting @xmath146 at the initial time @xmath147 ; the goal of the present discussion is also to understand when this initial time can be set . as we will see , the explicit solution for @xmath144 is not needed . in terms of the vector @xmath148 , the action becomes @xmath149 , \nonumber\\ { \tilde \omega}^2 & \equiv & r \left ( \omega^2 + k^t \ , k \right ) r^t.\end{aligned}\ ] ] we then introduce the orthogonal matrix @xmath150 satisfying @xmath151 and we decompose @xmath152 , \label{deco - psi}\ ] ] where @xmath153 destroys / creates a quantum with the frequency @xmath154 . these operators satisfy the algebra @xmath155 = \delta^{(3 ) } \left ( \vec{k } - \vec{p } \right ) \ , \delta_{ij}.\ ] ] for the systems that we study , in the deep sub - horizon regime @xmath156 @xmath157 with @xmath158 constant at leading order in slow roll . therefore , provided that @xmath159 , we can set the initial conditions in the adiabatic vacuum @xmath160 where @xmath161 is a constant arbitrary orthogonal matrix ( at the practical level , eq . ( [ h - in ] ) is an approximate solution of the equations of motion at early times ; the closer @xmath162 is to @xmath163 , the better this approximation is . this determines how early @xmath147 needs to be taken ) . the matrix @xmath161 is unphysical , as it drops from the equations of motion for the modes , and from the physical correlators ( which , as we will see , are given in terms of @xmath164 ) . the freedom associated to @xmath161 is the generalization to @xmath75 fields of the freedom of changing by a constant phase the wave function in the single field case . in the final expression in ( [ h - in ] ) we have used the freedom associated to @xmath161 to set the wave functions to be real at the initial time . combining ( [ psi - delta ] ) and ( [ deco - psi ] ) , we have @xmath165 using the fact that @xmath166 at @xmath147 , so that initially @xmath167 , we arrive at @xmath168 we start from these initial condition and evolve the equations of motion for the modes following from ( [ formal ] ) @xmath169 in this discussion , all the expressions are given in fourier space . let us denote by @xmath170 the original fields in real space , @xmath171 we have the two point correlation functions @xmath172 where the power spectra are given by @xmath173 \nonumber\\ & = & \frac{k^3}{2 \pi^2 } \ , { \rm re } \left [ \left ( { \cal m } \ , { \cal d } \ , { \cal d}^\dagger \ , { \cal m}^t \right)_{ij } \right ] . \label{power}\end{aligned}\ ] ] these correlators are the theoretical prediction , to be confronted with the statistical average of the corresponding quantities . the power spectra ( [ power ] ) are the generalization to a system of @xmath75 fields @xmath170 of the standard power spectrum of single field inflation . to summarize , starting from the original fields @xmath174 in momentum space , we perform ( [ y - delta ] ) to have a canonical kinetic term for @xmath175 . we then decompose this field in terms of annihilation / creation operators of the physical quanta in the system ( the particles of frequencies @xmath154 ) . we work in terms of the coefficients @xmath176 of this decomposition . we can set the initial conditions ( [ delta - in ] ) for these coefficients , provided that the initial time is chosen sufficiently early such that @xmath177 . starting from these initial conditions , the coefficients @xmath176 evolve according to ( [ delta - eom ] ) . we will see that for some choice of parameters some of the perturbations become exponentially large on a timescale @xmath178 , signaling an instability of the linearized theory . for values of parameters leading to stable solutions , the coefficients enter in observable quantities through the power spectra ( [ power ] ) . we introduce the two doublets @xmath179 for the left and right helicities , where @xmath180 the action for the tensor modes splits into two separate actions , one for the left and one for the right helicity doublet , which are formally of the type ( [ formal ] ) . the action for the left - helicity doublet is characterized by @xmath181 and @xmath182 \nonumber\\ \omega_{12}^2 & = & a k \ , \frac{2 g q^2}{m_p } + \frac{\cal h}{m_p } \left ( q ' + { \cal h } \ , q \right ) - \frac{g \lambda q^2 a \chi'}{f m_p } \nonumber\\ & & \simeq a k h { \rm o } \left ( \frac{1}{\sqrt{\lambda } } \right ) + a^2 h^2 { \rm o } \left ( \frac{1}{\sqrt{\lambda } } \right ) \nonumber\\ \omega_{22}^2 & = & k^2 - a k \left ( 2 g q + \frac{\lambda}{f } a \chi ' \right ) + g \frac { \lambda } { f } q a \chi ' \nonumber\\ & & \simeq k^2 - a k h \ , { \cal a}+ a^2 h^2 { \cal b } \label{left - action}\end{aligned}\ ] ] where the first expression for each coefficient is exact , while the second one is obtained from the slow roll approximation ( [ slow ] ) ( we note that @xmath183 ) . the quantities @xmath184 and @xmath185 are both of @xmath186 . using the slow roll result ( [ eqthree ] ) with @xmath83 given by ( [ qmin ] ) , they evaluate to @xmath187 the action for the right - helicity doublet is related to ( [ left - action ] ) by @xmath188 , signaling the breaking of parity invariance induced by the evolution of the pseudo - scalar inflaton . the interactions between the gravity wave and vector field tensor perturbations are slow - roll suppressed . however , the effective frequency squared @xmath189 of @xmath190 turns negative for an intermediate interval of time next to horizon crossing.dependence becomes negligible , and the mass term for the two vector perturbations @xmath191 reproduces the value @xmath3 given in @xcite . ] this leads to a tachyonic growth of @xmath192 in this interval of time , and , correspondingly , to a growth of @xmath193 . the same growth does not occur in the right - helicity sector due to the opposite sign of the linear term in @xmath194 . the situation is analogous to that first studied in @xcite , where the interaction @xmath195 between a vector field and a pseudo - scalar rolling inflaton results in a tachyonic growth of the vector modes of a given helicity . we note from ( [ ab ] ) that the length of the tachyonic region increases with increasing @xmath196 ( the last expression is the slow roll approximation ) . this corresponds to a larger tensor mode production with growing @xmath197 . this corresponds to decreasing @xmath72 in the numerical examples that we show below . the vector modes @xmath198 are non - dynamical , and can be integrated out . namely , they enter in the quadratic action of the perturbations without time derivatives , and therefore their equations of motion are algebraic equations in these variables ( recall that we are in momentum space ) . when we solve these equations , we obtain an expression for the non - dynamical modes in terms of the two dynamical modes @xmath199 and @xmath200 . we then insert this expression back in to the quadratic action for the vector modes , and obtain an action for @xmath201 only . in other words , the non - dynamical modes do not introduce additional degrees of freedom , but are completely determined by the dynamical ones . after integrating out the non dynamical modes , we define @xmath202 where @xmath203 ( with the upper @xmath204 sign corresponding to @xmath205 and the lower @xmath206 sign corresponding to @xmath207 ) . the modes @xmath208 are the canonical modes of the system , and are decoupled from each other at the linearized level : @xmath209 . \end{aligned}\ ] ] an explicit computation gives @xmath210 } { m_p^2 p^2 \pm 2 g m_p^2 q p + 2 g^2 q^2 \left ( m_p^2 + q^2 \right ) } \label{om2vpm}\ ] ] where @xmath211 is the physical momentum , and the coefficients @xmath212 are functions of background quantities . we note that @xmath213 we also note that the numerator in ( [ om2vpm ] ) is always positive . the exact expressions for @xmath212 are readily obtained , but they are not illuminating . we report here the leading expressions in the slow roll @xmath32 expansion : @xmath214 \nonumber\\ c_3 & = & h^3 \left [ \frac { g^2 q^2}{h^2 } \left ( 6 g \frac{q}{h } + \frac{\lambda}{f } \frac{\dot{\chi}}{h } \right ) + { \rm o } \left ( \frac { 1 } { \lambda } \right ) \right ] \nonumber\\ c_2 & = & h^4 \left [ \frac { g^2 q^2}{h^2 } \left ( 3 + 4 g \frac{\lambda}{f } \frac{\dot{\chi}}{h } + 8 g^2 \frac{q^2}{h^2 } \right ) + { \rm o } \left ( \frac { 1 } { \lambda } \right ) \right ] \nonumber\\ c_1 & = & h^5 \left [ \frac { g^4 q^4}{h^4 } \left ( 4 g \frac{q}{h } + 6 \frac{\lambda}{f } \frac{\dot{\chi}}{h } \right ) + { \rm o } \left ( \frac { 1 } { \lambda } \right ) \right ] \nonumber\\ c_0 & = & h^6 \left [ 4 g^5 \frac{q^5}{h^5 } \frac{\lambda}{f } \frac{\dot{\chi}}{h } + { \rm o } \left ( \frac { 1 } { \lambda } \right ) \right ] \nonumber\\\end{aligned}\ ] ] using ( [ slow ] ) , it is immediate to see that , in each of the above expression , the dominant term in the square parenthesis that we have written explicitly is a @xmath215 coefficient . moreover , the second term on the right hand side of ( [ om2vpm ] ) evaluates to @xmath216 . therefore , at leading order in slow roll , @xmath217 in the deep sub - horizon and super - horizon regime , respectively . we conclude that the vector sector is stable . we note that the super horizon limit of ( [ omega - v ] ) is precisely the slow roll expression of @xmath218 in the @xmath219 limit . the two dynamical vector modes originate from the perturbations of the vector field , and the expression for the mass in this limit coincides with what found in @xcite . after gauge fixing , the system of scalar perturbations comprises of @xmath220 mode from the inflaton , @xmath221 , @xmath222 from the su(2 ) field , @xmath223 , and @xmath224 from the metric , @xmath225 and @xmath226 . among , these @xmath227 modes , @xmath228 and @xmath229 are dynamical , while the other three modes are nondynamical ; namely they enter in the quadratic action without time derivatives , and they can be integrated out ( see the discussion at the start of the previous section ) . integrating out the metric perturbations makes the algebra extremely involved ; for this reason , in the study presented in the main text we make the approximation of setting @xmath230 from the start . we then integrate out @xmath231 , and we denote the resulting quadratic action by @xmath232 . we also performed the full computation , which we present in appendix [ app - dg - scalar ] . we denote by @xmath233 the quadratic action obtained by including all the modes and by integrating out the @xmath222 nondynamical ones . both @xmath232 and @xmath233 are functionals of the @xmath222 dynamical modes . we expand all entries in these actions in slow roll . as we show in appendix [ app - dg - scalar ] , all entries of the matrices of the two actions agree at the leading order in this expansion at all scales ( namely , for all values of @xmath234 ) , with a single exception . the exception is the the @xmath235 coefficient of the frequency matrix @xmath142 , for which the agreement is excellent only up to @xmath236 . the discrepancy that takes place afterwards is surely completely irrelevant for the stability study that we perform here ( as we will see , the instability , if present , manifests itself deeply inside the horizon ) . not surprisingly , the metric perturbations do not affect the stability of the background solution . very likely , this disagreement has also no significant consequence for the power spectra that we show below , since it manifests itself only after the power has frozen ( see the appendix for a detailed study ) . we therefore set @xmath230 , and integrate out @xmath231 . we then define @xmath237 in terms of the multiplet @xmath238 , the quadratic action for the scalar modes acquires , up to a total derivative , the form ( [ formal ] ) . we denote as @xmath239 and @xmath240 the @xmath241 matrices entering in this action . these matrices have the following entries @xmath242 and @xmath243 \nonumber\end{aligned}\ ] ] @xmath244 we stress that these expressions are obtained by disregarding the scalar metric perturbations , but that they are otherwise exact . we can also verify that the eigenvalues of the @xmath245 part of @xmath246 are @xmath247 in the super horizon regime and for @xmath219 , in agreement with @xcite . we solve the theory specified by this action following the steps outlined in subsection [ subsec : quantization ] . we assume that , after inflation , only the inflaton field provides a sizable contribution to reheating ( we note that the energy in @xmath248 is much smaller than the inflaton energy during inflation ) . in this limit , we have the curvature perturbation @xmath249 , with the power spectrum @xmath250 where @xmath251 is given in ( [ power ] ) , with @xmath174 being the three fields on the left hand side of ( [ canonical ] ) . in figure [ fig : pkt ] , we then present the time evolution of @xmath252 for a single mode ( a given @xmath194 ) and for the same choices of background parameters that we used in figure [ fig : bck ] for the background evolution . for definiteness , we considered in all cases the mode that leaves the horizon @xmath87 e - folds before the end of inflation , and we denoted the corresponding comoving momentum by @xmath253 . we choose the initial time of the evolution such that the mode is deeply inside the horizon at the start , and the last term in ( [ omega ] ) is negligible . we observe that the choice @xmath95 leads to an instability of the linearized theory , while the other three cases are stable , and are characterized by a power that freezes outside the horizon , as in the standard inflationary models . the instability of the @xmath95 choice manifests itself at some scale inside the horizon , namely @xmath254 . even the stable solutions show a different evolution at this scale ( note in the figure the small change in the slope of @xmath255 for the stable cases at the same scale at which the @xmath95 solution becomes unstable ) . we can see this analytically , by considering the approximated expressions of the scalar system inside the horizon . we will obtain the scale @xmath256 times a numerical factor , where @xmath257 ( in the numerical examples that we have shown , @xmath258 ) . let us discuss the approximation in more details . first of all , using the slow roll conditions ( [ slow ] ) , we find @xmath259 in all cases that we have studied . therefore , we can disregard @xmath260 in comparison to @xmath261 inside eqs . ( [ ks - dg0 ] ) and ( [ os - dg0 ] ) . from this , and from the slow roll approximations ( [ slow ] ) , we obtain , in the sub - horizon regime @xmath262 and @xmath263 we want to solve the evolution equations ( [ delta - eom ] ) until @xmath264 . we can do so by rewriting them as @xmath265 by inserting only the terms written explicitly in ( [ k - s - sub ] ) and ( [ o - s - sub ] ) , and by treating these terms as constant . all these approximations amount in considering only terms that contribute to the dynamics at @xmath266 . we note for instance that @xmath267 in this regime , so that it is consistent to set @xmath268 in the approximated equation , while retaining @xmath269 and the dominant term of @xmath270 . we also note that the fastest evolving coefficient in the matrices is @xmath271 that evolves on a @xmath272 timescale ; therefore time derivatives of the terms in ( [ k - s - sub ] ) and ( [ o - s - sub ] ) introduce at most terms with an additional factor @xmath4 , which are therefore suppressed in the @xmath273 and @xmath274 regimes . performing these approximations , the system ( [ app - sca - eq ] ) reduces to a set of linear second order equations with constant coefficients . the equations split in three separate groups , one for the complex unknowns @xmath275 ( with @xmath101 ) , one for the the complex unknowns @xmath276 , and one for the three complex unknowns @xmath277 . in each group we need to solve three second order differential equations , and therefore we have @xmath227 possible solutions ; we note that the three groups have the identical set of equations , and they differ only in the initial conditions . therefore the solutions are of the type @xmath278 where the matrices @xmath279 are integration constants ( so to match the initial conditions ) and @xmath280 are constant numbers . the system is unstable if any of the @xmath280 has a real and positive part . by solving the system , one can see that the only coefficient that can possibly be real is @xmath281^{1/2 } \label{xi}\end{aligned}\ ] ] and therefore @xmath282 where , using the slow roll condition , @xmath283 if the expression in parenthesis is negative the background solution is stable . otherwise , there is an instability at sufficiently large wavelengths . therefore @xmath284 for any fixed @xmath72 , this condition is violated at sufficiently small @xmath285 . this provides an upper bound to the amount of inflation in the model ( for any given choice of parameters ) . one needs to verify that this upper bound is comparable with the required amount of inflation . for the evolutions shown in figure [ fig : bck ] , the inflaton field , at @xmath87 e - folds before the end of inflation , evaluates to @xmath286 for @xmath95 , to @xmath287 for @xmath288 , to @xmath289 for @xmath290 , and to @xmath291 for @xmath292 . correspondingly , the rhs of the condition ( [ stability - param ] ) evaluates to @xmath293 for @xmath95 , to @xmath294 for @xmath288 , to @xmath295 for @xmath290 , and to @xmath296 for @xmath292 . the criterion ( [ stability - param ] ) therefore indicates that the choice @xmath95 does not lead to a stable inflationary solution of @xmath87 e - folds ( recall that @xmath1 and @xmath297 are fixed to @xmath298 and to @xmath299 , respectively , in this example ) , while the other choices do . this is in perfect agreement with figure [ fig : pkt ] . ( [ stability - k ] ) confirms that the instability , if it exists , takes place at @xmath300 times an order one factor . for example , for the unstable @xmath95 choice , the criterion ( [ stability - k ] ) gives an instability starting at @xmath301 , corresponding to @xmath302 , in excellent agreement with the evolution seen in the figure . finally , let us verify that the instability is extremely fast . let us assume that we have a background solution with @xmath303 . at any given moment , modes with @xmath304 have @xmath305 , and are therefore stable . moreover , the instability is also negligible at very large scales , given that @xmath306 for @xmath307 . however , for @xmath308 , a quick study of eq . ( [ xi ] ) shows that @xmath309 for @xmath271 smaller than , but parametrically equal to @xmath310 . ( for @xmath311 , the maximum of @xmath312 is obtained for @xmath313 ) . this corresponds to a very short instability time @xmath314 . therefore , each mode experiences a strong instability while still inside the horizon , and we conclude that the background solution is unstable whenever the condition ( [ stability - param ] ) is violated . we conclude this section by showing in figure [ fig : spe ] the power spectrum @xmath315 for the stable @xmath316 configurations , which we discuss in the next section . we performed a complete study of the linear order quantum fluctuations for the chromo - natural inflation model . we separated the metric , gauge and axion fluctuations into scalar , vector and tensor modes , verifying explicitly that they decouple at the linearized level . we computed their equations of motion and worked out the quantization of the system . the tensor sector consists of the two gravity wave polarizations plus two modes from the gauge field . the gauge mode of one helicity becomes tachyonic for some finite interval of time next to horizon crossing , sourcing one gravity wave helicity . the vector sector consists of two dynamical modes that originate from the vector field , and that remain perturbatively small . the scalar sector contains three dynamical modes , one originating from the inflation and two from the gauge field . we showed that , for some parameter choice , one of the eigenfrequencies of this system , that we denoted by @xmath312 , can become imaginary inside the hubble horizon , leading to a fast instability . let us compare this with a standard result in inflation . specifically , let us consider the gravity wave amplitude @xmath317 . the canonical variable @xmath318 obeys @xmath319 with the dispersion relation @xmath320 . as a consequence @xmath321 outside the horizon , corresponding to a frozen amplitude of @xmath322 therefore , although the frequency of the canonical mode becomes tachyonic outside the horizon , its magnitude is not large enough to lead to a physical instability . the situation is analogous for a test scalar field or for the scalar perturbations in the standard case . for the case at hand , the eigenfrequency @xmath312 is given in eq . ( [ xi ] ) . in this discussion , for illustrative purposes , let us approximate the full expression obtained in eq . ( [ xi ] ) with @xmath323 , where @xmath324 is defined in ( [ lambd ] ) , and @xmath325 an order one factor . no instability appears if @xmath325 is negative . otherwise , an instability takes place for modes of physical momentum @xmath326 . the timescale of the instability is @xmath327 , ( or , equivalently , the amplitude of the scalar modes grows proportionally to a large power of the scale factor ) which indeed corresponds to a fast instability . this heuristic discussion reproduces the results obtained from the precise form of @xmath312 ( see section [ sec : scalars ] for the precise computation ) . moreover , we also performed a fully numerical and exact study of the scalar perturbations that confirms these analytical results . the parameters of the model can be chosen so that @xmath312 remains real inside the horizon ( equivalently , @xmath328 in the heuristic expression ) so that the instability is avoided . this corresponds to choosing @xmath329 , where @xmath83 is the vev of the vector field , and @xmath36 the su(2 ) coupling . quite interestingly , the quantity @xmath3 coincides with the mass of the fluctuations of the vector field in the @xmath219 regime @xcite , and it approximates well this mass for @xmath330 . therefore , the inflationary solution is stable if and only if the vector field is sufficiently heavy . we have obtained the boundary of the stable region @xmath219 analyzed in @xcite and our formalism can be readily employed to study the phenomenology of the boundary region . this study is beyond the purposes of the present work . for illustrative purposes , we have however computed the power spectrum of @xmath331 for some sample choices of parameters . specifically , figure [ fig : spe ] shows the power spectrum for an axion decay constant @xmath332 , for @xmath34 ( we stress that the main motivation for the model is to provide inflation for a sub - planckian axion decay constant , and that this can be obtained for sufficiently large @xmath297 ) , and for three choices of @xmath72 . we note that , among those shown in figure [ fig : spe ] , only the power spectrum obtained for @xmath292 is sufficiently flat to meet the observational bounds @xcite . we numerically found that the spectral tilt is a decreasing function of @xmath72 in the @xmath333 interval ( equivalently , it is an increasing function of @xmath35 in this interval ) . this behavior can also be seen in the analytic relation given in @xcite in the large @xmath3 regime . therefore , both the requirement of stability and of flatness of the power spectrum pose a lower bound on @xmath35 . on the other hand , the discussion around eq . ( [ mgh ] ) leads to the conclusion that a large @xmath35 leads to a detectable or ruled out gravity wave signal . the enhanced signal is parity violating , which results in nonvanishing tb and eb correlations . a measure of the net handedness of the tensor modes is @xmath334 where @xmath335 is the power spectrum of the left / right - helicity gravity wave modes . the corresponding observational bounds have been studied in @xcite . in figure [ fig : r - dchi ] we compare the bounds presented in @xcite with the value of @xmath336 and @xmath337 obtained in this model , for our choice of @xmath332 and @xmath34 , and for different values of @xmath72 in the @xmath338 range ( the tensor power spectra are obtained from the @xmath235 element of ( [ power ] ) for the left and right helicity tensor sectors , following the procedure outlined in subsection [ subsec : quantization ] ) . greater values of @xmath72 do not lead to a visible gravity wave signal even in a cosmic variance limited experiment , while lower values are ruled out by the current limit @xmath339 @xcite . we point out that all values of @xmath72 considered in this plot are actually ruled out because they lead to a too small value of @xmath340 ( the largest value @xmath341 is obtained for @xmath342 , and @xmath340 then further decreases at larger @xmath72 ) . we stress that our choice of @xmath332 and @xmath34 is only dictated by the requirement of @xmath343 , so that the coupling to the gauge field plays a relevant role for the inflation dynamics ( which , in turns , is the initial purpose of the model @xcite ) . obviously , these values can be changed , and our phenomenology discussion has the only purpose of understanding what kind of limits can be imposed on the model . we expect that a viable region will exist at larger @xmath1 ( and smaller @xmath297 ) , as the model becomes closer to a free inflaton model slowly rolling on a flat potential . at the theoretical level , it is interesting to note that the model can result in an observable gravity wave signal in the near future , even if the inflaton spans a range which is some order of magnitudes smaller than the planck scale . this evades the lyth bound @xcite , that states that @xmath344 is possible only for an excursion of the inflaton of @xmath345 during the last @xmath346 e - folds of inflation . we note that the model we have studied evades the bound because of the @xmath8 interaction . quite interestingly , the only other examples that we are aware of where @xmath347 can be achieved even if the inflation evolution is orders of magnitude below the planck scale have been constructed where the inflationary range is smaller than , but still of @xmath345 @xcite . ] are those studied in refs . @xcite , which are characterized by the same pseudo - scalar interaction . as we already mentioned , ref . @xcite is also characterized by a @xmath348 evolution due to the damping from gauge field production . in this case , an interesting parity - violating gravity wave signal can be generated @xcite from the produced gauge quanta , but one needs to evade the simultaneous generation of too many non - gaussian scalar density perturbations @xcite . @xcite achieves this by considering a large number of gauge fields ( this reduces non - gaussianity by the central limit theorem ) or by introducing a curvaton field . @xcite shows that @xmath349 can be obtained , and the non - gaussianity limit can be respected , if the rolling - scalar is not the inflaton . an important difference is that , however , in the mechanism of @xcite the tensor modes are produced at the non - linear level by the vector fields produced by the rolling inflaton . for the model @xcite , the production occurs already at the linear level , due to its mixing with the vector mode induced by the vector vev and the non - abelian structure ( @xmath350 ) . quite likely , also in this model the vector modes @xmath192 can source a significant amount of scalar density perturbations at the non - linear level . this may reduce @xmath336 from the level studied here , although a too large @xmath3 will still likely be ruled out by the significant gravity wave production . it is possible that , for regimes resulting in acceptable @xmath336 , the sourced scalar modes will lead to interesting levels of non - gaussianity and primordial black holes as those obtained in @xcite . it is also possible that , for some choice of parameters , the vector field production will `` self - regulate '' ( with a consequent decrease of @xmath336 ) due to its backreaction on the inflationary dynamics , as visible in the background evolutions studied in @xcite ( a too large production may slow down the inflaton , and this may in turns decrease the vector production ) . all these interesting possibilities remain to be studied . in the first version of this manuscript , we pointed out that ( i ) the inflationary solution of @xcite is stable if and only if @xmath351 , and that ( ii ) the tensor - to - scalar ratio @xmath336 in this model is enhanced with respect to the case of a free inflaton , leading to violation of the lyth s bound @xcite for some choice of parameters . both these claims are confirmed by the present analysis . the stability study is unchanged with respect to the first version . the study of the tensor modes , and the phenomenology considerations that follow from it , have instead been updated , to include the helicity violating terms in the tensor action that were erroneously missing in the first version . such terms result in a further increase of @xmath336 and in a helicity violating gravity wave signal . the relevance of these terms was pointed out in @xcite , that appeared on the archive between the first and the current version of this manuscript . @xcite agrees with our limit @xmath351 for the stability of the inflationary solution , and our revised analysis agrees with the effect of the helicity violating effects found in @xcite . mp acknowledges nicola bartolo and sabino matarrese for useful conversations . this work was supported in part by doe grant de - fg02 - 94er-40823 at the university of minnesota . mp would like to thank the university of padova , and infn , sezione di padova , for their friendly hospitality and for partial support during his sabbatical leave . in section [ sec : scalars ] we studied the system of linear perturbations for the model , disregarding the perturbations of the metric . here we summarize the results for the full system , and we confirm the accuracy of the approximation made in the main text . we start from the full set of @xmath227 scalar perturbations ( after gauge fixing ) , including the @xmath224 modes from the metric perturbations ( @xmath225 and @xmath226 ) that we have ( artificially ) set to zero in the main text . both these modes are non - dynamical , and we integrate them out , together with the other non - dynamical perturbation @xmath231 . we are left with three dynamical modes , and we `` rotate '' them as in eq . ( [ canonical ] ) of the main text . in this way , we obtain the full quadratic action for the perturbations . it is formally of the type @xmath352 \label{formal - s2sf}\end{aligned}\ ] ] namely , the modes ( [ canonical ] ) are not the exact canonical scalar variables . however , as we now show , they provide a very good approximation to the canonical modes in the slow roll regime . more in general , the matrices of the full action ( [ formal - s2sf ] ) are extremely involved . we studied them in slow roll approximation . specifically , using the slow roll approximation ( [ slow ] ) , we can cast all the elements of these matrices in the form @xmath353 where the coefficients @xmath212 and @xmath354 only depend on the parameters of the model , and on slowly evolving background quantities . we computed the leading order expression for these coefficients in the slow roll approximation . for example , we obtain @xmath355 } \nonumber\\ & = & 1 + \frac{{\rm o } \left ( \frac{f^2}{m_p^2 y^{3/2 } \lambda^3 } \right ) h^4}{2 h^2 p^2 + { \rm o } \left ( \frac{1}{y^{3/2 } } \right ) \left [ 1 + { \rm o } \left ( \frac{1}{\lambda } \right ) \right ] h^4 } \nonumber\\\end{aligned}\ ] ] this expression is extremely close to one , since , parametrically , the second term is @xmath356 outside the horizon , and even more suppressed inside the horizon . in fact , the kinetic matrix differs from the identity only up to slow roll suppressed quantities : @xmath357 therefore , up to very small slow roll corrections , the modes ( [ canonical ] ) are also the canonical variable of the full scalar system . performing the same procedure on @xmath358 , we obtain @xmath359}{p^2 + m_g^2 \left [ 1 + { \rm o } \left ( \frac{1}{\lambda } \right ) \right ] } \nonumber\\ & = & \left ( \frac{k_{\rm s}}{a } \right)_{12 } \left [ 1 + { \rm o } \left ( \frac{\lambda^{-1 } } { 1 + y^{3/2 } \ , p^2 / h^2 } \right ) \right ] \nonumber\\ \left ( \frac{k_{\rm s , f}}{a } \right)_{13 } & = & - \frac{g^2 \lambda q^3}{\sqrt{2 } f \sqrt{p^2+m_g^2 } } \ , \frac{\left ( p^2 + m_g^2 \right ) \left [ 1 + { \rm o } \left ( \frac{1}{\lambda } \right ) \right ] } { p^2 + m_g^2 \left [ 1 + { \rm o } \left ( \frac{1}{\lambda } \right ) \right ] } \nonumber\\ & = & \left ( \frac{k_{\rm s}}{a } \right)_{13 } \left [ 1 + { \rm o } \left ( \lambda^{-1 } \right ) \right ] \nonumber\\ \left ( \frac{k_{\rm s , f}}{a } \right)_{23 } & = & \frac{h^4}{\left ( p^2+m_g^2 \right)^{3/2 } } \left [ { \rm o } \left ( \frac{1}{y^{9/4 } \lambda } \right ) \frac{p^2}{h^2 } + { \rm o } \left ( \frac{1}{y^{15/4 } \lambda } \right ) \right ] \nonumber\\ & \sim & { \rm o } \left ( \frac{h}{\sqrt{p^2+h^2 } } \frac{f}{m_p } \frac{1}{\lambda^{3/2 } } \right ) \times \left ( \frac{k_{\rm s , f}}{a } \right)_{12 } \end{aligned}\ ] ] where we recall that @xmath360 . the @xmath97 and @xmath361 entries are in excellent agreement with those given in the main text . we recall that the @xmath361 entry is much smaller than the @xmath97 entry inside the horizon , and it is negligible in the stability study . we note that , for the full system , the @xmath96 element is nonvanishing , while @xmath362 . however , this element is strongly slow roll suppressed with respect to the other two , and completely negligible . proceeding in the same way ( for brevity , we omit here the powers of @xmath72 ) , we obtain @xmath363 and we see that , for all these entries , the expressions of @xmath364 and of @xmath365 agree at all scales ( namely , for any value of @xmath234 ) up to subdominant @xmath366 corrections . for the @xmath235 entry , we obtain @xmath367 also on this entry , the two matrices are in perfect agreement during the full sub - horizon regime . however , while for the other entries the agreement continues also in the super - horizon regime , the @xmath235 entries differ from each other for @xmath368 . by evaluating the coefficients in ( [ om11-compa ] ) , we found that the disagreement starts only at @xmath369 for all the evolutions studied in the main text . to conclude , all the matrix elements , up to one exception , of the system of scalar perturbations studied in the main text ( in which we made the approximation of disregarding the scalar metric perturbations ) agree at all scales ( namely , for any value of @xmath234 ) with the corresponding entries of the full system up to subdominant terms in a slow - roll expansion . the single exception is the @xmath235 entry of @xmath142 , for which the agreement persists during the entire sub - horizon regime , at horizon crossing , and also up to @xmath236 , but not further . this proves that the stability study performed in the main text is valid also when the metric perturbations are included , given that the instability , when present , manifests itself deeply inside the horizon . most likely , this guarantees that also the power spectra shown in the main text are accurate , since the disagreement manifests itself only after the powers have frozen ( see figure [ fig : pkt ] ) . r. kallosh , lect . notes phys . * 738 * , 119 ( 2008 ) [ hep - th/0702059 [ hep - th ] ] ; r. kallosh , n. sivanandam and m. soroush , phys . d * 77 * , 043501 ( 2008 ) [ arxiv:0710.3429 [ hep - th ] ] . e. silverstein and a. westphal , phys . rev . d * 78 * , 106003 ( 2008 ) [ arxiv:0803.3085 [ hep - th ] ] ; l. mcallister , e. silverstein and a. westphal , phys . d * 82 * , 046003 ( 2010 ) [ arxiv:0808.0706 [ hep - th ] ] ; n. kaloper and l. sorbo , phys . rev . lett . * 102 * , 121301 ( 2009 ) [ arxiv:0811.1989 [ hep - th ] ] . c. germani and a. kehagias , phys . * 106 * , 161302 ( 2011 ) [ arxiv:1012.0853 [ hep - ph ] ] ; j. ohashi and s. tsujikawa , jcap * 1210 * , 035 ( 2012 ) [ arxiv:1207.4879 [ gr - qc ] ] ; d. maity , arxiv:1209.6554 [ hep - th ] . m. m. anber and l. sorbo , phys . rev . d * 81 * , 043534 ( 2010 ) [ arxiv:0908.4089 [ hep - th ] ] . s. mohanty and a. nautiyal , phys . d * 78 * , 123515 ( 2008 ) [ arxiv:0807.0317 [ hep - ph ] ] ; l. visinelli , jcap * 1109 * , 013 ( 2011 ) [ arxiv:1107.3523 [ astro-ph.co ] ] . e. dimastrogiovanni , n. bartolo , s. matarrese and a. riotto , adv . astron . * 2010 * , 752670 ( 2010 ) [ arxiv:1001.4049 [ astro-ph.co ] ] ; k. dimopoulos , int . j. mod . d * 21 * , 1250023 ( 2012 ) [ erratum - ibid . d * 21 * , 1292003 ( 2012 ) ] [ arxiv:1107.2779 [ hep - ph ] ] ; j. soda , class . * 29 * , 083001 ( 2012 ) [ arxiv:1201.6434 [ hep - th ] ] ; a. maleknejad , m. m. sheikh - jabbari and j. soda , arxiv:1212.2921 [ hep - th ] . n. bartolo , s. matarrese , m. peloso and a. ricciardone , arxiv:1210.3257 [ astro-ph.co ] . m. -a . watanabe , s. kanno and j. soda , phys . lett . * 102 * , 191302 ( 2009 ) [ arxiv:0902.2833 [ hep - th ] ] ; s. kanno , j. soda and m. -a . watanabe , jcap * 1012 * , 024 ( 2010 ) [ arxiv:1010.5307 [ hep - th ] ] ; k. murata and j. soda , jcap * 1106 * , 037 ( 2011 ) [ arxiv:1103.6164 [ hep - th ] ] ; k. yamamoto , phys . d * 85 * , 123504 ( 2012 ) [ arxiv:1203.1071 [ astro-ph.co ] ] ; k. -i . maeda and k. yamamoto , arxiv:1210.4054 [ astro-ph.co ] ; h. funakoshi and k. yamamoto , arxiv:1212.2615 [ astro-ph.co ] ; k. dimopoulos , m. karciauskas and j. m. wagstaff , phys . rev . d * 81 * , 023522 ( 2010 ) [ arxiv:0907.1838 [ hep - ph ] ] . k. dimopoulos , m. karciauskas and j. m. wagstaff , phys . b * 683 * , 298 ( 2010 ) [ arxiv:0909.0475 [ hep - ph ] ] ; j. m. wagstaff and k. dimopoulos , phys . d * 83 * , 023523 ( 2011 ) [ arxiv:1011.2517 [ hep - ph ] ] ; r. namba , phys . d * 86 * , 083518 ( 2012 ) [ arxiv:1207.5547 [ astro-ph.co ] ] . j. a. r. cembranos , c. hallabrin , a. l. maroto and s. j. n. jareno , phys . d * 86 * , 021301 ( 2012 ) [ arxiv:1203.6221 [ astro-ph.co ] ] ; j. a. r. cembranos , a. l. maroto and s. j. n. jareno , arxiv:1212.3201 [ astro-ph.co ] . b. himmetoglu , c. r. contaldi and m. peloso , phys . lett . * 102 * , 111301 ( 2009 ) [ arxiv:0809.2779 [ astro - ph ] ] . b. himmetoglu , c. r. contaldi and m. peloso , phys . d * 79 * , 063517 ( 2009 ) [ arxiv:0812.1231 [ astro - ph ] ] . b. himmetoglu , c. r. contaldi and m. peloso , phys . d * 80 * , 123530 ( 2009 ) [ arxiv:0909.3524 [ astro-ph.co ] ] . d. baumann _ et al . _ [ cmbpol study team collaboration ] , aip conf . * 1141 * , 10 ( 2009 ) [ arxiv:0811.3919 [ astro - ph ] ] . d. h. lyth , phys . * 78 * , 1861 ( 1997 ) [ hep - ph/9606387 ] . a. e. gumrukcuoglu , b. himmetoglu and m. peloso , phys . d * 81 * , 063528 ( 2010 ) [ arxiv:1001.4088 [ astro-ph.co ] ] . s. saito , k. ichiki and a. taruya , jcap * 0709 * , 002 ( 2007 ) [ arxiv:0705.3701 [ astro - ph ] ] . v. gluscevic and m. kamionkowski , phys . d * 81 * , 123529 ( 2010 ) [ arxiv:1002.1308 [ astro-ph.co ] ] . r. keisler , c. l. reichardt , k. a. aird , b. a. benson , l. e. bleem , j. e. carlstrom , c. l. chang and h. m. cho _ et al . _ , astrophys . j. * 743 * , 28 ( 2011 ) [ arxiv:1105.3182 [ astro-ph.co ] ] . i. ben - dayan and r. brustein , jcap * 1009 * , 007 ( 2010 ) [ arxiv:0907.2384 [ astro-ph.co ] ] . s. hotchkiss , a. mazumdar and s. nadathur , jcap * 1202 * , 008 ( 2012 ) [ arxiv:1110.5389 [ astro-ph.co ] ] . p. d. meerburg and e. pajer , arxiv:1203.6076 [ astro-ph.co ] . a. linde , s. mooij and e. pajer , arxiv:1212.1693 [ hep - th ] . n. barnaby , e. pajer and m. peloso , phys . d * 85 * , 023525 ( 2012 ) [ arxiv:1110.3327 [ astro-ph.co ] ] . p. adshead , e. martinec and m. wyman , arxiv:1301.2598 [ hep - th ] .

the earliest extant correspondence between pierre rmond de montmort and a member of the bernoulli family is a letter from montmort to johann bernoulli dated february 27 , 1703 , concerning a paper on calculus that the latter had written for the acadmie royale des sciences in paris @xcite . they corresponded sporadically over the next few years . on april 29 , 1709 , montmort sent bernoulli a copy of his book on probability , _ essay danalyse sur les jeux de hazard _ , that he recently had published @xcite . the book is the first in a series of books in probability published by several others over the years 1708 to 1718 in what hald [ ( @xcite ) , page 191 ] calls the `` great leap forward '' in probability . bernoulli replied with a gift of a copy of his nephew s doctoral dissertation @xcite , the second book in hald s `` great leap forward '' ; nicolaus bernoulli s book dealt with applications of probability . once johann bernoulli received his copy of _ essay danalyse _ , he sent , on march 17 , 1710 , a detailed set of comments on the book . in the letter bernoulli included another set of comments on _ essay danalyse _ , this one by his nephew nicolaus ( @xcite , pages 283303 ) . thus began a series of correspondence between montmort and nicolaus bernoulli on problems in probability . montmort included much of this correspondence in part v of the second edition of _ essay danalyse _ @xcite . the correspondence between montmort and nicolaus bernoulli after 1713 , left unpublished and largely ignored by historians , contains scientific news and further discussion of problems in probability . the major topic is a continuing discussion of issues related to the card game le her . next , in terms of ink spilt on probability , are discussions of the `` problem of the pool , '' or waldegrave s problem , generalized to more than three players , and of the game les trennes ( which may be translated as `` the gifts '' ) . the correspondence also contains discussions of various problems in algebra , geometry , differential equations and infinite series . as an aristocrat , montmort s network included both political and scientific connections . his letters to bernoulli contain some references to his political activities that sometimes kept him from replying promptly . his brother , nicolas rmond , was chef de conseil for phillipe duc dorlans , who became regent of france after his uncle louis xiv died in 1715 ( @xcite , page 599 ) . among the mathematicians of the era , montmort corresponded with isaac newton , gottfried leibniz , brook taylor and abraham de moivre , in addition to the bernoullis as well as several others . as a talented amateur mathematician , his work was well regarded by the mathematicians of his day . he was generous to his scientific friends . he received as guests to the chteau de montmort nicolaus bernoulli , brook taylor and one of the sons of johann bernoulli . he also sent gifts of cases of wine and champagne to both newton and taylor . le her is a game of strategy and chance played with a standard deck of fifty - two playing cards . the simplest situation is when two players play the game , and the solution is not simply determined even in that situation . montmort calls the two players pierre and paul . pierre deals a card from the deck to paul and then one to himself . paul has the option of switching his card for pierre s card . pierre can only refuse the switch if he holds a king ( the highest valued card ) . after paul makes his decision to hold or switch , pierre now has the option to hold whatever card he now has or to switch it with a card drawn from the deck . however , if he draws a king , he must retain his original card . the player with the highest card wins the pot , with ties going to the dealer pierre . the game can be expanded to more than two players . montmort [ ( @xcite ) , pages 186187 ] originally described the problem for four players and posed the question : what are the chances of each player relative to the order in which they make their play ? because of the winning conditions , it is obvious that one would want to switch low cards and keep high ones . the key is to find what to do with the middle cards , such as seven and eight , when two players are playing the game . in other cases , cards are clearly too low to keep or too high to switch , being much below or above the average in a random draw . naturally , the threshold would be lower with more than two players . in part essay danalyse _ , only the game with two players is considered . initially , montmort and nicolaus bernoulli wrote back and forth about the problem and came to the same solution . however , two of montmort s friends contended that this solution was incorrect . these were an english gentleman named waldegrave and an abbot whose abbey was only a league and a half ( about 5.8 kilometers ) from chteau de montmort ( @xcite , page 338 ) . montmort identified waldegrave only as the brother of the lord waldegrave who married the natural daughter of king james ii of england . lord waldegrave is henry waldegrave , 1st baron waldegrave , and his wife is henrietta fitzjames , daughter of james ii , and his mistress arabella churchill . the abbot is the abb dorbais ; montmort also refers to him as the abb de monsoury . the reason for the two appellations for the abbot is that his full name is pierre cuvier de montsoury , abb dorbais . he has been described as `` un prodige de bon coeur , durbanit et de science '' ( @xcite ) . for the spelling choice between montsoury and monsoury , it should be noted that montmort often spelled his name `` monmort . '' two other problems were discussed extensively in the correspondence . the first is the problem of the pool , a problem that waldegrave suggested to montmort and solved himself ( @xcite , page 318 ) . in _ essay danalyse _ the problem is solved for three players . it is often called waldegrave s problem @xcite . the `` pool '' is a way of getting three or more players to gamble against one another , when the game put into play is for two players only . in the situation for three players ( montmort uses the names pierre , paul and jacques ) , all three begin by putting an ante into the pot . then pierre and paul play a game against each other . the winner plays against jacques and the loser puts money into the pot . the game continues until one player has beaten the other two in a row . that player takes the pot . the game can be expanded to more than three players , but that situation was not fully treated in _ essay danalyse_. the second is the problem of solving the game les trennes ( or `` estreine , '' an alternative old french spelling ) . as described by montmort [ ( @xcite ) , pages 406407 ] , this is a strategic game between a father and his son . the father holds an odd or even number of tokens in his hand , which his son can not see . when the son guesses even , he receives a gift of four cus ( silver coins ) if he is correct and nothing if wrong . when the son guesses odd , he receives one cu if he is correct and nothing if wrong . the discussion of this game in the correspondence is only brought in to enlighten le her whose strategic nature is in some important respects essentially similar . montmort concludes the last letter ( to bernoulli ) that appears in _ essay danalyse _ with a remark that waldegrave had volunteered to take care of getting the book printed in paris . montmort s letter was dated november 15 , 1713 , and was written from paris . what is also of concern to us are the letters after this date and how these letters relate to earlier discussions . the unpublished correspondence begins with a letter from montmort to bernoulli dated january 25 , 1714 , in which he says that he has sent bernoulli two copies of the second edition of _ essay danalyse_. montmort was still in paris , where he claimed to have been for three months . he was staying at a hotel in rue des bernardins , which in modern paris is only a walk of 350 meters to the printer , jacques quillau in rue galande . presumably , waldegrave s help consisted mainly in dealing with the printer and the proof sheets as they came off the press , thus relieving montmort of some tedious work . to understand the discussion of le her after 1713 , it is necessary to describe the treatment of the game as it appears in the second edition of _ essay danalyse_. hald [ ( @xcite ) , pages 314322 ] provides a detailed description of the mathematical calculations involved in assessing the game . yet he devotes little space to elucidating the discussions among bernoulli , montmort and waldegrave , as well as the abb dorbais , concerning the issues surrounding these mathematical calculations . it is the substance of these discussions that are of interest to us . hald s only comment on the discussion over le her concerns a comment made by waldegrave and abb dorbais to the effect that bernoulli s reasoning in obtaining his mathematical solution is faulty . after pointing out their observation that bernoulli s solution fails to account for a player s probability of playing in a certain way , hald [ ( @xcite ) , page 315 ] claims : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ it is no wonder that bernoulli does not the implications of this remark , since the writers themselves have not grasped the full implication of their point of view . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ it is indeed true that there was some confusion on bernoulli s side which he deftly tried to hide . henny [ ( @xcite ) , page 502 ] comments that he is amazed to find expressed in the letters many concepts and ideas that appear in modern game theory . at the same time , he is surprised to find waldegrave defending his position so strongly against bernoulli who was the superior mathematician . henny states further that waldegrave did not have the necessary mathematical skills to provide a mathematical proof of his results . as we will show in a review of the treatment of le her in _ essay danalyse _ and the subsequent unpublished correspondence , both hald s and henny s insights fall short of the mark . , though obviously correct for the limited aspect in which he viewed the problem , is unsatisfactory to common sense , which suggests that in all circumstances there must be , according to the degree of our knowledge , at least one rule of conduct which shall be not less satisfactory than any other ; and this his discussion fails to provide . '' our discussion below will show that this assessment is misinformed . ] one reason they fall short is that they do not consider the full range of the various events that were under discussion and their associated probabilities . two events are natural for a probabilist to consider . the first is the distribution of the cards to pierre and paul . the second is the randomizing device used to come up with the mixed strategy prescribing when the players should hold and when they should switch . the randomizing device considered in _ essay danalyse _ is a bag containing black and white counters or tokens ( the old french word used is `` jetton '' ) . the third event that montmort , waldegrave and bernoulli consider ( but not hald or henny ) is difficult , or perhaps impossible , to quantify . this is the possibility that paul , say , is a poor player and does not follow a strategy that is mathematically optimal , or the possibility that paul , say , is a very good player who tries to trick pierre into making a poor choice . this kind of event unfolds regularly in modern poker games . another reason for which hald and henny see some confusion in the discussions among bernoulli , montmort and waldegrave is that what we are seeing in the correspondence is the complete unfolding of a problem from its initial statement , and discussions around it , to a complete solution . this is different from a `` textbook '' statement of a problem followed by a solution . in the latter case , the problem and solution are both well laid out . in the former case , there is some grappling with the problem until it becomes clear how to proceed . we begin with the correspondence in _ essay _ _ danalyse _ where le her is first mentioned . in johann bernoulli s 1710 letter to montmort in _ essay danalyse _ , he suggests more efficient methods to reach montmort s conclusions for a variety of problems and in some cases generalizes montmort s results . there is only one reference to the problem of le her , which is the second of four problems proposed in montmort [ ( @xcite ) , pages 185187 ] : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the second and the third [ problem ] seem to me amenable , but not without much difficulty and work , that i prefer to defer to you and learn the solution , than to work long at the expense of my ordinary occupations that leave me scarcely any time to apply myself to other things . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in his reply to this letter , which is dated november 15 , 1710 ( @xcite , pages 303307 ) , montmort makes no reference to this passage . nicolaus bernoulli s first letter to montmort , dated february 26 , 1711 , makes no reference to the game le her . it is a note in montmort s reply to nicolaus bernoulli , dated april 10 , 1711 ( @xcite , pages 315323 ) that initiates the discussion of montmort s second problem : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ i started some time ago to work on the solution of problems that i propose at the end of my book ; i find that in le her , when there are only two players left , pierre and paul , paul s advantage is greater than 1 in 85 , and less than 1 in 84 . this problem has difficulties of a singular nature . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in a postscript to this letter , montmort makes an additional remark : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ as there are few copies of my book left , there will soon be a new edition . when i have decided , i will ask you permission , and your uncle , to insert your beautiful letters which will make the principal embellishment . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ it is this announcement that may have motivated nicolaus bernoulli to continue his correspondence with montmort and to send him much interesting material . publishing mathematical material outside a scientific society or without a patron to cover the costs was an expensive proposition , one that montmort could afford . because of the specialized type that was used and the accompanying necessary skill of the typesetter , the cost of a mathematical publication was well above the norm for less technical books . bernoulli could get his results in print at no cost to himself . bernoulli responded with a long letter , dated november 10 , 1711 ( @xcite , pages 323337 ) . in this letter , he announces , among many other things , that he has also solved the two - person case for le her ( @xcite , page 334 ) : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ i also solved the problem on le her in the simplest case ; here is what i found . if we suppose that each player observes the conduct that is most advantageous to him , paul must only hold to a card that is higher than a seven and pierre to one that is higher than an eight , and we find under this supposition that the lot of pierre will be to that of paul as 2697 is to 2828 . supposing that paul also holds to a seven , then pierre must hold to an eight , and their lots will still be as 2697 to 2828 . nevertheless it is more advantageous for him not to hold to a seven than to hold to it , which is a puzzle that i leave you to develop . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ this passage is carefully worded , yet it will be misinterpreted by montmort and waldegrave . as we will see , a key aspect that is neglected by montmort and waldegrave is the antecedent of bernoulli s conditional statement starting with `` if we suppose that each player observes the conduct '' montmort s reply , dated march 1 , 1712 ( @xcite , pages 337347 ) , highly praises bernoulli s prior letter . he complains that , being in paris , he has had no time and peace to think on his own and , as a consequence , the main object of his letter is to report progress made by his two friends , the abb dorbais and waldegrave , on a problem proposed by bernoulli , and on the problem of le her . on the latter , montmort reports that `` they dare however not submit to your decisions '' ( @xcite , page 338 ) . however , as he says in a passage that is key to understanding the forthcoming controversy , the abb dorbais also previously disagreed with montmort : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ when i worked on le her a few years ago , i told m. labb de monsoury what i had found , but neither my calculations nor my arguments could convince him . he always maintained that it was impossible to determine the lot of pierre and paul , because we could not determine which card pierre must hold to , and vice versa , which results in a circle , and makes in his opinion the solution impossible . he added a quantity of subtle reasonings which made me doubt a little that i had caught the truth . that is where i was when i proposed that you examine this problem ; my goal was to make sure from you of the goodness of my solution , without having the trouble of recalling my ideas on this which were completely erased . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ montmort then claims that bernoulli s solution confirms what he had found , a decision that prompts a reply from waldegrave objecting to bernoulli s solution , quoted at length in montmort ( @xcite ) , pages 339340 . according to waldegrave and the abb dorbais , it is not true that paul must hold only to an eight and pierre to a nine . rather , that paul should be indifferent to hold to a seven or to switch , and that pierre should be indifferent to hold to an eight or to switch . waldegrave wrote the following to montmort ( @xcite , page 339 ) : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ we argue that it is indifferent to paul to switch or hold with a seven , and to pierre to switch or hold with an eight . to prove this , i must first explain their lot in all cases . that of paul having a seven , is @xmath0 when he switches , and when he holds on to it his lot is @xmath1 if pierre holds on to an eight , and @xmath2 if pierre switches with an eight . the lot of pierre having an eight is @xmath3 if he holds on to it , and @xmath4 if he switches in the case that paul only holds on to a seven ; and @xmath5 by holding on to it , and @xmath6 by switching in the case that paul holds on to a seven , so here they are . the lots of paul @xmath7 , those of pierre @xmath8 or @xmath9 . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ based on the numbers he obtains , waldegrave observes that `` 720 being more below 780 than 816 is above , it appears that paul must have a reason to switch with 7 '' ( @xcite , page 339 ) . the differences , @xmath10 and @xmath11 , are in the ratio @xmath12 , or @xmath13 , a ratio which later enters the discussion . in the rest of his argument , waldegrave talks of a weight instead of a reason . he first lets the weight that leads paul to switch be @xmath14 , and the weight that leads pierre to switch be @xmath15 . and he argues that the same weights lead paul and pierre to both strategies . @xmath14 leads paul to switch with 7 and , as a consequence , it also leads pierre to switch his 8 ; but what leads pierre to switch his 8 must lead paul to hold with 7 . so , @xmath14 leads paul to both switch with a 7 and hold on to it . the same goes for pierre . therefore , `` it is false that paul must only hold on to an 8 , and pierre to a 9 , '' which was bernoulli s claimed solution . the word `` probability '' comes up only once in this discussion , in the conclusion of the excerpt from waldegrave s letter to montmort . waldegrave writes ( @xcite , page 340 ) : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ apparently mr . bernoulli was simply looking at the fractions that express the different lots of pierre and paul , without paying attention to the probability of what the other will do . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ montmort leaves the discussion there without further comment . 240pt@lcc@ + & * switch the 8 * & * hold the 8 * + & * ( and under)*&*(and over ) * + switch the 7 & & + ( and under ) & & + hold the 7 & & + ( and over ) & & + upon receiving montmort s letter , bernoulli agrees with these figures , saying that `` the lots they found for pierre and paul are very right '' ( @xcite , page 348 ) . and yet , when bernoulli proposes his solution , and when montmort eventually publishes a table of probabilities as an appendix to _ essay danalyse _ ( @xcite , page 413 ) , the numbers are different . the bernoulli montmort probabilities are shown in table [ table1 ] , which appears in hald ( @xcite ) , page 318 . none of the parties in this debate actually explain their calculations . waldegrave s probabilities are justified in todhunter ( @xcite ) , pages 107110 ; the bernoulli montmort probabilities are in hald ( @xcite ) , pages 315318 . the difference in the probabilities is that waldegrave s probabilities are conditional on paul having a seven in his hand and the bernoulli montmort probabilities are the marginal probabilities for all cards that paul may hold . in a letter dated june 2 , 1712 , bernoulli replies to waldegrave s argument by accusing him of committing a fallacy . he argues that if we suppose that @xmath14 leads paul to switch with a seven , and so leads pierre to switch with an eight ( if pierre knows paul switches with seven ) , then it also leads paul to hold on to a seven . therefore , @xmath14 both leads paul to switch with a seven and to hold on to a seven . his conclusion is that ( @xcite , page 348 ) : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ we are supposing two contradictory things at the same time ; that is , that paul knows and ignores at the same time what pierre will do , and pierre what paul will do . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ bernoulli explains that if we do not commit this fallacy regarding what paul and pierre know about the other s intent , we are led to reasoning in a circle , which shows that waldegrave s argument can not show anything . this argument is peculiar , and seems to suggest that bernoulli does not understand waldegrave s point . it might , however , be simply a misinterpretation of waldegrave s argument , for it is expressed in terms of weight rather than in terms of probability . the word `` weight '' or `` poids '' in french offers more opportunity for misinterpretation . moreover , bernoulli admits having written his letter hastily , as he was preparing for a long trip through the netherlands and england . as a result of this travel , some subsequent letters are delayed , and the arguments they contain do not follow the chronological order of when the letters were written . a letter to bernoulli , dated september 5 , 1712 ( @xcite , pages 361370 ) , announces that waldegrave and the abb dorbais have seen bernoulli s reply in which he accuses them of committing a fallacy . montmort includes a note from the abb dorbais in which he claims that waldegrave has written a beautiful and precise reply to bernoulli s objection ; the rebuttal , however , is not included . in this note , the abb dorbais also enjoins montmort to take a side in this dispute between them . this suggests that , even if montmort thanked bernoulli for his solution , which he claimed agreed with his own , montmort has not yet made up his mind as to whether bernoulli really solved the problem . the next letter concerning le her is from bernoulli to montmort , dated december 30 , 1712 ( @xcite , pages 375394 ) . adding important pieces to the puzzle , it contains a three - page discussion of le her ( bernoulli mentions having just received the june 2 letter , since it was sent from switzerland to holland , then to england , and finally back to switzerland ) . bernoulli insists that , despite waldegrave s arguments , paul does not do as well by abiding to the maxim of holding to a seven , than that of switching with a seven . bernoulli then says ( @xcite , page 376 ) : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ if it were impossible to decide this problem , paul having a seven would not know what to do ; and to rid himself [ from deciding ] , he would subject himself to chance , for example , he would put in a bag an equal number of white tokens and black tokens , with the intent of holding to a seven if he draws a white one , & to switch with a seven if he draws a black one ; because if he put an unequal number he would be lead more to one party than to the other , which is against the assumption . pierre with an eight would do the same thing to see whether he must switch or not . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ this comment introduces with clarity the idea of chance by `` the way of tokens '' ( as they will say later ) . what bernoulli says here seems to confirm that , at first , when he accused waldegrave of committing a fallacy , he did not interpret waldegrave s weights as probabilities . nonetheless , he suggests that the only probability allocation compatible with the supposed state of ignorance of the players is that each player chooses a strategy with probability @xmath16 . under these choices , he computes the lot of paul ( which is then @xmath17 ) and concludes that it would be a bad thing for paul to randomize in this way , since he could guarantee himself a lot of @xmath18 . therefore , bernoulli concludes paul must always switch with a seven . as bernoulli says ( @xcite , page 376 ) , `` it is better to make the choice where we risk less . '' he then explains the reasoning that he had left out of his hastily written letter from june 2 . in contemporary terms , he calculated the unconditional probability of winning under each pure strategy profile ( without assuming that any card has been dealt yet ) . he displays a refined version of the reasoning that led to accusing waldegrave of a fallacy , yet it does not do full justice to waldegrave s idea . eight months later , on august 20 , 1713 , montmort [ ( @xcite ) , pages 395400 ] finally replies to bernoulli , complaining that he has , despite his philosophical inclinations , been involved in political activities , and so he did not have the leisure for intellectual work . thus , his letter only contains scientific news . there is only one brief mention of le her ; he tells bernoulli that , despite his last effort to provide a thorough and precise argument , waldegrave and the abb dorbais are still unconvinced by his claimed solution . shortly after , in a letter dated september 9 , 1713 , bernoulli also asks montmort to explain his own views on the dispute . montmort obliges him in his letter dated november 15 , 1713 . this is the last letter published in the second edition of _ essay danalyse _ ( @xcite , pages 403413 ) . the letter also contains an excerpt of a letter from waldegrave and a table of the lots of paul and pierre for the four crucial combinations of strategies , which are summarized in table [ table1 ] . here , then , is montmort s understanding of the situation . to begin with , he agrees with bernoulli that it is not indifferent to paul to switch or hold with a seven , and to pierre to switch or hold with an eight , because of bernoulli s calculations of the unequal chances for each strategy . ( this shows that bernoulli and montmort use `` indifferent '' in the sense of having the same probability of winning . for waldegrave and dorbais , however , `` indifferent '' seems to mean , perhaps more awkwardly , that no strategy dominates the other in probability . ) this being said , montmort nonetheless disagrees with bernoulli that this establishes the strategy as a maxim , that is , as a rule of conduct that must be obeyed invariably to obtain the best results . rather , he thinks that it is impossible to establish such a maxim ( @xcite , page 403 ) : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ [ t]he solution of the problem is impossible , that is , we can not prescribe to paul the conduct that he must adopt when he has a seven , and to pierre the conduct he must adopt when he has an eight . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ he grants that , if one is to choose a fixed and determined maxim , then switching on seven , for paul , will be better than any other , yet paul can hope to make his lot better . why , then , would a solution be impossible ? would the solution not be the optimum that one can reach in paul s hope of making his lot better ? montmort claims that , whereas he used to think that the use of black and white tokens to randomize strategies could avoid the `` circle , '' he does not think that anymore . he gives a general formula to find the probability of winning with a certain probability allocation for what we call a mixed strategy : @xmath19 where @xmath20 is paul s probability of switching with seven , @xmath21 is paul s probability of holding the seven , @xmath22 is pierre s probability of switching with an eight , and @xmath23 is pierre s probability of holding on to an eight . but how should the probabilities be chosen ? montmort claims that any argument will only inform us of what paul must do conditionally to what pierre does and vice versa , which leads us into a circle once again . he concludes that bernoulli s arguments to show that a circle does not occur are wrong , and instead formulates this thesis ( @xcite , page 404 ) : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ [ w]e must suppose that both players are equally subtle , and that they will choose their conduct only based on their knowledge of the conduct of the other player . however , since there is here no fixed point , the maxim of a player depends on the yet unknown maxim of the other , so that if we establish one , we draw from this supposition a contradiction that shows that we must not have established it . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ montmort also disagrees with bernoulli that , under pain of contradiction , if we are to use white and black tokens to randomize , we must use an equal number of tokens . instead , he thinks that the probability of winning calculated for the fixed and determined maxims shows that paul must switch more often with a seven than hold on to it . yet , he maintains ( @xcite , page 405 ) : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ but how much more often must he switch rather than hold , and in particular what he must do ( here and now ) is the principal question : the calculation does not teach us anything about that , and i take this decision to be impossible . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ thus , montmort believes , it seems , that there is no optimal probability allocation . but he has another reason for believing that the solution of the game is impossible . he has in mind the game les trennes ( @xcite , pages 406407 ) . montmort also believes that it is impossible to prescribe any strategy of play in les trennes because the players might always try , and indeed good players will try , to deceive other players into thinking that they will play something they are not playing , thus trying to outsmart each other ( `` jouer au plus fin '' is the phrase used in french ) . as he was finishing his letter , montmort received one from waldegrave and quoted extensively from it to bernoulli . _ essay danalyse _ essentially concludes with waldegrave s letter . waldegrave refers to a formula , which is not included by montmort ; it presumably is the formula displayed above . he explains that , if @xmath24 and @xmath25 ( so that the probability of paul switching with a seven is @xmath26 ) , then the lot of pierre is going to be @xmath27 no matter what @xmath22 and @xmath23 are . this shows that @xmath27 is paul s minimum lot . he can only adopt another conduct in the hope of making his lot better . this shows , he claims , that both bernoulli and ( formerly ) himself were wrong to claim that the lots of paul was to that of pierre as 2828 is to 2697 ; if both players play in the most advantageous way , paul s lot is @xmath28 . waldegrave is convinced that this is something that both bernoulli and montmort will agree to , now that it is agreed that one can use a randomized strategy . he also explains that , if pierre uses @xmath29 and @xmath30 , then @xmath27 will also be paul s maximum lot . waldegrave also asserts that it is impossible to establish a maxim ; he grants , however , that it is impossible for him to show this with the same level of evidence . this is often taken incorrectly as evidence of a lack of waldegrave s mathematical abilities . waldegrave is instead referring to the situation in which players may try to outsmart each other . waldegrave agrees that if paul does not use @xmath24 and @xmath25 , then it is possible for paul to do better than @xmath31 provided that pierre does not play in the best way . on the other hand , it would be worse if pierre plays correctly . furthermore , waldegrave remarks ( @xcite , page 411 ) : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ what means are there to discover the ratio of the probability that pierre will play correctly to the probability that he will not ? this appears to me to be absolutely impossible , and thus leads us into a circle . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ as with montmort , his main concern is that it is always possible for the players to try to outsmart each other ( `` jouer au plus fin '' ) . examining the detailed arguments provided by montmort , bernoulli and waldegrave reveals a picture that contrasts with the judgment that they were essentially confused on the fundamental concepts and methods required to solve a strategic game such as le her . in fact , we maintain that they understood most of the aspects of the problem with clarity . there are , however , a number of important outstanding issues left unresolved in the correspondence on le her as it appears in _ essay danalyse_. let us review them briefly . it is true that the letters reveal a certain type of misunderstanding ; however , it is not conceptual confusion , but rather mutual misinterpretation due to using terms differently . an instance of this is whether it is indifferent to paul to switch or hold to a seven . on the one hand , both montmort and bernoulli claim that it is not indifferent to paul because the chances of winning are not identical . on the other hand , waldegrave claims that it is indifferent , and the reason for that seems to be that neither pure strategy dominates the other in probability . another instance of this is the disagreement they appear to have on the existence of a circularity in the analysis of the game . montmort and waldegrave assert that there is a vicious circle that prevents one from establishing a maxim ; the circle they discuss , however , is really a regression _ ad infinitum _ , that is , to establish a maxim , we always need to go one step further in the `` @xmath14 must know what @xmath15 does '' loop ( bernoulli agrees with this point ) . however , bernoulli claims that there is a circle in waldegrave s argument , in the sense that either his argument is contradictory or a _ petitio principii _ ( but bernoulli is not considering randomizing strategies at this point ) . again , they are only contradicting each other in the wording , not in the idea . finally , a third instance is that montmort and waldegrave claim that the solution of the game is impossible , whereas bernoulli does not . here again , they disagree on what it means to `` solve '' the game le her . bernoulli claims that the solution is the strategy that guarantees the best minimal gain what we would call a minimax solution and that as such there is a solution . however , despite understanding this `` solution concept , '' montmort and waldegrave refuse to affirm that it `` solves '' the game , since there are situations in which it might not be the best rule to follow , namely , if a player is weak and can be taken advantage of . clearly , the concept of solution they have in mind differs from the minimax concept of solution . this latter concept , in addition to the probability of gain with a pure strategy and the probability allocation required to form mixed strategies , requires that we know the probability that a player will play an inferior strategy . but , they assert , this can not be analyzed by calculations , so the game can not be solved . this being said , there are a number of things that are said that suggest a certain level of confusion at a conceptual level . the two most important are these . first , bernoulli appears to have some difficulty with the relation between the knowledge of the players and the probabilities involved in mixing strategies . his circularity objection to waldegrave is awkward and somewhat mystifying . moreover , his argument that , if we allow randomized strategies with black and white tokens , it must be because neither player knows what the other player will do , and that as a result the only acceptable probability allocation of @xmath16 is problematic . this kind of mistaken argument has been repeated over the centuries by some of the greatest minds in probability , statistics and game theory . second , montmort understands very well the idea of randomizing strategies , but he nonetheless claims that there is no optimal probability allocation that can be calculated . this claim , however , was made before consulting waldegrave s letter in which he reveals the optimal probability . referring to a letter from bernoulli to montmort dated february 20 , 1714 , @xcite , in his treatment of le her , mentions only that bernoulli accepted waldegrave s solution to the problem . however , bernoulli had other things to say about le her in that same letter . henny also refers to a letter of january 9 , 1715 , from waldegrave to bernoulli in which waldegrave seemingly admits to bernoulli that he does not have the mathematical skills to actually prove his results . what henny leaves out is that the letter was written in reply to a detailed criticism of the solutions to le her that bernoulli had sent earlier to montmort . after some personal news and apologies for not writing sooner , in his letter of february 20 , 1714 , to , bernoulli initially thanks montmort for correcting , editing and making clearer his letters that montmort had printed in _ essay danalyse_. then follows the discussion of le her that @xcite only briefly mentions . initially , bernoulli suggests that the controversy is essentially over : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ concerning le her , i seem to have foreseen that in the end we would all be right . however , i congratulate mr . de waldegrave who has the final decision on this question , and i willfully grant him the honor of closing this affair _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ despite this , bernoulli still claims that he disagrees on a few minor points , and these point directly to the outstanding issues we mentioned above . the main concern is the relation between `` establishing a maxim '' and solving the problem of le her posed by montmort in his book . bernoulli states : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ one can establish a maxim and propose a rule to conduct one s game , without following it all the time . we sometimes play badly on purpose , to deceive the opponent , and that is what can not be decided in such questions , when one should make a mistake on purpose . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ this point was raised before by montmort and waldegrave , but they do not consider that such a play would be necessarily a mistake . whether or not this kind of play is a mistake , we saw that from the same consideration , montmort and waldegrave conclude that solving the problem is impossible . however , bernoulli now phrases things more carefully : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ mr . de waldegrave wrongs me on p. 410 by claiming that i once said that the lot of paul is to that of pierre as @xmath32 . if you carefully read my letter from oct 10 , 1711 , you will find that i did not say it absolute and without restriction . i beg you to consider those words : once we have determined or rather supposed what are the cards to which the players will hold , etc . and the following words . you will see that i there supposed that the players want to hold to a fixed and determined card , and indeed i had not thought about the way of tokens , which , as mr . de waldegrave said , is not among the ordinary rules of the game . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ bernoulli essentially says that he was misinterpreted and that he only computed the best odds of winning with a pure strategy , not that he established what a player should do in an actual game . moreover , if we grant his supposition , then he has found the most advantageous maxim . after this correction , bernoulli thinks the discussion is over , saying , `` we are thus all agreeing , and we have made peace ; _ canamus receptui _ [ sing retreat ] . '' in his response to bernoulli , dated march 24 , 1714 , montmort concurs by writing , `` i am quite pleased that we are all together by and large agreeing . '' in this letter montmort claims that he disagreed with bernoulli on some aspect of the corrected interpretation of his position , but he leaves it to a later letter to explain . however , in his next letter to bernoulli , november 21 , 1714 , montmort does little to clarify . he says , `` if it is ever permitted to say to two persons maintaining contradictory claims that they are both right , it is assuredly at this occasion in our dispute . '' montmort emphasizes that what he seeks is the correct advice that should be given to the players , but the discussion does not go much further . on august 15 , 1714 , montmort sent a letter to bernoulli containing a two - page `` supplement '' that reignites the debate . he makes six points . first , he claims that telling paul always to switch with a seven is bad advice , since his minimum lot is then 2828 . second , that it would be better advice to tell him to do whatever he pleases with a seven , so that he can look at both options indifferently . third , we can not say that this would be the best advice either , for knowing that , pierre would switch with an eight , in which case paul should certainly have held on to a seven . this leads to a vicious circle . fourth , if we admit the way of tokens , the best advice that he knows is to tell paul to have the ratio @xmath33 for switching with a seven . but even then , he does not think that we can demonstrate that it is the best advice . fifth , he claims that it is impossible at this game to determine the lot of paul , because one can not determine what manner of playing is the most advantageous to each player , even when we admit a randomized strategy . this point makes explicit for the first time montmort s ( and presumably waldegrave s ) idea that you can only claim that you have found the lot of a player ( which is what montmort s problem in _ essay danalyse _ demanded ) if we can determine what is the best way to play . moreover , determining the best way to play demands knowing more than the optimal token ratio for the randomized strategy . he adds that , of course , some methods of playing are better than others , as informed by the chances that have previously been calculated . he concludes , sixth , that he would not know what advice to give paul if he had to . this letter sharpens the debate , in that it makes explicit the connection between `` solving '' a game and giving advice for play in actual situations . in a long letter to montmort dated august 28 , 1714 , along with a `` supplement '' dated november 1 , 1714 , bernoulli replies to montmort point by point . he asks montmort a question that is meant to dismiss his argument : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ if , admitting the way of tokens , the option of 3 to 5 for paul to switch with a seven is the best you know , why do you want to give paul other advice in article 6 ? it suffices for paul to follow the best maxim that he could know . it is not enough to claim that there is still a circle despite my reasons , one must fight my reasons . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ and he continues : `` it is not impossible at this game to determine the lot of paul . '' to counter montmort s previous argument , he once again insists that either paul knows what pierre will do , in which case his maxim is clear , or he does not , in which case paul should use the probability @xmath16 in the randomized strategy to determine what to do . as he admits , this is the exact same position he had at the beginning of the discussion , supported by the exact same argument . thus , it seems that bernoulli has missed the point montmort made explicit in his august 15 letter . it is at this point that waldegrave reenters the debate at montmort s request . in a letter dated january 9 , 1715 , waldegrave reiterates the six points that montmort had laid out for bernoulli in his letter of august 15 . for each of the six points , waldegrave s arguments are longer and more detailed than what montmort had previously given . it is not until march 22 , 1715 , that montmort replies to bernoulli on this dispute . it is part of a very long letter that also contains the main topic for their further correspondence , infinite series . in this letter , montmort writes once again about his views . they are the same as what we have seen already . however , montmort stresses that a lot of what remains under discussion is based on inconsistent terminology and . in essence , he believes that the outstanding disagreements are only apparent contradictions . nonetheless , he introduces one more element to clearly his view . he distinguishes between the advice that he would put in print , or give to paul publicly , and the advice he would give so that only paul hears it . montmort claims that , for the former , he would choose the mixed strategy with @xmath24 and @xmath25 , since it is the one that demonstrably brings about the lesser prejudice . however , he explains that , in practice , if paul is playing against an ordinary player and not a mathematician , he would quietly give different advice that could allow paul to take advantage of his opponent s weakness . as he explains , the objective of this sort of analysis is not only to provide a maxim to otherwise ignorant players , but also to warn them about the potential advantages of using finesse . however , this latter part is not possible to establish , and it is in this sense that there is no possible solution to this problem . the next letter , sent by bernoulli to montmort on may 4 , 1715 , disregards montmort s nuance . to begin , bernoulli `` is forced to admit that he does not precisely know on what point [ they ] contradict each other . '' nonetheless , bernoulli explains that , in his view , the distinction between public and private advice , the possibility of using finesse , or something similar , does not alter the fact that @xmath24 and @xmath25 is the best solution , and that it determines the lot of paul ( so that not only is the game solvable , but it is indeed solved ) . despite bernoulli s explanation , montmort s next letter , dated june 8 , 1715 , once again reiterates that `` you have badly solved the proposed question , or you have not solved it at all . '' he makes explicit what he takes the proposed question to be : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the question is and has always been to know whether we can establish the lots and as a result the advantage of playing first under the supposition not that pierre and paul follow this or that maxim ( this would have no utility , no difficulty ) , but that both of them having the same skills , each follow the conduct that is the most advantageous . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ montmort then says that this dispute is beginning to bore him . he considers that furthering it will not make them learn anything new and that in the end the dispute must be about some other thing . our presentation of the correspondence makes it clear that they are using different concepts of solution ; bernoulli s in essence is the concept of the minimax solution , whereas montmort s further depends on the probability of imperfect play ( i.e. , on the skill level of the players ) . around this time , montmort s interest shifts from probability and its applications to infinite series . in fact , most of the remaining correspondence with bernoulli turns to that topic . at the same time , montmort began an extensive correspondence with brook taylor , also mainly on infinite series ( st . john s college library , cambridge , taylorb / e4 ) . although the dispute with bernoulli seems to have petered out , montmort was not yet done with it . in a letter dated july 4 , 1716 , montmort asked taylor to examine his dispute with bernoulli about le her and to express his opinion on who was right . he referred taylor only to the correspondence that appears in _ essay danalyse_. taylor apparently wrote back but with the wrong impression about what montmort wanted . montmort replied to taylor on august 4 , 1716 , that he did not want any new research into the problem but only to examine , at his leisure , which of bernoulli or montmort was right . in a letter to taylor dated november 10 , 1717 , montmort thanked taylor for his opinion on the dispute and concluded the letter by saying that waldegrave would write him about le her as well as another game . unfortunately , neither taylor s reply expressing his opinion nor waldegrave s letter to taylor are extant . compared to the discussion of le her , the remaining discussion in the post-1713 correspondence with regard to probability problems is relatively minor . for example , after the remarks on le her that bernoulli made in his letter of february 20 , 1714 , to montmort , bernoulli comments that he thinks there is an error in montmort s solution to a problem related to the jeu du petit palet in _ essay danalyse_. he asks montmort to check his solution . the problem appears to be problme iv in montmort ( @xcite ) , page 254 . the jeu du petit palet is a game in which players toss coins or flat stones ( the `` palets '' ) toward a target set on the ground or a table . the player with the most coins or stones on the target wins . the english equivalent game is called chuck - farthing or chuck - penny . what takes up much of the discussion , other than le her , is news about abraham de moivre s work . de moivre corresponded with both montmort and bernoulli until about 1715 when he ceased corresponding with either of them . prior to this discussion , bernoulli had sent de moivre a general solution to the problem of the pool on december 30 , 1713 ( @xcite , pages 106107 ) . a report on de moivre s activities in probability takes up part of a letter from bernoulli to montmort dated april 4 , 1714 . bernoulli mentions that de moivre has sent him a long letter with reports of new solutions that will appear in a much expanded version of his treatise _ de mensura sortis _ ( de moivre , @xcite ) . de moivre s new work , which was entitled _ the doctrine of chances _ , did not appear until 1718 ( de moivre , @xcite ) . no details are given to montmort other than that de moivre has made inroads in three areas . first , de moivre used his own method for the solution of the problem of the pool to generalize it to more than three players . second , he developed a new kind of algebra to solve probability problems . finally , bernoulli reports that de moivre considered that nearly all problems in probability can be reduced to series summations . not only did de moivre report that he had generalized the problem of the pool , but he also sent bernoulli his solution to the problem . at the time of his writing to montmort , bernoulli had not read the solution and did not pass the solution on to montmort . the new algebra is probably the one that de moivre developed for finding probabilities of compound events . see , for example , hald [ ( @xcite ) , pages 336338 ] for a modern discussion of this topic . this part of the letter ends with what might be interpreted as a nasty comment about de moivre : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ i will share here in confidence what he wrote to me concerning you . here is what he told me about your comments that i had sent him . ` i can not stop myself etc . our society etc . i just received etc . kind [ regards ] . ' after the letter i find written there these words : ` in a sense , ' that made me laugh . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ it is difficult to know what exactly bernoulli is saying here . it appears that he sent de moivre montmort s severe criticism of _ de mensura sortis _ that montmort published in _ essay danalyse _ ( @xcite , pages 363369 ) . later that month , montmort reported back to bernoulli that he received a very polite and fair letter from de moivre in which de moivre announced that he had found a new solution to the problem of the duration of play . see bellhouse [ ( @xcite ) , pages 111114 ] for a discussion of the publication of this solution . de moivre sent reports about more of his results in probability to montmort and montmort sent on a pr ' ecis of these results to bernoulli in a letter dated august 15 , 1714 . many of the results that montmort mentions found their way into _ the doctrine of chances _ , including what is called woodcock s problem discussed in bellhouse [ ( @xcite ) , pages 125126 ] . on august 28 , 1714 , bernoulli finally wrote to montmort enclosing a copy of de moivre s general solution to the problem of the pool . in the letter , bernoulli asks montmort to tell him what he thinks of the solution . he further states that it appears that de moivre is using bernoulli s approach to the solution for three and four players that appears in _ essay danalyse _ ( @xcite , pages 380387 ) . at the same time he is using an analytical approach rather than infinite series ( de moivre actually used a recursive method for his general solution ) . montmort replied on march 22 , 1715 , that he agrees with bernoulli s assessment . on returning from a trip to england , montmort reported to bernoulli in a letter dated june 8 , 1715 , that one of bernoulli s solutions to the problem of the pool had just been printed in the _ philosophical transactions _ ( bernoulli , @xcite ) . bernoulli had sent de moivre two solutions ; de moivre claimed he had found an error in the first solution . many in the past have tried unsuccessfully to identify the waldegrave who solved the problem of le her and who suggested the problem of the pool , often called waldegrave s problem . @xcite reviewed these attempts at identification and narrowed the field down to charles , edward or francis waldegrave , the three brothers of henry waldegrave , 1st baron waldegrave . bellhouse argued for charles waldegrave , but in view of new information his choice was incorrect . key to the proper identification is that several waldegraves siblings , cousins and at least one uncle of henry followed king james ii into exile in france after james was deposed in 1688 . the proper identification of the waldegrave of interest may be found in legal papers in the archive nationales de france in conjunction with a letter from waldegrave to nicolaus bernoulli ; the letter to bernoulli is signed only `` waldegrave '' and is the only known letter in waldegrave s hand that is extant ( universittsbibliothek basel l ia 22 , nr . other waldegrave signatures to compare to the one on bernoulli s letter can be found on various legal documents , two in france ( archives nationales de france mc / et / xvii/486 and 514 ) and one in england ( house of lords record office hl / po / jo/10/1/439/481 ) . see figure [ table2 ] . from the signatures , it is obvious that francis is the waldegrave of interest . from these records , it is also apparent that charles waldegrave handled the family s affairs in england while francis waldegrave took charge of them in france . what little is known of the life of francis waldegrave comes mostly from montmort s correspondence with brook taylor and nicolaus bernoulli . montmort reported to taylor one of waldegrave s political activities . waldegrave was planning to take part in the jacobite uprising in england in 1715 . he was to be part of an invasion force led by the son of james ii , james stuart . the uprising in england fizzled out , james stuart remained in france and waldegrave fell ill just prior to the time when the planned invasion was to occur . montmort called waldegrave s illness apoplexy ; it was probably a stroke . from time to time , montmort commented to taylor and bernoulli about waldegrave s illness , recovery and setbacks . at one point , for a cure or a rest , waldegrave took the waters at a spa in france . he also spent time at montmort s chateau . though ill , he was alive in france in 1719 when montmort died so that the flow of information to bernoulli and taylor about waldegrave stopped . presumably , waldegrave died in france . how waldegrave obtained his mathematical training is unknown . in whatever way he was educated , he was an adept amateur mathematician . this is contrary to henny s interpretation of waldegrave s skills . for example , henny [ ( @xcite ) , page 502 ] claims that waldegrave did not have the mathematical skills to work out a general method of calculation in le her . on the contrary , there is a hint of the fairly high level of waldegrave s mathematical abilities in a letter from montmort to bernoulli dated march 24 , 1714 . there montmort says that he is getting waldegrave to read lhpital s ( @xcite ) calculus book _ analyse des infiniment petits _ and that waldegrave has a natural aptitude for mathematics . at the time that montmort was sending _ essay danalyse _ to his publisher , francis waldegrave was living in rue princesse near glise saint - sulpice in paris . in modern paris , it is a @xmath34 kilometer walk to montmort s publisher in rue galande . in section [ sec1 ] it was mentioned that montmort was staying only @xmath35 meters from his publisher . this was not the only time that montmort enlisted a colleague to take on some of the tedious parts of getting results to print . after montmort sent brook taylor a number of theorems about infinite series , they decided that the results should be published in the royal society s _ philosophical transactions _ in a letter dated june 15 , 1717 , montmort gave taylor complete editorial control over the paper that included having taylor translate the results from french into latin ( st . john s college library , cambridge taylorb / e4 ) . taylor replied august 9 , 1717 , saying that he had made many changes and corrections to the paper ( st . john s college library , cambridge , taylorb / e5 ) . the unpublished letters between bernoulli and montmort reveal a much more complex story than either @xcite or @xcite have described . the entire group bernoulli , montmort and waldegrave were for the most part clear about the issues at the conceptual level . in the end it came down to a disagreement about what it meant to solve a problem . further , henny recognized many modern game theory concepts , but we show that the group s understanding of the modern notions is deeper than what henny realized . apart from the technical and conceptual aspects of le her and other probability problems , we also get a glimpse into the social side of a rich amateur mathematician at work . montmort was a good mathematician , but mathematics was his hobby and at times he did not have time to pursue his hobby . there is a bit of quid pro quo in his relationships with bernoulli , taylor and waldegrave . montmort acquires some status through his connections to artists , philosophers and scientists . he can impose on his scientific friends to do some of the more menial work for him in getting his research to print . on the other side , his scientific friends enjoy his hospitality , his gifts and the benefits of any political and scientific connections that he may have . traditionally , the mixed strategy solution with @xmath24 and @xmath25 for le her has been attributed to waldegrave . it certainly appears to be the correct attribution based on the correspondence in the second edition of _ essay danalyse_. however , in the long letter from montmort to bernoulli dated march 22 , 1715 , that covers discussions of le her , de moivre and other topics , montmort appears to claim priority of solution . as part of the discussion of le her , he says , `` although i first found the determination of the numbers @xmath20 and @xmath21 , @xmath22 and @xmath23 '' montmort s suggestion of priority could have come about as a result of a conversation between waldegrave and montmort , with waldegrave putting pen to paper . this illustrates fasolt s ( @xcite ) claims about the limits of history . our data from the past is what has been written , not what has been spoken . further , we can never know the tone behind what was written , such as bernoulli s apparently nasty comments to montmort about de moivre in his letter of april 14 , 1714 . instead of coming up in conservation , montmort may be claiming priority because he found the general formula in @xmath20 , @xmath21 , @xmath22 and @xmath23 ; the numbers were only a special case . or it could be something else . like le her itself , depending on how the problem is approached , the assignment of priority is a problem with no solution . we would like to thank dr . fritz nagel of universitt basel for giving us access to the correspondence between nicolaus bernoulli and pierre rmond de montmort . we also thank kathryn mckee and jonathan harrison of st . john s college library , cambridge , for providing us with copies of the correspondence between brook taylor and montmort . the originals of the letters of bernoulli , montmort and waldgrave are in universittsbibliothek basel . the letters from montmort to bernoulli are catalogued handschriften l ia 22:2 nr.187206 and from bernoulli to montmort are l ia 21:2 bl.209275 . the letter from bernoulli to waldegrave is catalogued l ia 21:2 bl.229v232r and the letter from waldegrave to bernoulli is l ia 22 , nr . 261 . when referencing these letters , we have to do so by the date , writer and recipient , rather than the catalog numbers

of the many intriguing avenues for investigation afforded by transiting exoplanets ( for a review , see * ? ? ? * ) , perhaps the most exciting is the opportunity to study directly the planetary atmosphere without the need to spatially resolve the light from the planet from that of its host star . at visible wavelengths , where the emitted light from the planet is negligible , the technique of transmission spectroscopy has been employed from space - based platforms to study atoms , molecules , and clouds in the atmospheres of hd 209458b @xcite and hd 189733b @xcite . ground - based studies had yielded only upper limits , albeit valuable ones @xcite until the recent breakthrough detection by @xcite of atomic sodium in the atmosphere of hd 189733b . searches for reflected starlight have delivered increasingly stringent constraints on the wavelength - dependent geometric albedo and phase function @xcite . at infrared wavelengths , the planet - to - star contrast ratio improves dramatically , permitting a direct study of the emitted planetary radiation through the modulation of the total light from the star and planet spanning times of secondary eclipse , when the planet is occulted by the star . ground - based searches for this effect have been frustrated to date by telluric variability @xcite , although @xcite have recently claimed a tentative detection for ogle - tr-113b . exemplifying the power of a small - aperture telescope in an ideal location , the _ spitzer space telescope _ @xcite has revolutionized this field by yielding a flurry of detections of the emitted radiation from 6 exoplanets @xcite , including recent determinations of the emitted spectra @xcite of hd 189733b and hd 209458b . numerous ongoing programs promise additional exciting results to come . although _ spitzer _ will exhaust its cryogen in the spring of 2009 , two photometric channels will continue to operate at full strength and will provide a singular window into the atmospheres of these distant worlds @xcite , if the warm phase of the _ spitzer _ mission is approved . each of the techniques described above is facilitated for large planets that orbit nearby , bright stars at small orbital separations . the discovery of the exoplanet hd 189733b @xcite provided a golden opportunity for practitioners of such techniques for two reasons . first , it is at a distance of only 19.3 pc , and hence at infrared wavelengths it is the brightest star ( @xmath6 ) known to host a transiting exoplanet . second , the planet - to - star surface area ratio is the second largest of the 28 transiting exoplanets described in the literature , surpassed by only tres-3 @xcite . it is primarily for these reasons of accessibility that we selected hd 189733b for intensive study with _ spitzer_. for the purposes of this paper , we adopt the values for orbital inclination @xmath7 , the planet radius @xmath8 , the stellar radius @xmath9 , and the time of transit center @xmath10 as determined by @xcite , and the orbital period @xmath11 as determined by @xcite . we note that the target star has an m - dwarf companion at a projected physical separation of 216 au @xcite . this companion appears in the field of view of the data we present below , and we comment upon its impact upon our photometric results when appropriate . the recent _ spitzer _ detections of the flux emitted by several exoplanets have spurred numerous theoretical efforts to model their atmospheres and predict their emitted spectra ( see for example * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? the constraints on the theoretical predictions become much more stringent with detections at numerous wavelengths , from which a broadband spectrum may be inferred . in this paper , we describe our estimate of planet - to - star contrast ratio in bands centered at @xmath12 . we then combine these results with our re - analysis of the earlier measurement at @xmath13 @xcite to consider what constraints may be placed on the temperature and chemistry of the planetary atmosphere , and the efficiency with which it redistributes the incident stellar radiation from the day side to the night side . these observations are particularly timely , given the recent study by @xcite , which included data presented in deming et al . ( 2005b ) , of the broad band spectrum of hd 209458b in 5 of the above bandpasses . in the assembled spectrum , features that had been predicted in absorption were observed in emission , a clear sign of an atmospheric temperature inversion @xcite . @xcite and @xcite have recently proposed a division between the predicted atmospheric temperature - pressure profiles for hot jupiters dependent upon the level of stellar insolation , in which only the most strongly irradiated planets show a temperature inversion . according to the criterion proposed by @xcite , hd 209458b is a member of the pm class and hence is expected to show a temperature inversion , whereas hd 189733b is a member of the pl class and should not . intriguingly , hd 209458b and hd 189733b also differ in that the radius of the former ( @xmath14 ; * ? ? ? * ) is much larger than that of the latter ( @xmath15 ; * ? ? ? * ) and is at odds with the predictions of models that do not include an additional energy source interior to the planet . whether the large radius and the existence of a temperature inversion share a common physical origin is unclear , but surely worthy of further investigation . our study of the broad band spectrum of hd 189733b provides an ideal opportunity to conduct a comparative study with hd 209458b . we observed hd 189733b over a period of 4.9 hours on ut 2005 november 24 , spanning a single secondary eclipse , using the infrared array camera ( irac ; * ? ? ? * ) on the _ spitzer space telescope_. we observed in subarray mode with an exposure time of 0.1 s and cycled between the four irac channels in order to obtain estimates of the depth of the eclipse at 3.6 , 4.5 , 5.8 , and 8.0 simultaneously . we obtained a total of 21,760 @xmath16 pixel images in each channel . images are taken in sets of 64 , and four sets of images ( @xmath17 images total ) are obtained in each channel before re - pointing the telescope to position the star on the subarray for the next channel . the position of the star on the subarray is still varying through the first set of 64 images after each re - pointing , and we chose to discard this initial set of images , leaving a total of 16,320 usable images in each of the four channels . the total size of this pointing drift is 0.3 pixels in the first set of 64 images , and 0.1 pixels or less in the following three sets of images . for each image , we calculate the jd value for the time at mid - exposure , and apply a correction to convert these jd values to the appropriate hjd taking into account the position of _ spitzer _ in the solar system at each point during the observations . because the two shortest wavelength irac channels ( 3.6 and 4.5 ) use insb detectors and the two longer wavelength channels ( 5.8 and 8.0 ) use si : as detectors , there are fundamental differences between the properties of the data taken with these two types of detectors . we describe our analysis for each type of detector separately below . the specific methods we employ our similar to those described in @xcite . because hd 189733 is a bright star and the background at these shorter wavelengths is minimal , we calculate the flux from the star in each image using aperture photometry with a circular aperture with a radius of five pixels . we determine the position of the star in each image as the position - weighted sum of the flux in a @xmath18 pixel box centered on the approximate position of the star . we estimate the background in each image by selecting a subset of pixels from the corners of the image where the point spread function of the star is faintest and avoiding the region around the m dwarf companion , making a histogram of the flux values in these pixels , and fitting a gaussian function to the center of this distribution . fluxes measured at these two wavelengths show a strong correlation with the changing position of the star on the array , at a level comparable to the depth of the secondary eclipse . this effect is due to a well - documented intra - pixel sensitivity @xcite , and can be removed by fitting the data with a quadratic function of the sub - pixel position of the center of light of the stellar image , according to : @xmath19\ ] ] where @xmath20 is the flux from the star prior to this distortion , @xmath21 is the measured flux , @xmath22 and @xmath23 denote the location in pixels of the center of the star on the subarray , and @xmath24 are the five free parameters in the fit . we fit the correction for the intra - pixel sensitivity of the array and the eclipse curve simultaneously to the data using a markov chain monte carlo method ( see for example * ? ? ? * ; * ? ? ? * ) with @xmath25 steps . we set the uncertainty on individual points equal to the standard deviation of the out - of - eclipse data after correction for the intra - pixel variations , and remove outliers of 5@xmath26 or more as calculated using the residuals from the best - fit light curve . we allow both the depth and timing of the secondary eclipse to vary independently for the eclipses at each of the two observed wavelengths , and fix the values of @xmath7 , @xmath8 , @xmath9 , and @xmath10 to those given in @xcite , the value for @xmath11 to that found by @xcite . we calculate our eclipse curve using the equations from @xcite for the case with no limb - darkening . after running the chain , we search for the point in the chain where the @xmath27 value first falls below the median of all the @xmath27 values in the chain ( i.e. where the code had first found an excellent fit ) , and discard all the steps up to that point . we take the median of the remaining distribution as our best - fit parameter , with errors calculated as the symmetric range about the median containing 68% of the points in the distribution . the distribution of values was very close to symmetric in all cases , and there did not appear to be any strong correlations between variables . figure [ eclipse_plot_no_norm ] shows the binned data with the best fit to the detector effects overplotted , and figure [ eclipse_plot_with_norm ] shows the binned data once these trends are removed , with best - fit eclipse curves overplotted . the best - fit eclipse depths and times are given in table [ eclipse_depths ] . at longer wavelengths the flux from the star is smaller and the zodiacal background is larger ; as a result we chose to use a smaller circular aperture with a radius of 3.5 pixels in order to minimize the noise contribution from this increased background . as before , we calculate the position of the star individually in each image as the position - weighted sum of the fluxes in a @xmath18 pixel box , and estimate the background using a gaussian fit to a histogram of the pixels in the corners of the array . fluxes in the first 10 images and the 58th image in each set of 64 are consistently below the median value for the set by as much as 10% , with the lowest values at the beginning of each set , so we chose to exclude these 11 images from each set of 64 in our analysis . there is no known intra - pixel sensitivity at these wavelengths @xcite , but there is another well - documented detector effect @xcite , which causes the effective gain ( and thus the measured flux ) in individual pixels to increase over time . this effect has been referred to as the `` detector ramp '' , and has also been observed @xcite in the irs 16 array , which is made from the same material . the size of this effect depends on the illumination level of the individual pixel , such that the time - dependent gain of different pixels form a well - defined family of functions . the most strongly - illuminated pixels ( @xmath28250 mjy sr@xmath29 in the 8 channel ) converge asymptotically to a constant value within the first hour of observations . the time scale for this convergence increases with decreasing illumination , and the limiting case ( for pixels that receive little illumination ) is that of a linear increase in the measured flux over time , with a slope that varies inversely with the logarithm of the illumination level . this effect is important for two reasons . first , it means that the observed 3% increase in the measured background flux at 8 over the period of the observations is most likely not the result of a real change in the zodiacal background , but is instead another example of this detector ramp . although the increased noise and smaller size of the background at 5.8 obscures this effect , there appears to be a similar upward trend . thus , rather than calculating the background in each image individually and subtracting that value , we subtract a constant background of 4.81 mjy sr@xmath29 per pixel from all the 8 images , and 0.96 mjy sr@xmath29 per pixel from all 5.8 images . this background is calculated as the median background value during the last hour of observation , when presumably the background is closest to the value it would ultimately obtain . we note that this choice has a negligible effect on our final eclipse depths , as the background constitutes only 0.2% and 1.7% of the signal in our aperture at 5.8 and 8.0 , respectively . this effect also produces a 1% increase in the measured flux from the star at 8.0 over the period of these observations ( see figure 1 ) . there does not appear to be any ramp visible at 5.8 ; instead this data shows a general downward trend . unlike the detector ramp at low illumination levels , the ramp for higher illuminations has an asymptotic shape , with a steeper rise in the first 30 minutes of observations . we discard the first 30 minutes of data in both the 5.8 and 8.0 channels and fit the remaining binned time series from our 3.5 pixel aperture with a quadratic function of @xmath30 , where @xmath31 is the change in time from the start of the observations . @xcite used this same functional form ( with additional degrees of freedom ) to describe the detector ramp in the 8 channel over a period of 33 hours of continuous observations of hd 189733 , and it accurately captures the behavior of this ramp for a range of illumination levels . unlike @xcite , we do not attempt to correct each of the pixels in the images individually for this ramp , as this is not necessary for our analysis . moreover , the lower fluxes , shorter time frame ( 5 hours instead of 33 ) and reduced cadence of our data ( from cycling between the four detectors ) make it difficult to characterize this effect accurately at the pixel level . instead we assume the effect of detector ramp in the binned flux in our 3.5 pixel aperture will have a shape that is representative of the flux - weighted average of the ramp effect for individual pixels . we fit both the quadratic function of @xmath30 and the theoretical eclipse curve to the data simultaneously using a markov chain monte carlo method as described in [ short_norm ] . as before , the distribution of values was very close to symmetric in all cases , and there did not appear to be any strong correlations between variables . best - fit eclipse depths and times from these fits are given in table [ eclipse_depths ] , and the time series both before and after correcting for detector effects are shown in figures [ eclipse_plot_no_norm ] and [ eclipse_plot_with_norm ] , respectively . as a check and to ensure a consistent methods between different bandpasses , we re - analyze the raw time series gathered with the _ spitzer _ irs instrument ( houck et al . 2004 ) at 16 ( prior to normalization using the time series for the m dwarf companion ) from @xcite . importantly , these data were obtained on ut 2005 november 17 , which falls between the times of our observations with mips ( 2.3 ) and irac ( 2.1 ) . we fit the 16 time series with the same quadratic function of @xmath30 as the 5.6 and 8.0 @xmath32 m data . as before , we fit this normalization and the eclipse curve to the data simultaneously using a markov chain monte carlo method with the error for individual points set equal to the standard deviation of the points after the end of the eclipse , calculated after fitting this section of data with a linear function of time to remove the overall slope . this data shows the same type of detector ramp as the 8.0 data , and we exclude from our fit the first 30 minutes of data where this ramp is steepest . our estimate of the eclipse depth , @xmath5% , is slightly shallower and has a smaller uncertainty than the value reported by @xcite , @xmath33% . we also performed a separate analysis of the time series assuming a simple quadratic function of time , which we found yielded the same eclipse depth and uncertainty as that stated in deming et al . we note that the analysis performed by those authors differed from ours in that they first divided the time series of hd 189733 by a fourth - order polynomial - fit in @xmath30 to the time series of the m - dwarf companion , and subsequently fit a quadratric function of time to the out - of - eclipse portion of the corrected time series . they also reported the results of a linear fit to the corrected time series , for which they obtained an eclipse depth of 0.521% , very similar to our final value , stated above . as discussed in [ long_norm ] , we feel that fitting with a quadratic function of @xmath30 represents both a more robust normalization than a quadratic function of time , as well as being a better description of the shape of the detector ramp . the reduced @xmath27 is 1.07 for the fit that uses a quadratic function of time , and falls to 1.00 when we adopt a quadratic function of @xmath30 , indicating that the latter function is indeed a better fit to the data . importantly , the difference in eclipse depths between the two normalizations is less than @xmath34 . we observed hd 189733 for 5 hours spanning a time of secondary eclipse on ut 2005 november 11 , and obtained 1548 images using the _ spitzer _ mips 24 @xmath32 m array @xcite . the mips observing sequence dithers the target position amongst 14 distinct positions on the detector array ( see 8.2.1.2.1 of the _ spitzer observer s manual _ ) . due to the variations in apparent sensitivity at each position , we elected to treat the series of images at each position as independent data sets . after we produce a photometric time series for each position , we normalize each time series to a common scale and subsequently combine them to form a single time series , as described below . we estimate the sky background in each image from a pixel box centered on the star , trimming all pixels more than @xmath35 away from the median value and fitting a gaussian function to the central region of a histogram of the remaining pixels . we find the average background over the duration of our observations is @xmath36 . we then subtract this background from each image individually , and fit the remaining flux in a pixel subarray ( which is large enough to encompass the first airy ring ) centered on the approximate position of the star . we found that varying the size of the subarray we used did not affect the final time series . we use a mips model point spread function ( psf ) for a 5000 k point source and allow both the @xmath22 and @xmath23 position of the psf as well as a constant scaling factor to vary in our fits . we correct for the effects of transient hot pixels in our subarray by making a histogram of flux values at a given pixel position over the set of images ( of which there were @xmath37 ) at a given nod position and flag @xmath38 outliers . we then assign a statistical weight of zero to these values in the fitting process described above ; @xmath39 of our images have one or fewer of these hot pixels . we discard images with 5 or more hot pixels in the subarray ; there are only 10 of these images in the entire time series . we note that the m - dwarf companion to hd 189733 , is included within our 10 pixel subarray ; we give zero statistical weight to the values within a 3 @xmath40 3 pixel box centered on the position of the companion in our fits . we note that the measured background periodically drops by @xmath41 , corresponding to the first image in each bundle of images spanning the sequence of nod positions . @xcite noted the same feature in their data , and chose to divide the measured flux by the background in order to remove the effect of this drop . this procedure would introduce significant noise in our data , because ( unlike the case of hd 209458 ) our target star has a flux twice the value of the background . thus we elected to simply discard the 136 images with these low sky backgrounds . for times later than 1.5 hours after the start of observations , the final time series at each of the 14 nod positions is well - described by a combination of the theoretical eclipse curve and a linear function of time . at earlier times , there appears to be a slight upward trend , possibly corresponding to the detector ramp described in [ long_norm ] . we fit the entire time series , but find that clipping this initial segment of the data in our fits changes the final measured eclipse depth by only @xmath42 . we fit a single eclipse curve and 14 linear functions of time ( one for each of the 14 nod positions ) to the data simultaneously using a markov chain monte carlo method as described in [ short_norm ] , trimming @xmath35 outliers in our fit . a binned version of the combined time series appears in figures [ eclipse_plot_no_norm ] and [ eclipse_plot_with_norm ] , and we state our estimates of the eclipse depth and its time of center in table 1 . with the 6 distinct band - integrated flux ratio measurements in hand , we proceeded to compare these values with model predictions . following barman et al . ( 2005 ) , two model atmospheres for hd 189733b were constructed , one assuming that the absorbed stellar flux is confined to the day side while the other model uniformly distributes the stellar flux over the entire planet . figure [ modelfig ] compares these models and the best - fit planck curve ( corresponding to a temperature of 1292 k ) to the measured day side secondary eclipse depths . we refer the reader to barman ( 2008 ) for a more detailed examination of atmospheric models of hd 189733b in light of our data , but we discuss several of the salient points below . our data show clear deviations from the planck curve . the greatest discrepancy occurs at 3.6 , for which the planet - to - star contrast ratio is more than twice as great as would be expected from pure blackbody emission . this emission feature is a prediction that is common to nearly all published theoretical models ( e.g. barman et al . 2005 ; burrows et al . 2006 ; fortney et al . 2006 ; seager et al . it arises from a hole in the atmospheric opacities near 4 ; significant absorption from both water and carbon monoxide in adjacent wavelengths regions causes flux to be squeezed out through this wavelength span of relatively low opacity . furthermore , the data exhibit a clear trough from @xmath43 m , which is consistent with water absorption . this finding is at odds with the results from grillmair et al . ( 2007 ) , who found that the spectrum of hd 189733b was remarkably flat across the 7.5@xmath4414.7 bandpass and , in particular , that it did not show the downturn at blue wavelengths expected from water absorption . however , the 8.0 and 16 photometric data in figure 3 are themselves only modestly inconsistent with the relatively noisy spectrum from grillmair et al . ( 2007 ) . it may be that while water is indeed present and seen in absorption at shorter wavelengths , additional spectral features are present in the @xmath45 region that mask the expected absorption due to water . fortney & marley ( 2007 ) have noted that the grillmair et al . ( 2007 ) result is itself difficult to reconcile with previously published 8.0 photometry ( knutson et al . 2007a ) , since it would require the presence of physically implausible opacity sources . the presence of water in absorption is in stark contrast to recent findings that water is in emission for hd 209458b ( knutson et al . 2008 ; burrows et al . 2007a ) suggesting fundamental atmospheric differences between these two planets . several authors ( e.g. fortney et al . 2008 ; burrows et al . 2008 ; see also hubeny et al . 2003 ) have discussed how the atmospheres of hot jupiters might bifurcate into two groups , with the presence or absence of a temperature inversion ( as deduced from the presence or absence of emission features in the spectrum ) arising under different levels of external irradiation . as noted by these authors , a candidate for the physical origin of this inversion is the presence of gaseous tio and vo in the upper atmospheres of the more strongly irradiated planets . interestingly , hd 209458b has an equilibrium temperature of @xmath46 k and hence is significantly more strongly illuminated than hd 189733b , which has an equilibrium temperature of only @xmath47 k ( torres et al . 2008 ) . while it is surely too early to state that the presence of a thermal inversion in the atmosphere of hd 209458b and the apparent lack of such an inversion in hd 189733b indeed result from their differing levels of irradiation , the connection is intriguing and worthy of further study . moreover , not only do the overall bolometric irradiances of the two planets different , but the g0v spectral type of hd 209458 ensures a substantially bluer stellar spectral energy distribution than that of the k2v spectral type of hd 189733 . we speculate that the enhanced short - wavelength irradiation for hd 209458b might generate dramatically different photochemical products and hazes than are present on hd 189733b ( see marley et al . 2007 for a discussion of photochemical hazes on hot jupiters ) . since the presence of hazes and their approximate size distribution may be inferred from transmission spectroscopy ( e.g.pont et al . 2008 ) , we urge future modeling efforts to consider the constraints offered by those studies in conjunction with the emission spectroscopy results we present here . we note that within the solar system , planetary atmospheres exhibit a wide variety of photochemical products and distinct physical mechanisms for temperature inversions , and we should expect no lesser diversity amongst the atmospheres of planets orbiting other stars . the planets also differ substantially in terms of their physical structure : hd 209458 is larger but less massive than hd 189733 , such that the densities of the two planets , @xmath48 for hd 209458b and @xmath49 for hd 189733 ( torres et al . 2008 ) differ by nearly a factor of 3 . the larger - than - predicted radii of some , but not all gas - giant exoplanets is one of the greatest conundrums in the field today . many authors have sought to explain the radii of the puffy planets by invoking an additional source in the energy budget of those planets ( e.g. bodenheimer et al . 2003 ; showman & guillot 2002 ; but see burrows et al . 2007b for a dissenting opinion ) . as the atmosphere is the energy gateway for the planet , it is natural to ponder whether there exists a physical connection between the observed properties of the atmospheres of these planets and the significant differences in their physical structure . the best avenue for progress on the observational front is to gather data for a planet that represents an extreme member of one of these two classes . we note that tres-4 ( mandushev et al . 2007 ) has the lowest density ( @xmath50 ) of the known transiting exoplanets , and , with an equilibrium temperature of 1785 k it is also one of the most strongly irradiated . hence it may be the ideal target with which to confirm or refute this proposed connection . the @xmath51 m irac observations clearly favor a model that assumes no redistribution of extrinsic energy to the night side ( upper curve in figure 3 ) , which would predict very cold night side temperatures . this inference is in contradiction to the recent 8 @xmath32 m light curve measurement from knutson et al . ( 2007a ) . those authors determined that the minimum hemisphere - integrated brightness temperature was @xmath52 k and the maximum hemisphere - integrated brightness temperature was not much larger , @xmath53 k. this relatively modest day - night contrast indicated that the planet was efficiently transporting energy from the day side to the night side . the essence of the problem is one of energy budgeting : our irac data indicate a very hot dayside , and the knutson et al . ( 2007a ) values indicate a relatively flat phase curve . if one assumes that the planet exhibits a similar lack of a day - night contrast at all wavelengths , the overall emission will exceed the energy it absorbs even under the most favorable case ( an albedo of 0 ) . although our 16 and 24 values are in agreement with a model with a small day - night variation ( the lower curve in figure 3 ) , the emission at these long wavelengths represents only a small fraction of the bolometric flux of the planet and thus can not , in itself , alleviate this concern . a partial reconciliation may be that the degree of energy redistribution is depth - dependent , and hence the emission phase curve of the planet is strongly wavelength dependent , such that the planet radiates relatively little energy from its night side at shorter wavelengths . a hint of this effect is provided by a recent upper limit on the day side @xmath54-band flux ( barnes et al . 2007 ) implying much cooler atmospheric temperatures across the deeper near - ir photosphere than predicted by the no - redistribution model . alternately , it could be that the planet possesses an internal energy source and hence , like jupiter , simply radiates more energy than it receives . one contender for such an energy source is the tidal dissipation of the orbital eccentricity @xmath55 , which is known to be small but not zero ( knutson et al . however , the predicted dissipation rate ( bodenheimer et al . 2003 ) based on the knutson et al . ( 2007a ) value is far too small to make this a viable explanation . barman ( 2008 ) considers the energy budget of this planet in detail . an additional constraint on the day / night temperature contrast may come from the recent detection by swain et al . ( 2008a ) of methane ( which , at colder temperatures , becomes the dominant carbon species over co ) in the transmission spectrum of hd 189733b . while a cool night side would naturally result in a higher methane abundance , we note that non - equilibrium chemistry as recently explored by cooper & showman ( 2006 ) could also lead to a homogenization of the co and methane abundances across the planet . the estimate of the eclipse depth in the knutson et al . ( 2007a ) 8.0 data was @xmath56% , which is 2.3@xmath26 less deep that the value we present here ( table 1 ) . this raises the unnerving possibility that the dayside emission from the planet may be varying in time , as discussed by rauscher et al . ( 2007a ) . this would be both a blessing , in that it would afford us the opportunity to study weather on this planet , and a curse in that it would significantly complicate the interpretation of broadband spectra assembled from measurements across several secondary eclipses . star spots can not account for this change : although the star displays a well - documented periodic brightness variation due to star spots ( winn et al . 2007a ) , we find upon scaling the variation observed at visible wavelengths to our 8.0 bandpass that star spots would alter the measured eclipse depth by only 0.0013% , a negligible effect . the observed arrival times of the secondary eclipses provides an exquisite constraint on @xmath57 , where @xmath55 is the orbital eccentricity and @xmath58 is the argument of periastron ( see e.g. charbonneau 2003 ) . after accounting for light travel time across the system , knutson et al . ( 2007a ) found that the secondary eclipse of hd 189733 occurred @xmath59 s later than the prediction for a circular orbit , indicating a residual orbital eccentricity of at least @xmath60% . our data are not of sufficient precision to confirm this offset . we find that in 5 of the 6 bandpasses , the eclipse center occurs at a time that is consistent with the expectation of a circular orbit ( assuming the value of @xmath11 from winn et al . 2007a and the value of @xmath10 from knutson et al . 2007a ; see table 1 ) . intriguingly , the eclipse at 3.6 was observed to arrive @xmath61 minutes later than expected . since this offset is chromatic , is surely is nt due to orbital eccentricity , but instead may arise from strong brightness variations across the visible hemisphere of the planet ( williams et al . 2006 ; rauscher et al . despite the seemingly high significance of this offset , we caution that this band is both the one with the largest instrumental corrections arising from intra - pixel sensitivity variations ( figure 1 ) and the largest residual systematic noise ( figure 2 ) . nonetheless , it may be that with greater cadence and more photons , one could detect the perturbations at ingress and egress that result from a brightness variation across the dayside hemisphere . by electing to observe in a single band and obviating the need to continually re - point the telescope as was necessary for the current data , one could realize an improvement of more than a factor of 4 in both cadence and total photons . importantly , such observations could be accomplished during the post - cryogenic phase of the _ spitzer _ mission . one of the salient features of the population of hot jupiter exoplanets is its diversity . these objects occupy what might first appear to be a rather limited range of orbital phase space . however , much hard work over the past 8 years has found 28 such planets that transit , providing precise constraints on their masses and radii . the scatter in their physical properties is astounding , and has challenged a community of theorists that must , after all , attempt to unify the bestiary . similarly , with our first detailed comparative studies of the atmospheres of these exoplanets , we are once again confronted by the differences amongst objects in seemingly similar environments . this diversity is both shocking and invigorating to the theoretical enterprise . we have only one more year to gather observations longward of 4.5 , at which time _ spitzer _ will exhaust its cryogen and the unique window into the atmospheres of these exoplanets will close . fortunately , we have learned that much of the key information regarding these atmospheres will be , in principle , contained in the two bandpasses afforded by the post - cryogenic phase of the _ spitzer _ mission . if it is approved , the _ spitzer _ warm mission will therefore be the pre - eminent tool for investigating the atmospheres of these distant worlds until the launch of _ jwst _ in the middle of the next decade . this work is based on observations made with the _ spitzer space telescope _ , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa . support for this work was provided by nasa through an award issued by jpl / caltech . hak was supported by a national science foundation graduate research fellowship . arribas , s. , gilliland , r. l. , sparks , w. b. , lpez - martn , l. , mediavilla , e. , & gmez - alvarez , p. 2006 , , 118 , 21 bakos , g. . , pl , a. , latham , d. w. , noyes , r. w. , & stefanik , r. p. 2006 , , 641 , l57 barman , t. s. , hauschildt , p. h. , & allard , f. 2005 , , 632 , 1132 barman , t. 2007 , , 661 , l191 barman , t. 2008 , , 676 , l61 barnes , j. r. , barman , t. s. , prato , l. , segransan , d. , jones , h. r. a. , leigh , c. j. , collier cameron , a. , & pinfield , d. j. 2007 , , 382 , 473 bodenheimer , p. , laughlin , g. , & lin , d. n. c. 2003 , , 592 , 555 bouchy , f. , et al . 2005 , , 444 , l15 bozorgnia , n. , fortney , j. j. , mccarthy , c. , fischer , d. a. , & marcy , g. m. 2006 , , 118 , 1249 burrows , a. , budaj , j. , & hubeny , i. 2008 , , 678 , 1436 burrows , a. , hubeny , i. , budaj , j. , knutson , h. a. , & charbonneau , d. 2007a , , 668 , l171 burrows , a. , hubeny , i. , budaj , j. , & hubbard , w. b. 2007b , , 661 , 502 burrows , a. , sudarsky , d. , & hubeny , i. 2006 , , 650 , 1140 burrows , a. , hubeny , i. , & sudarsky , d. 2005 , , 625 , l135 charbonneau , d. 2003 , scientific frontiers in research on extrasolar planets , 294 , 449 charbonneau , d. , noyes , r. w. , korzennik , s. g. , nisenson , p. , jha , s. , vogt , s. s. , & kibrick , r. i. 1999 , , 522 , l145 charbonneau , d. , brown , t. m. , noyes , r. w. , & gilliland , r. l. 2002 , , 568 , 377 charbonneau , d. , et al . 2005 , , 626 , 523 charbonneau , d. , brown , t. m. , burrows , a. , & laughlin , g. 2007 , in protostars and planets v , ed . b. reipurth , d. jewitt , & k. keil ( tucson : univ . of arizona press ) , 701 collier cameron , a. , horne , k. , penny , a. , & leigh , c. 2002 , , 330 , 187 cooper , c. s. , & showman , a. p. 2006 , , 649 , 1048 deming , d. , brown , t. m. , charbonneau , d. , harrington , j. , & richardson , l. j. 2005a , , 622 , 1149 deming , d. , seager , s. , richardson , l. j. , & harrington , j. 2005b , , 434 , 740 deming , d. , harrington , j. , seager , s. , & richardson , l. j. 2006 , , 644 , 560 deming , d. , richardson , l. j. , & harrington , j. 2007a , , 378 , 148 deming , d. , harrington , j. , laughlin , g. , seager , s. , navarro , s. b. , bowman , w. c. , & horning , k. 2007b , , 667 , l199 deming , d. , agol , e. , charbonneau , d. , cowan , n. , knutson , h. , & marengo , m. 2007c , american institute of physics conference series , 943 , 89 demory , b .- o . , et al . 2007 , , 475 , 1125 fazio , g. g. , et al . 2004 , , 154 , 10 ford , e. 2005 , , 129 , 1706 fortney , j. j. , marley , m. s. , lodders , k. , saumon , d. , & freedman , r. 2005 , , 627 , l69 fortney , j. j. , cooper , c. s. , showman , a. p. , marley , m. s. , & freedman , r. s. 2006 , , 652 , 746 fortney , j. j. , & marley , m. s. 2007 , , 666 , l45 fortney , j. j. , lodders , k. , marley , m. s. , & freedman , r. s. 2008 , , 678 , 1419 grillmair , c. j. , charbonneau , d. , burrows , a. , armus , l. , stauffer , j. , meadows , v. , van cleve , j. , & levine , d. 2007 , , 658 , l115 harrington , j. , hansen , b. m. , luszcz , s. h. , seager , s. , deming , d. , menou , k. , cho , j. y .- k . , & richardson , l. j. 2006 , science , 314 , 623 harrington , j. , luszcz , s. , seager , s. , deming , d. , & richardson , l. j. 2007 , , 447 , 691 houck , j. r. , et al . 2004 , , 154 , 18 hubeny , i. , burrows , a. , & sudarsky , d. 2003 , , 594 , 1011 knutson , h. a. , et al . 2007a , nature , 447 , 183 knutson , h. a. , charbonneau , d. , deming , d. , & richardson , l. j. 2007b , , 119 , 616 knutson , h. a. , charbonneau , d. , noyes , r. w. , brown , t. m. , & gilliland , r. l. 2007c , , 655 , 564 knutson , h. a. , charbonneau , d. , allen , l. e. , burrows , a. , & megeath , s. t. 2008 , , 673 , 526 leigh , c. , collier cameron , a. , udry , s. , donati , j .- f . , horne , k. , james , d. , & penny , a. 2003a , , 346 , l16 leigh , c. , cameron , a. c. , horne , k. , penny , a. , & james , d. 2003b , , 344 , 1271 mandel , k. & agol , e. 2002 , , 580 , l171 marley , m. s. , fortney , j. , seager , s. , & barman , t. 2007 , protostars and planets v , 733 morales - calderon , m. , et al . 2006 , , 653 , 1454 moutou , c. , coustenis , a. , schneider , j. , queloz , d. , & mayor , m. 2003 , , 405 , 341 moutou , c. , coustenis , a. , schneider , j. , st gilles , r. , mayor , m. , queloz , d. , & kaufer , a. 2001 , , 371 , 260 narita , n. , et al . 2005 , , 57 , 471 odonovan , f. t. , et al . 2007 , , 663 , l37 pont , f. , knutson , h. , gilliland , r. l. , moutou , c. , & charbonneau , d. 2008 , , 385 , 109 rauscher , e. , menou , k. , cho , j. y .- k . , seager , s. , & hansen , b. m. s. 2007a , , 662 , l115 rauscher , e. , menou , k. , seager , s. , deming , d. , cho , j. y .- k . , & hansen , b. m. s. 2007b , , 664 , 1199 reach , w. t. et al . 2005 , pasp , 117 , 978 redfield , s. , endl , m. , cochran , w. d. , & koesterke , l. 2008 , , 673 , l87 richardson , l. j. , deming , d. , & seager , s. 2003a , , 597 , 581 richardson , l. j. , deming , d. , wiedemann , g. , goukenleuque , c. , steyert , d. , harrington , j. , & esposito , l. w. 2003b , , 584 , 1053 richardson , l. j. , deming , d. , horning , k. , seager , s. , & harrington , j. 2007 , , 445 , 892 rieke , g. h. , et al . 2004 , , 154 , 25 rowe , j. f. , et al . 2006 , , 646 , 1241 rowe , j. f. , et al . 2008 , , submitted , astro - ph/0711.4111 seager , s. , richardson , l. j. , hansen , b. m. s. , menou , k. , cho , j. y .- k . , & deming , d. 2005 , , 632 , 1122 showman , a. p. , & guillot , t. 2002 , , 385 , 166 snellen , i. a. g. 2005 , , 363 , 211 snellen , i. a. g. , & covino , e. 2007 , , 375 , 307 swain , m. r. , vasisht , g. , & tinetti , g. 2008a , , 452 , 329 swain , m. r. , bouwman , j. , akeson , r. l. , lawler , s. , & beichman , c. a. 2008b , , 674 , 482 tinetti , g. , et al . 2007 , , 448 , 169 torres , g. , winn , j. n. , & holman , m. j. 2008 , , 677 , 1324 vidal - madjar , a. , lecavelier des etangs , a. , dsert , j .- m . , ballester , g. e. , ferlet , r. , hbrard , g. , & mayor , m. 2003 , , 422 , 143 werner , m. w. , et al . 2004 , , 154 , 1 williams , p. k. g. , charbonneau , d. , cooper , c. s. , showman , a. p. , & fortney , j. j. 2006 , , 649 , 1020 winn , j. n. , suto , y. , turner , e. l. , narita , n. , frye , b. l. , aoki , w. , sato , b. , & yamada , t. 2004 , , 56 , 655 winn , j. n. et al . 2007a , , 133 , 1828 winn , j. n. et al . 2007b , , 134 , 1707 lrrrrcrrrrr 3.6 & @xmath62 & @xmath63 & @xmath64 + 4.5 & @xmath65 & @xmath66 & @xmath67 + 5.8 & @xmath68 & @xmath69 & @xmath70 + 8.0 & @xmath71 & @xmath72 & @xmath73 + 16 & @xmath74 & @xmath75 & @xmath76 + 24 & @xmath77 & @xmath78 & @xmath79 +

the uniform sampling of bipartite , directed or undirected graphs ( without self - loops and multiple edges ) with fixed degree sequence has many applications in network science @xcite . it has proven difficult to generate truly unbiased samples of graphs with fixed degree sequence @xcite . in this paper we focus on markov chain approaches to this problem , where a graph is randomised by repeatedly making small changes to it . even though several markov chains have been shown to converge to the uniform distribution on their state space @xcite , the main question for both theoreticians and practitioners remains unanswered : that is , in all but some special cases it is unknown how many changes need to be made , i.e. how many steps the markov chains needs to take , in order to sample from a distribution that is close to uniform . the best known markov chain approach for sampling graphs with fixed degree sequence is the _ _ switching model _ _ @xcite . it finds an approximately uniform sample of bipartite graphs , undirected graphs or directed graphs with given vertex degrees , by repeatedly switching the ends of non - adjacent edge pairs . this simple yet flexible approach converges to the uniform distribution if implemented correctly . furthermore , this chain was proven as _ fully polynomial almost uniform sampler _ for the following classes of graphs : regular , half - regular and irregular with bounded degrees @xcite . however , even for these classes of graphs , the theoretically proven mixing time is much too large to use in practice , e.g. @xmath0 for regular graphs with degree @xmath1 @xcite . notice that the fully polynomial uniform sampler of jerrum et al . @xcite for perfect matchings can be used to sample all graphs with fixed degree sequence in polynomial time in transforming the fixed degree sequence problem in a perfect matching problem via an approach of tutte @xcite . however , the theoretical proven mixing times are much too large in practice and furthermore , this approach is more difficult to implement . in this paper we analyse and further develop a _ different _ markov chain approach : the curveball algorithm @xcite , which randomises bipartite graphs and directed graphs with self - loops . experimentally , this chain has been shown to mix much faster than the corresponding switching chain @xcite . the intuition behind the curveball algorithm mixing faster than the switching model can be understood when thinking of both algorithms as games in which kids trade cards . that is , think of the curveball algorithm as an algorithm that randomises the binary @xmath2 bi - adjacency matrix of a bipartite graph . imagine that each row of the adjacency matrix corresponds to a kid , and the @xmath3 s in each row correspond to the cards owned by the kid . then at each step in the curveball algorithm , two kids are randomly selected , and trade a number of their differing cards . using this same analogy for the switching model , in each step two cards are randomly selected and traded if firstly they are different and secondly they are owned by different kids . intuitively , the curveball algorithm is clearly a more efficient approach to randomise the card ownership by the kids . more formally , the curveball algorithm is also based on switches but instead of making one switch , several switches can be made in a single step . we show that this leads to possibly exponentially many graphs being reached in a single step , in contrast with the switching model where at most @xmath4 ( the maximum number of possible edge pairs ) graphs can be reached in a single step . several algorithms closely related to the curveball algorithm were discovered independently by verhelst @xcite . in particular , verhelst already made the critical change from switches to trades . the curveball algorithm is briefly mentioned by verhelst as a variation on his non - uniform sampling algorithms . however , he prefers a metropolis - hastings approach to obtain uniform samples , since intuitively it mixes faster . it is unclear if the added complexity of a single trade in this algorithm causes the overall algorithm to run faster . verhelst furthermore introduces an algorithm similar to the curveball algorithm that fixes the position and number of self - loops , and hence can be used to randomise directed graphs . here , we propose two extensions of the curveball algorithm : the directed curveball algorithm , which samples directed graphs and the undirected curveball algorithm , which samples graphs . our proposed algorithm for directed graphs differs from verhelst s algorithm in the way it deals with induced cycle sets @xcite . by introducing these extensions , we show that , just like the switching model , the curveball algorithm offers a flexible framework that can be used to randomise several classes of graphs . furthermore , we propose a modification to the curveball algorithm and the directed curveball algorithm , that further increases the number of states that can be reached in a single step . we refer to these algorithms as the global curveball algorithm and the global directed curveball algorithm respectively . in the card game analogy , our modification corresponds to letting _ all _ kids trade cards in pairs _ simultaneously _ instead of letting only one pair of kids trade . we prove that both extensions of the curveball algorithm , as well as our global directed curveball algorithms , converge to the uniform distribution . we do so by showing that their markov chains are ergodic ( the underlying state graph is non - bipartite and connected ) and the transition probabilities are symmetric ( see @xcite for an overview on random sampling ) . our proofs follow the approach in @xcite where the original curveball algorithm was proven to converge to the uniform distribution . we show experimentally that the introduced curveball algorithms all tend to mix magnitudes faster than the respective switching models . however , even though experimentally it is clear that the curveball algorithm outperforms the switching model , we do not have a theoretical justification for this . in fact , it turns out that the techniques used to prove fast mixing for the switching chain can not be transferred to the curveball algorithm . hence we present the question of fast mixing for curveball algorithms as an interesting open problem . in our opinion , the curveball algorithm provides a big opportunity and step forward to fast mixing markov chains for the sampling of graphs with fixed degree sequence . the remainder of this paper is organised as follows . section [ sec : cb_ext ] first discusses the original curveball algorithm in terms of adjacency lists , it then introduces two extensions of the curveball algorithm : the directed curveball algorithm that randomises directed graphs , and the undirected curveball algorithm that randomises graphs . furthermore it introduces our modification to the curveball algorithm and directed curveball . in section [ sec : unbiased ] we prove that under mild conditions , all proposed algorithms converge to the uniform distribution . section [ sec : mixing_times ] presents our experimental results on the run - times of all proposed algorithms . furthermore we analyse why the proof of rapid mixing for the switching model can not be used for the curveball algorithm . finally we discuss our conclusions and recommendations for further research in section [ sec : concl ] . we start with a formal definition . given two lists @xmath5 and @xmath6 of non - negative integers , the _ realization problem for bipartite fixed degree sequence _ asks whether there is a labeled bipartite graph , @xmath7 , such that all vertices @xmath8 with @xmath9 have degree @xmath10 and all vertices @xmath11 with @xmath12 degree @xmath13 . analogously , the _ realisation problem for directed fixed degree sequence _ asks for a labeled directed graph for a list @xmath14 , and the _ realisation problem for undirected fixed degree sequence _ for a graph for given list @xmath15 for an overview about these problems see @xcite . the corresponding graphs or lists for each problem are called _ realisations _ or _ degree sequences _ , respectively . the curveball algorithm , as introduced in @xcite , is a markov chain approach to the uniform random sampling of a realisation with bipartite fixed degree sequence . given one such realization , the curveball algorithm finds others by repeatedly making small changes to the _ adjacency list representation of the bipartite graph . _ the adjacency list representation @xcite of a bipartite graph @xmath16 is a set of lists @xmath17 , one for each vertex @xmath9 . the list @xmath17 contains the indices @xmath18 corresponding to the neighbours @xmath19 of @xmath20 ( see figure [ fig : adj_lists](a ) ) corresponding to neighbours of the vertex @xmath21 . depending on the degree sequence , the curveball algorithm may run faster on this representation . ] . the adjacency list representation of all graphs discussed in this paper are in fact sets of _ sets_. we will therefore from now on refer to this representation as the _ adjacency set representation_. the adjacency set representations of ( a ) a bipartite graph , ( b ) a digraph , and ( c ) a graph.,width=453 ] the curveball algorithm randomises the adjacency sets of a bipartite graph using the following steps . ( a ) select two sets @xmath17 and @xmath22 at random . ( b ) let @xmath23 be all indices that are in @xmath17 but not in @xmath22 , i.e. @xmath24 . similarly define @xmath25 . ( c ) create new sets @xmath26 by removing @xmath23 from @xmath17 and adding the same number of elements randomly chosen from @xmath27 . combine @xmath28 with the remaining elements of @xmath27 to form @xmath29 . ( d ) reiterate step ( a)-(c ) @xmath30 times , for a certain fixed number @xmath30 . we follow @xcite and refer to one iteration of steps ( a)-(c ) as a _ trade _ and the number of exchanged indices @xmath31 as the _ size of the trade_. notice that trades can be of size zero and such trades correspond to repeating the current state in the markov chain . furthermore , note that each switch in the switching model for directed graphs equals a trade of size one in the curveball algorithm , as was shown in @xcite . however , the curveball algorithm in addition allows trades of larger size . as discussed in @xcite , the curveball algorithm can be used to sample directed graphs with self - loops ( without multiple edges ) with fixed in- and out - degrees . the adjacency set representation of a digraph with self - loops consists of sets @xmath17 corresponding to out - neighbours of a vertex @xmath20 ( figure [ fig : adj_lists](b ) ) correspond to the in - neighbours of @xmath20 . depending on the digraph that is being randomised , the curveball algorithm may be more efficient when using the in - neighbour representation . ] . we now introduce two extensions of the curveball algorithm : the directed curveball algorithm randomises directed graphs for a fixed degree sequence @xmath32 @xmath33 @xmath34 , @xmath35 , @xmath36 , @xmath37 ( realisation problem for directed graphs ) and the undirected curveball algorithm randomises graphs for a fixed degree sequence @xmath38 ( realisation problem for graphs ) @xcite . the directed curveball algorithm randomises directed graphs by randomising their adjacency set representation . the adjacency set representation of a digraph is the same as that of directed graphs with self - loops , except that it has the property @xmath39 for all @xmath40 , since directed graphs do not contain self - loops . the _ directed curveball algorithm _ differs from the curveball algorithm in step ( b ) only . in the directed curveball algorithm , the set @xmath23 is defined as all elements in @xmath17 not in @xmath22 _ and not equal to @xmath18 _ , i.e. @xmath41 . the set @xmath42 is defined analogously by @xmath43 . this small change ensures no self - loops are introduced while randomising directed graphs . figure [ fig : simple_cb ] illustrates the directed curveball algorithm . notice that for the directed curveball algorithm trades can be of size zero and such trades correspond to repeating a state in the markov chain . the lemma below shows that all switches in the switching model for directed graphs equal trades of size one in the directed curveball algorithm . but , the directed curveball algorithm in addition allows trades of larger size . [ lem : subgraph_simple_dir ] any switch in a digraph is a trade of size one in step ( c ) of the directed curveball algorithm . let @xmath44 and @xmath45 be edges in a digraph @xmath46 that are allowed to be switched . then @xmath47 can not be equal to @xmath48 and @xmath49 can not be equal to @xmath50 since otherwise this switch would introduce a self - loop . furthermore @xmath51 and @xmath52 since otherwise the resulting digraph would have multiple edges . in particular this implies @xmath53 and @xmath54 . now if row @xmath47 and row @xmath49 are selected for a trade , then @xmath55 and @xmath56 are possible sets in step ( c ) that lead to exactly the two new edges @xmath57 and @xmath58 . illustration of the directed curveball algorithm . in this example , the vertices @xmath59 and @xmath60 are selected in step ( a ) . in step ( b ) their entries are compared to create the sets @xmath61 and @xmath62 . notice that 3 is removed from @xmath63 since trading it would introduce a self - loop at vertex @xmath59 . in step ( c ) the new sets @xmath64 and @xmath65 are constructed by randomly redistributing the elements @xmath66 and @xmath67 . step ( d ) repeats steps ( a)-(c ) @xmath30 times . the dashed arrow shows the result of the single trade made in steps ( a)-(c).,width=453 ] the undirected curveball algorithm samples graphs with fixed degree sequence . the adjacency set representation of a graph @xmath68 is a set of sets @xmath17 . the set @xmath17 now contains the indices of the neighbours of vertex @xmath20 ( figure [ fig : adj_lists](c ) ) . the symmetry of a graph is reflected in its adjacency set representation . that is , @xmath40 is an element of @xmath22 if and only if @xmath18 is also an element of @xmath17 . furthermore , these sets have the property that @xmath39 for all @xmath40 , since graphs do not contain self - loops . we introduce the undirected curveball algorithm . this algorithm randomises the adjacency set representation of a graph while maintaining its symmetry and ensuring no self - loops are introduced . the undirected curveball algorithm is defined by the following steps . ( a ) randomly select two sets @xmath17 and @xmath22 . ( b ) let @xmath23 be the set of elements in @xmath17 not in @xmath22 and not equal to @xmath18 , i.e. @xmath41 . analogously define @xmath43 . ( c ) create a new set @xmath26 by removing @xmath23 from @xmath17 and adding the same number of elements randomly chosen from @xmath27 . combine @xmath28 with the remaining elements of @xmath27 to form @xmath29 . ( c@xmath69 ) for each index @xmath70 , replace @xmath18 by @xmath40 in @xmath71 , similarly for each @xmath72 , replace @xmath40 by @xmath18 in @xmath73 . ( d ) reiterate step ( a)-(c@xmath69 ) @xmath30 times , for a certain fixed number @xmath30 . figure [ fig : simple_undir_cb ] illustrates the undirected curveball algorithm . step ( b ) ensures no self - loops are introduced . step ( c@xmath69 ) is well - defined since @xmath70 implies @xmath74 and @xmath75 . this in turn implies that @xmath76 and @xmath77 by symmetry of the adjacency sets . thus we can replace @xmath18 by @xmath40 in @xmath78 to obtain @xmath71 . similarly @xmath72 implies that @xmath40 is an element of @xmath79 and that @xmath18 is not . and thus , replacing @xmath40 by @xmath18 in @xmath79 is well - defined . step ( c@xmath69 ) thus ensures that @xmath80 represents a graph . illustration of the undirected curveball algorithm . in this example the sets @xmath81 and @xmath82 are selected in step ( a ) . in step ( b ) their entries are compared to create the sets @xmath83 and @xmath84 . in step ( c ) the new sets @xmath85 and @xmath64 are constructed by randomly redistributing the elements @xmath3 and @xmath86 . step ( c ) updates the sets corresponding to indices involved in the trade . in this case we need to update @xmath87 by removing @xmath88 and inserting @xmath89 , conversely @xmath90 . step ( d ) repeats steps ( a)-(c ) @xmath30 times . the dashed arrow shows the result of the single trade illustrated by steps ( a)-(c).,width=453 ] notice that the undirected curveball algorithm includes trades of size zero which correspond to repeating the current state in the markov chain . furthermore , lemma [ lem : switch_equals_trade_undir ] shows that any switch in the switching model for graphs corresponds to a trade of size one in the undirected curveball algorithm . in fact , figure [ fig : undircb ] shows that for each switch in the switching model , there are two different trades of size one in the undirected curveball algorithm . [ lem : switch_equals_trade_undir ] let @xmath91 be graphs that differ by a switch . there are two trades of size one in the undirected curveball algorithm from @xmath46 to @xmath92 . without loss of generality we may assume that @xmath46 and @xmath92 differ by a switch from @xmath93 and @xmath94 to @xmath95 and @xmath96 ( see figure [ fig : undircb ] ) . let @xmath97 be the adjacency set representation of @xmath46 , then @xmath53 since the edge @xmath93 is an edge of @xmath46 , the edge @xmath96 is not and @xmath50 can not be equal to @xmath49 since @xmath98 and @xmath92 has no self - loops . similarly we find that @xmath54 and hence the trade that swaps @xmath50 and @xmath48 between rows @xmath47 and @xmath49 results in the graph @xmath92 . similarly , there is a second trade which generates @xmath92 , namely the trade that exchanges @xmath47 and @xmath49 between sets @xmath99 and @xmath100 . a switch corresponds to a trade of size one in the undirected curveball algorithm . notice that each switch can be realized by two distinct trades : the switch from @xmath93 and @xmath94 to @xmath95 and @xmath96 can be realized by selecting @xmath101 and @xmath102 or by selecting @xmath99 and @xmath100 . , width=453 ] analogous to the other versions of the curveball algorithm , the undirected curveball algorithm in addition allows trades of larger size , corresponding to making several switches at once . we now introduce the global curveball algorithm and the global directed curveball algorithm as a modification of the curveball algorithm and the directed curveball algorithm respectively . the number of graphs that can be reached by a single step in the markov chain of these _ global _ curveball algorithms is even higher than for the regular curveball algorithms . this modification is motivated by our desire to improve the curveball algorithm in situations where it can reach the exact same number of graphs as the switching model . this happens for instance when a biadjacency matrix represents a perfect matching , i.e. row and column sums are one . in this case , each trade has at most size one , and hence corresponds to either a switch or a repeated state . as explained in the introduction , instead of attempting trades between two sets ( kids ) in the adjacency set representation , the global algorithms allow each of the sets to trade in pairs . more formally , the global curveball algorithm and the global directed curveball algorithm are defined as follows . we replace step ( a ) in the curveball algorithm and directed curveball algorithm by the following step : for all lists @xmath103 choose uniformly at random a 2-partition @xmath104 , @xmath105 , @xmath106 , @xmath107 for even @xmath108 , or , @xmath109 , @xmath105 , @xmath106 , @xmath110 , @xmath111 for odd @xmath108 . for each pair @xmath112 apply steps ( b)-(c ) . step ( d ) again reiterates steps ( a)-(c ) @xmath30 times . we refer to one iteration of steps ( a)-(c ) as a _ global trade _ in analogy to the term trade for the curveball algorithm . in the appendix we present algorithm [ algorithmtwopartition ] for determining a uniform @xmath88-partition in @xmath113 asymptotic runtime . in our example of a perfect matching above , the global curveball algorithm has an exponential number of possible transitions for each realisation : each pair of the @xmath114 possible partitions allows @xmath115 different global trades , since all @xmath116 adjacency list pairs allow two trades ( a switch or a trade of size zero ) . this exponential number of possible transitions is in contrast to only a quadratic number of transitions in the curveball algorithm : each transition corresponding to a single switch for each of the @xmath117 possible adjacency list pairs . a similar modification to the undirected curveball algorithm is not possible . a trade between two sets affects additional sets in step ( @xmath118 ) , and hence a trade between a pair of sets is not independent of trades between other pairs of sets . a markov chain can be seen as a _ random walk _ @xcite on a set @xmath119 of combinatorial objects , the so - called _ states_. two states @xmath120 are connected via a transition edge @xmath121 , when @xmath47 can be transformed into @xmath50 via a small local change . for the switching chain such a ` local change ' corresponds to a switch , and in the curveball algorithms to one trade . for both algorithms , the states of @xmath119 are all realisations of a fixed degree sequence . this definition induces a so - called directed _ state graph _ @xmath122 , representing the states and how they are connected by local changes . a step from @xmath47 to @xmath50 in a random walk is done with _ transition probability _ @xmath123 in @xcite the curveball algorithm was proven to converge to the uniform distribution by applying the fundamental theorem for markov chains ( see for example @xcite ) . [ thm : mc_stationary ] a finite markov chain converges to its unique stationary distribution if its state graph @xmath122 is connected and non - bipartite . if there exists a probability distribution @xmath124 $ ] such that the detailed balanced equations , @xmath125 , are satisfied for all @xmath126 then @xmath127 is the unique stationary distribution . markov chains which fulfill these properties are called _ ergodic_. this theorem implies that an ergodic markov chain converges to the uniform distribution if @xmath128 for all @xmath120 . we now derive the conditions for which the directed curveball algorithm , the undirected curveball algorithm , the global curveball algorithm , and the global directed curveball algorithm converge to the uniform distribution on their respective state spaces , i.e. the set of all possible solutions of the realisation problem . we start by deriving the transition probabilities of the directed curveball algorithm . [ lem : subgraph_simple_dir ] let @xmath129 and @xmath80 be two adjacency set representations of directed graphs with equal degree sequence . the transition probability @xmath130 from @xmath129 to @xmath80 , in the directed curveball algorithm , is given by @xmath131 where @xmath132 and @xmath133 . hence , @xmath134 for all @xmath135 . the probability of transitioning from a state @xmath129 to another state @xmath80 that differs in a trade between sets @xmath17 and @xmath22 can be found as follows . the probability of selecting set @xmath17 and set @xmath22 equals the inverse of the number of pairs of sets in the adjacency set representation , i.e. @xmath136 , where @xmath108 equals the number of sets . the probability that shuffling @xmath27 results in state @xmath80 equals the inverse of the number of ways you can select @xmath137 unordered elements from the set @xmath27 in step ( c ) of the algorithm . this probability equals @xmath138 since @xmath139 . to show that the probabilities @xmath130 and @xmath140 are equal for all adjacency sets @xmath129 and @xmath80 we only need to show that this is true in the non - trivial case when the adjacency sets differ exactly in two sets , say @xmath17 and @xmath22 . @xmath141 equals @xmath142 , and @xmath143 equals @xmath144 since trades do not add or remove elements . we find @xmath145 since @xmath146 , and @xmath147 . similarly @xmath148 , and indeed we find @xmath134 . we next discuss irreducibility of the directed curveball algorithm . notice that lemma [ lem : subgraph_simple_dir ] implies that the state graph of the switching model for simple directed graphs is a subgraph of the state graph of the directed curveball algorithm because each switch is a trade of size one . hence , since both markov chains have the same state space , whenever the switching model for directed graphs has irreducible markov chain then so does the markov chain of the directed curveball algorithm . this leads to the following theorem . [ thm : concergence_directed_notconnected ] if the state graph corresponding to the switching chain for directed fixed degree sequences is connected , then the markov chain of the directed curveball chain converges to its stationary distribution , which is the uniform distribution . on the other hand , if the markov chain of the switching model is not connected , then applying the curveball algorithm to any component converges to the uniform distribution on all states in this component . the state graph of the switching model with respect to directed graphs is a subgraph of the state graph of the directed curveball algorithm ( lemma [ lem : subgraph_simple_dir ] ) . hence , a connected state graph of the switching chain implies a connected state graph of the directed curveball chain . the state graph of the directed curveball chain is always non - bipartite , since there is a non - zero transition probability @xmath149 of repeating each state @xmath129 in step ( c ) , due to trades of size zero . finally @xmath134 for all states @xmath129 and @xmath80 ( see lemma [ lem : subgraph_simple_dir ] ) . hence convergence to the uniform distribution follows from theorem [ thm : mc_stationary ] . it is well - known that the switching model for directed graphs can have a reducible markov chain @xcite . the simplest example being a directed cycle on three vertices , its opposite orientation can not be achieved by switches , since no switch is possible without introducing self - loops . we know of two approaches to mitigate this problem for switching chains : one is to introduce an additional move which reorients directed cycles of length three ( hexagonal move in @xcite ) . this is the approach taken in @xcite to sample directed graphs . however , we follow a second approach that uses a pre - sampling step @xcite . we prefer this approach because the corresponding markov chain runs on a ( potentially much ) smaller state graph compared with the triangle reorientation chain and hence should be faster . furthermore , this approach is much easier to transfer to the directed curveball algorithm . to discuss this approach we need the definition of _ induced cycle sets _ @xcite for a directed graph sequence @xmath150 . an induced cycle set consists of three indices , @xmath151 , @xmath152 and @xmath153 , for pairs in @xmath150 such that the vertices @xmath154 form a directed cycle in _ each _ directed graph realisation of @xmath150 . let @xmath155 be the state graph of the switching model for directed graphs with fixed degree sequence @xmath150 . berger et al @xcite prove that @xmath155 is non - connected if and only if @xmath150 contains an _ induced cycle set_. in fact , they prove that if @xmath150 contains @xmath156 induced cycle sets , then @xmath155 consists of @xmath157 isomorphic components where each component corresponds to a specific orientation for all @xmath156 cycles . hence , instead of introducing a triangle reorientation , berger et al . @xcite choose one of the isomorphic components uniformly at random prior to running the switching model on this component . notice that they also show that all induced cycle sets are disjunctive , i.e. at most @xmath158 such cycle sets are possible . the state graph of the directed curveball algorithm decomposes in @xmath157 isomorphic components , where @xmath159 is the number of induced cycle sets . if it is not connected @xmath160 , then applying the directed curveball algorithm on any component leads to the uniform distribution of all states in this component . the state graph for the directed curveball decomposes in at most @xmath157 isomorphic components because the state graph of the switching chain for directed sequences is a subgraph ( lemma [ lem : subgraph_simple_dir ] ) of the state graph of the directed curveball algorithm , and the state graph of the switching chain decomposes in @xmath157 isomorphic components @xcite . all these components contain realisations which only differ in the orientation of triangles of induced cycle sets . the directed curveball algorithm basically applies switches , and is not able to change these triangles . hence , the state graph of the directed curveball chain has exactly @xmath157 components , consisting of exactly the same states as the components of the switching chain state graph . therefore trades must be identical in each component leading to isomorphic components . using the non - bipartitness of each component ( proof of theorem [ thm : concergence_directed_notconnected ] ) and @xmath161 with lemma [ lem : subgraph_simple_dir ] , it follows with theorem [ thm : mc_stationary ] that the directed curveball algorithm on a component converges to the uniform distribution on all states in the component . this theorem implies that choosing one component uniformly at random and applying the directed curveball algorithm on this component leads to a uniform distribution of all states . hence , we propose the following adjusted directed curveball algorithm for a directed graph @xmath46 with degree sequence @xmath150 : ( 1 ) identify all @xmath156 induced cycle sets in @xmath150 . ( 2 ) choose a random orientation for each induced cycle set in @xmath46 , leading to a realisation @xmath92 of @xmath150 . ( 3 ) use the directed curveball algorithm starting with @xmath92 . let @xmath46 be a directed graph with directed graph sequence @xmath150 . the adjusted directed curveball algorithm converges to the uniform distribution on all directed graph realisations of @xmath150 . we now propose a linear - time algorithm for the identification of all induced cycle sets in ( 1 ) of the adjusted directed curveball algorithm . this approach follows a result of lamar @xcite , which we write in a different form and simplify . first we give a classical definition of _ corrected ferrers matrices _ for a degree sequence @xmath162 , @xmath106 , @xmath163 in non - increasing lexicographical order . a corresponding @xmath164 _ ferrers matrix _ @xmath165 is a matrix with row sums @xmath166 . each row consists of consecutive @xmath3 s followed by consecutive @xmath167 s with the exception that the diagonal elements @xmath168 are always @xmath169 this leads to column sums @xmath170 of @xmath171 the classical result of chen - fulkerson - ryser states that @xmath150 has a realisation if and only if @xmath172 for all @xmath173 for a comprehensive discussion we recommend the paper of berger @xcite . we define @xmath174 as the function with @xmath175 for @xmath176 and @xmath177 in all other cases . lamar @xcite stated the following result . [ theoremlamar ] let @xmath178 be a lexicographical non - increasing degree sequence with a directed graph as realisation , and @xmath170 the column sums of its corrected ferrers matrix . + let @xmath179 be a permutation of @xmath150 which was generated by exchanging the component order in all pairs and sorting it in non - increasing lexicographical order . let @xmath180 the column sums of its corrected ferrers matrix . + indices @xmath181 form an induced cycle set in @xmath150 if and only if 1 . @xmath182 2 . @xmath183 3 . @xmath184 for @xmath185 4 . @xmath186 for @xmath187 we state a simpler version of this theorem which is based on the observation that items ( 2 ) and ( 4 ) follow directly from items ( 1 ) and ( 3 ) . thmtheoreminducedcycles [ theoreminducedcycles ] let @xmath178 be a lexicographical non - increasing degree sequence with a directed graph as realisation , and @xmath170 the columnn sums of its corrected ferrers matrix . indices @xmath181 form in @xmath150 an induced cycle set if and only if 1 . @xmath182 2 . @xmath188 for @xmath185 the proof is given in the appendix . this results in the following algorithm for detecting all induced cycle sets in linear time ( which was also the case for lamar s theorem [ theoremlamar ] ) . ( 1 ) sort @xmath150 in non - increasing lexicographical order , ( 2 ) determine the set @xmath189 of all triples @xmath190 fulfilling ( 1 ) in theorem [ theoreminducedcycles ] , ( 3 ) construct the corresponding ferrers matrix for @xmath150 , and ( 4 ) determine @xmath191 for @xmath173 if @xmath192 for triple @xmath193 then @xmath194 is an induced cycle set . step ( 1),(2),(4 ) can be done in @xmath113 time . the construction of the ferrers matrix needs @xmath195 time where @xmath196 denotes the number of @xmath3 s in the matrix . in summary this algorithm leads to an asymptotic linear time . we now discuss the conditions under which the undirected curveball algorithm converges to the uniform distribution . we start by deriving its transition probabilities . [ lem : undir_cb_trans_prob ] let @xmath129 and @xmath80 be two adjacency set representations of graphs with equal degree sequence . the transition probability @xmath130 from @xmath129 to @xmath80 , in the undirected curveball algorithm , is given by @xmath197 with @xmath132 , @xmath133 , @xmath198 and @xmath199 . in particular , @xmath134 for all states @xmath135 . when the adjacency sets @xmath129 and @xmath80 differ by a trade of size one between sets @xmath17 and @xmath22 involving indices @xmath156 and @xmath200 , then they also differ by a trade of size one between sets @xmath78 and @xmath79 involving indices @xmath40 and @xmath18 ( see lemma [ lem : switch_equals_trade_undir ] ) . hence , we need to add the probabilities of selecting either one of these trades . when @xmath129 and @xmath80 differ in trade of size larger than one , there is a unique pair of sets that corresponds to this trade , hence we find the usual transition probability . to see that @xmath134 , observe that just like in the directed curveball algorithm , a trade between sets @xmath17 and @xmath22 to form @xmath26 and @xmath29 implies that @xmath201 and @xmath202 since trades leave common elements invariant and do not alter the number of elements in eachs set . for any graph @xmath46 , the markov chain of the undirected curveball algorithm starting at @xmath46 converges to the uniform distribution on all graphs with the same degree sequences as @xmath46 . the state graph of the switching chain for graphs with fixed degree sequence is a subgraph of the state graph of the undirected curveball algorithm on the same states ( lemma [ lem : switch_equals_trade_undir ] ) . the state graph of the switching model was shown to be connected in @xcite which implies the connectance of the state graph of the undirected curveball algorithm . the state graph of the undirected curveball algorithm is always non - bipartite , since there is a non - zero probability of repeating each state , due to trades of size zero . finally @xmath134 , see lemma [ lem : undir_cb_trans_prob ] . hence by theorem [ thm : mc_stationary ] the undirected curveball algorithm converges to the uniform distribution on its state space . we start by deriving the transition probabilities for the global curveball algorithm of subsection [ sec : modified_curveball ] . first we calculate the number of possible global trades for _ one _ @xmath88-partition @xmath203 , and develop the number of possible transitions for _ all _ @xmath88-partitions . this value will be taken as a basis for calculating the transition probabilities . in the following we denote by _ even partition _ of a set @xmath204 a @xmath88-partition @xmath205 , and by _ odd partition _ @xmath206 . [ lem : number_global_trade ] let @xmath129 be the adjacency set representations of a bipartite graph with degree sequence @xmath150 , and @xmath203 be a @xmath88-partition of @xmath207 with @xmath208 for even @xmath108 , and @xmath209 for odd @xmath108 . the number @xmath210 of global trades for @xmath203 in the global curveball chain is @xmath211 with @xmath212 and @xmath213 . the number of global trades for one partition @xmath203 in the global curveball algorithm is the product of the number of trades for each of the randomly chosen pairs @xmath214 , since trades for these pairs are applied independently . the number of trades for each pair of rows is the same as that in the original curveball algorithm . this number was derived in @xcite and our result now follows . we now discuss an example that shows that two different @xmath88-partitions and corresponding global trades may result in the same change in the adjacency set representation . let @xmath129 be the following adjacency set representation of a digraph with self - loops : @xmath215 $ ] , @xmath216 $ ] , @xmath217 $ ] and @xmath218 $ ] . consider the following two different @xmath88-partitions @xmath219 and @xmath220 . it is easy to see that the set of global trades for both partitions is exactly the same . this example leads us to derive the transition probabilities of the global curveball algorithm as follows . lemmamodifiedcurveballtransition [ lem : modified_curveball_transitions ] let @xmath129 and @xmath80 be two adjacency set representations of bipartite graphs with equal degree sequence , and @xmath221 the set of all @xmath88-partitions of @xmath207 . the transition probability @xmath130 from @xmath129 to @xmath80 , in the global curveball algorithms , is given by @xmath222 where @xmath210 is the number of global trades for one partition @xmath203 in lemma [ lem : number_global_trade ] . the number of @xmath88-partitions @xmath223 is given by @xmath224 for even partitions , and by @xmath225 for odd partitions . in particular , @xmath134 for all @xmath135 . we prove the result for the bipartite and directed case simultaneously . each partition in @xmath221 corresponds to at most one global trade ( see lemma [ lem : number_global_trade ] ) between @xmath129 and @xmath80 . the probability of selecting a partition that corresponds to a global trade between @xmath129 and @xmath80 equals @xmath226 . for each of these partitions , @xmath203 , the probability of selecting the corresponding global trade between @xmath129 and @xmath80 equals @xmath227 . we derive the number of @xmath88-partitions @xmath223 in the appendix . using the same arguments as in proof of lemma [ lem : subgraph_simple_dir ] we get @xmath134 for all @xmath135 because @xmath227 is the probability of @xmath116 independent trades in the directed curveball algorithms . recalling that global trades correspond to a number of independent trades in the curveball algorithm it follows that the state graph of the directed version of the global curveball algorithm decomposes in @xmath157 isomorphic components whenever induced cycle sets are contained in degree sequence @xmath228 all results from subsection [ sec : unbiased_simple ] can be applied analogously leading to an _ adjusted version of the global directed algorithm _ which samples uniform at random one isomorphic component and uses global trades to sample within this isomorphic component uniformly at random . we do not repeat all details from subsection [ sec : unbiased_simple ] . instead we state the following theorem . [ thm : concergence_directed_modified ] if the state graph corresponding to the switching chain for directed fixed degree sequences is connected , then the markov chain of the global directed curveball chain converges to its stationary distribution , which is the uniform distribution . on the other hand , if the markov chain of the switching model is not connected , then applying the global directed curveball algorithm to any component converges to the uniform distribution on all states in this component . the most important question for practitioners as well as theoreticians is how many steps the ( global ) curveball algorithms have to run from an initial probability distribution ( where an initial state is taken from ) to sample from a probability distribution which is close to the uniform distribution . this number is defined as the _ total mixing time _ , i.e. the number @xmath30 of reiteration steps in the curveball algorithms . the curveball algorithm has experimentally been shown to run much faster than the switching algorithm @xcite . although we do not know if the total mixing time of the curveball algorithm is smaller than that of the switching model , we show that all of our proposed curveball algorithms ( see section [ sec : cb_ext ] ) tend to run faster in experiments than the respective switching models . to establish the run - time of a markov chain we need to know both its mixing time , that is the approximate number of steps needed to reach its stationary distribution , and its step run - time , i.e. the the average time it takes to execute a single step in the chain . all algorithms were implemented in the r programming language and are publicly available @xcite . the perturbation scores @xcite between directed graphs in the markov chain and the original directed graph provides a dissimilarity measure . we use the point where this perturbation score stabilizes as an estimate for the mixing time of the markov chains . figure [ fig : cb_converge_directed](a ) indicates that the directed curveball algorithm mixes in fewer steps than the switching model : the perturbation score stabilizes much faster . furthermore figure [ fig : cb_converge_directed](b ) shows that the steps in the directed curveball algorithm take much less time to run than those in the switching model . ( a)the perturbation score at each step in the directed curveball algorithm ( solid ) and the switching model with respect to simple directed networks ( dashed ) as compared to the initial network . we randomise five erds - rnyi random networks @xmath229 with different edge densities @xmath230 . the perturbation score stabilizes much earlier for the directed curveball algorithm than for the switching algorithm . ( b)the time in seconds , that it takes to run 1000 steps in the directed curveball algorithm and the switching model.,width=453 ] we found similar results for the undirected curveball algorithm and the corresponding switching model , as shown in figure [ fig : cb_converge_undirected ] . our experiments indicate an improvement in speed both in terms of the mixing time and the step run - time . the improved step run - time of the curveball algorithm might be due to the use of the adjacency set representation of networks , as this is an efficient method for storing networks @xcite . ( a)the perturbation score at each step in the undirected curveball algorithm ( solid ) and the switching model with respect to undirected networks ( dashed ) as compared to an initial network . we randomise five erds - rnyi random networks @xmath229 with different @xmath230 . ( b)the time in seconds , that it takes to run 1000 steps in the undirected curveball algorithm and the switching model.,width=453 ] even though there is experimental evidence that the mixing time of the curveball algorithms is much faster than that of switching models , there is currently no theoretical proof . there are few theoretical results about the rapid mixing of the switching model . for the special case of regular and semi - regular networks @xcite , the polynomial upper bound for the mixing time was found using a multi - commodity flow argument @xcite . these proofs rely on defining a special class of paths between all states in the state graph . paths are chosen in such a way that the load on each edge ( the number of paths it takes part in ) is relatively small . the mixing time can then be bounded from above in terms of a product of these edge loads and the inverse of their transition probabilities . unfortunately the multi - commodity flow method can not be used to prove rapid mixing for the curveball algorithm . the same class of paths could be used , but the argument breaks down when estimating the transition probabilities . the reason for this is that the transition probabilities in the curveball algorithm can be exponentially small with respect to the number of vertices @xmath108 in a network , leading to an exponential factor in the upper bound . we do not believe that the small transition probabilities are an _ actual _ obstruction to fast mixing of the curveball algorithms , since each state also has a corresponding exponential number of neighbouring states . the fastest mixing markov chain on @xmath30 states has the complete graph as its state graph , with all transition probabilities equal to @xmath231 . with an exponentially large state space , these probabilities are also exponentially small . intuitively , the curveball algorithm is much closer to this optimal situation than the switching method . it appears that an altogether different method is needed to find a theoretical upper bound for the mixing time of the curveball algorithm . this is a difficult , but important open problem . the curveball algorithm seems to be a step in the right direction for the fast generation of random directed networks . in this paper we introduced two extensions of the curveball algorithm : the directed curveball algorithm and the undirected curveball algorithm . these algorithms were developed to randomise undirected and simple directed networks while fixing their degree sequence . it is important for random network models to sample without bias . we proved that both the directed curveball algorithm and the undirected curveball algorithm converge to the uniform distribution . furthermore , experimental evidence shows that they do so much faster than the well - known switching models . we recommend the use of these models over that of the switching model , especially for large networks . we pointed out why current techniques can not be used for formal proof of rapid mixing of the curveball algorithm . developing new techniques and proving rapid mixing is an interesting open problem . 10 https://github.com/queenbne/curveball . https://github.com/queenbne/switchingmodelfamily . y. artzy - randrup and l. stone . generating uniformly distributed random networks . , 72:056708 , 2005 . a. berger . the connection between the number of realizations for degree sequences and majorization . , a. berger . a note on the characterization of digraphic sequences . , 314:38 41 , 2014 . a. berger and m. mller - hannemann . uniform sampling of digraphs with a fixed degree sequence . in _ proceedings of the 36th international conference on graph - theoretic concepts in computer science _ , pages 220231 . springer - verlag , 2010 . full version available as preprint in arxiv:0912.0685v3 . c. j. carstens . a uniform random graph model for directed acyclic networks and its effect on motif - finding . , 2:419430 , 2014 . c. j. carstens . proof of uniform sampling of binary matrices with fixed row sums and column sums for the fast curveball algorithm . , 91:042812 , 2015 . c. cooper , m. dyer , and c. greenhill . corrigendum : sampling regular graphs and a peer - to - peer networks . , 2012 . r. b. eggleton and d. a. holton . simple and multigraphic realizations of degree sequences . in _ combinatorial mathematics viii _ , pages 155172 . springer berlin heidelberg , 1981 . l. erds , i. mikls , and l. soukup . towards random uniform sampling of bipartite graphs with given degree sequence . , 2010 . p. l. erds , i. mikls , and z. toroczkai . new classes of degree sequences with fast mixing swap markov chain sampling . , abs/1601.08224 , 2016 . nj gotelli and gl entsminger . ecosim : null models software for ecology . version 7 . acquired intelligence inc . & kesey - bear . jericho , vt 05465 , 2009 . c. greenhill . a polynomial bound on the mixing time of a markov chain for sampling regular directed graphs . , 18(1):p234 , 2011 . c. greenhill . the switch markov chain for sampling irregular graphs : extended abstract . in _ proceedings of the twenty - sixth annual acm - siam symposium on discrete algorithms _ , pages 15641572 , 2015 . m. jerrum . . birkhuser verlag , basel , switzerland , 2003 . m. jerrum and a. sinclair . approximating the permanent . , 18(6):11491178 , 1989 . m. jerrum , a. sinclair , and e. vigoda . a polynomial - time approximation algorithm for the permanent of a matrix with nonnegative entries . , 51:671697 , 2004 . d. jungnickel . . springer verlag , heidelberg , 1999 . r. kannan . markov chains and polynomial time algorithms . in _ foundations of computer science , 1994 proceedings . , 35th annual symposium on _ , pages 656671 , 1994 . o. d. king . comment on `` subgraphs in random networks '' . , 70(5):058101 , 2004 . donald e. knuth . . addison wesley longman publishing co. , inc . , redwood city , ca , usa , 1998 . m. d. lamar . on uniform sampling simple directed graph realizations of degree sequences . , abs/0912.3834 , 2009 . d. a. levin , y. peres , and e. l. wilmer . . american mathematical society , providence , rhode island , 2009 . l. lovsz . random walks on graphs : a survey . in _ combinatorics , paul erds is eighty _ , volume 2 , pages 353397 . jnos bolyai mathematical society , 1996 . s. maslov and k. sneppen . specificity and stability in topology of protein networks . , 296:910913 , 2002 . i. mikls , p. l. erds , and l. soukup . towards random uniform sampling of bipartite graphs with given degree sequence . , 20(1):p16 , 2013 . i. mikls and j. podani . randomization of presence - absence matrices : comments and new algorithms . , 85:8692 , 2004 . m. molloy and b. reed . a critical point for random graphs with a given degree sequence . , 6(2 - 3):161180 , 1995 . m. e. j. newman . . oxford university press , 2010 . m. e. j. newman , s. h. strogatz , and d. j. watts . random graphs with arbitrary degree distributions and their applications . , 64(2):026118 , 2001 . a. r. rao , r. jana , and s. bandyopadhyay . a markov chain monte carlo method for generating random ( 0 , 1)-matrices with given marginals . 58:225242 , 1996 . h. j. ryser . combinatorial properties of matrices of zeros and ones . , 9:371377 , 1957 . a. sinclair and m. jerrum . approximate counting , uniform generation and rapidly mixing markov chains . , 82(1):93133 , 1989 . g. strona , d. nappo , f. boccacci , s. fattorini , and j. san - miguel - ayanz . a fast and unbiased procedure to randomize ecological binary matrices with fixed row and column totals . , 5:4114 , 2014 . r. taylor . , chapter constrained switchings in graphs , pages 314336 . springer berlin heidelberg , berlin , heidelberg , 1981 . the factors of graphs . , ( 4):314328 , 1952 . n. d. verhelst . an efficient mcmc algorithm to sample binary matrices with fixed marginals . , 73(4):705728 , 2008 . in the following we give an algorithm for computing a @xmath88-partition @xmath203 of a set @xmath207 uniformly at random . the basic idea is that there is always a pair of integers in each partition containing the minimum number of a set @xmath232 the algorithm creates one pair with this number @xmath40 and chooses the partner @xmath18 randomly from @xmath233 it remains to find a @xmath88-partition of a set @xmath234 which does nt contain @xmath40 and @xmath235 set @xmath204 . uniform sampled @xmath88-partition @xmath205 of @xmath234 ( for even @xmath108 ) . uniform sampled @xmath88-partition @xmath236 ( for odd @xmath108 ) . initialize @xmath237 choose @xmath238 at random and set @xmath239 and @xmath240 . choose the smallest number @xmath40 in @xmath234 . choose a random element @xmath18 in @xmath241 @xmath242 @xmath243 the while - loop in step ( 5 ) will be used at most @xmath116 times . a careful implementation with @xmath234 as an initial increasing array of numbers @xmath244 requires for step ( 6 ) and deleting @xmath40 in ( 9 ) , @xmath245 time , for step ( 7 ) @xmath245 time to choose @xmath18 @xcite and to delete it in ( 9 ) . this leads to @xmath113 time . we show that conditions 1 . ) and 2 . ) imply that @xmath181 is an induced cycle set . for this we give an interpretation of 2 . ) . a well - known classical connection using the principle of double - counting gives for the ferrers sequence @xmath170 , @xmath246 . let us now consider an arbitrary adjacency matrix @xmath129 for a directed graph with degree sequence @xmath228 condition 2 . ) with @xmath247 states that all @xmath3 s of the first @xmath248 columns of @xmath129 equal the number of the @xmath3 s in the first @xmath248 columns of its ferrers matrix . the same can be find for the first @xmath249 columns when considering @xmath250 in 2 . ) . for @xmath251 , @xmath252 , and @xmath250 it states that @xmath129 contains in the @xmath40th column one more @xmath3 than @xmath165 , in the @xmath253th column the same number , and in the @xmath249nd column one @xmath3 less . more formally , condition 2 . ) states for matrix @xmath129 a ) for each @xmath254 all @xmath255 @xmath3 s of row @xmath18 are contained in the first @xmath248 columns , b ) for each @xmath256 ( i.e. the induced cycle set ) we have in rows @xmath18 @xmath3 s in the first @xmath248 columns and a @xmath3 in one of the columns @xmath40 , @xmath253 and @xmath249 for rows @xmath40 , @xmath253 and @xmath253 . furthermore , it states c ) that all others @xmath257 , and that all corresponding rows possess @xmath3 s in their first @xmath249 columns with the exception of the diagonal element . + let us now assume 1 . ) and 2 . ) are fulfilled but @xmath181 is not an induced cycle set . then it exists a realisation @xmath46 with adjacency matrix @xmath80 such that @xmath181 is an induced cycle in @xmath46 which can be destroyed by a switch in either deleting an arc or adding an additional arc . we consider the case that @xmath258 can be switched . all other cases can be treated analogously . then there exists another switching arc @xmath259 such that @xmath260 are non - arcs , and all vertices @xmath261 are pairwise disjunctive . we have @xmath262 , otherwise we have a contradiction to b ) due to @xmath263 is a non - arc . if @xmath264 ( @xmath265 and @xmath266 not possible due to assumption ) than we have in row @xmath48 of @xmath80 a @xmath3 in column @xmath249 and a @xmath167 in column @xmath253 which contradicts c ) or a ) since @xmath267 it remains @xmath268 since @xmath269 is no arc in @xmath80 , we have in row @xmath48 a @xmath167 in column @xmath253 and a @xmath3 in column @xmath270 which contradicts a ) or c ) . hence , @xmath46 can not exist which means that an induced cycle of vertices @xmath40 , @xmath253 and @xmath249 can never be destroyed by a switch . the fist part of this lemma was already given in section [ sec_modified_curveball ] . we proof the formula for @xmath271 by induction . for @xmath272 and @xmath273 we have only one partition and @xmath274 in both cases . for @xmath275 we have in a @xmath88-partition @xmath276 three possibilities to choose @xmath153 , and @xmath151 and @xmath152 are forced by this choice . this results in @xmath89 possible @xmath88-partitions , and indeed @xmath277 for @xmath278 in a @xmath88-partition @xmath279 , @xmath151 can be fixed as @xmath280 because the number @xmath3 must be in one pair . then we have @xmath89 possibilities to choose @xmath152 , and after this choice @xmath153 and @xmath281 are settled . hence , the number of @xmath88-partitions is @xmath89 , which corresponds to @xmath277 let us assume that the claim is true for all @xmath282 for a given @xmath204 in an even partition we can choose @xmath283 , because @xmath108 has to be in one of these pairs . for @xmath152 we have @xmath284 possible numbers @xmath285 from @xmath3 to @xmath284 . we delete @xmath108 and @xmath285 in @xmath234 and get @xmath284 subsets @xmath286 for @xmath287 for each @xmath288 we can apply the induction hypothesis . each of the @xmath284 @xmath289 can be combined with @xmath290 leading to a partition of @xmath234 . hence , we get for @xmath234 , @xmath291 . in the odd case we first need to choose an element @xmath292 randomly , and then we apply for the remaining even set @xmath293 the formula for the even case .

detached , double - lined eclipsing binaries provide a largely model - independent means to precisely and accurately measure fundamental stellar properties , particularly masses and radii . in the best - observed systems the precision of these can be at the @xmath7 per cent level , and thus place stringent constraints on stellar evolution models ( e.g. @xcite ) . despite this , as far as we are aware , there are at present only four known systems with one or more components between @xmath8 and the hydrogen burning limit on the main sequence ( i.e. old , field stars ) : cm dra @xcite , cu cnc b @xcite , lp 133 - 373 @xcite , and sdss j031824 - 010018 @xcite . although jw 380 @xcite , nsvs 02502726b @xcite , and the ngc 1647 system of @xcite also satisfy this mass criterion , these objects are still contracting on the pre main - sequence . of the known systems , the only two with parameters determined to better than @xmath9 per cent are cm dra and cu cnc . for sdss j031824 - 010018 , the knowledge of the parameters is limited largely by radial velocity errors , since this system is extremely faint ( sdss @xmath10 ) and has an extremely short orbital period ( @xmath11 ) , meaning long integrations can not be used to obtain better signal - to - noise . hence there is little possibility for substantial improvement in the parameters in the near future . it is clear that , in order to better - constrain the stellar mass - radius relation on the main sequence , more bright , low - mass eclipsing binaries are needed to yield extremely precise masses and radii . furthermore , observations of these systems have indicated significant discrepancies with the stellar models . this is particularly the case below @xmath12 , and the components of cm dra ( the lowest - mass system with @xmath13 observational errors ) have radii @xmath14 per cent larger than the theoretical predictions from state - of - the - art stellar evolution models . it has been suggested ( e.g. @xcite ) that the reason for this discrepancy may be that close binaries are not actually representative of single - stars at this level of precision . the effect of the close companion and tidal locking is likely to significantly increase activity levels in close binaries , and these authors suggest that it could be this effect that is responsible for the inflated radii of cm dra . @xcite examined the available sample of single - star and eclipsing binary measurements to search for correlations of radius with activity levels and metallicity , finding that such a correlation of activity with radius does appear to exist for members of close binary stars . this result was based on a small number of measurements for late m - dwarfs , and the conclusions would be strengthened by the availability of additional precise estimates of the parameters for binaries with a range of activity levels and orbital periods , to explore the available parameter space . we present the discovery of a new bright , low - mass eclipsing binary system , gj 3236 . the components show high activity levels , as evidenced by h@xmath6 emission lines in the spectra of both stars , and show rotational modulations in the light curves that are synchronized with the binary orbital period of @xmath15 . this system thus has the potential to yield an additional precise test of stellar evolution models for high activity and short orbital periods . eclipses in gj 3236 were detected during 2008 january in the first two weeks of routine operations of the newly - commissioned mearth observatory , a system designed primarily to search for transiting super - earth exoplanets orbiting around the nearest @xmath16 mid to late m - dwarfs in the northern hemisphere ( @xcite ; @xcite ) . exposure times on each field observed by mearth are tailored to achieve sensitivity to a particular planet size for the assumed stellar parameters of the target star , and were @xmath17 for gj 3236 . after the initial detection of the eclipses , we switched to a follow - up mode , observing at the highest possible cadence ( i.e. continuously , resulting in a cadence of approximately @xmath18 including overheads ) during eclipse and for an additional @xmath19 window surrounding the eclipse , and at the normal cadence for the mearth survey of approximately @xmath20 for the remainder of the time . in this way , we sample as well as possible the rapid flux decrement during eclipse , and the out of eclipse portions of the light curve . we note that sampling the latter is of crucial importance for obtaining accurate parameters from light curve analyses of binary systems showing out of eclipse modulations , such as the present one . we obtained data both in the initial season after the eclipses were first detected ( 2008 january to may , inclusive ) , and following the summer monsoon ( 2008 october onwards ) . during this time , there were a number of software improvements ( including the ability to update the telescope pointing after taking each science image , resulting in substantially improved light curves due to minimizing the drift across the detector , and hence the effect of flat fielding errors , fringing , etc . ) and our filter system was changed from cousins @xmath21 to a custom long - pass filter with transmission from @xmath22 , and limited at the long wavelength end by the tail of the ccd quantum efficiency curve ( see @xcite ) . we have elected to use only the latter data - set for modeling , despite this having the disadvantage of being observed in a non - standard bandpass , because it is approximately contemporaneous with our @xmath23-band follow - up photometry . we see evidence for evolution of the amplitude and phase of the out of eclipse modulation between the two observing seasons , so this simplifies the analysis by allowing the use of a single set of spot parameters to describe both bandpasses . a total of @xmath24 observations were obtained between ut 2008 october 2 and ut 2008 december 11 , including 6 primary and 3 secondary eclipses sampled at high cadence . data were reduced using the standard mearth reduction pipeline , which is at present identical to the monitor project pipeline described in @xcite . we used a smaller aperture radius of @xmath25 ( @xmath26 ) than the usual value of @xmath27 adopted for bright stars , due to the presence of a nearby star approximately @xmath28 from gj 3236 at the epoch of our observations , and @xmath29 fainter in the mearth bandpass . this star does not share a common proper motion with gj 3236 and is therefore not physically associated . reducing the aperture size in this way reduces the flux contributed by this star in the light curve of gj 3236 to a negligible level ( we estimate a @xmath30 error in the measured fluxes , and thus also in the eclipse depths ) . the high proper motion of the gj 3236 system allows us to constrain the contribution of any additional background stars in the photometric aperture that are not co - moving with gj 3236 itself , using previous epochs of imaging . we show in figure [ third_light_images ] a series of three images centered on the position of the photometric aperture used for gj 3236 in the mearth images , demonstrating that there are no such objects . if any additional third light is indeed present , it must therefore be co - moving with gj 3236 , meaning it is highly likely to be physically associated , given the extremely small probability of such a chance alignment . ( radius ) photometric aperture used to derive our light curves . data are from the first and second epoch palomar sky surveys as provided by the digitized sky survey ( top and center panels ) , and the mearth stacked master image ( bottom panel ) . the approximate epochs of the images are @xmath31 ( poss-1 ) , @xmath32 ( poss-2 ) , and @xmath33 ( mearth ) . all three panels have the same center , scale and alignment on - sky , with north up and east to the left , covering @xmath34 in the horizontal direction.,width=268 ] the photometry was calibrated using observations of standard star fields from @xcite taken automatically each night by the mearth observation scheduling software . by fitting all the standard star observations from several photometric nights , we derived the following color equation to convert between the mearth bandpass and the standard johnson - cousins system : @xmath35 the full mearth light curve is reproduced in table [ mearth_lc_table ] . lrrrrrrr @xmath36 & @xmath37 & @xmath38 & @xmath39 & @xmath40 & @xmath41 & @xmath42 & @xmath43 + @xmath44 & @xmath45 & @xmath46 & @xmath47 & @xmath48 & @xmath49 & @xmath50 & @xmath51 + @xmath52 & @xmath53 & @xmath46 & @xmath54 & @xmath55 & @xmath56 & @xmath57 & @xmath58 + @xmath59 & @xmath60 & @xmath46 & @xmath61 & @xmath62 & @xmath63 & @xmath64 & @xmath65 + @xmath66 & @xmath67 & @xmath46 & @xmath68 & @xmath69 & @xmath70 & @xmath71 & @xmath72 + observations centered around the primary eclipse of ut 2008 september 25 and the secondary eclipse of ut 2008 september 27 were obtained using the keplercam instrument on the flwo @xmath73 telescope . we used the standard binning @xmath74 readout mode , since the plate scale of @xmath75 per unbinned pixel significantly oversamples the typical seeing at flwo . the resulting plate scale was @xmath76 per summed pixel . we used the @xmath23 filter and an exposure time of @xmath77 . we also used observations of standard star fields from @xcite taken on ut 2008 september 28 to calibrate the keplercam photometry onto the standard johnson - cousins system . a total of @xmath78 observations were taken , where possible starting @xmath79 before first contact and finishing @xmath79 after last contact to sample the out - of - eclipse portions of the light curve and thus allow the eclipse depths to be properly measured . the end of the primary eclipse was missed due to twilight . these photometric data were reduced using the same pipeline as described in [ mearth_obs_sect ] . we used an aperture radius of @xmath80 binned pixels , corresponding to @xmath81 on - sky ( recalling from [ mearth_obs_sect ] that the size of this aperture is important due to the nearby star ) . the difference in magnitude between gj 3236 and the fainter star is only @xmath82 in @xmath23-band , which leads to approximately a @xmath83 error in the measured fluxes of gj 3236 , and thus a similar error in the measured eclipse depths . since the point spread functions on our images are not very well - behaved it will be challenging to reduce this using psf - fitting photometry , and it is still substantially smaller than the other sources of error in the present light curve models , so we have not pursued this issue further . the full @xmath23-band light curve is reproduced in table [ v_lc_table ] . lrrrrrrr @xmath84 & @xmath85 & @xmath86 & @xmath87 & @xmath88 & @xmath89 & @xmath90 & @xmath91 + @xmath92 & @xmath93 & @xmath94 & @xmath95 & @xmath96 & @xmath97 & @xmath98 & @xmath99 + @xmath100 & @xmath101 & @xmath94 & @xmath102 & @xmath103 & @xmath104 & @xmath105 & @xmath106 + @xmath107 & @xmath108 & @xmath94 & @xmath109 & @xmath110 & @xmath111 & @xmath90 & @xmath112 + @xmath113 & @xmath114 & @xmath94 & @xmath115 & @xmath116 & @xmath117 & @xmath118 & @xmath119 + observations in the near - infrared @xmath120-band were obtained using the robotic peters automated infrared imaging telescope ( pairitel ) from ut 2008 february 17 to ut 2008 march 3 ( inclusive ) . we scheduled observations around the times of primary and secondary eclipse , obtaining @xmath121 data points spread over @xmath122 . exposure times were @xmath123 . the observations were automatically scheduled , collected , and reduced by the fully robotic pairitel observing system @xcite . we then produced differential photometry using a set of comparison stars chosen from the 2mass catalog . we estimate a photometric error of approximately @xmath124 per data point from the scatter of the out - of - eclipse measurements . the full @xmath120-band light curve is reproduced in table [ j_lc_table ] . lrrrrrrr @xmath125 & @xmath126 & @xmath127 + @xmath128 & @xmath129 & @xmath127 + @xmath130 & @xmath131 & @xmath127 + @xmath132 & @xmath133 & @xmath127 + @xmath134 & @xmath135 & @xmath127 + spectroscopic observations were obtained using the tres fiber - fed chelle spectrograph on the flwo @xmath136 tillinghast reflector . we used the medium fiber ( @xmath137 projected diameter ) throughout , yielding a resolving power of @xmath138 . table [ spec_obs_table ] summarizes these measurements , including the radial velocities of both components of the binary derived therefrom . the tres instrument is extremely stable , so we acquired thar wavelength calibration exposures before or after the target exposures , rather than using the simultaneous calibration fiber , which can lead to contamination of the target spectrum in the red - most orders from the very strong ar lines in the thar lamp spectrum . a second fiber was placed on sky , but this was not used due to a problem causing extremely poor throughput in the tres sky fiber , which has since been resolved . lrrrrrr @xmath139 & & & @xmath140 & @xmath141 & @xmath142 & @xmath143 + @xmath144 & & & @xmath145 & @xmath146 & @xmath147 & @xmath148 + @xmath149 & & & @xmath150 & @xmath151 & @xmath152 & @xmath153 + @xmath154 & @xmath155 & @xmath156 & @xmath157 & @xmath158 & @xmath159 & @xmath160 + @xmath161 & @xmath162 & @xmath163 & @xmath164 & @xmath165 & @xmath166 & @xmath167 + @xmath168 & @xmath169 & @xmath170 & @xmath171 & @xmath172 & @xmath173 & @xmath174 + @xmath175 & @xmath176 & @xmath177 & @xmath178 & @xmath179 & @xmath180 & @xmath181 + @xmath182 & & & @xmath183 & @xmath184 & @xmath185 & @xmath186 + @xmath187 & & & @xmath188 & @xmath189 & @xmath190 & @xmath191 + @xmath192 & & & @xmath193 & @xmath194 & @xmath195 & @xmath196 + @xmath197 & & & @xmath198 & @xmath199 & @xmath200 & @xmath201 + @xmath202 & @xmath203 & @xmath204 & @xmath205 & @xmath206 & @xmath207 & @xmath208 + @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath213 & @xmath214 & @xmath215 + @xmath216 & @xmath217 & @xmath218 & @xmath219 & @xmath220 & @xmath221 & @xmath222 + @xmath223 & @xmath224 & @xmath225 & @xmath226 & @xmath227 & @xmath228 & @xmath229 + @xmath230 & & & @xmath231 & @xmath232 & @xmath233 & @xmath234 + @xmath235 & & & @xmath236 & @xmath237 & @xmath238 & @xmath239 + @xmath240 & & & @xmath241 & @xmath242 & @xmath243 & @xmath244 + @xmath245 & @xmath246 & @xmath247 & @xmath248 & @xmath249 & @xmath250 & @xmath251 + @xmath252 & @xmath253 & @xmath254 & @xmath255 & @xmath256 & @xmath257 & @xmath258 + @xmath259 & @xmath260 & @xmath261 & @xmath262 & @xmath263 & @xmath264 & @xmath265 + @xmath266 & @xmath267 & @xmath268 & @xmath269 & @xmath270 & @xmath271 & @xmath272 + @xmath273 & @xmath274 & @xmath275 & @xmath276 & @xmath277 & @xmath278 & @xmath279 + @xmath280 & @xmath281 & @xmath282 & @xmath283 & @xmath284 & @xmath285 & @xmath286 + @xmath287 & @xmath288 & @xmath289 & @xmath290 & @xmath291 & @xmath292 & @xmath293 + @xmath294 & @xmath295 & @xmath296 & @xmath297 & @xmath298 & @xmath299 & @xmath300 + @xmath301 & @xmath302 & @xmath303 & @xmath304 & @xmath305 & @xmath306 & @xmath307 + @xmath308 & @xmath309 & @xmath310 & @xmath311 & @xmath312 & @xmath313 & @xmath314 + @xmath315 & @xmath316 & @xmath317 & @xmath318 & @xmath319 & @xmath320 & @xmath321 + @xmath322 & @xmath323 & @xmath324 & @xmath325 & @xmath326 & @xmath327 & @xmath328 + @xmath329 & @xmath330 & @xmath331 & @xmath332 & @xmath333 & @xmath334 & @xmath335 + @xmath336 & & & @xmath337 & @xmath338 & @xmath339 & @xmath340 + spectra were reduced using standard procedures in iraf @xcite from the echelle package to extract the spectra from the target fiber and sky fiber simultaneously . the tres ccd is read out using two amplifiers , which necessitates combining them before tracing and extraction . this was done using the mscred package in iraf , applying a multiplicative gain correction to equalize the difference in gain between the two readout electronics chains . we then divided by a normalized flat field exposure to correct for fringing in the red - most regions of the spectrum , which has a significant effect on the orders we used for the radial velocity analysis . at the time of our observations , the tres detector suffered from a very high energetic particle hit rate , the source of which is under investigation . the particle hits were removed by median - combining each set of multiple exposures for the flat fields , target and calibrations , and then using the statistics of the median frame to apply an upper envelope clip to the target frames themselves . this allowed us to avoid having to combine our target exposures to eliminate the particle hits , which would smear out the radial velocity variations slightly as a result of the extremely short orbital period of gj 3236 . the stability of the tres instrument allowed us to define the aperture trace for each fiber using the high signal - to - noise flat fields , which were then used to extract the target and calibration spectra , employing the `` optimal '' ( in a least squares sense ) weighting scheme of @xcite . the spectra were wavelength - calibrated using the standard ecidentify procedure , which employs the ( known ) dispersion relation of the chelle to solve for a wavelength solution over all orders of the spectrum simultaneously . this is necessary due to the relative paucity of sufficiently intense lines from the thar lamp in each order of the spectrum at wavelengths @xmath341 . table [ photparams ] summarizes the known system properties , from the literature ( principally the proper motion survey of @xcite used to select the target stars for the mearth survey ) and our own @xmath23 and mearth photometry , converted to the standard johnson - cousins system . lr @xmath342 & @xmath343 + @xmath344 & @xmath345 + @xmath346 & @xmath347 + @xmath348 & @xmath349 + + mearth & @xmath350 + + @xmath351 & @xmath352 + @xmath353 & @xmath354 + + @xmath355 & @xmath356 + @xmath357 & @xmath358 + @xmath359 & @xmath360 + the average color of @xmath361 indicates an average spectral type of approximately m4 using the color to spectral type relation of @xcite for young galactic disk stars . for detached eclipsing binaries with circular orbits , the radial velocity ( rv ) and light curve models are largely independent . we therefore carried out a preliminary analysis of the available light curves before starting to obtain spectroscopic observations . the principal purpose of doing this was to determine an extremely precise orbital period to better - target the radial velocity observations , but we can also constrain the system eccentricity by using the phase of the secondary eclipses relative to the primary eclipses ( related to @xmath362 ) . a simple geometric argument can then be used to obtain a limit on the eccentricity itself , assuming no _ a priori _ knowledge of @xmath363 . we used jktebop ( @xcite ; @xcite ) , a modified version of ebop ( eclipsing binary orbit program ; @xcite ; @xcite ) , to obtain these parameters by fitting the mearth light curve . the program was modified to fit simultaneously for the eb model , and a simple form for the synchronized out - of - eclipse modulations , assuming they can be approximated by a truncated fourier series : @xmath364 where @xmath365 , and @xmath366 is the orbital period of the binary . the normalization term is omitted since this is already taken into account by the standard ebop model . the revised code yields a fit with very small residuals for the present system , and the use of the form in eq . ( [ ooe_eq ] ) keeps the number of parameters required to describe the modulations to a minimum compared to a full spot model . doing so will not necessarily reproduce the correct eclipse depths for a given set of physical parameters ( or vice versa ) , but should be adequate for deriving an accurate system ephemeris . we relax this assumption in [ wd_sect ] , where we adopt a full spot model for determining the geometric and radiative parameters of the system . the following parameters were allowed to vary in the fit : the sum of the radii divided by the semimajor axis , @xmath367 ( where @xmath368 , @xmath369 is the stellar radius of star @xmath370 and @xmath371 is the semimajor axis ) , orbital inclination @xmath372 , @xmath362 , central surface brightness ratio @xmath120 ( essentially equal to the ratio of the primary and secondary eclipse depths ) , orbital period @xmath366 , orbital phase zero - point @xmath373 , light curve normalization , and the parameters @xmath374 , @xmath375 , @xmath376 and @xmath377 describing the out of eclipse modulations . single - band light curves showing grazing eclipses constrain the ratio of the radii only very weakly ( or equivalently , the luminosity ratio ) , so these parameters and the mass ratio ( used for computing the oblateness of the stars and the reflection effect ) were fixed at @xmath378 for the initial analysis . the results for the orbital parameters were found not to change significantly upon revising these to the measured values from the spectroscopy once they were available . our light curves are not of sufficient quality to fit for the stellar limb darkening , so we assumed a square root limb - darkening law with coefficients appropriate to the @xmath21 ( cousins ) passband from @xcite using the phoenix model atmospheres . given the lack of a spectroscopic constraint on the effective temperature , surface gravity and metallicity for either star , we have instead assumed solar metallicity , and derived temperatures and surface gravities of @xmath379 and @xmath380 , and for the secondary , @xmath381 and @xmath382 by iterating the modeling process . the assumed temperatures were derived using the masses from the combined orbital and light curve solution , interpolating between the compilation of values for field stars by @xcite . see also [ kin_sect ] where these temperatures are verified using the final model . a gravity darkening exponent of @xmath383 , a value appropriate for stars with convective envelopes @xcite , was also assumed , and the option in jktebop to calculate the reflection effect was used rather than fitting for it . we report only the orbital parameters ( @xmath366 , @xmath373 and @xmath362 ) from the ebop analysis , and have adopted these for the remainder of this work . the parameter uncertainties were derived using a monte carlo algorithm built - in to jktebop @xcite . briefly , this algorithm uses the best fit to generate a synthetic light curve , injecting gaussian noise with amplitude determined by the observational errors ( which were scaled such that the reduced @xmath384 of the fit was equal to unity ) to produce a simulated light curve , which is then fit to determine a new set of parameters . the parameter uncertainties can be estimated using the distribution of the parameters from a large number of realizations ( here , @xmath385 ) of this process . see also @xcite for more details . times of minimum light derived for the @xmath386 well - sampled eclipses in our @xmath23 and mearth observations are reported in table [ minlight ] . these were estimated using the method of @xcite over a @xmath387 region in normalized orbital phase around each eclipse . we subtracted the truncated fourier series in eq . ( [ ooe_eq ] ) from the light curves before applying this analysis to reduce the effect of any systematic errors arising from the out - of - eclipse modulation . ll @xmath366 & @xmath388 + @xmath373 & @xmath389 + + @xmath362 & @xmath390 + @xmath391 & @xmath392 ( 95% cl ) + & @xmath393 ( 99% cl ) + + @xmath6 & @xmath394 + + + + @xmath395 & @xmath396 + @xmath397 & @xmath398 + @xmath399 & @xmath400 + + @xmath401 & @xmath402 + @xmath403 & @xmath404 + @xmath405 & @xmath406 + @xmath407 & @xmath408 + + @xmath409 & @xmath410 + @xmath411 & @xmath412 + + + + @xmath395 & @xmath413 + @xmath397 & @xmath414 + @xmath399 & @xmath415 + + @xmath401 & @xmath416 + @xmath403 & @xmath417 + @xmath405 & @xmath418 + @xmath407 & @xmath419 + + @xmath409 & @xmath420 + @xmath411 & @xmath421 + + + + @xmath395 & @xmath422 + @xmath397 & @xmath423 + @xmath399 & @xmath424 + + @xmath401 & @xmath425 + @xmath403 & @xmath426 + @xmath405 & @xmath427 + @xmath407 & @xmath428 + + @xmath409 & @xmath420 + @xmath411 & @xmath429 + lrrrl @xmath430 & @xmath431 & @xmath432 & @xmath23 + @xmath433 & @xmath434 & @xmath435 & @xmath23 + @xmath436 & @xmath437 & @xmath438 & mearth + @xmath439 & @xmath440 & @xmath441 & mearth + @xmath442 & @xmath443 & @xmath444 & mearth + @xmath445 & @xmath446 & @xmath447 & mearth + @xmath448 & @xmath449 & @xmath450 & mearth + @xmath451 & @xmath452 & @xmath453 & mearth + @xmath454 & @xmath455 & @xmath456 & mearth + @xmath457 & @xmath458 & @xmath459 & mearth + @xmath460 & @xmath461 & @xmath462 & mearth + radial velocities were obtained using the two - dimensional cross - correlation algorithm todcor @xcite , which uses templates matched to each component of a spectroscopic binary to simultaneously derive the velocities of both stars , and importantly for the present application , an estimate of their luminosity ratio ( @xmath6 ) in the spectral bandpass . we performed both a standard absorption line based cross - correlation analysis , and an emission line analysis using the strong h@xmath6 emission observed in both components of gj 3236 . the signal to noise ratio of the latter is substantially higher in our data , but we caution that the source of the h@xmath6 emission in our target stars is not well - understood . the emission line analysis results could suffer systematic errors depending on the relative velocities of the regions of the chromosphere emitting the h@xmath6 and the stellar photospheres , and this is exacerbated in the present system because it is highly likely that the stellar spin and binary orbit are synchronized , both from the light curves and expectations from tidal theory @xcite . we therefore conservatively adopt the results from the absorption line analysis despite the larger errors . the absorption line analysis used as a template spectrum a single epoch observation of barnard s star ( @xcite ; also known as gj 699 ) taken on ut 2008 october 20 . we used a wavelength range of 8700 to 8850 in order 50 of the spectrum to derive the velocities , since this region contains a number of reasonably strong metallic lines in m - dwarfs , and is free of telluric absorption lines . the ca ii infra - red triplet at 8498 , 8542 , and 8662 was deliberately avoided since the absorption in these lines is saturated , and they are therefore intrinsically very broad , and emission cores are often observed in these lines in active stars , which would severely complicate the use of an inactive field star as a template . we assumed a barycentric radial velocity of @xmath463 for barnard s star , derived from presently unpublished cfa digital speedometer measurements spanning @xmath464 . for the emission line analysis , we used as templates spectra of two m4 dwarfs that were found to display h@xmath6 emission , gj 856a and gj 3379 , observed as part of another program using the same instrument . gj 856a was observed on ut 2008 october 17 and gj 3379 was observed on ut 2008 october 19 . we used a wavelength range of 6520 to 6645 in order 37 for the emission line cross - correlations , noting that the cross - correlations are dominated by the h@xmath6 lines . barycentric radial velocities of gj 856a and gj 3379 were determined by cross - correlation with barnard s star , using the 8700 to 8850 region as for our absorption line analysis of gj 3236 . given that todcor provides a measure of the correlation between the observed and template spectra , it is possible to run cross - correlations against a series of templates and use the correlation value as an indicator of the template match . both the effective temperature and the rotational velocity of the templates affect the solution , so it is vital to explore this parameter space . due to the availability of a limited range of observed templates , the effective temperature could not be varied , but we were able to rotationally broaden the template spectra in order to create primary and secondary templates with @xmath465 . we used linear limb darkening coefficients @xmath466 and @xmath467 from @xcite for the sdss @xmath468 passband ( a close match to the effective wavelength of the spectroscopy ) , which correspond to our assumed temperatures for the primary and secondary ( see [ ebop_lc_sect ] ) . we ran todcor on the grid of @xmath469 template combinations , and sought the maximum of the resulting correlation values to determine the most likely @xmath470 values for the two components of the binary system . barnard s star is a very slow rotator , so the rotation of the template is negligible given our spectral resolution . our analysis indicates @xmath471 , which is consistent with the expected value of @xmath472 if the spin and orbit are synchronized . @xmath473 was poorly - constrained , with the maximum correlation occurring at @xmath474 , whereas we would expect @xmath475 assuming synchronization . we have simply adopted the latter value , since the light curves and primary velocity indicate that the system is indeed synchronized . using the templates with @xmath471 and @xmath476 , todcor indicates @xmath477 . for the emission line analysis , we assumed the stellar rotation values from the absorption line analysis , rotationally broadening the observed spectra of gj 856a and gj 3379 . the intrinsic rotational broadening in both of these templates is again negligible for our purposes . todcor gave a maximum correlation for @xmath478 , which provides a measure of the relative h@xmath6 emission luminosities of the two stars . since the orbital period and time of primary eclipse are extremely well - determined from the light curves , we held these values fixed in the final radial velocity orbital solution . we also assumed a circular orbit . figure [ rv_fit ] shows the resulting radial velocity curves . the parameters derived from the radial velocity analysis are given in table [ orbparams ] . we assumed equal observational errors on each radial velocity point , scaled such that the reduced @xmath384 of the fit was equal to unity . radial velocities of the components of double - lined spectroscopic binaries are prone to a `` peak pulling '' effect as they approach the @xmath395 velocity ( at times of minimum separation of the two components in the spectrum ) . the todcor method can largely eliminate this , provided the template spectra match the target sufficiently well . in our experience , this effect is particularly sensitive to errors in the assumed rotational broadening : if the template contains too much or too little rotation , the velocities would be pushed toward or away from @xmath395 to compensate . examining the residuals of our solution in figure [ rv_fit ] indicates little or no `` peak pulling '' , which further justifies our assumed rotational velocities and the assumption of negligible rotation in the templates . in order to properly account for spots on the components of gj 3236 , we use the most recent version ( wd2007 ) of the standard @xcite code to derive the geometric and radiative parameters of the system . this code incorporates a full physical model including spots and using model atmospheres . wd normally uses the @xcite model atmospheres to calculate the emergent flux from surface elements on each binary component . these models span an effective temperature range of @xmath479 , and for temperatures below @xmath480 , wd interpolates between a black body of the appropriate temperature and the @xmath480 model atmosphere . this is weighted in such a fashion as to produce a pure black body at @xmath481 , and for intermediate temperatures there is a smooth ramp toward a pure model atmosphere at @xmath480 . molecular sources of opacity increasingly begin to dominate the optical spectra of m - dwarfs in this temperature range , so a black body is clearly a poor approximation for the emergent flux . since both components of gj 3236 appear to be cooler than @xmath480 , we have modified our copy of wd to substitute phoenix nextgen model atmospheres @xcite in place of these interpolations for effective temperatures cooler than @xmath480 . our assumptions for the wd light curve model are detailed in table [ wd_pars ] . we have assumed zero orbital eccentricity , synchronized spin and orbit for both components of the binary , and no third light , which are justified in [ mearth_obs_sect ] and [ ebop_lc_sect ] . llll & detached + + @xmath373 & hjd0 & @xmath482 ( hjd ) & phase zero - point + @xmath366 & period & @xmath483 & orbital period + @xmath484 & dpdt & @xmath485 & first derivative of period + + @xmath403 & sma @xmath486 & @xmath404 & projected semimajor axis + @xmath401 & rm & @xmath402 & mass ratio + @xmath395 & vga & @xmath396 & systemic velocity + + @xmath391 & e & @xmath485 & orbital eccentricity + @xmath487 & f1 & @xmath378 & primary rotation parameter + @xmath488 & f2 & @xmath378 & secondary rotation parameter + + & hla & varied & light curve normalization + + @xmath489 & tavh & @xmath490 & primary effective temperature + @xmath491 & tavc & varied & secondary effective temperature + @xmath492 & phsv & varied & primary surface potential + @xmath493 & pcsv & varied & secondary surface potential + + @xmath372 & incl & varied & orbital inclination + @xmath494 & el3 & @xmath485 & third light + + @xmath495 & alb1 & @xmath496 & primary bolometric albedo + @xmath497 & alb2 & @xmath496 & secondary bolometric albedo + @xmath498 & gr1 & @xmath499 & primary gravity darkening exponent + @xmath500 & gr2 & @xmath499 & secondary gravity darkening exponent + + @xmath501 & xlat1 & @xmath502 & spot 1 latitude + @xmath503 & xlong1 & varied & spot 1 longitude + @xmath504 & radsp1 & varied & spot 1 radius + @xmath505 & temsp1 & 0.96 & spot 1 temperature contrast + + @xmath506 & xlat2 & @xmath502 & spot 2 latitude + @xmath507 & xlong2 & varied & spot 2 longitude + @xmath508 & radsp2 & varied & spot 2 radius + @xmath509 & temsp2 & 0.96 & spot 2 temperature contrast + + @xmath6 & & @xmath394 & light ratio + the limited phase coverage and lack of repeat observations in our @xmath23-band photometry ( leading to an uncertainty in any corrections for color - dependent `` second - order '' atmospheric extinction ) , and poor precision ( large scatter ) in the @xmath120-band photometry means in practice that these are not useful in aiding the determination of the geometric ( and thus physical ) parameters of the system . our experiments with the wd model indicate that these provide minimal additional constraints on any of the parameters when compared to using the mearth data alone . we have therefore elected to use only the mearth data for fitting for the system parameters , and then evaluated this model with respect to the other passbands as a check . doing so necessitates fixing one of the component effective temperatures , since the relative eclipse depths measured in a single passband merely determine the relative temperatures of the two components . we have therefore assumed @xmath379 following [ ebop_lc_sect ] . in practice even given multi - band photometry it is usually necessary to fix this parameter : although the color - dependence of the eclipse depths does in principle provide information on the value of @xmath489 , this is normally poorly constrained due to the difficulty of measuring differential colors to high precision . in order to reproduce the out - of - eclipse variations seen in our optical light curves , at least two spots must be introduced into the model . the configuration of these spots affects the measured eclipse depths , depending on whether a spotted or unspotted part of the photosphere is eclipsed . this predominantly affects the derived ratio of surface brightnesses as measured by the ratio of depths of the primary and secondary eclipses ( and thus the ratios of the component radii and effective temperatures inferred from this ) , and a combination of @xmath510 and the orbital inclination inferred from the total eclipse depth . since our @xmath23-band data do not cover the portions of the light curve out of eclipse , and the scatter in the @xmath120-band data is too large to usefully constrain the model , the spot modeling must be done using single - band light curve information . spots are introduced in the wd code using a standard four - parameter model , where each spot is parametrized using the spot latitude @xmath511 and longitude @xmath512 , radius @xmath513 and temperature contrast @xmath514 . the latter quantity gives the ratio of effective temperatures of the spotted and unspotted photosphere . for a single - band light curve , the spot longitude is well - constrained by the phase of the out of eclipse modulation , but the combination of @xmath511 , @xmath513 and @xmath514 is largely degenerate ( these quantities determine the amplitude and shape of the variation ) . moreover , we can place each spot on either star . although the data do rule out some of the possible spot locations , there are still a range of possibilities , which introduces a systematic error into the geometric parameters of the system . in principle , measuring the color - dependence of the out - of - eclipse modulation would allow this degeneracy to be reduced . given the presently available data , we instead probe this systematic error by considering three characteristic spot configurations that are consistent with the light curve . these are : ( 1 ) one close to polar spot on each star , with @xmath515 , with both spots located in the same hemisphere as the chord traversed during the eclipse , resulting in the spot on the primary being eclipsed by the secondary during primary eclipse , and vice versa during secondary eclipse ; ( 2 ) likewise , only with the spots located in the opposite hemisphere , resulting in no spots being eclipsed ; and ( 3 ) a case intermediate between these , with a `` symmetric '' spot configuration of two spots on each star , at latitudes of @xmath502 with one located in each hemisphere . in each model , the spot latitudes and temperature ratios were held fixed , allowing only longitude and radius to vary , and in configuration ( 3 ) we enforced symmetry of the spot pattern about the equator , i.e. both spots on one star were forced to have the same longitudes and radii . this means that all three models have the same number of parameters ( @xmath516 ) . to derive robust error estimates , including the correlations between parameters , we have implemented a markov chain monte carlo ( mcmc ) algorithm around the wd light curve model . following @xcite , we used the metropolis - hastings algorithm @xcite to estimate the _ a posteriori _ joint probability distribution of the fitted model parameters . the fitting statistic was the traditional @xmath517 , where the observational error estimates were scaled such that each light curve s reduced @xmath384 was equal to unity . we briefly summarize the metropolis - hastings algorithm here . starting from an initial point in parameter space , the algorithm takes the most recent set of parameters and perturbs one or more parameters by a random gaussian deviate . if the perturbed parameter set has a lower @xmath384 than its progenitor , it is accepted as a new point in the chain . if it has a larger @xmath384 , it is accepted with a probability @xmath518 . if it is not accepted , the original point is repeated in the chain . the size of the perturbations were adjusted so that @xmath519 of the proposed points were accepted . we included the spectroscopic light ratio between the primary and secondary components as an observation in our @xmath384 statistic , which in practice was implemented by adding an extra contribution to @xmath384 of the form : @xmath520^{2}\ ] ] where @xmath521 and @xmath522 are the observed light ratio and its error , taken from table [ wd_pars ] , and @xmath523 is the value calculated from the wd model . this is necessary for systems with near - circular orbits and grazing eclipses because the light curves only very weakly constrain this quantity ( or equivalently , the ratio of the component radii ) . uninformative ( uniform ) priors were assumed on all other parameters . the final parameters and error estimates were determined by taking the @xmath524 confidence interval ( @xmath525 and @xmath526 percentiles , corresponding to @xmath527 standard deviation for a gaussian distribution ) of the samples generated by our mcmc procedure . the chains were run until they contained at least @xmath528 samples , and we discarded the first @xmath529 of the points in each chain in order to ensure they had converged . these parameters are reported in table [ params ] , and figure [ lc_model ] shows the light curves with the best - fitting model overplotted . llll & @xmath530 & @xmath531 & @xmath532 + @xmath533 & @xmath534 & @xmath535 & @xmath536 + @xmath492 & @xmath537 & @xmath538 & @xmath539 + @xmath493 & @xmath540 & @xmath541 & @xmath542 + @xmath372 & @xmath543 & @xmath544 & @xmath545 + + @xmath503 & @xmath546 & @xmath547 & @xmath548 + @xmath504 & @xmath549 & @xmath550 & @xmath551 + @xmath507 & @xmath552 & @xmath553 & @xmath554 + @xmath508 & @xmath555 & @xmath556 & @xmath557 + + @xmath510 & @xmath558 & @xmath559 & @xmath560 + @xmath561 & @xmath562 & @xmath563 & @xmath564 + + @xmath565 & @xmath566 & @xmath567 & @xmath568 + @xmath569 & @xmath570 & @xmath571 & @xmath572 + @xmath573 & @xmath574 & @xmath575 & @xmath576 + @xmath577 & @xmath578 & @xmath579 & @xmath580 + + @xmath371 & @xmath581 & @xmath582 & @xmath583 + @xmath584 & @xmath585 & @xmath586 & @xmath587 + @xmath588 & @xmath589 & @xmath590 & @xmath591 + @xmath592 & @xmath593 & @xmath594 & @xmath595 + @xmath596 & @xmath597 & @xmath598 & @xmath599 + + @xmath600 & @xmath601 & @xmath602 & @xmath603 + @xmath604 & @xmath605 & @xmath605 & @xmath605 + @xmath606 & @xmath607 & @xmath608 & @xmath609 + @xmath610 & @xmath611 & @xmath612 & @xmath611 + and @xmath120 passbands ( top , middle and bottom panels , respectively ) . in each panel , the upper sub - panels show the light curve ( black points ) with the wd model @xmath613 ( see table [ params ] ) overplotted ( solid lines ) , and the lower sub - panels show the residuals after subtracting the model from the data . the uppermost panel for the mearth data shows a binned light curve ( in 250 bins each spanning @xmath614 in normalized orbital phase ) to allow the features to be more clearly distinguished given the large number of data points . , title="fig:",width=316 ] and @xmath120 passbands ( top , middle and bottom panels , respectively ) . in each panel , the upper sub - panels show the light curve ( black points ) with the wd model @xmath613 ( see table [ params ] ) overplotted ( solid lines ) , and the lower sub - panels show the residuals after subtracting the model from the data . the uppermost panel for the mearth data shows a binned light curve ( in 250 bins each spanning @xmath614 in normalized orbital phase ) to allow the features to be more clearly distinguished given the large number of data points . , title="fig:",width=316 ] and @xmath120 passbands ( top , middle and bottom panels , respectively ) . in each panel , the upper sub - panels show the light curve ( black points ) with the wd model @xmath613 ( see table [ params ] ) overplotted ( solid lines ) , and the lower sub - panels show the residuals after subtracting the model from the data . the uppermost panel for the mearth data shows a binned light curve ( in 250 bins each spanning @xmath614 in normalized orbital phase ) to allow the features to be more clearly distinguished given the large number of data points . , title="fig:",width=316 ] the most serious limitations in the present analysis result from the use of only single - band light curves to model the out of eclipse variations . this is clear from comparing the results for the three solutions reported in table [ params ] , where we find significant ( @xmath615 standard deviation ) differences in the orbital inclination @xmath372 , ratio of effective temperatures @xmath533 , and in @xmath510 , which depend on the assumptions made regarding the locations of the spots on the components of the binary . given our radial velocity errors , the orbital inclination uncertainty has little effect on the final physical parameters , as evidenced by the component masses and semimajor axis reported in table [ params ] , which are essentially identical to within the observational errors . the dominant uncertainty in the final solution is therefore in the component radii . of the three models considered , the symmetric spot distribution in model ( 3 ) seems the most physically realistic given the known properties of spots on low - mass stars . these are often found to be polar in light curve models ( e.g. @xcite ) . moreover , doppler imaging studies ( e.g. @xcite ) indicate that in reality , the surfaces of low - mass stars probably have many small spots distributed over a range of latitudes . this situation is largely indistinguishable from a small number of large , polar spots in one - dimensional light curve data , although it is important to note that the effects on the measured eclipse depths _ could _ be different . there is no a - priori reason to expect the spots to be concentrated into one hemisphere given the close to edge - on inclination of the binary orbit . we therefore favor model ( 3 ) , but report all three solutions to provide a realistic evaluation of the systematic error in our results , and reiterate that the data do not distinguish between the three possibilities , with all of these having comparable @xmath384 values and residuals . as a check , we compare the model fit to the mearth data alone with our @xmath23 and @xmath120-band data in figure [ lc_model ] . in order to do this , we have refit the model for these bands allowing only the light curve normalization parameter ( hla ) to vary , with all other parameters fixed to the values determined from the mearth data . the model is consistent with the @xmath120-band data within the scatter , and is reasonably consistent with the @xmath23-band data , considering that the early parts of the secondary eclipse curve were taken during quite non - photometric conditions , and that we have not accounted for color - dependent or differential atmospheric extinction effects , which are expected to be larger in this band than the redder mearth or @xmath120 bandpasses , and can not readily be determined due to the lack of repeat observations . the presence of spots and proximity effects in very close binary systems have a small influence on the shape of the radial velocity curve , causing it to depart from a keplerian orbit as we assumed in [ rv_sect ] . we have investigated the influence of these effects on our results by comparing radial velocity curves generated using the wd model and the simple keplerian model , finding that the corrections to radial velocities taken out of eclipse are dominated by the spots , the properties of which are largely unknown . however these are very small ( @xmath616 ) compared to the uncertainties in our radial velocity measurements ( approximately @xmath617 ) , so we are therefore justified in neglecting them for the present analysis . provided the effective temperature of one component of an eclipsing binary is known , the combined light curve and radial velocity parameters are then sufficient to determine intrinsic , bolometric luminosities of both components of the system . this follows directly from the definition of the effective temperature in terms of the stefan - boltzmann law : @xmath618 by using bolometric corrections and the measured system magnitudes , we can then infer the distance , provided the reddening can be constrained or assumed to be zero . in the present case , the latter is a reasonable assumption , since our target is very nearby ( @xcite give a `` photometric parallax '' of @xmath619 , and @xcite gives @xmath620 ; note that both of these assume the system is a single star , which means they underpredict the distance to a near equal luminosity binary such as the present one by a factor of approximately @xmath621 ) . in order to determine the effective temperatures , we assume the intrinsic colors and bolometric corrections for m - dwarfs compiled by @xcite and @xcite to convert the integrated system @xmath622 color presented in table [ photparams ] into @xmath489 , assuming the measured effective temperature ratio and radii for both components from the light curve model . we used the polynomial fits of @xcite , which cover the required spectral type range , and assume an uncertainty of @xmath623 in these fits as stated in her appendix c. we also assume a @xmath624 systematic uncertainty in the effective temperature scale , which has been included and propagated in the errors we report . table [ kinparams ] gives our derived parameters for the gj 3236 system . llll @xmath489 & @xmath625 & @xmath626 & @xmath625 + @xmath491 & @xmath627 & @xmath628 & @xmath629 + @xmath630 & @xmath631 & @xmath632 & @xmath633 + @xmath634 & @xmath635 & @xmath636 & @xmath637 + + @xmath638 & @xmath639 & @xmath640 & @xmath641 + @xmath642 & @xmath643 & @xmath644 & @xmath645 + + @xmath646 & @xmath647 & @xmath648 & @xmath649 + @xmath650 & @xmath651 & @xmath652 & @xmath653 + + @xmath654 & @xmath655 & @xmath656 & @xmath657 + @xmath23 & @xmath658 & @xmath659 & @xmath660 + @xmath661 & @xmath662 & @xmath663 & @xmath662 + the refined value of @xmath489 reported here differs by approximately @xmath664 from the assumed value in [ ebop_lc_sect ] used to determine limb darkening coefficients . this discrepancy is @xmath665 , and should have a negligible effect on the parameters determined from the light curve analysis , so we elected not to repeat this using the updated value of @xmath489 given the computational expense involved . given the full system kinematic information ( position , proper motions , @xmath395 velocity and parallax from the eb analysis ) we can also calculate the @xmath666 components of the space velocity relative to the sun . we use the method of @xcite , but adopt the definition that positive @xmath654 values are away from the galactic center ( a left - handed coordinate system ) for better consistency with the literature . these are included in table [ kinparams ] . comparing to the velocity ellipsoids derived by @xcite , gj 3236 appears to be a member of the galactic disc , and lies within the bounds of the young galactic disc population as defined by @xcite . comparing to the @xmath666 space motions of various solar neighborhood moving groups ( e.g. @xcite ) , the velocities are consistent with with those of the hyades moving group , which has @xmath667 @xcite . despite this possible membership in the hyades group , we note that the dispersion in stellar parameters for the group members does not provide a very useful constraint on age or metallicity for the present system . @xcite obtain a metallicity of @xmath668 } = -0.09 \pm 0.17 $ ] for the hyades group . figures [ mrrel ] and [ mtrel ] show the position of gj 3236 relative to other known objects and theoretical models of the m - dwarf mass - radius and mass - effective temperature relationships . . black points with error bars show the two components of the gj 3236 system , for all three solutions reported in table [ params ] . the gray points are known systems from the literature ( @xcite ; @xcite ; @xcite ; @xcite ) . the black line shows the theoretical mass - radius relationship from @xcite for an age of @xmath669.,width=316 ] , but showing the mass - effective temperature relationship . again , all three solutions are shown for gj 3236 , but the differences are relatively minor in this plane . we also include several more massive m - dwarf ebs in this figure for comparison : yy gem @xcite , gu boo @xcite , and ogle bw3 v38 @xcite . the black line shows the theoretical relationship from @xcite for an age of @xmath670.,width=316 ] although many of the best - observed systems are found to exhibit larger radii than the models predict , the slope of the mass - radius relation between the components of each system ( or equivalently , the ratio of the component radii ) is generally found to agree well between the observations and the theoretical models . examining the positions of our solutions for gj 3236 , our models 1 and 3 for the out of eclipse modulations are reasonably consistent with this expectation , whereas the slope defined by the components for solution 2 is substantially different . although this indicates the assumptions used in model 2 may be unrealistic , the present observational data for gj 3236 do not distinguish between the three possibilities . using this as a constraint in the modeling would be dangerous because the results would then no longer yield a completely independent test of the theoretical models , so instead we suggest that a better solution to the issue would be obtaining improved multi - band photometry to better - constrain the spot parameters . therefore , within the present observational uncertainties , the radii of gj 3236 appear to be consistent with the theoretical predictions at approximately @xmath5 , when we account for the systematic uncertainties represented by the three solutions we have presented . furthermore , the central values appear to lie above the theoretical curve . this is largely consistent with the results for the well - known systems in the literature , where the radii are typically found to be @xmath671 larger than the theoretical predictions . in the effective temperature versus mass plane ( figure [ mtrel ] ) , the effective temperatures of the components of gj 3236 are found to be cooler than the models predict , by approximately @xmath672 . this is in good agreement with many of the well - known systems in the literature as shown in the figure , such as cm dra . compared to the well - known systems cm dra and cu cnc , gj 3236 has a shorter orbital period , and the h@xmath6 emission in the spectroscopy and out - of - eclipse modulations we observe in the light curve are indicative of high activity levels , with both the photometric period and the spectroscopic line broadening apparently consistent with the stellar spin being synchronized to the binary orbit , as expected from tidal theory ( e.g. @xcite ) . @xcite propose two hypotheses to explain the observed radius discrepancies between theoretical models and eclipsing binaries : ( 1 ) that the inflated radii result from reduced convective efficiency due to high rotation rates and large magnetic fields ; or ( 2 ) that magnetic spot coverage of the surfaces leads to reduced heat flux , and thus larger radii and cooler effective temperatures . hypothesis ( 2 ) has been found to be consistent with some of the well - known eclipsing binary systems , e.g. yy gem @xcite , where the radius discrepancy ( and the discrepancy in effective temperature ) can be explained by the presence of starspots covering approximately @xmath673 of the stellar surface ( if the spots are cooler than the photosphere by @xmath674 ; @xcite ) . gj 3236 shows out of eclipse modulations of comparable amplitude to yy gem , a compatible radius discrepancy within the present observational errors , and similarly has effective temperatures somewhat cooler than the models predict , so this is an attractive hypothesis . by obtaining precise , multi - band photometry covering the entire orbital phase , it may be possible to constrain the spot temperatures , and hence observationally test this argument . the main limitations in the orbital and geometric modeling of the system arise from the lack of out - of - eclipse data in multiple passbands , which leads to large errors ( @xmath675 ) on the radii when we take the degeneracies in the spot configuration into account . the precision of the mass estimates ( @xmath4 ) is limited by the error in the radial velocity measurements , and an important contribution to the uncertainties in the radii will be from the spectroscopic light ratio if the uncertainty in spot parameters can be resolved . further high - precision photometric measurements and high - resolution spectroscopy are therefore needed , and combined with careful analysis these should allow the precision of the mass and radius measurements for this system to be improved , potentially to beyond the @xmath124 level as for the well - known systems cm dra and yy gem . the mearth team gratefully acknowledges funding from the david and lucile packard fellowship for science and engineering ( awarded to dc ) . this material is based upon work supported by the national science foundation under grant number ast-0807690 . sq and dwl acknowledge support from the nasa kepler mission under cooperative agreement ncc2 - 1390 . gt acknowledges partial support from the nsf through grant ast-0708229 . we thank eric mamajek for helpful discussions regarding solar neighborhood kinematics , and the `` astro - comb '' team at cfa : claire cramer , david phillips , alexander glenday , chih - hao li , dimitar sasselov and ronald walsworth , for their generous donation of engineering time to obtain spectra of our target . the referee is thanked for comments that helped to improve the manuscript . this research has made extensive use of data products from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center / california institute of technology , funded by nasa and the nsf , nasa s astrophysics data system ( ads ) , and the simbad database , operated at cds , strasbourg , france . the digitized sky surveys were produced at the space telescope science institute under u.s . government grant nag w-2166 . the images of these surveys are based on photographic data obtained using the oschin schmidt telescope on palomar mountain and the uk schmidt telescope . the plates were processed into the present compressed digital form with the permission of these institutions . the peters automated infrared imaging telescope ( pairitel ) is operated by the smithsonian astrophysical observatory ( sao ) and was made possible by a grant from the harvard university milton fund , the camera loan from the university of virginia , and the continued support of the sao and u.c . the pairitel project and jsb are further supported by nasa / swift guest investigator grant nng06gh50 g . we thank m. skrutskie for his continued support of the pairitel project . andersen , j. 1991 , , 3 , 91 baraffe , i. , chabrier , g. , allard , f. & hauschildt , p. h. 1998 , , 337 , 403 barnard , e. e. 1916 , , 29 , 181 barnes , j. r. , collier cameron , a. , james , d. j. & donati j .- f . 2000 , , 314 , 162 barnes , j. r. & collier cameron , a. 2001 , , 326 , 950 bessell , m. s. 1991 , , 101 , 662 bessell , m. s. 1995 , in the bottom of the main sequence any beyond , ed . c. tinney ( springer , berlin , germany ) , p.123 bessell , m. s. & brett , j. m. 1988 , , 100 , 1134 blake , c. h. , et al . 2005 , , 435 , 181 blake , c. h. , torres , g. , bloom , j. s. , & gaudi , b. s. 2008 , , 684 , 635 bloom , j. s. , starr , d. l. , blake , c. h. , skrutskie , m. h. , & falco , e. e. 2006 , in asp conf . 351 , astronomical data analysis software and systems xv , ed . c. gabriel et al . ( san francisco : asp ) , 751 cakirli , o. , ibanoglu , c. & gungor , c. 2009 , new astronomy , in press . arxiv : 0902.3327 chabrier , g. , gallardo , j. , baraffe , i. 2007 , , 472 , 17 chiba , m. , & beers , t. c. 2000 , , 119 , 2843 claret , a. 2000 , , 363 , 1081 claret , a. 2004 , , 428 , 1001 delfosse , x. , forveille , t. , mayor , m. , burnet , m. , & perrier , c. 1999 , , 341 , 63 eggen , o. j , & sandage , a. 1967 , , 148 , 911 etzel , p. b. 1980 , ebop user s guide , 3rd edn . , ucla astronomy & astrophysics ford , e. b. 2005 , , 129 , 1706 gliese , w. & jahreiss , h. 1991 , preliminary version of the third catalogue of nearby stars , astron . rechen - institut , heidelberg hastings w. k. 1970 , biometrika , 57 , 97 hauschildt , p. h. , allard , f. & baron , e. 1999 , , 512 , 377 hebb , l. , wyse , r. f. g. , gilmore , g. , & holtzman , j. 2006 , , 131 , 555 hewett , p. c. , irwin , m. j. , bunclark , p. , bridgeland , m. t. , kibblewhite , e. j. , he , x. t. , smith , m. g. 1985 , , 213 , 971 hillenbrand , l. a. 1997 , , 113 , 1733 irwin , j. , et al . 2007a , , 380 , 541 irwin , j. , irwin , m. , aigrain , s. , hodgkin , s. , hebb , l. , & moraux , e. 2007b , , 375 , 1449 irwin , j. , charbonneau , d. , nutzman , p. , & falco e. 2008 , to appear in the proceedings of iau symposium 253 : `` transing planets '' , ed . f. pont et al . ( cambridge : cambridge university press ) . arxiv : 0807.1316 johnson d. r. h. , & soderblom , d. r. 1987 , , 93 , 864 joy a. h. & sanford r. f. 1926 , , 64 , 250 kopal , z. 1959 , semi - detached binaries and stellar evolution , astronomical society of the pacific leaflets , vol . 8 , p.81 kurucz , r. l. 1979 , , 40 , 1 kurucz , r. l. 1993 , vizier on - line data catalog : vi/39 . kwee , k. k. & van woerden h. 1956 , , 12 , 327 lacy , c. h. 1977 , , 218 , 444 landolt , a. j. 1992 , , 104 , l340 leggett , s. k. 1992 , , 82 , 351 lpine , s. 2005 , , 130 , 1680 lpine , s. , & shara , p. 2005 , , 129 , 1483 lpez - morales , m. & ribas , i. 2005 , , 631 , 1120 lpez - morales , m. 2005 , , 660 , 732 lucy , l. b. 1967 , zeitschrift fr astrophysik , 65 , 89 maceroni , c. & montalbn j. 2004 , , 426 , 577 metcalfe , t. s. , mathieu , r. d. , latham , d. w. , & torres , g. 1996 , , 456 , 356 metropolis , n. , rosenbluth a. w. , rosenbluth m. n. , teller a. h. , teller e. 1967 , , 21 , 1087 morales , j. c. , et al . 2009 , , 691 , 1400 nutzman , p. & charbonneau , d. 2008 , , 120 , 317 popper , d. m. & etzel , p. b. 1981 , , 86 , 102 pra , a. & zwitter , t. 2005 , , 628 , 426 ribas , i. 2003 , , 398 , 239 ribas , i. , morales , j. c. , jordi , c. , baraffe , i. , chabrier , g. & gallardo , j. 2008 , , 79 , 562 rodon , m. 1986 , , 1986 , 165 , 135 soderblom , d. r. & mayor , m. 1993 , , 105 , 226 southworth , j. , maxted , p. f. l. , & smalley , b. 2004a , , 351 , 1277 southworth , j. , zucker , s. , maxted , p. f. l. , & smalley , b. 2004b , , 355 , 986 southworth , j. , smalley , b. , maxted , p. f. l. , claret , a. , & etzel , p. b. 2005 , , 363 , 529 tody , d. 1993 , in asp conf . 52 , astronomical data analysis software and systems ii , ed . hanisch , r. j. v. brissenden , j. barnes ( san francisco : asp ) , 173 torres , g. & ribas , i. 2002 , 567 , 1140 vaccaro , t. r. , rudkin , m. , kawka , a. , vennes , s. , oswalt , t. d. , silver , i. , wood , m. , smith , j. a. 2007 , , 661 , 1112 van gent , h. 1926 , , 3 , 121 wilson , r. e. & devinney e. j. 1971 , , 166 , 605 zahn , j .- p . 1977 , , 57 , 383 zhao , j. , zhao , g. , & chen , y. 2009 , , 692 , 113 zucker , s. & mazeh , t. 1994 , , 420 , 806

the problem of heat conduction is a well - studied field . more than two centuries ago , joseph fourier summarized the behavior of heat conduction by the law that bears his name . this law describes phenomenologically that the heat current is proportional to the temperature gradient . the detailed atomistic theories of heat conduction appeared only much later . for heat conduction in gas , the simple kinetic theory gives the result @xmath3 , where @xmath4 is specific heat , @xmath5 is average velocity and @xmath6 is mean - free path . peierls theory of heat conduction in insulating solids @xcite is a classic on this subject . these early theories deal with mostly the relevant three dimensions . it turns out that low dimensional systems are more interesting and in some sense strange . an analysis of a simple one - dimensional ( 1d ) harmonic oscillator model shows @xcite that there is no well - defined temperature gradient , the thermal conductivity diverges with system sizes as @xmath7 or @xmath1 , depending on the boundary conditions . there is a general argument , for momentum conserving systems , that the thermal conduction in 1d is necessarily divergent @xcite . there have been many analytical and numerical studies of 1d heat conduction ( see ref . @xcite and @xcite for review ) . we ll mention some of the most relevant papers to current work . the work of lepri _ et al _ @xcite by mode - coupling theory and molecular dynamics suggests a divergent thermal conductivity exponent of @xmath8 , i.e. , @xmath9 for a 1d chain model with fermi - pasta - ulam ( fpu ) interactions . mode - coupling theory is usually applied in the dynamics of liquids @xcite . the first use of this theory in the context of heat conduction appears only recently , mostly due to lepri and his collaborators @xcite . perverzev analyzed the same problem with the peierls theory for phonon gas and gave the same conclusion of @xmath8 exponent @xcite . the result of @xmath8 is also supported by numerical simulation from several groups @xcite . these results are supposed to be universal to some extent . however , it is challenged by a different result of @xmath10 by narayan and ramaswamy @xcite , based on fluctuating hydrodynamics and renormalization group analysis . the numerical result for this 1/3 law is not convincing , as for the same model the hard - particle gas model some obtained 1/3 @xcite , while others obtained different value 1/4 @xcite . but for the fpu model , there is no good evidence for an exponent of 1/3 @xcite . when momentum conservation is broken such as the one with on - site potential , the heat conduction can become normal again like the frenkel - kontorova model @xcite and the @xmath11 model @xcite . in order to understand the underlying microscopic dynamical mechanism of the fourier law , a different class of models billiard channels have been introduced and studied in recent years @xcite . various exponent values are found in such systems . thus , it is believed that a universal constant does not exist at all . instead , the divergent ( convergent ) exponent of the thermal conductivity is found related to the power of super ( sub ) diffusion @xcite . besides the theoretical significance of heat conduction research in low dimensional systems , it is also of practical importance . recent development of nanotechnology will enable us to manufacture devices with feature sizes at molecular level . the understanding of heat conduction mechanism will allow us to control and manipulate heat current , and eventually to design novel thermal devices with certain function @xcite . to this end , more realistic physical models are necessary . among many others , nanotubes and polymer chains are most promising . there have been a number of numerical works to compute the thermal conductivity of the carbon nanotubes @xcite . recent molecular dynamics ( md ) study of carbon nanotubes with realistic interaction potential suggested a divergent thermal conductivity for narrow diameter tubes @xcite . the quantum effect of such systems is also very interesting @xcite . we study the heat conduction of a 1d solid , as a classical system . a brief version of this paper is reported in ref . @xcite . in the rest of the paper , we introduce the quasi - one - dimensional chain model in sec . [ sec : chain ] . we discuss the basis of the mode - coupling theory , the projection method in sec . [ sec : projection ] . the mode - coupling approximations and their numerical and analytical solutions are discussed in sec . [ sec : mode - coupling ] and [ sec : solution ] . the basic output of the mode - coupling analysis is the dependence of damping of the modes with the wave - vector of the modes . we find that the transverse modes are diffusive , with @xmath12 , while the longitudinal modes are super - diffusive , @xmath13 , where @xmath14 is the decay rate for mode with momentum or lattice wave number @xmath15 . we discuss the relationship between the damping of the modes with the heat conductivity through green - kubo formula in sec . [ sec : green - kubo ] . our mode - coupling theory predicts that the heat conductance diverges with the @xmath10 exponent when the transverse motions are important , while @xmath8 is recovered if the transverse motions can be neglected . in sec . [ sec : md ] , we present nonequilibrium molecular dynamics results ( with heat baths ) of heat conductance and compare with mode - coupling theory . we conclude in the last section . most of the previous studies considered only strictly 1d models , with the fermi - pasta - ulam ( fpu ) model as the most representative . the strictly 1d models may not be applicable to real systems such as the nanotubes . real systems of nanotubes or wires live in three - dimensional space . the added transverse motion and the flexibility of the tube at long length scales will certainly scatter phonons , and thus should have a profound effect on thermal transport . while a direct simulation of a realistic system such as a polyethelene chain with empirical force fields such as that in refs . @xcite or nanotube with tersoff potential @xcite is possible , we think it is useful to consider a simplified model which captures one of the important features of the real systems transverse degrees of freedom . therefore we propose to study the following chain model in two dimensions @xcite : @xmath16 where the position vector @xmath17 and momentum vector @xmath18 are two - dimensional ; @xmath19 is lattice constant . the minimum energy state is at @xmath20 for @xmath21 to @xmath22 . if the system is restricted to @xmath23 ( corresponding to @xmath24 ) , it is essentially a 1d gas with harmonic interaction . the coupling @xmath25 is the spring constant ; @xmath26 signifies bending or flexibility of the chain , while @xmath27 is the bond angle formed with two neighboring sites , @xmath28 , and unit vector @xmath29 , @xmath30 . unlike the fpu model which does not have an energy scale , the second bond - angle bending term introduces an energy scale . in this work , we take mass @xmath31 , spring constant @xmath32 , and the boltzmann constant @xmath33 , thus the most important parameters are @xmath26 and temperature @xmath34 . we follow the formulation of the projection method in ref . @xcite . let @xmath35 be a column vector of @xmath36 components of some arbitrary functions of dynamical variables @xmath37 . each of the function @xmath38 can be complex . later , we shall choose @xmath39 to be the canonical coordinates of the system . we use @xmath40 to denote the hermitian conjugate of @xmath41 . the equation of motion for @xmath41 is @xmath42 where @xmath43 is the liouville operator @xmath44 equation ( [ eqmotion ] ) can be viewed as a partial differential equation with variables @xmath37 and time @xmath45 . @xmath46 , i.e. , the quantity @xmath41 at time @xmath45 when the initial condition at @xmath47 is @xmath37 . quantity without a subscript @xmath45 will be understood to be evaluated at time @xmath47 , e.g. , @xmath48 . a formal solution to eq . ( [ eqmotion ] ) is simply @xmath49 . the projection operator on a column vector @xmath50 is defined by @xmath51 where @xmath52 and @xmath53 are @xmath54 matrices . the angular brackets denote thermodynamical average in a canonical ensemble at temperature @xmath34 . the comma separating the two terms is immaterial , but we use a notation of inner product . one can verify that @xmath55 is indeed a projection operator , i.e. , @xmath56 . if we apply the projection operator @xmath57 and @xmath58 to the equation of motion , we get two coupled equations . solving formally the second equation associated with @xmath59 , and substituting it back into the first equation , we obtain an equation for the projected variable that formally resembles a langevin equation : @xmath60 where @xmath61 is the complex unit , and @xmath62 this set of equations is deterministic and formally exact . the only assumption made is that equilibrium distribution can be realized . what is more important is the correlation functions , which are physical observables . we define the normalized correlation function ( correlation matrix ) as @xmath63 it is an identity matrix at @xmath47 and has the property that @xmath64 . from eq . ( [ langevin ] ) we have , @xmath65 this equation can be solved formally using proper initial condition with a fourier - laplace transform , @xmath66 = \int_0^\infty e^{-i z t } g(t)\ , dt,\ ] ] and similarly for @xmath67 . the solution is @xmath66 = \bigl ( i(z - \omega ) + \gamma[z ] \bigr)^{-1}. \label{eqgw}\ ] ] to simplify notation , we have used parentheses ( e.g. , @xmath68 ) for functions in time domain , and square brackets with the same symbol ( e.g. , @xmath69 $ ] ) for their corresponding fourier - laplace transform in frequency domain . the information about the system is in the memory kernel @xmath67 . however , such a correlation function is difficult to calculate , since the evolution of the `` random force '' @xmath70 does not follow the dynamics of the original hamiltonian system . for example , it is impossible to compute directly @xmath70 from molecular dynamics . for this reason , a `` true force '' can be introduced which obeys the normal evolution , i.e. , @xmath71 the correlation function of the true force , @xmath72 and that of the random force are related in fourier space as @xcite @xmath73^{-1 } = \gamma_f[z]^{-1 } - \bigl ( i ( z - \omega ) \bigr)^{-1}. \label{eqf2r}\ ] ] this completes the formal theory of projection due originally to zwanzig @xcite and mori @xcite . these results are formal and exact . they give us relation between correlation of the `` force '' and correlation of dynamical variables . they are the starting point for mode - coupling theory . in the next subsections , we apply it to our chain model and introduce a series of approximations to solve it . we now apply the projection method to our chain model . we choose the normal modes as the basic quantities @xmath39 that we are going to project out . there are several reasons for choosing the normal modes . to zeroth order approximation , each mode is nearly independent . the slowest process corresponds to long wave - length modes . the effect of short wave - length modes can be treated as stochastic noise ( the random force @xmath70 ) . we choose @xmath41 to be the complete set of canonical momenta and coordinates : @xmath74 where @xmath75 we have defined the position vector @xmath76 , so that @xmath77 and @xmath78 are deviations from zero - temperature equilibrium position . because the fourier transform is a periodic function , the index @xmath79 is unique only modulo @xmath1 . as a result , we can also let @xmath79 vary in the range @xmath80 . we also note that @xmath81 . with these definitions , we can compute the matrix @xmath82 and expression for the true force @xmath83 in the general theory . we find that for @xmath84 , the components are @xmath85 we have used equal - partition theorem for the average kinetic energy expression , and the last equation merely defines the effective frequencies @xmath86 for each mode . note that due to translational invariance , correlation between different @xmath79 modes vanishes . correlation between transverse and longitudinal modes also vanishes due to the reflection symmetry of @xmath87 for the chain . thus equal - time correlation for @xmath41 is diagonal , @xmath88 we have defined @xmath89 as a @xmath90 diagonal matrix with elements @xmath91 ; @xmath92 is a @xmath93 identity matrix . similarly , the correlation @xmath94 is found from @xmath95 the second equation is from a general virial theorem @xcite . we have @xmath96 the expression for the true force @xmath97 is then @xmath98 note that only the momentum sector has a nonzero value , and @xmath99 is given approximately by eq . ( [ eqmotion1 ] ) and ( [ eqmotion2 ] ) below . with this special form of @xmath83 , the damping matrix @xmath100 is also diagonal and is nonzero only in the @xmath101 components . with these results , eq . ( [ eqgw ] ) becomes @xmath66 = \begin{pmatrix } iz d & - \tilde \omega^2 d \cr d & ( iz + \tilde \gamma[z ] ) d \cr \end{pmatrix } , \label{general - gz}\ ] ] where @xmath102)^{-1}$ ] is a @xmath90 diagonal matrix , and @xmath103 $ ] is the fourier - laplace transform of the correlation @xmath104 , which is also diagonal ( we ll denote as @xmath105 ) . in particular , we have the usual expression for the normalized coordinate correlation , @xmath106 & = & { i z + \gamma_k^\mu[z ] \over \tilde \omega_k^\mu{}^2 - z^2 + iz \gamma_k^\mu[z]}. \label{eqvtog}\end{aligned}\ ] ] for simplicity , we drop the @xmath107 subscript for the coordinate correlation for the rest of the paper . finally , the relation between the random force correlation and true force correlation , eq . ( [ eqf2r ] ) , becomes , @xmath108 } = \frac{1}{\gamma_{f , k}^\mu[z ] } - { i z \over \tilde \omega_k^\mu{}^2 - z^2 } . \label{eqvftov}\ ] ] the force correlation @xmath109 can be computed from the correlation function @xmath110 . it is convenient to separate the linear term in the force from the nonlinear contribution . so we write @xmath111 where @xmath112 is effective angular frequency and @xmath113 is bare angular frequency of the mode . we have dropped the indices @xmath79 and @xmath114 since these equations apply for any of the modes . @xmath115 is at least quadratic in @xmath116 . we note that the correlation of @xmath117 is linearly related to the coordinate correlation function @xmath118 $ ] , @xmath119 = { - iz(\tilde\omega^2-z^2 ) \over \tilde\omega^2 } + { ( \tilde\omega^2 - z^2)^2 \over \tilde\omega^2 } g[z ] . \label{eqgtogamf}\ ] ] this is simply a consequence of eqs . ( [ eqvtog ] ) and ( [ eqvftov ] ) , but can also be derived directly from the definition . the second derivative of @xmath120 in frequency domain is @xmath121 $ ] multiplied by @xmath122 . using the fact that @xmath123 - \dot q(0 ) - iz q(0),\ ] ] we can also derive eq . ( [ eqgtogamf ] ) with the understanding that @xmath124 is an average over the initial conditions . since @xmath118 $ ] is finite or at least should not diverge precisely at @xmath125 , this implies @xmath126 = 0 $ ] . with the above results , we can derive an expression of the true force correlation in terms of nonlinear force correlation , @xmath127 where @xmath128 . the mixed term can be expressed in terms of @xmath118 $ ] by noting that @xmath129 . the two mixed terms @xmath130 and @xmath131 are equal due to time - reversal symmetry . we find @xmath132 - iz.\ ] ] finally , we have @xmath119 = { \delta \tilde\omega^2 \over \tilde\omega^2 } \bigl((2z^2-\omega_0 ^ 2-\tilde\omega^2)g[z ] + 2iz \bigr ) + \gamma_n[z].\ ] ] we can also express @xmath118 $ ] in terms of the nonlinear part of the force correlation , @xmath133 ( @xmath134 ) : @xmath135 = { iz(2\omega_0 ^ 2 - \tilde\omega^2-z^2 ) + \tilde\omega^2 \gamma_n[z ] \over ( z^2-\omega_0 ^ 2)^2 } . \label{eqvntog}\ ] ] again , since @xmath118 $ ] can not diverge precisely at @xmath136 , this implies that @xmath137 $ ] must take a special form to cancel the apparent divergence . thus , if we do not take care of these superficial divergences , we will not be able to make correct prediction for the correlation function . we can also relate the original damping function @xmath100 to the nonlinear one , @xmath73 = { -i(\tilde \omega^2 - \omega_0 ^ 2)^2 z + \tilde\omega^2 ( \tilde\omega^2 - z^2 ) \gamma_n[z ] \over \omega_0 ^ 4 - \tilde \omega^2 z^2 - iz\tilde \omega^2 \gamma_n[z]}. \label{eqvntov}\ ] ] this last equation is useful for approximating the damping function . all of these relations are exact . this ends our formal application of the projection method to the chain model . to make some progress for analytic and numerical treatment , we have to make some approximations . first , we ll consider small oscillations valid at relatively low temperatures . an approximate hamiltonian for small oscillations near zero - temperature equilibrium position , keeping only leading cubic non - linearity in the hamiltonian , is then given by @xmath138 where @xmath139 are the ` bare ' dispersion relations , and @xmath140 the absence of @xmath141 term in hamiltonian ( [ h - perturb ] ) is due to the quadratic nature of the potential , while the absence of the terms of the form @xmath142 and @xmath143 is due to the reflective symmetry about @xmath144-axis of the hamiltonian . we view @xmath79 and @xmath145 as independent component when taking the derivatives . a slightly modified hamilton s equation ( because of the use of complex numbers ) describes the dynamics : @xmath146 this gives the following equations of motion : @xmath147 where @xmath148 with these expressions , we are ready to compute the force correlation function in terms of dynamic variables @xmath149 . we write @xmath150 where @xmath151 is the difference between bare dispersion relation and effective dispersion relation . the second term @xmath152 denotes the rest of the nonlinear force ( we take only the quadratic terms in @xmath116 ) . due to translational and reflective symmetries , the correlation matrix formed by @xmath153 is diagonal without any approximation . the time - displaced correlation for the diagonal terms defines true force correlation . the nonlinear part of the contributions is @xmath154 in order to have a closed system of equations for the normalized correlation functions , we use the standard mean - field type approximation , @xmath155 . owing to the @xmath156 correlation between different @xmath79 , the double summation can be reduced to a single one . we introduce @xmath157 and similarly @xmath158 associated with @xmath159 . in terms of @xmath160 , we obtain @xmath161 where @xmath162 equations ( [ eqmodekp - full - para ] ) and ( [ eqmodekp - full - perp ] ) together with the relations among @xmath163 , @xmath159 , and @xmath164 , eqs . ( [ eqvtog ] ) , ( [ eqvntov ] ) , and ( [ eqgammatonu ] ) , form a system of close equations , which can be solved in principle . however , because of the singular nature at @xmath136 in eq . ( [ eqvntog ] ) , any approximation to @xmath133 will destroy a subtle cancellation of the singularity , rendering the problem impossible to solve . damping functions of the slowest modes , i.e. , real part of @xmath165 $ ] ( solid line ) and @xmath166 $ ] ( dashed line ) vs @xmath167 , computed from molecular dynamics for parameters of set e : @xmath32 , @xmath168 , @xmath169 , @xmath170 , @xmath31 . the top figure is for size @xmath171 and bottom figure for @xmath172 . ] the damping function @xmath173 $ ] is the central function that a successful theory needs to be able to calculate . in fig . [ fig - mdgm ] we present examples of such functions determined from equilibrium molecular dynamics ( md ) simulation in a microcanonical ensemble with periodic boundary condition . a more correct comparison of md with mode - coupling theory should use an ensemble of initial conditions distributed according to canonical weight . this may be unimportant when @xmath1 is large . we compute @xmath174 $ ] from the ratio of two correlation functions , @xmath175/g_{qp}[z ] = iz + \gamma[z]$ ] . for small systems , there are a lot of peak structures associated with the low frequency modes ; the feature appears washed out for large systems . to make the problem tractable , we make a bold approximation . instead of eq . ( [ eqvntov ] ) , we take @xmath176 \approx \nu_{n , k}^\mu[z ] . \label{our - approximation}\ ] ] this is equivalent to say @xmath177 . this appears justified for the longitudinal modes at sufficiently low temperatures but problematic for the transverse mode , as @xmath178 is linear in wave number @xmath179 but @xmath113 is quadratic in @xmath179 . we can consider the limit of small @xmath179 . in such limit , the difference between @xmath180 and @xmath100 is dropped . lepri s treatment @xcite is similar to the above approximation . we also note that in the work of scheipers and schirmacher for damping of anharmonic crystals @xcite , effective cubic coupling is used instead of the ` bare ' coupling , @xmath181 . the standard operation of projecting the random force onto bilinear form @xmath182 to get the mode - coupling equations also agrees with the approximation , eq . ( [ our - approximation ] ) , but with somewhat different `` vertex functions '' replacing eq . ( [ eqkpara ] ) and ( [ eqkperp ] ) . in order to obtain a numerical solution , besides the model parameters ( mass @xmath183 , lattice spacing @xmath19 , couplings @xmath25 and @xmath26 ) , we also need the effective dispersion relation . we used md data for this purpose . it turns out that a two - parameter fit of the form @xmath184 characterizes the effective dispersion relation very well , where @xmath185 is sound velocity at small @xmath79 and @xmath186 is maximum frequency at @xmath187 . we used a fast fourier transform to solve the equations iteratively . given an initial @xmath188 $ ] , correlation function in frequency domain is calculated by eq . ( [ eqvtog ] ) . inverse transform gives us @xmath189 , which is used to calculate the discrete sum in eqs . ( [ eqmodekp - full - para ] ) and ( [ eqmodekp - full - perp ] ) to obtain @xmath190 . we transform back to frequency domain for the next iteration . real part of @xmath165 $ ] ( solid line ) and @xmath166 $ ] ( dashed line ) vs @xmath167 from a full mode - coupling theory for parameters of set e ( same as that in fig.[fig - mdgm ] ) at @xmath171 and 64 . the input effective dispersion relation parameters , @xmath191 , and @xmath192 , are given in table [ tb-1 ] . ] real part of @xmath165 $ ] ( solid line ) and @xmath166 $ ] ( dashed line ) vs @xmath167 from a full mode - coupling theory for parameters of set e , but at different temperature @xmath34 . other parameters are the same as in fig . [ fig - mkgm ] . ] figure [ fig - mkgm ] is our theoretical calculation with parameters exactly at set e. this figure should be compared with the md data in fig . [ fig - mdgm ] . we observe that the predicted results are too smooth without much structure . the mode - coupling theory predicts a damping that is a factor of two larger for the longitudinal modes and much too large for the transverse modes . by adjusting the temperature to a lower value ( one half for @xmath171 and a quarter for @xmath172 ) , we obtain curves which closely resemble the md results . however , the mode - coupling theory still overestimates the value of transverse damping . it is clear that @xmath193 is considerably smaller in actually system , but our mode - coupling theory can not produce a small enough @xmath193 . the results presented in fig . [ fig - mkgmh ] give us some hints that the mode - coupling equations are essentially correct , but the parameters of the model need to be adjusted . in fact , simple perturbative expansion of the hamiltonian can not correctly predict @xmath194 . according to md result , @xmath195 , but the leading contribution from perturbation calculation should give the ` bare ' frequency , @xmath196 . the adjusting of the parameters can be made more rigorous @xcite with an introduction of an effective hamiltonian , @xmath197 the form of the hamiltonian is dictated by symmetry . translational invariance requires that @xmath198 . the system is symmetric under reflection about @xmath144-axis or @xmath199-axis . thus the hamiltonian should be invariant under the transformation @xmath200 or @xmath201 . in addition , @xmath202 is symmetric under permutation of @xmath15 and @xmath179 , and @xmath203 is symmetric under permutation of all three indices . we also have @xmath204 and similarly for @xmath203 . although the original hamiltonian does not have @xmath205 term , such a term can present . one of the reasons that such term can present is due to the non - analytic behavior of the potential ( because of the absolute value @xmath206 ) . ldddd set e & bare & n=8 & n=64 & n=1024 + @xmath207 & 2 & 1.435 & 1.341 & 1.329 + @xmath208 & 0 & 0.621 & 0.669 & 0.674 + @xmath209 & 2 & 1.495 & 1.543 & 1.553 + @xmath210 & 1.095 & 1.087 & 1.182 & 1.194 + @xmath211 & 2 & 0.7113 & 0.6333 & 0.6337 + @xmath212 & 0 & -0.1060 & -0.0221 & -0.0076 + @xmath213 & -0.6 & -0.1099 & -0.1242 & -0.1293 + @xmath214 & 0 & 0.6362 & 0.8965 & 0.9312 + @xmath215 & 0 & -0.0424 & -0.1346 & -0.1469 + we determine the parameters of the effective hamiltonian , @xmath216 , @xmath217 , @xmath218 , and @xmath205 , by fitting them to the observed time - independent correlation functions from md . in order to be able to carry out the calculation analytically , we treat the interactions as perturbations , i.e. , @xmath218 and @xmath205 are assumed small . all the equilibrium averages are approximated by the leading contribution from the perturbation , @xmath219 where @xmath220 is thermodynamical average with respect to the non - interacting harmonic oscillators ( product of gaussian integrals ) . under the above approximation , we find @xmath221 and the interaction parameters @xmath222 the actual form of the @xmath89 is fitted according to eq . ( [ omega_tilde_fit ] ) . for @xmath218 and @xmath205 we fit into a functional form @xmath223 where @xmath224 , @xmath225 , and @xmath226 . the values of @xmath211 to @xmath215 , together with @xmath185 and @xmath227 are listed in table [ tb-1 ] . it is surprising that not only is the @xmath205 term present , but also is its magnitude comparable to @xmath218 . the column marked ` bare ' are the parameters corresponding to the ` bare ' hamiltonian , eq . ( [ h - perturb ] ) . the bare parameters are reached only at very low temperatures , such as @xmath228 . since we have factored out the leading size dependence , we expect the parameters weakly depending on size @xmath1 . before presenting our results with the effective parameters , it is worth pointing out that the standard procedure of deriving mode - coupling equations gives identical result as our approximation , eq . ( [ our - approximation ] ) . the standard procedure is to project the random force , @xmath229 , onto a bilinear form of the basic variable @xmath230 . applying the general projection operator , eq . ( [ projector ] ) , to our case , this bilinear projector is @xmath231 where @xmath232 or @xmath233 is a pair of indices , e.g. , @xmath234 , and @xmath235 is @xmath236 . in evaluating the four - point correlation functions , we have neglected the perturbation term . it turns out that the last term projects out only the @xmath237 component . thus , the specific values of @xmath238 are not needed . after applying projector @xmath239 onto @xmath240 , we neglect the difference between normal dynamics @xmath241 and anomalous dynamics for @xmath242 , @xmath243 . we have @xmath244 the mode - coupling equations are then , @xmath245 where @xmath246 note that the vertex coupling @xmath247 is proportional to the three - point correlation independent of the specific form of random force . this version of mode - coupling equation need not have the limitation of small oscillations , and can be applied to high temperatures as well . real part of @xmath165 $ ] ( solid line ) and @xmath166 $ ] ( dashed line ) vs @xmath167 from a full mode - coupling theory for set e , using effective parameters given in table [ tb-1 ] . ] figure [ fig - mkvgm ] is a calculation of the real part of @xmath248 $ ] for @xmath249 . comparing with fig . [ fig - mkgm ] and [ fig - mkgmh ] , we see that using effective parameters brings into much better agreement with the md data . in fact , comparing to fig . [ fig - mdgm ] , most of the features are reproduced , such as the locations and heights of the peaks . the most important improvement is the ratio of parallel to perpendicular damping . the normalized correlation functions @xmath250 ( upper part ) and @xmath251 ( lower part ) on a @xmath252 system for data set e. the solid curves are from full mode - coupling theory , while the dotted dashes are from equilibrium molecular dynamics . ] in figure [ fig - mkgt ] , we compare the normalized dynamic correlation functions for size @xmath252 and the slowest mode @xmath249 . the frequencies of oscillations are slightly different in md and mode - coupling calculation , thus there are phase shifts at long times . the longitudinal damping agrees with each other extremely well . a fit of the logarithm of amplitudes ( maxima and minima ) versus time for the md data gives the decay rate @xmath253 , @xmath254 , while mode - coupling predicts @xmath255 and @xmath256 , respectively . mode - coupling theory with effective parameters still overestimates the transverse damping by a factor of 2 to 3 . the decay rate @xmath257 for each mode @xmath79 for system size @xmath258 . circles ( @xmath259 ) and triangles ( @xmath260 ) are from molecular dynamics by fitting the correlation function , while the continuous curves are from a full mode - coupling theory for set e , using effective parameters given in table [ tb-1 ] . ] in fig . [ fig - gamma - compare ] , we show the damping of the mode in terms of the decay constant @xmath257 , which is defined by the exponential decay in @xmath261 . the symbols are obtained from md simulation , while the curves are from the full mode - coupling theory with effective parameters . for md data , @xmath257 is obtained by a least - square fit to the amplitudes . for mode - coupling theory , we use an approximation , @xmath262 $ ] . the bar means we average over a window centered around @xmath263 with width about @xmath257 . we observe that excellent agreement between md data and mode - coupling theory is obtained for the parallel component . the mode - coupling theory overestimates the transverse mode damping by some constant factor . in any case , the slopes of the curves agree very well with the expected results ( to be discussed in the next subsection ) , @xmath264 and @xmath265 . to simplify the equations further , we consider the large size limit . since the large size asymptotic behavior is the most interesting , such limiting results are justified . we thus consider the limit of @xmath266 and keep only leading contributions in lattice momentum @xmath267 . in this limit the three kernel functions @xmath268 become constants . in addition , we define a @xmath79-independent @xmath269 function by taking the limit @xmath270 . the leading @xmath79-dependence in @xmath271 is @xmath272 . when this term is factored out , we expect that @xmath273 becomes independent of @xmath79 in the limit of large @xmath1 and small @xmath15 . with these simplification and approximation , and converting the discrete summation to integral , we have in the limit of large sizes : @xmath274 where @xmath275 if we use the perturbation expansion hamiltonian , or @xmath276 if we use the effective hamiltonian . we also linearize the dispersion relation so that @xmath277 , @xmath278 or @xmath208 are effective sound velocities for the longitudinal and transverse modes . we shall refer to eqs . ( [ eqsimple - mode - para ] ) and ( [ eqsimple - mode - perp ] ) as simple mode - coupling theory and the finite @xmath1 version as the full mode - coupling theory . assuming that the simplified mode - coupling equations represent the essence of the physics regarding the damping and time - dependent correlation functions , we now consider analytic solution to eqs . ( [ eqsimple - mode - para ] ) and ( [ eqsimple - mode - perp ] ) . first we notice some constraints on the functions . because @xmath279 is a real function , we must have @xmath188^ * = \nu[-z]$ ] . since @xmath280 , we must have @xmath281 , for @xmath282 or @xmath283 ( assuming @xmath284 ) . this implies that the fourier - laplace transform must be integrable : @xmath285 \,dz = \frac{k}{a}.\ ] ] next , the large @xmath167 behavior can be obtained by integrating by part few times , using the boundary conditions @xmath286 , @xmath287 , @xmath288 , and @xmath289 . for simplicity , we have omitted the lattice momentum index @xmath179 and mode index @xmath290 or @xmath283 , since it is true for any @xmath291 . with these results , we find @xmath292 & = & \int_0^\infty \!\!\nu(t)\ , e^{-izt } dt \nonumber\\ & = & \frac{k}{2\pi } \int_{-\pi / a}^{\pi / a}\!\!\!\!\!dq \int_0^\infty\!\ ! g_q^1(t)\ , g_q^2(t)\ , e^{-izt } dt \nonumber \\ & = & \frac{k}{iaz } + \frac{k}{2\pi iz^3 } \int_{-\pi / a}^{\pi / a}\!\!\!\!\!\!dq(\tilde \omega_1 ^ 2 + \tilde \omega_2 ^ 2 ) + o(\frac{1}{z^{5}}),\end{aligned}\ ] ] where for @xmath293 , @xmath294 and @xmath295 are all @xmath296 , while for @xmath297 , it is @xmath298 and @xmath296 . we can establish that @xmath299 \propto ( iz)^{-1/2 } , \quad \nu^\perp[z ] \propto { \rm const},\ ] ] in the limit of small @xmath193 . we proceed as follows : first we assume that @xmath300 = \nu_0^\perp$ ] or @xmath301 . we then derive an expression for @xmath302 and @xmath303 , and show that indeed @xmath303 approaches a @xmath156-function in a proper limit . when @xmath300 $ ] is a constant , the inverse fourier transform for the correlation function @xmath118 $ ] can be performed exactly , to give the transverse correlation function as @xmath304 where @xmath305 , and @xmath306 . we have linearized the dispersion relation , @xmath307 . we only look at the dominant term in time dependence of @xmath302 . for @xmath308 , since large @xmath179-mode decays rapidly , we ignore the term with amplitude proportional to @xmath309 , and also drop the oscillatory term and approximate , @xmath310 . after integrating over @xmath179 ( we also extend the limit as from @xmath311 to @xmath312 ) , we obtain the leading @xmath45 dependence , as @xmath313 if the oscillatory terms are kept , they only contribute an exponentially small term , @xmath314 ; and the @xmath309 terms give a contribution that decays much faster , as @xmath315 . the contribution from @xmath316 term can be neglected , because it decays as @xmath317 ( due to eq . ( [ gpara - of - t ] ) ) . but this term does provide a crossover from @xmath318 for intermediate @xmath167 to the asymptotic @xmath319 for very small @xmath167 . the fourier - laplace transform gives @xmath299 \approx \frac{k^\parallel}{4 } \sqrt { \frac{1}{i\nu_0^\perp z } } = b ( iz)^{-1/2}. \label{nu - para - z}\ ] ] we now get an expression for the longitudinal correlation function . formally , it is given by the inverse transform @xmath320 \over z^2 - c^\parallel{}^2 q^2 - i z \gamma_q^\parallel[z ] } e^{izt } dz,\ ] ] where @xmath321 = q^2 \nu^\parallel[z ] = bq^2/\sqrt{iz}$ ] , @xmath322 . the dominant contribution is from @xmath167 when the denominator is close to zero . the integral can be approximately estimated by the residue theorem . by location the poles @xmath323 we obtain the dispersion relation for @xmath324 . in the limit of small @xmath179 , we find @xmath325 where @xmath326 . therefore , @xmath327 similarly with the dispersion relation for the transverse mode , @xmath328 , we can compute @xmath329 we have neglected the @xmath330 term . since @xmath331 is smaller than @xmath332 , we drop the second term . we symmetrically extend the function . note that @xmath303 can be casted into the functional form @xmath333 which has the property that @xmath334 thus @xmath335 behaves as a @xmath156-function as @xmath336 : @xmath337 this allows us to identify the constant @xmath338 . the condition for @xmath336 is the same as @xmath339 or @xmath340 . since @xmath341 ( c.f . eq.([kratio ] ) ) , small @xmath193 also implies small @xmath208 . although the above derivation suggests that the result is valid only for the special limit . but in fact , this is also the asymptotic result in the limit @xmath342 . to show this , we note that the asymptotic behavior is picked up by the scaling @xmath343 \approx \nu[z]$ ] . if this is true , then @xmath188 \propto z^{-\delta}$ ] . this scaling limit coincides with the limit of @xmath340 . for the simple mode - coupling theory that we have already taken the large @xmath1 limit , we can also apply the fast fourier transform method . however , we find that a direct numerical integration in frequency space is much more accurate . for example , @xmath300 $ ] is expressed as @xmath344 & = & { k^\perp \over ( 2\pi)^3 } \int_{-\infty}^{+\infty}\!\!\!\!\!\!\ ! d\omega \int_{-\infty}^{+\infty}\!\!\!\!\!\!\ ! d\omega ' \,2\,f_\perp(\omega , \omega ' ) \nonumber \\ & & \left ( { \rm p } { 1 \over i(z-\omega-\omega ' ) } + \pi \delta(z-\omega-\omega ' ) \right ) , \label{principle - value - int}\end{aligned}\ ] ] where p stands for cauchy principal value and @xmath345 q^2 \over \omega^2 - ( c^{\parallel,\perp})^2 q^2 - i\omega \nu^{\parallel,\perp}[\omega ] q^2 } .\end{aligned}\ ] ] when the linear approximation is made to the dispersion relations , the integral over @xmath179 can be performed analytically . we only need to do a two - dimensional integral over @xmath346 and @xmath347 . the principal value integral is taken care by locating exactly the singularity . since the integration routine needs a smooth function @xmath188 $ ] , we fit the results by a pad approximation ( in variable @xmath348 or @xmath349 ) . the procedure is iterated several times for convergence . this is programmed in _ mathematica_. it turns out essential to have more than double precision accuracy in the integration routine in order for the singular integrals properly converged . some results of this calculation are already presented in ref . . real part of @xmath350 $ ] ( top ) and @xmath300 $ ] ( bottom ) for data set e. the smooth curves are from simple mode - coupling theory , the ( red ) curves with spikes are computed from full theory at @xmath258 , while the straight , dash lines are analytic asymptotic results . ] real part of @xmath350 $ ] ( solid line ) and @xmath300 $ ] ( dashed line ) for data set j ( @xmath351 , @xmath352 , and @xmath353 ) . the insert shows the effective exponent @xmath354 . ] in fig . [ fig - nu ] , we compare three levels of approximations of @xmath355 $ ] : the simple mode - coupling theory , the full theory computed on @xmath258 by @xmath355 = \gamma_1^\mu[z]/(2\pi/(na))^2 $ ] , and the asymptotic result of @xmath300 = \frac{1}{2 } k^\perp \bigl(1/|c^\parallel - c^\perp| + 1/|c^\parallel + c^\perp|\bigr)$ ] , and eq . ( [ nu - para - z ] ) . noticeable differences are seen between full theory and simplified . this is partly due to the fact that we are comparing a finite @xmath1 with an infinite @xmath1 result . we also note that the asymptotic slope of @xmath356 for @xmath357 $ ] is approached rather slowly . figure [ fig - nu - crossover ] shows a stronger crossover effect from slope of @xmath10 ( corresponding to @xmath9 , to be discussed later ) to that of the true asymptotic law of 1/2 . this set of curves corresponds to data set j. real part ( top ) and imaginary part ( bottom ) of @xmath358 vs @xmath167 for @xmath359 and @xmath360 ( solid line ) , 1/8 ( open circles ) , and 8 ( crosses ) . ] we now look at some of the scaling properties that are implied from eqs . ( [ eqsimple - mode - para ] ) and ( [ eqsimple - mode - perp ] ) . first , we consider the simple case of @xmath361 , @xmath284 , and @xmath362 . in this case the two equations degenerate into one equation given by lepri @xcite . by measuring frequency in terms of @xmath185 ( i.e. , consider variable @xmath363 ) we can scale away @xmath185 , thus the following equation is an exact scaling @xmath364 where @xmath269 is a function of @xmath167 with parameters @xmath365 , @xmath185 , and @xmath19 that we have written out explicitly . this equation tells us that out of the three parameters that characterize the solution , we can pick two as independent . the `` shapes '' of the solutions are all the same for @xmath366 . without loss of generality , we can fix @xmath367 . next , we consider a general scaling solution of the form @xmath368 substituting this scaling form into eq . ( [ principle - value - int ] ) , by requiring that the result must be consistent , we obtain the following conditions @xmath369 where @xmath370 is scaling exponent associated with integration variable @xmath179 , @xmath371 . a unique solution to the set of linear equations is obtained , @xmath372 , @xmath373 , @xmath374 , and @xmath375 . since eq . ( [ lepri - scaling ] ) is ( approximately ) valid for any @xmath376 , we can choose @xmath377 . this scaling solution implies that @xmath378 power - law behavior @xmath318 is obtained in the limit of small @xmath167 , relatively large @xmath185 , and small @xmath19 , if @xmath379 is finite . the crossover to other behavior occurs at large @xmath380 and @xmath381 . a simple dimension analysis also leads to the @xmath318 factor . let the dimension of length and time be @xmath382 and @xmath34 , respectively . then the dimensions of relevant quantities are @xmath383 = t^{-1}$ ] , @xmath384 = lt^{-1}$ ] , @xmath385 = l$ ] , @xmath386 = l^2 t^{-1}$ ] , and @xmath387 = l^3 t^{-2}$ ] . from the five quantities , we can construct three dimensionless variables @xmath388 if there is any relation between the @xmath389 s , it must be in the form @xmath390 ( buckingham pi theorem @xcite ) , or @xmath391 = z^{-1/3 } k^{2/3 } f\left ( { z a\over c } , { k \over a c^2 } \right),\ ] ] where @xmath392 is an arbitrary , dimensionless function . this result , of course , is consistent with eq . ( [ lepri - scaling ] ) . this also suggests that the scaling is not approximate , but an exact result . [ fig - lepri - scaling ] is a test of the above scaling by numerical solutions . for small @xmath167 , the power - law @xmath318 is verified to high accuracy . since this scaling is exact , the deviations are due purely to numerical errors in solving the equations . the case of the coupled longitudinal and transverse equations is somewhat difficult to analyze . we can require a very general scaling of the form @xmath393 and similarly for the perpendicular component . if we require for consistency of scaling for both longitudinal and transverse components , we find that the only scaling solution is the symmetric solution , i.e. , the scaling discussed above with identical longitudinal and transverse scaling exponents . if we abandon the exact scaling for the transverse component and look only at longitudinal component , we can require that @xmath394 as before we can require that there is no scaling for the coupling constant @xmath395 without loss of generality ( i.e. , @xmath396 ) , the above equations imply a relation @xmath397 in particular , if @xmath398 , we must have @xmath399 and @xmath400 . this is consistent with the fact that the scaling region of @xmath401 is obtained for small @xmath208 . a complete and clear scaling picture is still lacking . from the numerical solution , we observe that besides well characterized scaling regimes , there are also plenty of intermediate regions and crossovers . for very large @xmath208 and small @xmath193 , or small @xmath208 and large @xmath193 , the behavior of the solutions are difficult to characterize . in this section , we make the connection of the damping of modes with thermal conduction . the starting point is the green - kubo formula for transport coefficient : @xmath402 for the special case of heat conduction in one - dimensional chain , it is better to call @xmath403 heat conductance rather than heat conductivity . in analogous to electric circuit , the heat conductance relates the temperature gradient to energy current ( fourier law ) , @xmath404 the quantity @xmath405 is related to energy current by @xmath406 . the central quantity that we compute in equilibrium and nonequilibrium heat conduction is the ( total ) heat current @xmath407 . since the heat current is a macroscopic concept determined by the conservation of ( internal ) energy , a microscopic version of it is not unique . the current expression must satisfy the energy continuity equation in the long - wave limit . we derive an expression starting from @xmath408 , where the local energy per particle is @xmath409 . by regrouping some of the terms using translational invariance , we arrive at the heat current per particle : @xmath410 where @xmath411 , @xmath412 , @xmath413 is a unit vector in direction of @xmath414 . the total heat current in @xmath199 direction , @xmath415 is the quantity appearing in the green - kubo formula . it is equal to the macroscopic heat current density ( energy per unit area per unit time ) integrated over a volume . for theoretical analysis , we need to express @xmath407 in terms of the dynamical variables @xmath416 and @xmath230 . a general expression will be rather complicated . again as in mode - coupling approach , we ll consider small oscillation expansions , and consider only the leading contributions . neglecting the nonlinear contribution of @xmath417 , we obtain @xmath418 where @xmath419 is the bare dispersion relation given by eqs . ( [ bare-1 ] ) and ( [ bare-2 ] ) . more general expression in a quantum - mechanical framework including the cubic terms is given in ref . @xcite . an approximation for the correlation function of the current can be obtained , again using a dynamic mean - field approximation , as @xmath420 the above expression can be further simplified by the approximation @xmath421 . we find @xmath422 we have dropped the contribution from the perpendicular component , because it decays much faster ( due to an extra @xmath423 factor in the integrand ) . the green - kubo integrand from md ( circles ) and mode - coupling theory ( curve ) for set e with size @xmath258 . the straight has a slope @xmath424 . the insert shows the same data on a linear scale . ] we compare the mode - coupling result of green - kubo integrand with that of md in fig . [ fig - jt - compare ] . we have used eq . ( [ gk - mkp ] ) for the mode - coupling calculation . note that the correlation functions @xmath425 , @xmath426 , and @xmath427 are simply related in frequency domain through eq . ( [ general - gz ] ) . the agreement is reasonable with the largest deviation about a factor of 2 . the asymptotic behavior of @xmath317 is consistent with both sets of results . the finite lattice conductance defined by green - kubo formula at parameter set e. solid circles are from full mode - coupling theory ; the triangles are from molecular dynamics with periodic boundary condition ; the straight lines have slope of 1/2 . ] the heat conductance on a finite lattice is sensitive to boundary conditions . clearly , in the mode - coupling formulation , we have used periodic boundary condition . if we integrate over time from 0 to @xmath312 first on the second line of eq . ( [ mk - t - dependence ] ) , we obtain @xmath428 . on a finite lattice we have a lower momentum cut - off , @xmath429 . the size dependence of the conductance on finite lattice is then @xmath430 since @xmath401 , @xmath431 . this result is in agreement with that of deutsch and narayan for a model ( the random collision model ) where transverse motion is taken into account only stochastically @xcite . in fig . [ fig - kappa ] , the mode - coupling result of @xmath432 is compared with equilibrium molecular dynamics result with periodic boundary conditions . excellent 1/2 power is observed for the mode - coupling result . we found it is rather difficult to get converged value from molecular dynamics . but the results are consistent with a @xmath433 law . the relation of the mode - coupling theory with the results of nonequilibrium situation of low and high temperature heat - baths at the ends is not clear cut . the standard assumption is that the correlation should be cut off by a time scale of order @xmath1 due to the interaction with the heat baths . finite size result is obtained by integrating the power - law decay to a time of @xmath434 . this gives us the asymptotic behavior for the conductance at large @xmath1 as @xmath435 since we find @xmath401 for small @xmath193 , the thermal conduction diverges as @xmath0 . when @xmath316 is sufficiently large , we observe @xmath2 we note that the interaction potential is not smooth at the point when two particles overlap . this can cause numerical instability . thus , we replaced the original potential with a modified one , @xmath436 which smooths out the discontinuous derivatives at @xmath437 with a small correction of order @xmath438 . in addition to replacing the spring potential by @xmath439 , we also replace the cosine term by @xmath440 so that the value is well - defined for all positions @xmath441 . in actual simulation , we have used very small @xmath442 to @xmath443 so that its effect should be comparable to error caused by finite time - step @xmath444 . we also used large @xmath445 as a way of simulating a slightly different model to study the robustness of the results . we solve the system of equations @xcite @xmath446 where @xmath447 is the total force acting on the @xmath232-th particle , @xmath448 and @xmath449 obey the equation @xmath450 where @xmath451 and @xmath452 are the temperatures of the two heat baths at the ends . the summation is over the particles belonging to the heat bath . we have used 4 particles for the nos - hoover heat baths , with extra first two and last two particles at fixed positions . the coupling parameter @xmath453 is taking to be 1 . for the central part , we used a second - order symplectic algorithm ( or equivalently , the velocity verlet algorithm ) , while for the heat - bath , we used a simple difference scheme accurate to second order in @xmath444 for @xmath454 . although the total energy ( hamiltonian ) is no longer a conserved quantity when the heat - bath is introduced , the above equation still has a conserved quantity of similar character : @xmath455 this quantity can be used to monitor the stability of the algorithm . since we run for very long time steps of @xmath456 to @xmath457 , it is essential that the algorithm is stable over extended period of time . even with a symplectic algorithm , stability is not guaranteed by merely taking small @xmath444 @xmath458 . thus , we adjusted the conserved quantity , eq . ( [ conserved ] ) , to its starting value after certain number of steps . temperature of @xmath232-th particle , @xmath459 vs scaled position , @xmath460 , for data set e with @xmath461 ( plus ) , 256 ( dash line ) , and 1024 ( solid line ) . ] in fig . [ fig - temperature - scaling ] , we observe that good temperature profiles are established . due to relatively large temperature difference between low and high temperature heat baths , and perhaps also due to the nature of heat baths , the profile is not linear . however , the scaling with @xmath1 is approximately obeyed . @xmath462 vs @xmath1 for different parameters @xmath463 . set c : @xmath464 , the chain is compressed to have an average distance between particle @xmath465 ; set g : ( 10 , 0.2 , 0.4 , 0 ) ; set i : ( 0.1 , 0.3 , 0.5 , 0.2 ) ; set k : ( 0.5 , 1.2 , 2 , 0.4 ) ; set l : ( 25 , 1 , 1.5 , 0.2 ) . all of the sets have @xmath32 , mass @xmath31 , and lattice constant @xmath169 . the number indicates the slope of least - squares fit . ] in addition to the thermal conductance results reported in ref . @xcite , fig . [ fig - jl - extra ] gives additional data for variety of parameters . one of the aim of these additional runs is to check the robustness of the @xmath10 power law for thermal conductance . it is found in ref @xcite for set e exponent @xmath466 . this is an excellent confirmation of the mode - coupling theory . we note that at parameter set @xmath365 , the @xmath10 law is not destroyed by introducing large @xmath445 . thus we believe that the @xmath10 law is not specific to the potential used in ref . @xcite . in fact , the fermi - pasta - ulam model with transverse motion was studied @xcite and result is also consistent with the @xmath10 power law . at low temperatures for set i and c , we appear to observe exponent of 0.4 . what is particularly interesting for set c is that the chain is compressed to an average distance of 1.5 from the equilibrium distance of 2 . the behavior is not changed much comparing to uncompressed chain . set g and l represent very large angular coupling @xmath26 . we note that @xmath467 corresponds to a situation that equilibrium can not be established . as @xmath26 becomes larger , the equilibration becomes increasingly difficult . in ref . @xcite we reported a logarithmic behavior for @xmath468 ( set b ) . this logarithmic behavior does not maintain when @xmath26 is increased further . the exponent @xmath469 for set l and g is closed to 1/2 . this may be related the results of disordered harmonic chain . lepri _ et al _ first introduced mode - coupling theory into the problem of heat conduction in low dimensional systems to interpret the md results qualitatively . we have developed the theory further by introducing a ` full theory ' and effective parameters for theory . the full theory is solved numerically . at this level of approximation , we find that mode - coupling theory gives excellent prediction on a quantitative level . the longitudinal damping ( decay rate ) @xmath259 agrees with md data with few per cents of deviation . the mode - coupling theory somewhat overestimated the transverse damping @xmath260 by a factor of two to three . the full theory assumes that the damping function ( memory kernel ) @xmath470 $ ] is an explicit function of two variables , @xmath79 and @xmath167 . in a simple mode - coupling theory , we assume @xmath470 \approx ( 2\pi k/(na))^2 \nu[z]$ ] , valid for small @xmath471 . the simple theory is computationally efficient , and contains essential asymptotic features , i.e. , @xmath264 and @xmath472 . through detailed comparison between md and mode - coupling theory on memory kernel , decay of the modes , the time - dependent correlation functions , green - kubo integrand , and finally the finite - size conductance , we have strong evidence that the mode - coupling theory is correct , and it captures the essential features of the system . we have used parameter set e as the main data set for comparison . the reason for choosing this particular set of parameters is that it gives the cleanest power - law behavior . other parameters will be either qualitatively similar , or crossovers will be observed . in fact mode - coupling theory naturally predicts crossover , due to the presence of @xmath316 term . when this term is dominant , or the transverse effect can be neglected , we should see behavior of @xmath473 . thus we feel that the controversy of the result of lepri _ et al _ @xcite and that of narayan and ramaswamy ( @xmath474 ) @xcite can be reconciled within the current mode - coupling theory . w. is supported by an academic research grant of national university of singapore and by singapore - mit alliance . b. l. is supported by academic research grant of national university of singapore . b. li , l. wang , and b. hu , phys . lett . * 88 * , 223901 ( 2002 ) ; b. li , g. casati , and j. wang , phys . e * 67 * , 021204 ( 2003 ) ; b. li , g. casati , j. wang , and t. prosen , cond - mat/0307692 ; d. alonso , r. artuso , g. casati , and i. guarneri phys . lett . * 82 * , 1859 ( 1999 ) ; d. alonso , a. ruiz , and i. de . vega , phys . e * 66 * , 066131 ( 2002 ) . l. i. manevitch and a. v. savin , phys . e * 55 * , 4713 ( 1997 ) ; a. v. savin and l. i. manevitch , _ ibid . _ * 61 * , 7065 ( 2000 ) ; a. v. savin , l. i. manevitch , p. l. christiansen , and a. v. zolotaryuk , physics - uspekhi * 42 * , 245 ( 1999 ) .

recently , the theoretical and empirical foundations of the physics of traffic flow ( for a review see refs . @xcite ) has come into the focus of the physical community . the motion of individual cars has many peculiarities , since it is controlled by motivated driver behavior , together with some physical boundaries . nevertheless , on macroscopic scales the vehicle ensembles displays phenomena like phase formation and phase transitions widely met in physical systems ( see , e.g. , refs . so , the cooperative behavior of cars treated as active particles seems to be of a more general nature than the mechanical laws and constructing a consistent theory of traffic flow `` from scratch '' is up to now a challenging problem . to describe individual car dynamics a great variety of microscopic models have been proposed . these models differ in the details of the interaction between cars and the time update rule , ranging from differential equations to cellular automata @xcite . there has been a big deal of work on the macroscopic behavior emerging from the microscopic dynamics when exploring the behavior of systems of interacting cars . however , there is still a lot of controversy in both the macroscopic behavior when compared to reality @xcite , and in the microscopic foundations of the individual car dynamics itself @xcite . one currently adopted approach to specify the microscopic governing equations of the individual car motion is the so - called social force model . more details can be found in refs . @xcite ; here only the basic ideas are touched on . at each moment @xmath0 of time , a driver @xmath1 changes the speed @xmath2 of her car depending on the road conditions and the arrangement of the neighboring cars : @xmath3 the term @xmath4 describes the motion of car @xmath1 on the empty road , whereas the term @xmath5 allows for the interaction of car @xmath6 with car @xmath7 ( @xmath8 ) . the interaction is due to the necessity for driver @xmath6 to keep a certain safe headway distance between the cars . all the models mentioned above use various _ anstze _ for the last term . the most interesting special case , which covers the majority of all traffic flow situations , is that of single - lane traffic . here , all cars can be ordered according to their position on the road in the car motion direction @xmath9 , here @xmath10 . most models take into account solely nearest neighboring cars @xmath6 and @xmath11 , i.e. , @xmath12 for @xmath13 and , may be , @xmath14 only . however , more complicated models exist that can be described as models with anticipation @xcite or the so - called intelligent driver model @xcite . the earliest `` follow - the - leader '' models @xcite relate the acceleration @xmath15 of car @xmath1 to the velocity difference @xmath16 only , i.e. , @xmath17 where @xmath18 is the characteristic time scale of the velocity relaxation . in subsequent generalizations of this model @xmath18 became a function of the car motion state , in particular , of the current velocity @xmath19 and the headway @xmath20 ( for a review see refs . @xcite ) . here , @xmath21 is the car length . in refs . @xcite another approach called the optimal velocity model is proposed that describes the individual car motion by @xmath22\ , , \label{int.2}\ ] ] where @xmath23 is the steady - state velocity ( the optimal velocity ) chosen by drivers as function of the headway @xmath24 between the cars . it should be noted that this approach is related to much earlier safety distance models @xcite . concerning the fundamentals of approximations such as @xmath25 , it is noted that there are actually two stimuli affecting the driver behavior . one of them is the necessity to move at the mean speed of traffic flow , in the given case at the speed @xmath26 of car @xmath11 . so , first , driver @xmath1 should control the velocity difference @xmath16 . the other is the necessity to maintain a safe headway distance @xmath27 depending on the current velocity @xmath28 . the `` following - the - leader '' models mainly take into account the former stimulus . the optimal velocity model , conversely , allows for the latter stimulus only . more sophisticated approximations , e.g. , refs . @xcite to name but a few , allow for both stimuli . note also a simple _ ansatz _ called the combined model in ref . @xcite which is also related to the intelligent driver model @xcite : [ imodels ] @xmath29\\ & & { } -\frac{\kappa}{\tau_v}\left[v_\alpha -\vartheta _ { \text{opt}}(h_\alpha ) \right ] \ , . % -\vartheta _ { \text{opt}}(x_{\alpha + 1}-x_{\alpha } ) \right ] \ , , \label{cm}\end{aligned}\ ] ] this equation takes into account both stimuli via a phenomenological coefficient @xmath30 . this is also the case for the helly model @xcite which can be written as ( cf . ref . @xcite ) @xmath31\\ & & { } + \frac1{\tau_v l_h}\left [ ( x_{\alpha + 1}-x_\alpha)-h_{\text{opt}}(v_\alpha)\right ] \ , , \label{helmod}\end{aligned}\ ] ] where @xmath32 is a certain spatial scale and @xmath33 is the optimal headway distance chosen by drivers when moving at speed @xmath34 . ref . @xcite used a linear _ ansatz _ for @xmath33 . later @xcite , this model was generalized to allow for the dependence of the kinetic coefficients on the motion state . however , the question whether a collection of variables such as @xmath35 specifies the acceleration @xmath36 completely is not trivial . drivers are characterized by the motivated behavior rather than physical regularities . for example , memory effects may be essential and can destroy the direct relationship @xmath37 . up to now , memory effects in car following modelling have been treated only in a simplified version . this has been done by relating the current acceleration @xmath38 to the velocities @xmath39 , @xmath40 and the headway distance @xmath41 taken at a previous moment ( @xmath42 ) of time ( for a review of such an approach concerning with the `` following - the - leader '' models see refs . @xcite , regarding the optimal velocity model see refs . @xcite ) . here @xmath43 is the formal delay time in the driver response which is treated as a constant . such an approach , however , is rather formal , because , first , it is not clear why the memory effects relate only two moments of time instead of a certain interval as a whole . second , a simple physiological delay in the driver response as well as the mental driver estimation of the surrounding situation should contribute to the value of @xmath43 . if the latter contribution is essential the delay time @xmath43 is to depend substantially on the state of car motion . moreover , from our point of view the memory effects stem from the fact that the description of individual car motion is a boundary value problem rather than an initial value problem . indeed , at the current moment of time the headway distance and the car velocity are quantities given for the driver beforehand . correcting the car motion she should choose such a driving strategy that in a certain time interval the car velocity and headway distance attain their optimal stationary values , at least , approximately . these memory effects are the topic of this paper . we propose that drivers plan their behavior for a certain time in advance @xcite instead of simply reacting to the surrounding situation . similar ideas related to the optimum design of a distance controlling driver assistance system are discussed in ref . mathematically , the driver s planning of the further motion is just to find extremals of a certain priority functional that ranks outcomes of different driving strategies . the derivation of microscopic governing equations for systems with motivated behavior based on a certain `` optimal self - organization '' principle has been discussed recently @xcite . the assumption adopted in these works is that individuals try to minimize the interaction strength or , equivalently , to optimize their own success and to minimize the efforts required for this . the approach discussed here applies to the concepts of mathematical economics , namely , to the notion of preferences and utility ( see , e.g. , ref . @xcite ) . in ref . @xcite a specific form of the priority functional has been proposed . from that , a certain _ ansatz _ like the combined model ( [ cm ] ) or the helly model ( [ helmod ] ) could be _ derived_. nevertheless , the question how to find the priority functional `` from scratch '' remains open . in the present paper the priority functional is constructed by applying to general properties of the driver behavior for a simplified situation . a car following ( no overtaking allowed ) a lead car which is driven with constant speed @xmath44 is considered . the task is to derive governing equations for the following car motion specified by the time dependence of the velocity @xmath45 and the headway distance @xmath46 . assuming a driver to be able of comparing any two states @xmath47 , @xmath48 , the phase plane @xmath49 can be ordered by a preference relation @xmath50 . therefore , the existence of a cost function @xmath51 may be assumed such that : @xmath52 typically , @xmath53 is called the utility function and ( for it ) the relation @xmath50 matches the inequality @xmath54 . here , the use of condition ( [ 1.1 ] ) is preferred , because then the cost function @xmath51 is similar to the free energy of physical systems , with the minima as the stationary states . obviously , the cost function @xmath51 can not be specified completely because a composite function @xmath55 $ ] also meets condition ( [ 1.1 ] ) for any increasing function @xmath56 $ ] . therefore , at the current stage of the theory development any approximation or _ ansatz _ adopted for the cost function @xmath51 has no meaning . to overcome this problem and to fix a certain class of the cost functions the preference relation defined on a set of car motion paths @xmath57 is considered . a car moving from one point to another can meet different states of traffic flow . it enables one to unite various points on the phase plane in an element treated as a whole with respect to the driver preference , reducing the numerical uncertainty in their evaluation . these parts are supposed , first , to connect the same origin - destination pair @xmath58 and , second , to go through a traffic flow pattern whose structure is known to the driver beforehand . such a set obviously can be ordered by the driver preference and the corresponding cost functional @xmath59 is assumed to exist . here , only the fact that a car motion path can consist of various fragments of the phase plane is concerned with . analysis of the real car dynamics is postponed to the following section . and @xmath60 illustrating the property of the cost functional with respect to path partitioning . letters @xmath61 and @xmath62 stand for the origin and destination points , respectively , and the symbols @xmath63 denote the local states of traffic flow.[f2],width=264 ] figure [ f2 ] shows two paths @xmath1 and @xmath60 which are equivalent with respect to the driver preference . they differ from each other only in the order of passing through the traffic states , @xmath64 for the path @xmath1 and @xmath65 for the path @xmath60 . it demonstrates the fact that if a path goes through a certain small neighborhood of the point @xmath66 one or many times then only the cumulative time interval @xmath67 during which the system was located in the given neighborhood contributes to the cost functional . the corresponding weight coefficient @xmath51 obviously plays the role of a certain cost function for the given traffic state . the resulting cost functional for the whole path is @xmath68dt\ , . \label{1.2}\ ] ] if the driver has incomplete information about the traffic flow pattern ahead she plans the further motion only within some limits , spatial and temporal ones . this can be taken into account by introducing a certain cofactor in front of the function @xmath51 which depends on spatial coordinates or time . the initial introduction of the cost function remains the freedom of its choice too `` wide '' , as was noted above . the constructed form of the cost functional reduces this freedom substantially . namely , let @xmath69 and @xmath70 be two functions entering a cost functional similar to ( [ 1.2 ] ) with the prediction effect such that they preserve the preference relation with respect to the car motion paths on an one - lane road . then , as shown in appendix [ appp : a ] , these cost functions are related by the expressions @xmath71 depending on the spatial or temporal limitation in the driver prediction . here @xmath72 is an arbitrary positive constant and @xmath73 , @xmath74 are some functions of the headway distance @xmath24 or the time @xmath0 . in other words , relations ( [ cf.4 ] ) , ( [ cf.11 ] ) specify the family of the admissible type of cost functions preserving the driver preference relation and , in addition , correspond to linear transformation of the cost functional ( appendix [ appp : a ] ) . for example , @xmath75^{2}$ ] is also a cost functional . however , formula ( [ 1.2 ] ) enables us to fix its specific form being linear with respect to time integration along car motion paths . we note that this construction conforms with regarding the motion cost as the `` physiological '' cost of driving per time , i.e. , having the meaning of a rate . both of these families possess the same freedom in choosing a specific form of the cost function especially dealing with the extremals of the cost functional . in fact , the type ( [ cf.11 ] ) is chosen in the following . in the following , the priority function of the car motion state with respect to the velocity @xmath34 and the reciprocal value @xmath76 of the headway distance will be characterized . note , that the value @xmath77 is not the real car density on a highway . the present analysis ignores the car length , so we have preferred to use the value @xmath77 defined as above . the most preferable state is the motion on an empty road ( @xmath78 ) at maximum speed @xmath79 . this maximum speed is determined by external factors . so , at @xmath80 the cost function @xmath81 has its global minimum : @xmath82 set equal to zero keeping in mind the aforesaid about the freedom in specifying the cost function . since there is no other minimum for the motion on an empty road,@xmath83 can be fixed . because of the car construction a driver can visually control the headway distance within some value @xmath842 m. when the headway distance @xmath24 attains such values a driver has to stop her car because of possible collision even at sufficiently slow velocities . so , @xmath85 coincides with the characteristic headway distance in dense jams where the car density attains the possible maximum . in the following , all headway distances are related to the scale @xmath85 and the dimensionless variable @xmath86 is used . when not moving at all ( @xmath87 ) and the density of cars surrounding the given car does not come close to the limit values , i.e. , @xmath88 , it does not matter how many cars are located in the vicinity . so the assumption that at @xmath87 and for @xmath89 the cost function is independent of @xmath77 can be used : @xmath90 in all other cases , @xmath91 depends on certain combinations of @xmath77 and @xmath34 rather than on @xmath77 individually , at least when @xmath92 . considering the behavior of @xmath91 for a fixed speed @xmath34 it can be stated that driving with small values of @xmath24 ( large values of @xmath77 ) requires a lot of effort . thus , @xmath91 increases with @xmath77 . the opposite case of small values of @xmath77 ( large values of @xmath24 ) deserves special attention . without the possibility of overtaking no especially attractive headway distance can be marked . therefore , we assume that the cost function @xmath93 possesses the only one minimum attained at the boundary point @xmath94 provided the velocity @xmath34 is fixed . keeping in mind the aforesaid it is reasonable to write @xmath95 for @xmath96 . here the exponent @xmath97 is a constant and the function @xmath98 as @xmath99 . since the cost function attains its minimum at the boundary point we may set @xmath100 . moreover , the latter term on the right - hand side of exp . ( [ 1.6na ] ) can be interpreted as a certain `` interaction '' potential between the following and lead cars which is long - distant one for @xmath100 . a detailed analysis of effects caused by the value of the exponent @xmath97 requires an individual study . here the value @xmath101 is actually chosen as just a simple reasonable assumption . since the effect of the surrounding cars is depressed for small values of the car velocity @xmath34 the function @xmath102 is also to attain its minimum at @xmath103 . therefore , @xmath104 as @xmath105 should hold . in other words , inside a certain neighborhood of the origin @xmath106 the following expansion @xmath107 can be adopted . ) .[f3 ] ] taking into account these speculations about the behavior of the cost function @xmath108 caused by variations of both its arguments , the following simple _ ansatz _ will be used subsequently : @xmath109 figure [ f3 ] displays this function . it has only one global minimum at @xmath78 and @xmath110 . for a fixed velocity @xmath34 it attains a local minimum at the boundary @xmath78 . _ ansatz _ ( [ 1.7 ] ) generalizes the adopted assumptions about the cost function . the former term takes into account relations ( [ 1.3 ] ) , ( [ 1.4 ] ) , the latter one is based on approximation ( [ 1.6n ] ) . of course , the parabolic _ ansatz _ ( [ 1.7 ] ) is approximate only in the limit @xmath111 . since this is the main region of interest , this does not constrain its usefulness . to deal with the car dynamics we should construct the cost functional @xmath112 for the car motion paths @xmath113 treated now as continuous functions of time @xmath0 . we note that the time dependence of headway distance @xmath46 gives us the complete information about the car dynamics due to the relationship @xmath114 . leaping ahead , we say that facing this problem it is necessary to introduce additional notions . first , we should expand the phase space in describing the car motion state because transient processes are now the subject of consideration . second , drivers plan their behavior for a certain time in advance instead of simply reacting to the surrounding situation . so we should specify the region inside which a driver can monitor the traffic flow evolution and , thus , plan driving her car . a priority functional similar to eq . ( [ 1.2 ] ) must span the time interval corresponding to this region . beyond it the contribution of the path fragments to evaluating the car motion quality at the given moment of time has to be fairly minor . dealing with transient processes in the car motion we should consider once more the collection of phase variables characterizing the cost of car motion at the current moment of time . keeping in mind conventional driver experience , we will expand the current state of car motion given by headway @xmath24 and velocity @xmath34 with the car acceleration @xmath115 . this is essential because a driver can not change the position and velocity of her car immediately , they vary continuously in time and contain no sharp jumps . conversely , a driver controls the acceleration directly governing the car motion . besides , she can change the acceleration practically without delay because in the present analysis it is quite reasonable to ignore time scales related to physiological properties of the driver or to the mechanical properties of the car . so , the cost function @xmath116 for the motion state @xmath117 can be written : @xmath118 where the time scale @xmath119 s characterizes the acceleration capability of the car . in writing this expression we have assumed that driving without acceleration is preferable . then , the cost function @xmath120 had been expanded into a taylor series with respect to @xmath115 , keeping the leading term only . according to the result to be obtained the time scale @xmath121 entering expression ( [ 2.1 ] ) and the one used in equation ( [ cm ] ) are practically the same . it should be noted that confining ourselves to expansion ( [ 2.1 ] ) with respect to @xmath115 the difference between acceleration and deceleration processes has been lost . in reality , they are different . ignoring this difference leads to models where cars can crash . it is possible to take into account this effect using the approach under development , which , however , is worthy of individual investigation and will be done somewhere else . and its relationship with the limit angle of perception @xmath122.[f4 ] ] for a real driver , various thresholds in the driver recognition of hazards and obstacles exist @xcite . one of them is the distance @xmath123 at which a driver can recognize the behavior of other objects . this distance is usually related to the threshold of the visual angle @xmath122 subtended , for example , by vehicles ahead ( fig . [ f4 ] ) . the value of @xmath122 can be estimated as @xmath12430 min of arc @xcite . the critical angle @xmath122 , the characteristics height @xmath125 of cars , and the corresponding mean distance @xmath123 are related by @xmath126 this allows for the estimation of the recognition distance as @xmath127400 m setting @xmath128 m and @xmath12915 minutes of arc . as previously @xcite , we relate the driver anticipation with the region of size @xmath123 in front that is clearly observable and where she can recognize the car behavior . a driver plans her motion based on the information received by monitoring traffic flow inside the observable region . under normal conditions this region should enable her to govern the motion effectively , for example , to decelerate in advance avoiding a possible accident . therefore its size @xmath123 has to meet the inequality @xmath130 . leaping ahead , we introduce the value @xmath131 treated as a small parameter in the theory to be developed . in particular , for @xmath132 s , @xmath133 km / h , and @xmath134 m we have @xmath135 the size @xmath123 of the recognition distance can be estimated using another argumentation . if @xmath136 s is the typical deceleration time from the velocity @xmath79 attained on the given empty road to zero value then the estimate @xmath137 should hold . for @xmath138 km / h we have again @xmath139 m. now the cost functional of car motion can be written in an integral form as expression ( [ 1.2 ] ) containing the integrand ( [ 2.1 ] ) . the car is assumed to be located at point @xmath140 along the road and to move with speed @xmath34 at the current moment @xmath0 of time . the possible paths of further motion @xmath141 form the set @xmath142 on which the cost functional is defined . here and below gothic letters will be used to label the path variables . the function @xmath143 allows for a derivation of all the dynamical variables , the headway distance @xmath144 , the car velocity @xmath145 , and the acceleration @xmath146 of the trial car path . the cost functional evaluating the quality of the path @xmath143 can be written as @xmath147 \mathcal{f}^d(\mathfrak{h},\mathfrak{u},\mathfrak{a})\,d\mathfrak{t}\,,\\ \intertext{or } \mathcal{l}\{\mathfrak{h}\ } & = \int_t^\infty \exp\left[-\frac{v}{\lambda}(\mathfrak{t}-t ) \right ] \mathcal{f}^d(\mathfrak{h},\mathfrak{u},\mathfrak{a})\,d\mathfrak{t}\ , . \label{2.7}\end{aligned}\ ] ] forms ( [ 2.4n ] ) and ( [ 2.7 ] ) correspond to the driver prediction with spatial or temporal limitations , respectively . the cost functional of form ( [ 2.4n ] ) was used in @xcite . subsequently , functional ( [ 2.7 ] ) is used because its form simplifies the mathematical manipulations when the velocity @xmath44 of the lead car is fixed . in particular , dealing with functional ( [ 2.7 ] ) we can specify directly the set of trial paths @xmath148 using solely the time dependence of @xmath149 of the headway distance . the path variables @xmath150 , @xmath151 are completely determined by the dependence @xmath149 : @xmath152 furthermore , the cost functional ( [ 2.7 ] ) matches the driver prediction with temporal limitation , the mechanism which also allows for the interpretation of the recognition region size in terms of @xmath153 . however , both of the cost functionals lead practically to the same results for the analyzed situation . each one of trial paths originates at time @xmath0 and starts from @xmath154 on the phase plane determined by the headway distance and velocity of the car at the current time @xmath0 . since we investigate the car motion inside traffic flow but not processes of leaving it we assume that time variations @xmath46 of the headway distance are bounded . therefore the trial paths fulfil @xmath155 and do exhibit bounded variations only as time goes to infinity . in other words , there is a constant @xmath156 with @xmath157 in what follows , the driver behavior will be described by the extremals of the cost functional ( [ 2.7 ] ) . using the standard variational technique and taking into account conditions ( [ 2.5 ] ) , ( [ 2.6 ] ) the governing equation for these extremals can be derived : @xmath158 to study the spectral properties of eq . ( [ 2.8 ] ) it is linearized around the stationary solution @xmath159 . its eigenfunctions can be found by the _ ansatz _ @xmath160 the obtained eigenvalue equation is given by @xmath161 with the coefficients @xmath162 and the derivatives are taken at the stationary point @xmath159 . the inequality @xmath163 is supposed to hold beforehand , in particular , it is the case for the cost function ( [ 1.7 ] ) . equation ( [ 2.800 ] ) possesses four roots , one pair @xmath164 of them have positive real parts , the other @xmath165 have negative ones . since the eigenfunctions with the eigenvalues @xmath165 diverge as @xmath166 we must omit them by virtue of condition ( [ 2.6 ] ) . roughly speaking , these divergent eigenfunctions describe the process of a driver leaving traffic flow , for example , to stop the car . such processes are not under consideration . the former pair of eigenvalues are given by : @xmath167^{\tfrac{1}{2}}. \label{2.804}\ ] ] so , in a small neighborhood of the stationary point @xmath159 of equation ( [ 2.8 ] ) any extremal @xmath168 can be written as @xmath169 where @xmath170 and @xmath171 are some constants . equation ( [ 2.8 ] ) is of fourth order , so its general solution is specified by four conditions . solutions that diverge as time goes to infinity have to be omitted ( they are related to @xmath165 ) . their divergence is due to the exponential cofactor in integral ( [ 2.7 ] ) so it is retained beyond a small neighborhood of the stationary point @xmath172 . therefore , only two conditions are needed to specify the desired solution of eq . ( [ 2.8 ] ) . in particular , the initial headway distance @xmath24 and the car velocity @xmath34 determine it completely . the latter statement can also be proved for the more general form ( [ 2.4n ] ) of the cost functional @xcite . containing the current time @xmath0 as its argument . so the found optimal paths @xmath168 depend on the time @xmath0 when the driver plans the further motion . the right panel presents the ideal case where the driver is able to evaluate the motion cost precisely and the corresponding cost function @xmath173 does not contain the time @xmath0 . so , the optimal path family degenerates into one curve . it should be noted that the latter is correct even if the time @xmath174 enters the cost function via variations of the lead car velocity . [ fn ] ] in the previous section the dynamical cost functional was stated . at the next step the driver strategy based on this evaluation of the motion quality should be described ( fig . following @xcite , it is supposed that the driver , first , in planning the further motion chooses the path @xmath175 minimizing the motion cost functional ( [ 2.7 ] ) , @xmath176 to do this the driver `` solves '' equation ( [ 2.8 ] ) subject to conditions ( [ 2.5 ] ) , ( [ 2.6 ] ) and , thus , get the optimal path of further driving , @xmath177 . as noted in the previous section the optimal path choice is completely determined by the current car velocity @xmath34 and the headway distance @xmath24 . the terminal conditions , i.e. , the goal of reaching the steady - state motion is implied . since the driver controls the car motion through choosing the adequate value @xmath115 of acceleration she has to `` find '' the second derivative of @xmath168 with respect to the former argument @xmath174 ( see exp . ( [ 2.4 ] ) ) and , then , to `` calculate '' the result at the current time . therefore , @xmath178 relates the current car acceleration @xmath115 to the current headway distance @xmath24 and the car velocity @xmath34 . formula ( [ 2.61 ] ) describes the driver s choice at the current moment @xmath0 of time . to convert it into the governing equation of car motion we adopt the second assumption that the driver performs this choice continuously : @xmath179 all the car following models based on this equation , but with different particular forms of the cost function @xmath180 may be categorized as the rational car dynamics approximation . equation ( [ 2.62 ] ) holds even if the cost function @xmath181 depends explicitly on the time @xmath0 at which the driver plans her further motion . then , the driver evaluation of traffic flow state will change in time . therefore , the resulting path @xmath182 of the real car motion envelops the family of the optimal paths @xmath183 generated at different moments @xmath0 of time ( fig . [ fn ] , left panel ) . the present paper assumes driving to be perfect . the driver s choice is based on the precise knowledge of the cost function @xmath184 depending solely on the current headway distance @xmath24 , car velocity @xmath34 , acceleration @xmath115 , and , may be , the current time @xmath0 . it does not depend on the moment of time when the driver evaluates the possible paths of her further motion . this assumption leads to what is called the nash equilibrium in game theory @xcite . if , at time @xmath185 the driver has found the optimal path @xmath186 of the car motion ( which meets equation ( [ 2.8 ] ) and does not contain explicitly @xmath185 ) , then any re - computation a certain time @xmath187 later gives the same result , @xmath188 for @xmath189 . in other words , if at time @xmath185 the driver has chosen an optimal path @xmath190 then the further motion will be described by it independently of either the driver follows it without correction or optimizes the car motion continuously ( fig . [ fn ] , right panel ) . the nash equilibrium in the driver strategy enables us to conclude that equation ( [ 2.8 ] ) describes the real car dynamics , not only the imaginary paths existing in the driver s mind during her planing of the further motion . in particular , it can be rewritten in a form containing the real acceleration @xmath115 , velocity @xmath34 , and headway distance @xmath24 : @xmath191 in the vicinity of the stationary point @xmath192 its solution is actually given by _ ansatz _ ( [ 2.56 ] ) , which immediately leads us to the following expressions for the amplitudes [ 2.13 ] @xmath193 and relates the car acceleration @xmath115 to the car velocity @xmath34 and the headway distance @xmath24 : @xmath194 . \label{2.13aa}\ ] ] these results are analyzed individually . the stationary point @xmath195 of equation ( [ 2.9 ] ) gives the headway @xmath196 which the driver chooses in order to follow the lead car at speed @xmath44 . in particular , the expression @xmath197 specifies the relationship between the values of the headway @xmath24 and the car velocity @xmath34 for the stationary traffic flow imitated by the given car following problem . solving eq . ( [ 2.11 ] ) for the car velocity @xmath34 the optimal velocity approximation is obtained , @xmath198 . in particular , for the specific form ( [ 1.7 ] ) of the cost function @xmath51 we immediately get for @xmath199 @xmath200 where the spatial scale @xmath62 is given by expression @xmath201 relation ( [ 2.12 ] ) or similar sigmoid functions are widely used in current literature . in particular , for @xmath202 m and @xmath203 m exp . ( [ 2.51 ] ) gives the estimate @xmath204 m typically ascribed to the spatial scale @xmath62 . and the ratio @xmath205 vs @xmath206 . in drawing these curves exp . ( [ 2.804 ] ) with @xmath207 and @xmath208 was used . the solid curve is @xmath209 , the dashed curve is @xmath210 . [ fbrdm],title="fig:",width=264 ] + by introducing the time scale @xmath211 and the coefficient @xmath212 the car dynamics equation ( [ 2.13aa ] ) can be rewritten as @xmath213.\ ] ] equation ( [ lge ] ) plays a significant role in models that describe non - rational driver behavior @xcite , where expression ( [ lge ] ) specifies an optimal acceleration that could be chosen by rational drivers and the parameter @xmath214 was introduced phenomenologically . here , @xmath214 can be computed using the cost function ( [ 1.7 ] ) . it is plotted as well as the ratio @xmath205 as a function of the variable @xmath206 ( fig . [ fbrdm ] ) . this representation is due to the fact that the coefficient @xmath215 ( for the cost function ( [ 1.7 ] ) @xmath216 for @xmath217 ) , the coefficient @xmath218 by virtue of the adopted assumption ( [ 2.sigma ] ) , whereas the coefficient @xmath206 varies around unity . below , the latter will be justified and the dependence of the coefficient @xmath206 on the car motion state will be analyzed . as seen in fig . [ fbrdm ] the time scales @xmath18 and @xmath121 practically coincide with each other and the parameter @xmath214 is a small value . using again the cost function eq . ( [ 1.7 ] ) , and taking into account @xmath219 and @xmath216 , the optimal velocity dependence @xmath220 is given by eq . ( [ 2.12 ] ) . the parameter @xmath206 depends as follows on the car motion state @xmath221 it attains the maximum @xmath222 at @xmath223 . using @xmath224 s , @xmath138 km / h , @xmath225 m , and @xmath226 m , respectively , @xmath227 is obtained . of course , this is not a precise numerical value , but in general @xmath228 may be assumed . ( labelled by numbers in squares ) . the solid curves display the relaxation caused by a perturbation in the headway , the dashed one shows the same for a velocity perturbation . the variations in the headway distance and relative velocity are measured either in units of the initial values @xmath229 and @xmath230 or @xmath231 and @xmath232 , respectively . [ frc],title="fig:",width=264 ] + ( labelled by numbers in squares ) . the solid curves display the relaxation caused by a perturbation in the headway , the dashed one shows the same for a velocity perturbation . the variations in the headway distance and relative velocity are measured either in units of the initial values @xmath229 and @xmath230 or @xmath231 and @xmath232 , respectively . [ frc],title="fig:",width=264 ] + the precise value of @xmath233 matters . according to sec . [ sec : ee ] the car dynamics is characterized by two eigenfunctions with eigenvalues given by eq . ( [ 2.804 ] ) . when @xmath234 , the relaxation is a pure fading process characterized by time scales @xmath235 and @xmath236 . otherwise , @xmath237 , the relaxation process is characterized by the complex eigenvalues , @xmath238 , @xmath239 . however in this region their dependence on @xmath206 is sufficiently weak , @xmath240{\omega}$ ] for @xmath241 as follows from exp . ( [ 2.804 ] ) . therefore the car relaxation processes should be rather insensitive to the value of @xmath206 as it varies from unity to about ten . figure [ frc ] demonstrates this fact showing the relaxation curves induced by deviations from the stationary values of the headway and the car velocity individually . all the corresponding curves lie near each other , only for @xmath242 a larger deviation is visible . this is because for such values of @xmath206 the eigenvalues @xmath243 differ essentially . by contrast , the cost of acceleration , i.e. , the cost of correcting the car motion depends strongly on @xmath206 because this component of the cost function ( [ 2.1 ] ) varies linearly with @xmath206 to the first approximation . since the maximum @xmath233 of the parameter @xmath206 is attained at @xmath244 this means that the greater the value of @xmath206 , the larger is the cost of the car governing with respect to the cost of motion in its own accord . keeping in mind the aforementioned speculations let us assume that the composed parameter @xmath233 takes a certain fixed numerical value , @xmath245 . it leads immediately to a certain relationship between the initial quantities of the model , @xmath79 , @xmath123 , @xmath121 , @xmath85 . in this collection only the optimal velocity @xmath246 on empty road is determined solely by the driver behavior . the other parameters , namely , the size @xmath123 of recognition region , the characteristic time scale @xmath121 of the car ability to accelerate or decelerate , and the characteristic headway @xmath85 that drivers can control reliably and should be about the typical headway distance in dense jams are determined by other mechanisms , e.g. , car properties , driver physiology , visual conditions and the like . at first glance such relations seem to be impossible , because a driver chooses the optimal velocity on empty road depending on the road conditions , speed regulations , weather , _ etc_. nevertheless , there is an additional factor affecting the car motion on empty roads , the driver experience . if the driver preference for the empty road velocity stems from here experience of driving in low density traffic flow then such relations can exist . in the given model this assumptions lead to the expression @xmath247^{\tfrac12 } \frac{l^{\frac14}\lambda^{\frac34}}{\tau}\ ] ] naturally , the specific form of relation ( [ 2.vmax ] ) will change if , for example , another _ ansatz _ of the cost function is used , however , its qualitative features should be retained . these speculations , definitely , are not enough to regard relations similar to formula ( [ 2.vmax ] ) as an established theoretical result . we have presented it for discussion only . as discussed already there are two stimuli for a driver to change the current motion state . one is the velocity difference @xmath248 between the leading car and her car . the second is the difference between the current speed @xmath34 and the optimal speed @xmath23 as function of headway . models like ( [ imodels ] ) take into account both of them , leaving the determination of the weight coefficients to appropriate empirical or experimental data . the results obtained so far actually allow to predict these coefficients . of the combined model ( [ cm ] ) vs the stationary headway @xmath249 . for comparison , the dashed curve shows the optimal velocity , eq . ( [ 2.12]).[fkap],width=264 ] this can be demonstrated with the combined model ( [ cm ] ) . linearizing eq . ( [ cm ] ) around the stationary point @xmath172 and comparing the result with eq . ( [ 2.13aa ] ) we get @xmath250 ^{-1}. \label{2.cfp4}\end{aligned}\ ] ] this expression gives @xmath251 as function of the car motion state . figure [ fkap ] plots this dependence for the cost function ( [ 1.7 ] ) . as seen in fig . [ fkap ] the weight coefficient @xmath251 is small for all interesting values of the headway distance in the car following regime . only for small headways @xmath252 ( dense jams ) or for large headways @xmath249 exceeding @xmath62 substantially ( free flow ) @xmath251 approaches unity . as the headway distance @xmath249 varies in the interval @xmath253 the coefficient @xmath254 changes essentially , leading to qualitatively different car dynamics . as mentioned before , when @xmath255 the relaxation exhibits damped oscillations around the stationary point @xmath172 . this is due to the eigenvalues @xmath256 ( see exp . ( [ 2.804 ] ) ) having non - zero imaginary parts . since for @xmath257 their real and imaginary parts are of the same order the car motion relaxation is characterized by one time scale about one ( in units of @xmath121 ) . this is true also for @xmath258 because @xmath259 there , although the dynamics is a pure fading process now . for still smaller values of @xmath260 the ratio @xmath261 of these eigenvalues becomes smaller than one half , which may be defined as the two - scale regime of the dynamics , see fig . [ fro ] . of the corresponding eigenfunctions describing the car motion relaxation as a function of @xmath206 or the headway @xmath249 when @xmath262 is fixed . computations in this figure used cost function ( [ 1.7 ] ) and @xmath263.[fro],width=302 ] the change between the two - scale and the one - scale regime corresponds to a value of @xmath2641.5 @xmath62 . below this value , the dynamical behavior is one - scale , above it is two - scale and will be called `` fast - and - slow '' in the following . concerning with the `` fast - and - slow '' dynamics the velocity relaxation and the headway relaxation can be analyzed individually . according to exps ( [ 2.13 ] ) the initial difference @xmath265 contributes mainly to the eigenfunction with the eigenvalue @xmath239 . thereby , the velocity difference @xmath248 contributes mostly to the amplitude @xmath266 . the amplitude @xmath267 also contains the term of the same magnitude , however , the time scale @xmath236 on which the corresponding eigenfunction varies is much bigger than the time scale @xmath235 of the other eigenfunction . so , the velocity relaxation falls on the first eigenfunction . thus , the velocity difference @xmath248 disappears practically completely during the time @xmath235 , which forms the `` fast '' stage of the car relaxation . at the next `` slow '' stage of duration @xmath235 the headway deviation from the equilibrium value @xmath249 disappears . to summarize , the `` fast - and - slow '' car dynamics is a two - stage process where the velocity difference between the cars is eliminated first . later on , the headway is optimized . while this is being done , the resulting velocity difference is not essential and the driver can govern the car motion without the necessity of responding fast . . [ fol],width=264 ] another important characteristics , separating dense traffic and quasi - free flow can be derived as follows . the solution of the eigenvalue equation ( [ 2.800 ] ) depend actually on two values , @xmath268 and @xmath206 , since @xmath216 for the cost functional ( [ 1.7 ] ) . the parameter @xmath269 by virtue of the adopted assumption ( [ 2.sigma ] ) . the maximum of @xmath206 attained at @xmath270 is much larger than @xmath271 as it results from exp . ( [ 2.omegamax ] ) . however , as the headway distance increases the value @xmath254 decreases as @xmath272 ( see exp . ( [ 2.omega ] ) ) whereas @xmath273 . so there is a value @xmath274 of the headway distance at which both these terms contribute to the eigenvalues to the same extent . to find @xmath274 consider @xmath275 where @xmath276 and @xmath277 , thereby , @xmath278 and @xmath279 by virtue of formula ( [ 2.804 ] ) so , @xmath274 is specified by @xmath280 the value @xmath274 divides the headway distance region into two parts , see fig . [ fol ] . when @xmath281 the velocity of the following car is close to @xmath79 , so this type of car motion will be referred to as the quasi - free flow . in this case the eigenvalue equation ( [ 2.800 ] ) can not be simplified , so to describe the quasi - free flow the cost functional ( [ 2.7 ] ) only in its full form may be used . in the opposite case , @xmath282 , which will be called dense traffic mode , the term @xmath268 is ignorable , reducing the eigenvalue equation ( [ 2.800 ] ) to @xmath283 moreover , in the latter case the description of car dynamics can be reduced to the standard form of classical mechanics , enabling us to analyze the nonlinear stage of the car dynamics . in the dense traffic limit @xmath282 , the variational principle based on optimizing the cost functional ( [ 2.7 ] ) can be simplified essentially . in the given limit the characteristic time scales of the car dynamics are @xmath121 and @xmath284 , with both of them being small in comparison with @xmath285 . the latter follows from the adopted assumption ( [ 2.sigma ] ) and the condition @xmath286 for @xmath282 . in this case , as shown in appendix [ app : b ] , the cost functional ( [ 2.7 ] ) can be replaced by the following effective functional whose integrand does not contain a time dependent factor @xmath287 where the integrand , which will be called the lagrangian of the car dynamics is given by @xmath288 interestingly , @xmath289 contains the lead car velocity @xmath44 as a parameter and attains its extremal value with respect to @xmath24 at the point @xmath196 corresponding to the optimal driving with the velocity @xmath44 . the term @xmath290 has been introduced for the sake of convenience only , it does not affect the extremal equation but enables the lagrangian to attain a minimum with respect to the car velocity @xmath34 at the stationary point @xmath172 . both functionals ( [ 2.7 ] ) and ( [ 10.2 ] ) possess the same extremals to the first order in the small parameter @xmath291 . in particular , the extremals of functional ( [ 10.2 ] ) meet the equation @xmath292 it corresponds directly to the initial full equation ( [ 2.9 ] ) where the second and third terms in the former parentheses are ignored whereas the second term in the latter parentheses is replaced by its value taken at the stationary point . it is justified because these terms are due to variations of the time dependent cofactor . keeping in mind its following applications we rewrite lagrangian @xmath289 also in the form @xmath293 in particular , for the cost function ( [ 1.7 ] ) we get @xmath294 in obtaining formula ( [ 10.302 ] ) we have taken into account the expression ( [ 2.12 ] ) relating the velocity @xmath44 to the headway distance @xmath249 , expression ( [ 2.omega ] ) , and omitted some insignificant terms . the latter term in expression ( [ 10.301 ] ) can be also omitted because it has no effect on the extremal equation . the most essential feature of the effective cost functional ( [ 10.2 ] ) is the absence of a time dependent cofactor . this enables us to reformulate the car dynamics in terms of autonomous hamiltonian equations . finding the extremals of functional ( [ 10.2 ] ) can be done as follows . the phase space @xmath295 is expanded to @xmath296 by adding the relationship between the headway distance @xmath24 and the car velocity @xmath34 as an additional constraint : @xmath297 and the initial and final conditions , @xmath298 for the sake of simplicity the parameter @xmath44 has been omitted in the list of variables of @xmath299 , the current time is set to zero . by using a lagrange multiplier , the problem ( [ app.4 ] ) is rewritten in the standard form . namely , the extremals of @xmath300dt \label{app.7}\ ] ] are sought . they are defined on the extended set of variables @xmath301 and subject to conditions ( [ app.6a ] ) , ( [ app.6b ] ) . the extremals of functional ( [ app.7 ] ) obey the classical lagrange equations . in constructing the hamiltonian @xmath302 that produces the same extremal equations the pontryagin technique is used @xcite . introducing the new variable @xmath303 @xmath304 and solving for @xmath305 , i.e. , finding @xmath306 , the desired hamiltonian is written as @xmath307 it can be demonstrated directly that the desired extremals meet the standard hamiltonian system of equations @xmath308 again , the optimization of functional ( [ app.4 ] ) actually leads to a boundary value problem because the extremal is specified by both the initial and final conditions ( [ app.6a ] ) , ( [ app.6b ] ) . the hamiltonian approach ( [ app.h1])([app.h2 ] ) also shares this property . namely , the initial values of the variables @xmath24 and @xmath34 are given by conditions ( [ app.6a ] ) , whereas the initial values @xmath309 and @xmath310 of the quasi - momenta @xmath311 and @xmath303 should be chosen such that the system tend to the stationary point @xmath172 as @xmath312 . it is the same situation that we met in dealing with equation ( [ 2.9 ] ) however , the hamiltonian description has a certain advantage . first , the hamiltonian itself is conserved,@xmath313 second , an additional autonomous first integral of the system ( [ app.h1 ] ) , ( [ app.h2 ] ) , can be found , at least , for a certain stage of the car dynamics . fortunately , this is the case for the `` fast - and - slow '' car dynamics analyzed below . note , that this hamiltonian description of the car motion relaxation towards the stationary state does not contradict the conservation of phase volume being the general property of hamiltonian systems . the matter is that the relaxation process is described by only _ one _ path leading to the stationary point of the saddle type . other possible paths are not considered . the two eqns ( [ app.h1 ] ) can be understood readily . the equation for @xmath314 is just @xmath315 , while the equation for @xmath316 , or , what is the same , expression ( [ app.q ] ) shows that the quasi - momentum @xmath303 is proportional to the car acceleration @xmath317 : @xmath318 the physical meaning of the quasi - momentum @xmath311 is more complex . to clarify it , rewrite the equation line ( [ app.h2 ] ) in terms of partial derivatives of the lagrangian @xmath319 . namely , the definitions of the quasi - momentum @xmath303 ( [ app.q ] ) and the hamiltonian ( [ app.h ] ) enables us to represent these equations as @xmath320 in particular , expressions ( [ 10.100])([10.102 ] ) immediately lead , as it must , to equation ( [ 10.0 ] ) . expression ( [ 10.101 ] ) demonstrates that the quasi - momentum @xmath311 includes the rate @xmath321 of acceleration variations . in this way the variable @xmath321 implicitly enters the hamiltonian @xmath302 specifying a hyper - surface in the four - dimensional phase space @xmath322 . even more , by ignoring the dependence of @xmath319 on the headway distance @xmath24 , the velocity relaxation towards the stationary value @xmath44 is governed by the equation @xmath323 it results from the effective cost functional ( [ 10.2 ] ) where the car velocity @xmath34 is treated as a primary argument . as will be seen below , there is a stage of the car dynamics where this assumption is justified and the quasi - momentum @xmath324 is conserved . only the dependence of the lagrangian @xmath319 on the headway distance @xmath24 , see eq . ( [ 10.102 ] ) , causes variations in the quasi - momentum @xmath311 . since the velocity control is of primary importance in the driving strategy , the variable @xmath311 can be regarded as a certain measure of the necessity to control also the headway distance when eliminating the velocity difference . for @xmath325 the last term on the right - hand side of exp . ( [ 10.3b ] ) is small . since the headway @xmath24 enters the hamiltonian ( [ app.h ] ) exactly via this term time variations in the quasi - momentum @xmath311 are retarded by virtue of the first equation at line ( [ app.h2 ] ) . the same follows from eq . ( [ 10.102 ] ) . the other variables @xmath24 , @xmath34 , @xmath303 can vary substantially on scales independent of the value @xmath325 . then , the car dynamics can exhibit multi - scale relaxation . in particular , sec . [ sec : tcmd ] demonstrated this fact , namely , it has been shown that in the given limit the system dynamics comprises the fast and slow stages . exactly the time scale of the latter stages depends on the value @xmath326 . during the former stage the quasi - momentum @xmath311 should be practically constant , i.e. , it is approximately a first integral of the system ( [ app.h1 ] ) , ( [ app.h2 ] ) . this property , together with the conservation of hamiltonian ( [ app.h ] ) , allows the integration of the system of equations ( [ app.h1 ] ) , ( [ app.h2 ] ) . during the following slow stage a quasi - stationary approximation can be used to analyze the system evolution . the results below are exemplified with the cost function ( [ 1.7 ] ) . they can be easily generalized to other cost functions . at the zeroth approximation in the small parameter @xmath206 the quasi - momentum @xmath311 is a constant . during the fast stage mainly the velocity deference @xmath248 is eliminated , so eq . ( [ 10.101 ] ) formally describes the relaxation process of the car velocity @xmath34 to the stationary value @xmath44 . therefore , by virtue of exp . ( [ 10.301 ] ) , the quasi - momentum @xmath311 takes zero value , @xmath327 . in this limit the lagrangian component @xmath328 as well as the additive constant of the component @xmath329 can be omitted . so , the hamiltonian ( [ app.h ] ) becomes @xmath330 and the non - canonical variables @xmath295 are used . since the hamiltonian @xmath331 is conserved and the fast stage describes the velocity relaxation to @xmath44 the car dynamics obeys the equation @xmath332 this immediately gives the relationship between the acceleration @xmath115 and the velocity @xmath34 @xmath333 the sign in the latter equation has been chosen so to allow for the system relaxation . equation ( [ fs.1 ] ) is , in fact , of the general form and the linear form of eq . ( [ nad12 ] ) is due to the adopted quadratic _ ansatz _ of the cost function but not a consequence of linearization . besides , equality ( [ fs.1 ] ) can be read as follows . the comparison of the hamiltonian parts @xmath334 and @xmath335 with the corresponding components of lagrangian ( [ 10.3b ] ) demonstrates that the function @xmath334 actually measures the cost of the car acceleration and the function @xmath335 does the same with respect to the car motion relative to the optimal driving conditions . thereby , during the fast stage the driver corrects the car dynamics so that the cost of acceleration be equal to the cost of current motion measured relative to the stationary conditions . and the car velocity @xmath336 where the car dynamics is collisionfree ( based on equation ( [ nad13])).[fcol ] ] the headway distance @xmath24 does not enter explicitly the governing equation of the fast stage . however , it varies during the fast stage and finally attains a value @xmath337 differing from the initial value @xmath338 . in the adopted simple approximation of the cost function eq . ( [ nad12 ] ) enables to estimate easily the value @xmath337 . namely , the direct integration of eq . ( [ nad12 ] ) yields the relationship @xmath339 applying to formula ( [ nad1010 ] ) it can be seen that expression ( [ nad13 ] ) holds until the headway distance @xmath24 becomes too small and it is impossible to ignore the effect of the term @xmath340 , i.e. , when @xmath341{\omega}\ , h_v \ll h_v\,.\ ] ] the car dynamics in the region of small values of the headway distance when the probability of collision is high is worth of an individual consideration . here we only touch on this problem by assuming the collision to happen when the value @xmath337 given by exp . ( [ nad13 ] ) becomes equal to zero . this assumption is justified as a rough approximation due to estimate ( [ nad14 ] ) , see fig . the boundary of the collisionless region can be shifted to the right substantially by dropping the assumed quadratic dependence of the cost function on acceleration . this will be done elsewhere . when the system attains a quasi - equilibrium with respect to the car velocity @xmath34 its further evolution is due to the direct dependence of the hamiltonian @xmath342 on the headway distance @xmath24 . it enters the hamiltonian via @xmath328 . as a result , the velocity difference @xmath248 and the acceleration @xmath115 should be small . keeping this in mind , eq . ( [ 10.101 ] ) can be solved for @xmath311 . substitution of the obtained expression into eq . ( [ 10.102 ] ) , and the further linearization with respect to the variable @xmath248 and the component @xmath343 results in @xmath344 if @xmath345 is the characteristic time scale of the slow stage then @xmath346 these estimates together with eq . ( [ st.1 ] ) lead to @xmath347 in particular , the rate @xmath348 of time variations in the quasi - momentum @xmath303 scales as @xmath349 for @xmath350 . since the lagrangian @xmath299 has a minimum at @xmath351 the quasi - momentum @xmath311 can be estimated as @xmath352 by virtue of eq . ( [ 10.102 ] ) . in the limit @xmath325 the term @xmath348 can be ignored in comparison with the quasi - momentum @xmath311 . then , eq . ( [ 10.101 ] ) immediately leads to the relation @xmath353 which is no more than the standard expression for the momentum of a system described by the lagrangian @xmath354 with @xmath355 . by the same reasons all the terms in hamiltonian ( [ app.h ] ) containing the quasi - momentum @xmath303 may be omitted . then the substitution of ( [ st.3 ] ) into ( [ app.h ] ) yields the hamiltonian of the slow stage @xmath356 in the lagrangian component @xmath328 the car velocity @xmath34 has been replaced by its stationary value @xmath44 , and some negligible terms have been omitted . then the conservation of the hamiltonian @xmath357 during the system dynamics gets the form @xmath358 the conservation law ( [ st.6 ] ) can be understood in analogy to that for the fast stage . however , during the slow stage of the car dynamics the car velocity plays the role of the control parameter . the driver changes the speed to correct the headway distance . again the cost of deviation of the car velocity from the stationary value is chosen to be equal to the cost of the car motion measured relative to the optimal conditions . it should be pointed out that the slow stage of the car dynamics is governed by the conservation law , exp . ( [ st.6 ] ) , which contains only the headway distance @xmath24 and the velocity @xmath34 . at the first approximation , the acceleration @xmath115 does not enter at all . in this meaning , the slow stage is similar to other physical systems . however , the stationary point of the car dynamics is a saddle point rather than a minimum . up to now , the two different stages of car - following discussed already in sec . [ sec : tcmd ] have been identified with two different conservation laws . both of them can be unified into just another effective conservation law interpolating exps . ( [ fs.1 ] ) and ( [ st.6 ] ) : @xmath359 within the same accuracy it is possible to interpolate directly eqs . ( [ nad12 ] ) and ( [ st.7 ] ) , leading to @xmath360 . \label{nad1010}\ ] ] besides , the proposed interpretations of the conservation laws ( [ fs.1 ] ) and ( [ st.6 ] ) enable us to formulate a generalized principle of adequate control . it declares that the effective cost of correcting the car motion via changing the car acceleration ( fast stage ) or the car velocity ( slow stage ) is equal to the cost of the current state of car motion measured with respect to the optimal driving conditions . a variational approach to the description of car dynamics has been developed in case of a car following a lead car moving at a constant speed . to derive governing equations for the following car motion the driver preference has been used to construct a cost functional . its extremals specify the optimal paths of the further motion for the following car . applying to the general properties of the driver behavior we analyzed the basic properties of the cost function and proposed a simple parabolic _ ansatz _ , which , nevertheless , catches typical features of traffic properties . the concept of a rational driver is formulated . it comprises the assumptions that the driver follows the optimal paths and corrects the car motion continuously . in this case the optimal path is a nash equilibrium of the system , which is defined as follows . if the driver has chosen an optimal path at a certain moment of time then no further correction of the car motion is necessary because it leads to the same result . as the consequence of the nash equilibrium the extremals of the cost functional specify the real dynamics of cars with rational drivers although originally they determine only the imaginary paths in the driver s mind when planning the further motion . in this way we obtained several results . the optimal velocity approximation has been derived . the weight coefficient entering a car following model combining the `` following - the - leader '' model and the optimal velocity model has been found depending on the headway distance . as an important result it has been shown that the car dynamics can be categorized under different types according to its properties . first , we have shown that there can be two types of the car relaxation towards the stationary motion , the mono - scale dynamics and the `` fast - and - slow '' dynamics . second , we singled out the quasi - free motion and the dense traffic mode . the variational technique for the latter mode can be simplified essentially . namely , it is possible to reformulate the cost functional so that it does not contain a time dependent cofactor . as a result , the autonomous hamiltonian description for the car dynamics has been constructed . for the dense traffic mode the `` fast - and - slow '' dynamics has been analyzed for the nonlinear stage . in particular , different conservation laws for this stages have been found . a generalized principle of adequate car motion control has been proposed . we assumed that the lead car moves at a fixed speed . if it is not the case two different situations should be singled out . for perfect drivers who can predict rigorously the motion of the car ahead the main results still hold . otherwise the cost function will depend not only on the current time but also , what is crucial , on the time when the driver has started to evaluate the further car dynamics . this disturbs the nash equilibrium and the car correction should be carried out continuously . briefly this problem was studied in @xcite but it actually requires a more detailed investigation . real drivers have certain limitations . they are not capable of finding the optimal path precisely and they can not correct the car motion continuously . so , the concept of rational driver behavior is just the first approximation of the real situation and deviations from this perfect behavior should be analyzed consistently . a first step towards this problem can be found in ref . here we justify some assumptions adopted there and substantiated them in a different way . let us again concern the driver evaluation of the car motion quality . for a given path of the car s further motion @xmath361 the cost functional ( [ 2.7 ] ) is written as @xmath362 \,.\end{gathered}\ ] ] expression ( [ br.1 ] ) can be represented also in the form @xmath363 \nonumber\\ & + \frac{2v}{\vartheta_{\text{max}}^{2}}(h - h_{v})+\mathcal{l}_{0}\ , , \label{br.2}\end{aligned}\ ] ] where @xmath364 describes the deviation of the given path from the stationary car motion trajectory , @xmath365 here , @xmath366 is the cost of stationary car motion , @xmath367 in addition , when deriving expression ( [ br.1 ] ) we have used @xmath368 integrated various fragments of eq . ( [ br.1 ] ) by parts , assuming @xmath369 , and dropped terms like @xmath370 where ever possible . when the driver plan her further motion the current values of the headway @xmath24 and the velocity @xmath34 are regarded as the initial conditions . so , when choosing the optimal path she can minimize only the first term @xmath371 of expression ( [ br.2 ] ) . this optimization is implemented through the adequate control over the car acceleration @xmath115 , so , exactly the acceleration plays the role of the control parameter available for the driver actions . a real driver can only approximately evaluate the quality of motion . let us describe the threshold in the driver perception of the motion quality by @xmath372 where @xmath373 is a small constant , @xmath374 , because definitely a driver can recognize the difference in the state of staying and freely motion on an empty road . when the controllable part @xmath375 of the cost motion is much smaller then the threshold ( [ br.4 ] ) the driver actually has no information of how to govern the car motion . in this case it is natural to assume that he will not do anything with respect to the car driving and , so , will fix the car motion at the current state , including the current acceleration @xmath115 . thereby the inequality @xmath376 determine the region in the phase space @xmath295 inside which the driver can not control the car motion . in evaluating such a driver behavior with expression ( [ br.2 ] ) we can formally treat @xmath377 , @xmath378 , and @xmath379 as constants independent of one another . so setting @xmath380 , @xmath381 , and @xmath382 and taking into account the relation between @xmath196 and @xmath44 we get from condition ( [ br.5 ] ) the approximate boundary of this region in the following form @xmath383 we recall that the parameter @xmath214 entering eq . ( [ lge ] ) coincides with @xmath384 . so we reproduced here the expression for the rational driving boundary as has been introduced in paper @xcite by a different line of reasonings . these investigations were supported in part by rfbr grants 01 - 01 - 00389 , 00439 , ur grant 01.03.005/2 , intas grant 00 - 0847 , and russian program `` integration '' project b0056 , during a stay of one of the authors ( il ) at the institute for transport research of the german aerospace centre . a set @xmath385 of one - dimensional paths @xmath386 of car motion is considered . all the paths are supposed to originate from one point @xmath387 at the initial time @xmath388 , i.e. @xmath389 , and to go to infinity , @xmath390 , as @xmath312 such that the car velocity @xmath391 is bounded during the motion , @xmath392 ( here @xmath62 is some constant ) . it is assumed that the set @xmath385 is ordered by a certain preference relation @xmath393 and there is a functional @xmath394 of the form @xmath395 measuring numerically the preference relation , i.e. meeting the condition @xmath396 here , first , the integrand @xmath397 depends on the car position on the road @xmath398 , the velocity @xmath45 , and , may be , the current moment @xmath0 of time . it plays the role of cost function , i.e. evaluates the quality of the car motion state at current time . second , the weight factor takes one of the forms @xmath399 where @xmath400 and @xmath401 as @xmath402 or @xmath403 . the functions @xmath404 and @xmath405 describe the spatial or temporal limitations in the driver prediction , respectively . the examples @xmath406 where @xmath407 and @xmath44 are some constants , enable us to relate the cost functional ( [ appp1 ] ) to ones used in the main text . the specific form of @xmath404 and @xmath405 , however , does not affect the results to be obtained in this appendix . let @xmath408 be another cost functional of form ( [ appp1 ] ) with its own integrand @xmath409 that also describes the same preference relation . the purpose of the present appendix is to derive the relationship between the cost functions @xmath397 and @xmath409 . for this purpose the path set @xmath385 is divided into classes @xmath410 of equivalent paths , i.e. @xmath411 if @xmath412 or , what is the same , @xmath413 . the cost functional @xmath414 based on the integrand @xmath409 has to preserve this class partition . so the condition @xmath415 must hold for any infinitesimal perturbation @xmath416 of a path @xmath398 within the class @xmath417 $ ] . using the lagrange multiplier technique and taking into account the initial condition @xmath418 ( thus , @xmath419 ) the latter requirement is reduced to the equation @xmath420 where the function @xmath421 and @xmath422 is a certain constant for all the points @xmath423 on the phase plane @xmath424 related by the given curve @xmath398 . let us show that the value of @xmath425 is constant for _ all _ the points of the phase plane . the cost functions @xmath409 , @xmath397 are chosen , thereby , if the _ quasi - local _ equation ( [ appp4 ] ) holds , the value of @xmath426 must be determined by any fragment of the path @xmath398 under consideration , for example , by a fragment @xmath427 . for @xmath428 and @xmath429 any point @xmath430 can be joined to the fragment @xmath431 by a certain path @xmath432 which must be characterized by the same value of @xmath425 . the latter proves this statement because the time moments @xmath433 have been taken arbitrary . the cost function @xmath397 ( as well as @xmath409 ) and the weight factor @xmath434 describe different effects of the car driving , so they are not related to each other . thus the terms of eq . ( [ appp4 ] ) containing @xmath435 , @xmath436 , or @xmath437 should be equal to zero individually . thereby @xmath438 and @xmath439 the solutions of system ( [ appp5 ] ) , ( [ appp6a ] ) and system ( [ appp5 ] ) , ( [ appp6b ] ) have the forms @xmath440 respectively , where @xmath441 and @xmath442 are arbitrary functions of @xmath140 and @xmath0 . thereby the desired relationship of the cost functions @xmath397 , @xmath409 is given for @xmath443 by the expression @xmath444 where the constant @xmath445 must be positive , @xmath446 . formulae ( [ appp7a ] ) and ( [ appp7b ] ) specify the admissible family of cost functions entering cost functional ( [ appp1 ] ) that preserve the preference relation . as seen directly , integral ( [ appp1 ] ) with the integrands @xmath441 and @xmath442 is independent of a particular form of the path @xmath398 . so the transformations ( [ appp7a ] ) and ( [ appp7b ] ) correspond to the linear transformation of the cost functional . the problem of finding the extremals of the functional @xmath447 defined on a set of paths @xmath448 is considered . here , first , the integrand @xmath449 is a given function of the headway distance @xmath24 , the current velocity @xmath450 , and the acceleration @xmath451 . second , the weight factor @xmath452 is a function of time @xmath0 confined within some time interval @xmath453 , i.e. , @xmath454 , @xmath455 for @xmath456 , and @xmath457 sufficiently fast as @xmath403 . so , @xmath458 can be adopted without lost of generality . third , the trial paths @xmath182 meet the initial conditions @xmath459 and do not diverge as time goes on , i.e. , there is a number @xmath460 so that @xmath461 the properties of the functional extremals have been studied already . keeping in mind the results in sec@xmath462[sec:2 ] , a special case is analyzed in the following . let @xmath463 be the characteristic time scale of the velocity relaxation to the stationary value @xmath44 . in particular , the values of @xmath463 are estimated by the eigenvalues of the corresponding extremal equation linearized near the stationary point . the case @xmath464 is considered here to show how the functional ( [ appb.1 ] ) can be rewritten to eliminate the time dependent factor @xmath465 . in this way the problem of finding the extremals gets the classical form met in theoretical mechanics . the stationary point @xmath466 is determined by the properties of both the function @xmath449 and the factor @xmath465 ( see exp . ( [ 2.11 ] ) ) . as a result , at the stationary point ( see again exp . ( [ 2.11 ] ) ) @xmath467 the derivative @xmath468 does not contain factors similar to @xmath469 . so , @xmath465 can not be omitted directly . as a first step the function @xmath449 is replaced by the difference @xmath470 . this does not affect the extremals . the difference @xmath471 as a function of @xmath0 is practically confined within a time interval @xmath472 whose upper boundary @xmath473 and , so , @xmath474 . therefore , at the next step expression ( [ appb.1 ] ) is rewritten as @xmath475 which is justified to the first order in the small parameter @xmath476 by virtue of expressions ( [ appb.4 ] ) and the fact that the second term in exp . ( [ appb.5 ] ) plays a substantial role only when the path @xmath46 tends to the stationary point . here the symbol @xmath477 means that functional ( [ appb.5 ] ) possesses the same collection of the extremals as the initial functional @xmath478 within the adopted accuracy and the time dependent factor @xmath465 has been approximated by @xmath479 formula ( [ appb.6 ] ) can be regarded actually as the definition of the time scale @xmath480 ^{-1}$ ] . taking into account the relation @xmath481 and integrating the second term in exp . ( [ appb.5 ] ) by parts yields @xmath482 omitting the constant components of the integrand which does not affect the extremals the required result follows @xmath483 minimization of functional ( [ appb.8 ] ) gives the desired extremals and integral ( [ appb.8 ] ) does not contain a time dependent factor . the function @xmath484 can be regarded as a lagrangian of the car dynamics . r. wiedemann , simulation des straenverkehrsflues , ph.d .- thesis , heft 8 der schriftenreihe des ifv , karlsruhe , in german ( 1974 ) , c. l. barrett , m. wolinsky , and m. w. olesen , in : _ proceedings of international workshop of traffic and granular flow _ , editors d. e. wolf , m. schreckenberg , and a. bachem ( world scientific , singapore , 1996 ) , p. 169 .

one of the most important questions in condensed matter physics relates to how we should characterize and classify the many different quantum phases of matter . a large part of the story is the theory of symmetry breaking , which tells us that we should classify various phases based on the symmetries of the ground state wave function . yet with the discovery of the fractional quantum hall ( fqh ) states@xcite came also the understanding that there are many distinct and fascinating quantum phases of matter , called topological phases , whose characterization has nothing at all to do with symmetry . how should we characterize and systematically classify the different possible topological phases that may occur in a fqh system ? let us run through a few obvious possibilities . we know that the fqh states contain topology - dependent and topologically stable degenerate ground states , which allow us to introduce the concept of topological order in fqh states.@xcite such topology - dependent degenerate ground states suggest that the low energy theories describing the fqh states are topological gauge field theories in 2 + 1 dimensions , called chern - simons theories.@xcite so one possibility is that we may try to classify the different fqh phases by classifying all of the different possible chern - simons theories . but this is not a satisfactory approach for non - abelian fqh states because we do not have a good way of knowing which chern - simons theories can possibly correspond to a physical system made of electrons and which can not . another possibility comes through the connection between fqh wavefunctions and conformal field theory ( cft ) . it was discovered around 1990 that correlation functions in certain two - dimensional conformal field theories may serve as good model wavefunctions for fqh states.@xcite thus perhaps we may classify fqh states by classifying all of the different conformal field theories that may be used to construct a `` valid '' fqh wavefunction . while the connection between fqh wave functions and cft correlators has been extremely fruitful in both constructing new model wavefunctions and understanding their topological properties , the cft approach is incomplete in that there does not exist a complete classification of all possible conformal field theories that may be used to construct fqh wavefunctions . this precludes the possibility of systematically classifying fqh states . the cft approach also obscures the essential physics and the essential properties of the conformal field theory that allow one to obtain amenable fqh wavefunctions . in an attempt to obtain a systematic classification of fqh states , it was shown recently that a wide class of single - component fqh states and their topological excitations can be classified by their pattern of zeros , which describe the way ideal fqh wavefunctions go to zero when various clusters of particles are brought together.@xcite this analysis led to the discovery of some new non - abelian fqh states whose corresponding cft has not yet been identified . it also helped elucidate the role of cft in constructing fqh wavefunctions . the cft encodes the way the wavefunction goes to zero as various clusters of electrons are brought together . the order of these zeros must satisfy certain conditions and the solutions to these conditions correspond to particular cfts . thus in classifying and characterizing fqh states , one can bypass the cft altogether and proceed directly to classifying the different allowed pattern of zeros and subsequently obtaining the topological properties of the quasiparticles from the pattern of zeros . this construction can then even be thought of as a classification of the allowed cfts that can be used to construct fqh states . furthermore , these considerations give way to a natural notion of which pattern of zeros solutions are simpler than other ones . in this sense , then , one can see that the moore - read pfaffian quantum hall state@xcite is the `` simplest '' non - abelian generalization of the laughlin state . in this paper , we generalize the pattern - of - zeros classification to multilayer fqh wave functions . such a systematic classification leads us to the construction of non - abelian multilayer fqh states and also helps identify the simplest non - abelian generalizations of the halperin@xcite bilayer fqh states . therefore , in the same way that the pfaffian fqh state is the simplest non - abelian single - layer state and so one of the first non - abelian states expected to be realized experimentally , our construction allows us to identify the simplest non - abelian bilayer states and therefore some of the ones that deserve further consideration in future experimental and numerical work . we would like to point out that the `` 1d charge - density - wave '' characterization of single - component fqh states@xcite is closely related to the pattern - of - zeros approach . our pattern - of - zeros approach for multilayer fqh states can also be viewed as a way to generalize the `` 1d charge - density - wave '' characterization to multicomponent cases.@xcite this paper is organized as follows . we begin by describing the ideal fqh wave functions that we can characterize by the pattern of zeros and their relation to symmetric holomorphic polynomials . in section [ pozcharacterization ] , we describe the different ways to characterize the pattern of zeros . in section [ conditions ] we find the conditions that the pattern of zeros must satisfy in order to describe valid fqh wave functions . in section [ idealhamiltonians ] , we sketch how one may begin to construct ideal hamiltonians whose ground states will be fqh wave functions with a given pattern of zeros . after a brief summary of the pattern of zeros data and conditions in section [ summary ] , we explain in section [ cftsection ] the relation between the pattern - of - zeros approach and the cft approach to fqh wave functions . in section [ examples ] , we describe some example solutions of this systematic classification of multilayer fqh wave functions , which yields many non - abelian multilayer states . in section [ discussion ] we discuss some of the simplest of these non - abelian bilayer states that may be relevant for experiments on two - component quantum hall systems and that warrant further numerical study . the ground state wave function of a two - dimensional system of electrons in the lowest landau level can be written in the form @xmath4 where @xmath5 , @xmath6 are the coordinates of the @xmath7 electron , and @xmath8 is a holomorphic function of @xmath9 . since the electrons obey fermi statistics , @xmath10 is anti - symmetric under interchange of any two coordinates @xmath9 and @xmath11 when all of the electrons are identical . in many physical situations , the electrons may be distinguished by various quantum numbers , such as a spin index ( when the zeeman energy is not too high ) , a layer index ( in a multilayer two - dimensional electron system ) , a valley index ( such as in graphene and sige heterostructures ) , _ etc . in such cases , the ground state wave function in the lowest landau level may instead be written in the form @xmath12 where @xmath13 is a flavor index and @xmath14 is the number of different flavors . @xmath10 is then necessarily antisymmetric only under interchange of @xmath15 and @xmath16 for any @xmath17 and @xmath18 . given any antisymmetric polynomial @xmath19 , we can uniquely construct a symmetric polynomial : @xmath20 @xmath21 will also be a polynomial because @xmath22 must vanish when any two identical particles approach each other . thus the above division by the factor @xmath23 will never produce any poles in the resulting function . _ therefore to classify fqh phases of electrons , we can restrict our attention mainly to symmetric polynomials @xmath24 , where @xmath10 is invariant under the interchange of @xmath15 and @xmath16 for any @xmath17 and @xmath18 , but not necessarily invariant under the interchange of @xmath15 and @xmath25 if @xmath26 . in this paper we will often refer to @xmath27 as a layer index . in the following , @xmath10 will always refer to such a symmetric multilayer polynomial . we will introduce data , such as @xmath28 , @xmath29 , and @xmath30 , to characterize and classify bosonic fqh states ( symmetric polynomials ) @xmath24 . from the above discussion , we see that the same set of data also characterizes fermionic fqh states whose wave functions are given by @xmath31 . progress in understanding the various possible topological phases in the fqh effect has occurred largely because of the discovery of various kinds of model wave functions and the ideal hamiltonians that yield these wavefunctions as their ground states . in the pattern of zeros approach , we classify all of the possible ideal wave functions for fqh phases . in this section we will explain what we mean by `` ideal . '' for each topological phase in the fqh system , we want to have a certain representative wave function that captures the topological properties of the phase . the prototypical example of such an ideal wavefunction is the laughlin wave function at filling fraction @xmath32 : @xmath33 at every particle in @xmath34 , there are @xmath29 zeros , and there are no off - particle zeros . the exact ground state of the microscopic hamiltonian with the coloumb interaction does not have this simple property . nevertheless , the ideal laughlin wavefunction captures all of the essential topological properties of these phases . the laughlin wavefunctions are also ground states of certain ideal hamiltonians that contain interaction potentials that are constructed only from @xmath35-functions and their derivatives . for example , if the interaction potential is of the form @xmath36 then the following wavefunction is an exact ground state , with vanishing total potential energy : @xmath37 on the other hand , if the interaction potential between two electrons is of the form @xmath38 with @xmath39 and @xmath40 , then the zero - energy ground state will be @xmath41 more complicated ground states can be obtained through more complicated interactions . for example , consider the following three - body interaction between electrons : @xmath42 where @xmath43 is the total symmetrization operator between @xmath44 , @xmath45 , and @xmath46 . an exact zero - energy ground state of this interaction is the pfaffian wavefunction at @xmath47 : @xmath48 where @xmath49 is the total antisymmetrization operator between @xmath50 . since the pfaffian and laughlin wavefunctions are exact zero - energy ground states of interaction potentials that are constructed only from @xmath35-functions and their derivatives , these wavefunctions can be characterized completely by their pattern of zeros , i.e. by the order of the zeros in the wavefunction as different numbers of particles are taken together . these wave functions have certain ideal properties ; for example , the zeros that are bound to a particle lie precisely at the location of that particle , not slightly away from it . in this paper , we will classify ideal fqh wavefunctions , which are wave functions that are exact zero energy ground states of such ideal hamiltonians . the existence of these ideal hamiltonians is crucial . we can write down any arbitrary complex function of @xmath51 variables , but we can not know whether it corresponds to a valid topological phase of matter in the limit @xmath52 unless we also know that it is the ground state of some local , gapped hamiltonian . if this is the case , then it is conceivable that there exists some physical situation in which the low energy effective interactions between electrons yields a phase of matter that is in the same universality class as the ideal wave functions that we constructed . unfortunately , judging whether a many - particle interacting hamiltonian is gapped in the thermodynamic limit is in general an intractable problem . so here we limit ourselves to classifying ideal wave functions , which at least we believe can be realized as ground states of local hamiltonians . whether the corresponding hamiltonian can be gapped in the thermodynamic limit is a question that we must attempt to answer along the way . the ideal quasihole wavefunctions for these states are also zero energy eigenstates of their corresponding ideal hamiltonians . the true ground state is distinguished by being the unique translationally invariant state with lowest angular momentum . thus we can also classify topologically distinct quasiholes by their pattern of zeros , i.e. by the order of the zeros in the quasihole wavefunction as different numbers of electrons are taken to the quasihole . while the quasihole wavefunctions are also zero energy states , the quasiparticles will be gapped and will have the same topological properties as the quasiholes . therefore in the following , when we discuss the pattern of zeros of the quasiparticles , we are referring to the pattern of zeros of the quasihole wave functions . not all fqh phases have such ideal wave functions . the hierarchy states and the composite fermion states , for example , do not have ideal wave functions . while these phases do have their own model wave functions , they are not ideal in the sense that they can not be directly described by their pattern of zeros or written as a correlation function of conformal primary operators in a cft . therefore the pattern - of - zeros construction does not directly classify these phases . we will however discuss how they are related to the ideal multilayer fqh wave functions in section [ sec : n11 ] . the spirit of the pattern - of - zeros approach is to consider bringing together @xmath53 particles of type @xmath27 , for @xmath13 , and asking how @xmath10 goes to zero under such a procedure . the order of the zero will be denoted @xmath30 , where @xmath54 . in the following we will more precisely define @xmath30 and discuss some different yet equivalent ways of characterizing the pattern of zeros . this discussion is a straightforward generalization of the discussion in the single - layer case.@xcite consider a set of @xmath53 coordinates of each type @xmath27 , and set @xmath55 . define @xmath30 as the minimal power of @xmath56 in the polynomial @xmath10 . this means that if we set @xmath57 and we take @xmath58 , then @xmath59 where @xmath60 is a polynomial in @xmath61 and the remaining coordinates @xmath62 and @xmath63 . we refer to @xmath62 as the coordinate of an @xmath64-cluster . we assume that @xmath30 is independent of the choice of @xmath62 , which must be the case for translationally invariant wavefunctions . we also assume that @xmath30 is independent of the choice of @xmath65 and that different polynomials @xmath66 obtained from different choices of @xmath67 are linearly dependent . this is the assumption of _ unique fusion . _ we can immediately deduce some basic properties of @xmath30 . since @xmath10 has no poles , it is clear that @xmath68 . since @xmath10 must be single - valued under rotating @xmath69 in the complex plane by an angle @xmath70 , @xmath30 must be an integer . let @xmath71 be the minimal power of @xmath72 ; that is , @xmath73 . a translationally invariant @xmath10 will have @xmath74 , otherwise it will vanish everywhere . thus , for a translationally invariant polynomial , @xmath30 is a nonnegative integer that characterizes the order of zero that results when the size of an @xmath64-cluster goes to zero . in the previous section , we introduced the derived polynomials @xmath76 . as a consequence of the unique fusion condition , these polynomials are actually independent of @xmath77 . we may consider more general derived polynomials by bringing together other sets of coordinates in @xmath78 to obtain @xmath79 . then we may consider bringing together an @xmath64-cluster and a @xmath80-cluster : @xmath81 thus , @xmath82 characterizes the order of the zeros in the derived polynomials as a cluster of @xmath64 electrons are brought close to a cluster of @xmath83 electrons . the unique - fusion condition assumes that the derived polynomials obtained from different ways of fusion are always linearly dependent . the fact that @xmath10 is a single - valued , symmetric polynomial implies @xmath84 we can deduce a relation between @xmath82 and @xmath30 as follows . the order of the zero obtained by creating an @xmath85-cluster is @xmath86 . one way of creating such a cluster is by first creating an @xmath64-cluster , then creating a @xmath83-cluster , and finally bringing together the two clusters to create an @xmath87-cluster . the order of zero in this case will be @xmath88 . thus @xmath82 can be obtained from @xmath89 through the formula @xmath90 since @xmath91 , where recall @xmath92 is the unit vector in the @xmath27 direction , we also have @xmath93 from this recursion relation and from the fact that @xmath91 , we may obtain @xmath30 from the sequence @xmath82 . therefore we may equivalently label the pattern - of - zeros data using @xmath89 or @xmath94 . the integer @xmath30 has the following meaning . a polynomial with @xmath53 particles of @xmath95 kind ( @xmath53 particles in the @xmath95 layer ) has a total order of @xmath30 . in other words the total angular momentum of the quantum hall droplet is @xmath30 if the droplet has @xmath96 particles in the @xmath95 layer . if we remove one particle from the @xmath95 layer , then the total angular momentum of the quantum hall droplet will be reduced to @xmath97 . thus we can interpret @xmath98 as the angular momentum of the highest occupied orbital in the @xmath95 layer for a quantum hall droplet with @xmath99 particles in the @xmath100 layer . the @xmath14-dimensional sequence of vectors @xmath101 will be called the sequence of highest occupied orbitals ( hoo ) . we see that @xmath102 makes sense only when @xmath103 . we will set @xmath104 when it does not make sense . from , we also see that there is one - to - one correspondence between the sequence @xmath30 and @xmath105 . thus we can also use @xmath106 to characterize the pattern of zeros in the wave function . a fqh wave function @xmath107 defined on a sphere forms a representation of @xmath108 . in such a case , @xmath109 represents the stereographic projection onto the plane of a point on the sphere . a single particle in the lowest landau level can fill any of the @xmath110 orbitals , so the representation of @xmath108 formed in this case is the one with angular momentum @xmath111 . the su(2 ) lie algebra is generated by @xmath112 in the multilayer case , then , the angular momentum of a particle of type @xmath27 will be @xmath113 where @xmath114 is the total number of flux quanta through the sphere seen by the particles in the @xmath95 layer . note that here we allow the numbers of flux quanta in different layers to be different . the total angular momentum of an @xmath64 cluster in the @xmath109 direction will be the eigenvalue of the operator @xmath115 the operator @xmath116 counts the total power of a polynomial . since the minimum total power of @xmath117 is @xmath30 , the minimum total angular momentum of an @xmath64-cluster is given by @xmath118 . this means that the @xmath64-cluster carries an angular momentum of @xmath119 where @xmath120 and @xmath121 . we will use this relation later to construct ideal hamiltonians and to place conditions on the pattern of zeros for when they can correspond to rotationally invariant wave functions on the sphere . for the pattern of zeros to describe a valid fqh wavefunction , it must satisfy certain consistency conditions . we already encountered several such conditions above . for instance , we found that @xmath30 is a nonnegative integer , @xmath122 , and @xmath123 is even . in the following we find additional conditions that the pattern of zeros must satisfy . one of the most important conditions on the wave function is simply the condition that the wave function have no poles . this condition is remarkably restrictive on the allowed pattern - of - zeros sequences . consider a derived polynomial @xmath124 and fix all coordinates but @xmath62 , thus viewing it as a complex function @xmath125 . @xmath125 has zeros at isolated points , but no poles anywhere . some of the zeros are located at @xmath126 , @xmath127 , . those zeros are called on - particle zeros . the rest of the zeros are called off - particle zeros . if we imagine taking @xmath62 around @xmath126 without enclosing any off - particle zeros , then @xmath128 will pick up a phase @xmath129 . similarly , if we take @xmath62 around @xmath127 without enclosing any off - particle zeros , then @xmath128 will pick up a phase @xmath130 . now consider taking @xmath131 . under such a process , some nearby off - particle zeros will also be taken to @xmath127 . therefore , if we take @xmath62 around a contour that encloses both @xmath126 and @xmath127 in the limit that @xmath132 , the complex function @xmath128 must change by a phase that is greater than or equal to @xmath133 . the phase can never be less than this amount because that would require the existence of off - particle poles that get taken to @xmath127 in order to diminish the strength of the on - particle zeros . by definition , the phase change of @xmath128 under the above procedure is @xmath134 . therefore , the condition that the wavefunction have no poles leads directly to the following concavity condition on the integers @xmath82 : @xmath135 in cases where all of the zeros are located on the particles and there are no off - particle zeros , the above inequality is saturated . this occurs in the laughlin states @xmath136 , and their multilayer abelian generalizations , the halperin states @xmath137 in the following we will rewrite the concave condition as @xmath138 @xmath139 the cluster condition is a way to associate some kind of grading to the polynomials that is physically meaningful . let @xmath140 for @xmath13 be a set of vectors that generate an @xmath14-dimensional lattice , where @xmath14 is , as before , the number of flavors of particles ( or the number of layers ) . the cluster condition states that the concave condition is saturated , _ i.e. @xmath141 if either @xmath64 , @xmath83 , or @xmath142 lie on the lattice generated by @xmath143 . that is , if either @xmath64 , @xmath83 , or @xmath142 can be written as a linear combination with integer coefficients of the vectors @xmath144 . this means that a derived polynomial containing a @xmath145 cluster is non - zero unless @xmath146 coincides with the coordinates of another cluster ; viewed as a function of the single variable @xmath146 , it has no off - particle zeros . a consequence of this is that if all of the particles are fused to form @xmath147-clusters , then the resulting derived polynomial has the laughlin - halperin form ( see ( [ halperin ] ) ) and there are no off - particle zeros . _ the single - layer read - rezayi @xmath148 parafermion wave functions satisfy an @xmath28-cluster condition and they are exact ground states of hamiltonians with @xmath149-body interactions . for a fixed filling fraction , as @xmath28 increases , the number of topologically distinct quasiparticles , the ground state degeneracy on higher genus surfaces and the complexity of interactions necessary to realize the state all increase . this suggests that the energy gap typically decreases with increasing @xmath28 . wave functions that do not obey a cluster condition can be thought of as having infinite @xmath28 and are not expected to correspond to gapped phases . this intuition also comes from the cft approach to fqh wave functions ; infinite @xmath28 corresponds to an irrational conformal field theory , which does not yield a finite number of quasiparticles and a finite ground state degeneracy on the torus . in the multilayer case , we may use the volume of the unit cell spanned by @xmath140 as one way to measure the complexity of a given fqh state . the cluster condition is extremely powerful and simplifying because it allows us to determine the entire pattern - of - zeros sequence from knowledge of a `` small '' number of them . to see how this works , first observe using ( [ clustercond ] ) that @xmath150 where we have defined the matrix @xmath151 . so for any vector @xmath152 , where @xmath153 is an integer and @xmath154 , we have : @xmath155 the above equations imply @xmath156 where we have also defined the matrix @xmath157 . in terms of the sequence @xmath158 , this implies that for @xmath152 , where @xmath153 is an integer , @xmath159 therefore , all of the integers @xmath30 are specified by the points @xmath160 within the unit cell spanned by @xmath144 . in terms of the hoo squence @xmath106 , we have : @xmath161 finally , note that since @xmath162 is an even integer , we have : @xmath163 the density profiles of the single - particle states in the lowest landau level , @xmath164 , are in the shape of a ring , with a peak at a radius @xmath165 , where @xmath166 is the magnetic length . such a wave function has an angular momentum @xmath29 . when many of these orbitals are filled by particles , the total wavefunction will describe a uniform , rotationally symmetric state that goes to zero at a radius @xmath167 , where @xmath168 refers to the filled orbital with maximum angular momentum . therefore , a given quantum hall wavefunction will describe a qh droplet of area @xmath169 , and @xmath168 is given by the maximum power of @xmath44 ( or @xmath9 for any other fixed @xmath17 ) . @xmath168 is also equal to the number of flux quanta , @xmath170 . an important constraint on the multilayer quantum hall wavefunctions is that they must describe systems in which each layer occupies the same area , up to small corrections . the requirement that each layer occupies exactly the same area amounts to the requirement that each layer has exactly the same number of flux quanta , @xmath171 . however , it is reasonable to include states in which different layers occupy equal areas only up to @xmath172 corrections , where @xmath173 is the number of electrons . such a requirement of approximately equal area layers is summarized in the following equation : @xmath174 we wish to see how this condition translates into a condition on the pattern of zeros . the conditions are slightly different depending on whether we ultimately want to characterize gapped fqh phases of fermions or bosons . if we are interested in fermionic phases , we require that @xmath175 be a valid fqh wave function of fermions , which does not require that @xmath176 be a valid fqh wave function of bosons . in what follows we will explicitly analyze the bosonic case , where we require @xmath177 to be a valid multilayer fqh wave function of bosons . @xmath114 is equal to the maximal power of @xmath72 ; for the boson wavefunction @xmath10 , this is given by @xmath178 where recall @xmath179 is the number of particles of type @xmath27 . using the cluster condition , we find @xmath180 where we have set @xmath181 . requiring ( [ equalareacond1 ] ) , we obtain the following condition on the pattern of zeros : @xmath182 this can be seen most easily by ignoring the @xmath172 terms in ( [ nphieqn1 ] ) , taking @xmath183 , and inverting the result to obtain @xmath184 , which must be nonnegative . from this analysis , we learn that if @xmath185 is not invertible , then the pattern of zeros can not fix the ratio of particles @xmath186 in the different layers . therefore the corresponding fqh state has a gapless mode corresponding to the relative density fluctuations between the different layers . as a simple example of this analysis , consider the @xmath187 halperin bilayer state , which is known to have a gapless density mode and for which @xmath188 , which is not invertible . a macroscopic number of particles can freely go from one layer to the other without changing the area of the quantum hall droplets , signalling the existence of a gapless relative density mode . inverting ( [ nphieqn1 ] ) yields @xmath189 from which we can read off the filling fraction in each layer : @xmath190 the total filling fraction is the sum of the filling fraction of each layer : @xmath191 . for fermions , ( [ nphipoz ] ) is modified to @xmath192 due to the extra factor @xmath193 in @xmath194 . note that @xmath195 still describes the pattern of zeros of the symmetric polynomial @xmath10 . the result for fermions is therefore @xmath196 where @xmath197 is the @xmath198 identity matrix . if @xmath199 is not invertible in the fermionic case , then there are gapless relative density modes , which is why the filling fraction in each layer becomes undefined . consider a multilayer quantum state with @xmath179 particles in the @xmath95 layer . we want to put the quantum state on a sphere with @xmath200 flux quanta in the @xmath95 layer . we would like to know for which set of @xmath200 can the quantum hall state completely fill the sphere ? naively , one may expect @xmath200 and @xmath179 are related by the filling fraction in each layer @xmath201 . however the precise relation between the number of flux quanta and the number of electrons includes a shift , @xmath202 where @xmath43 is of order 1 in the large @xmath173 limit ( see eqn . ( [ nphieqn2 ] ) ) . more precisely , completely filling the sphere means that the quantum hall state is rotationally invariant with zero total angular momentum . using , we find that , for a bosonic fqh state characterized by @xmath30 with @xmath179 particles and @xmath203 flux quanta in the @xmath95 layer , the maximum total angular momentum is given by @xmath204 where @xmath205 and @xmath206 . for a fermionic fqh state characterized by @xmath30 with @xmath179 particles and @xmath203 flux quanta in the @xmath95 layer , the maximum total angular momentum is given by @xmath207 in the above , @xmath208 must satisfy @xmath209 in order for the wave function to fit into each layer . completely filling the sphere requires that @xmath210 and @xmath211 . we see that @xmath212 and @xmath208 must satisfy @xmath213 which implies ( for both bosons and fermions ) @xmath214 if a given @xmath212 does not satisfy , then the corresponding quantum hall state ( with @xmath179 particles on the @xmath95 layer ) can not completely fill the sphere . for @xmath212 that satisfies , the corresponding quantum hall state can completely fill the sphere and has zero total angular momentum if @xmath208 is given by . can generally be satisfied only if @xmath212 lies on the lattice spanned by @xmath144 . we would like to remark that it is easy to have different numbers of flux quanta on different layers in numerical calculations . the pattern of @xmath215 where the quantum hall state has zero total angular momentum on the sphere can be used as a fingerprint to identify different quantum hall states through numerical calculations ( for examples , see tables [ fluxnotable23 ] and [ fluxnotable45 ] ) . the analysis of the single - layer case in has suggested an additional condition : @xmath217 there , it was found that allowing @xmath218 allows for certain pattern - of - zeros sequences that either do not correspond to single - valued wavefunctions ( such as the square root of the pfaffian ) or could not correspond to translationally invariant wavefunctions . it was suggested that one way to rule out such possibilities is to impose ( [ delta3even ] ) . how should this condition be generalized to the multilayer situation ? one natural generalization is to impose @xmath219 for all @xmath64 , @xmath83 , and @xmath142 . however , we find that this condition is too restrictive . it rules out certain known fqh wavefunctions , such as the @xmath220 non - abelian spin singlet states.@xcite the need to relax this condition while still having it remain compatible with the single - layer situation suggests that we should impose ( [ delta3even ] ) only for choices of @xmath64 , @xmath83 , @xmath142 that are collinear through the origin . while it was found that allowing @xmath221 allows for pattern - of - zeros sequences that do not seem to correspond to valid translationally invariant , single - valued wavefunctions , there are known cases of cfts with @xmath221 that do seem to yield translationally invariant , single - valued wavefunctions . one such example is the so - called gaffnian wavefunction , which has @xmath222 and which can be constructed using the minimal model cft @xmath223.@xcite this cft however is non - unitary . it has been suggested that fqh wavefunctions constructed using non - unitary cfts correspond to gapless phases;@xcite whether this is always necessarily the case is currently an important open question in fqh theory . these considerations suggest that in order to restrict ourselves to pattern - of - zeros sequences that have a corresponding unitary cft , we should impose @xmath224 for those @xmath64 , @xmath83 , and @xmath142 that are collinear through the origin . in our search for pattern - of - zeros solutions , we will impose this condition and analyze the resulting states . the precise connection , if any , between this condition and valid fqh wavefunctions that correspond to unitary cfts remains to be clarified given a pattern - of - zeros sequence , it is important to be able to construct a local , gapped hamiltonian whose ground state wave function has the given pattern of zeros . if this is possible , then we know that the corresponding pattern - of - zeros sequence describes a topological phase of matter . whether this particular phase is realized in an experiment then depends on the particular types of low energy effective interactions between the electrons in the unfilled landau levels . we can go about constructing such a hamiltonian by noticing that on a sphere , the integers @xmath30 are directly related to the angular momentum of the @xmath64-cluster . for an electron system on a sphere with @xmath225 flux quanta for the @xmath27th layer , an electron of type @xmath27 will carry an angular momentum @xmath226 . for an @xmath227-cluster , the maximum angular momentum is therefore @xmath228 . however , for a polynomial @xmath229 described by a pattern of zeros @xmath89 , the maximum allowed angular momentum of the @xmath64 cluster is only @xmath230 . the pattern of zeros forbids the appearance of angular momentum @xmath231 for any @xmath227-clusters in @xmath232 . such a condition can be easily enforced by writing the hamiltonian as a sum of projection operators , @xmath233 . let @xmath234 be a projection operator that acts on the @xmath64-cluster hilbert space . @xmath233 projects onto the subspace of @xmath227-clusters that have total angular momenta greater than @xmath235 . now consider the hamiltonian @xmath236 where @xmath237 sums over all of the @xmath227-clusters for a fixed @xmath64 . the wave function described by @xmath195 will clearly be a zero - energy ground state of the above @xmath238 . in many cases , there is only one unique ground state wave function with minimal total angular momentum , but in general there can be many independent polynomials with the same pattern of zeros . in such a situation , the hamiltonian would need to be modified further to select for a particular polynomial with the given pattern of zeros . in order for the above hamiltonian to be local , @xmath239 must be limited to a small , finite number of @xmath64-clusters . but as a result , we can not guarantee in general that the ground state wave functions will all be described by the sequence @xmath89 for every @xmath64 , or even that they will obey the cluster condition . in many of the known cases , such as the laughlin , moore - read , and read - rezayi wave functions , the sum over @xmath64-clusters can indeed be terminated after the first few clusters while still yielding a unique zero energy ground state wave function with minimal angular momentum which is described by @xmath195 . the above construction for @xmath240 should therefore be viewed as a starting point for constructing an ideal hamiltonian that is local , gapped , and whose unique minimal angular momentum ground state wave function is described by @xmath158 . in some of the simplest cases , we know that this construction suffices . we have found that the polynomials @xmath241 , @xmath242 , that may correspond to stable fqh states are described by the following data : @xmath243 where @xmath28 and @xmath29 are @xmath198 matrices with integer entries that satisfy @xmath244 the above implies that @xmath185 is a symmetric matrix . furthermore , for the pattern of zeros to fix the relative densities of particles in each layer , we have @xmath245 otherwise , there are gapless relative density fluctuations . we also have @xmath246 the total filling fraction is @xmath247 . note that the @xmath158 need to be specified only for values of @xmath64 that are contained in the unit cell spanned by @xmath248 , where @xmath249 corresponds to the @xmath95 row of the matrix @xmath28 : @xmath250 . using the cluster condition , we can determine @xmath30 for all @xmath64 simply from @xmath29 , the fact that the @xmath91 , and from the values of @xmath251 for which @xmath83 lies in the unit cell spanned by @xmath144 : @xmath252 where @xmath145 . @xmath30 must satisfy : @xmath253 where @xmath254 @xmath255 finally we impose @xmath256 for those @xmath64 , @xmath83 , and @xmath142 that are collinear through the origin . consider two polynomials of @xmath257-cluster form , @xmath10 and @xmath258 , and consider their product : @xmath259 . the pattern of zeros of @xmath260 is the sum of the pattern of zeros of @xmath10 and @xmath258 : @xmath261 similarly , the data in terms of @xmath29 and @xmath262 are also additive . note that all of the conditions that we impose on the pattern - of - zeros are linear . thus the pattern of zeros of @xmath260 is also valid . however , notice that the condition for filling the sphere is not linear . there may be two fqh wavefunctions @xmath10 and @xmath258 that can fill the sphere but whose product @xmath260 can not fill the sphere . thus , we can divide the pattern - of - zeros solutions into primitive and non - primitive solutions . primitive solutions are solutions that can not be written as a sum of two other solutions . the pattern - of - zeros approach is closely related to the conformal field theory approach to constructing fqh wavefunctions . in the cft approach , the symmetric polynomial @xmath263 that describes a multilayer fqh state can be written as a correlation function of a set of electron operators @xmath264 in a cft:@xcite @xmath265 the operators @xmath264 are written in the form : @xmath266 where @xmath267 is a vertex operator in a @xmath268 cft . it has scaling dimension @xmath269 . @xmath270 is a simple current operator ; that is , it satisfies the following fusion relation : @xmath271 this abelian fusion rule is the cft version of the unique - fusion condition . the cluster condition implies that @xmath272 satisfies @xmath273 where @xmath250 . an @xmath64-cluster of electrons will be described by the operator @xmath274 thus we see that the cluster condition implies that an @xmath249 cluster is described by a vertex operator @xmath275 . if all of the particles are grouped into @xmath28-clusters , then the corresponding derived polynomial will be just a correlation function of vertex operators in a @xmath268 theory , which will have no off - particle zeros and will be of the laughlin - halperin form . let us denote the scaling dimension of the operator @xmath276 as @xmath277 where @xmath278 is the scaling dimension of the simple current @xmath279 and @xmath280 is the scaling dimension of the vertex operator . notice that since @xmath281 , the simple current scaling dimensions satisfy @xmath282 . the scaling dimension of the gaussian part is given by @xmath283 the pattern of zeros are related to the scaling dimensions through the relation @xmath284 this allows us to obtain the scaling dimensions from the pattern of zeros . using the cluster condition , some algebra shows that @xmath285 and so ( [ scalingdimga ] ) becomes @xmath286 the scaling dimensions of the simple - current part can also be determined from the pattern of zeros by using the fact that @xmath287 , @xmath288 , and applying ( [ pozsc ] ) iteratively . this yields : @xmath289 multiplying both sides by @xmath290 gives @xmath291 in a similar manner , one can obtain @xmath292 which determines @xmath278 in terms of the pattern - of - zeros . note that the correlation function of the gaussian part is , leaving the background charge implicit , @xmath293 thus the fqh wave function is of the form @xmath294 where @xmath295 @xmath296 arises from the correlation function of the simple current sector and is the `` non - abelian part '' of the wavefunction . in this manner , each pattern - of - zeros solution corresponds to the current algebra of a rational cft . the connection between the pattern of zeros approach and the cft approach can be thought of in the following way . the pattern of zeros describes the essential properties of the cft that yield valid fqh wavefunctions . so in order to classify ideal fqh wavefunctions , one can bypass the cft altogether and go directly to the heart of the matter : characterizing the allowed pattern - of - zeros solutions . furthermore , since each pattern - of - zeros solution corrresponds to a cft , the pattern - of - zeros classification can be viewed as a classification of the allowed cfts that can be used to construct fqh wave functions . in this formulation , the pattern of zeros classifies all those ideal fqh wavefunctions that can be formulated as a correlation function of _ conformal primary fields . there are also many fqh wavefunctions , such as the hierarchy states and the jain series , that can not be written in this way . these wavefunctions are outside of the pattern - of - zeros classification . however they may be closely related to the pattern - of - zeros construction . we comment on this connection elsewhere . _ using ( [ scalingdimsc ] ) , we can derive a formula for @xmath30 in terms of @xmath297 , @xmath28 , and @xmath29 : @xmath298 thus there is a one - to - one correspondence between the simple - current scaling dimensions and the sequence @xmath195 . this means that there is yet another way to label the pattern of zeros . earlier , we found that one convenient labeling of the pattern of zeros is with an @xmath198 matrix @xmath28 , an @xmath198 matrix @xmath29 , and the value of the non - negative integers @xmath30 for @xmath64 lying inside the unit cell spanned by the rows of @xmath28 . an alternative , equivalent labeling of the pattern of zeros is by specifying the following data : @xmath299 for @xmath64 lying inside the unit cell spanned by @xmath28 . this labeling is convenient because it makes close contact with the corresponding cft description ; if @xmath300 for some @xmath64 , then the cft has a non - trivial simple - current structure and therefore generally also has some form of non - abelian statistics . in the single - layer case , many of the pattern - of - zeros solutions were found to be closely related to the @xmath302 parafermion cfts of zamalodchikov and fateev.@xcite what are the relevant cfts to expect in the multilayer case ? the answer is that some of the corresponding cfts in the multilayer case will be closely related to the @xmath303 parafermion cfts that were constructed by gepner , where @xmath0 is a simple affine lie algebra at level @xmath304 and @xmath305 is the rank of the lie algebra @xmath306.@xcite the case @xmath307 is equivalent to the @xmath302 parafermion cfts of zamalodchikov and fateev . the simple - current algebra of the @xmath301 parafermion cft has the following structure . for every element @xmath308 of the @xmath305-dimensional root lattice of @xmath306 , associate a simple current operator @xmath309 . the simple - currents will have the fusion rules @xmath310 furthermore , @xmath311 if @xmath312 is an element of @xmath304 times the long root lattice of @xmath306 . the scaling dimension of @xmath309 is given by @xmath313 where @xmath314 is an integer equal to the minimum number of roots from which @xmath315 is composed . the inner product @xmath316 is defined with respect to the quadratic form matrix of @xmath306 . if we are considering quantum hall states with @xmath14 layers , then we would expect to see the appearance of these parafermion cfts with rank @xmath317 . therefore in the bilayer case , one class of states that we expect to see should be related to @xmath318 parafermion cfts where @xmath306 is a simple lie algebra of rank 2 . there are only three simple lie algebras of rank 2 : @xmath319 , @xmath320 , and @xmath321 . of these , only @xmath319 is simply laced , so the long root lattice is the same as the root lattice . this means that for the pattern - of - zeros solutions that correspond to @xmath322 , we expect to see solutions that correspond to @xmath323 cfts . in appendix [ su3para ] we will describe the @xmath220 cft in more detail . the parafermion cfts for @xmath324 and @xmath325 , on the other hand , are more complicated because the long root lattice is different from the root lattice . for example , @xmath326 cfts will generically be relevant for @xmath327 while @xmath328 cfts will generically be relevant for @xmath329 . in this section we examine explicitly several simple bilayer pattern of zeros solutions . we first fix the cluster structure @xmath28 to have a simple form . then we try to find all the solutions @xmath158 that satisfy the conditions listed in section [ summary ] . the simplest non - abelian states can be obtained from the simplest cluster structures @xmath330 and @xmath331 . these are the simplest non - abelian generalizations of the bilayer abelian laughlin - halperin states . note that by definition the ordering of the rows in @xmath28 is arbitrary ; we choose it so that @xmath332 . interchanging the layers yields the same physical system but corresponds to interchanging @xmath333 with @xmath334 and @xmath335 with @xmath336 , so two matrices @xmath28 and @xmath337 that are related by such an interchange are regarded as equivalent . to list the solutions for those simplest cases , we may use some known cfts to construct the appropriate simple - current algebra that corresponds to the pattern - of - zeros solutions . using this known cft , we can then write the wavefunction explicitly . however , the wave function that we write down may not be unique in some cases ; there may be several independent polynomials that have the same pattern of zeros . this corresponds to there being several distinct cfts whose simple - current algebra possesses the same pattern of zeros . in the following examples , we will make this choice when necessary so that we can explicitly write down a wave function with a certain pattern of zeros . these states are all abelian and correspond to the halperin @xmath339 states . the unit cell spanned by the @xmath249 consists only of the points @xmath340 and @xmath341 ; by translation invariance @xmath342 . thus these states are completely characterized by the matrix @xmath29 and are of the form @xmath343 the @xmath29 matrix here is exactly the @xmath344 matrix that describes all abelian fqh states.@xcite we also have @xmath345 for all of these states the cft has no simple - current part and corresponds to a @xmath346 gaussian cft . although the single - layer hierarchy states , such as the @xmath347 hierarchy state , do not have ideal single - layer wave functions , there are ideal multilayer states that have the same topological orders as those single - layer hierarchy states . for example , the topological order in the @xmath347 hierarchy state is described by the @xmath344-matrix @xmath348 ( in the symmetric basis).@xcite such a topological order can be represented by the ideal bilayer state with @xmath338 and @xmath349 . so although the pattern - of - zeros construction does not directly classify those single - layer hierarchy states , their topological orders can still be described by the pattern - of - zeros approach . this choice of @xmath28 requires that the electron operators in the cft must take the form @xmath351 where @xmath352 and @xmath272 has scaling dimension that is integer or half - integer . this latter fact can be obtained from the condition @xmath353 . one general way of constructing such a simple - current operator is by expressing it as a set of majorana fermions from several copies of the ising cft : @xmath354 where @xmath355 is the majorana fermion from the @xmath17th ising cft . such an operator has scaling dimension @xmath356 and gives rise to the following fqh wavefunction : @xmath357 @xmath358 represents the coordinates in both layers : @xmath359 @xmath360 is defined in ( [ phiga ] ) . note that the simple - current algebra in this case implies @xmath361 . that is , these states can all also be viewed as satisfying the cluster condition for @xmath362 , but with a different choice @xmath29 . for every pattern - of - zeros solution found here , there is an equivalent one for @xmath363 . later , we will list the solutions with the @xmath364 cluster structure . some of those solutions actually have @xmath365 . [ [ n - left - protectbeginmatrix-1 - 1 - 0 - 1-protectendmatrix - right ] ] here , the electron operators must be of the form @xmath367 the fact that @xmath368 however also forces @xmath369 . thus in fact the corresponding cfts do not have a simple - current part ; they all correspond to a @xmath346 guassian cft . all of these states are therefore abelian and correspond to the halperin @xmath339 states . thus , all pattern - of - zeros solutions where @xmath370 actually also satisfy the cluster condition for @xmath371 . using the cluster condition and the fact that @xmath372 , it is easy to see that the pattern of zeros for these states are completely characterized by the matrix @xmath29 and therefore that these states are identical to the halperin states . [ [ n - left - protectbeginmatrix-1 - 0 - 0 - 2-protectendmatrix - right ] ] in this case , the electron operators in the cft must take the form @xmath374 where @xmath352 . from @xmath375 , it follows that the scaling dimension of @xmath272 is integer or half - integer . this means that we can in general write it as a product of @xmath376 majorana fermion operators from @xmath376 independent copies of the ising cft . thus the pattern - of - zeros solution for this choice of @xmath28 includes only two classes of states : the abelian halperin states and the following version of the pfaffian states : @xmath377 such a state spontaneously breaks the discrete @xmath378 symmetry associated with interchanging the two layers . for this choice of @xmath28 , the corresponding cfts have two simple current operators , @xmath380 and @xmath381 , which each square to the identity : @xmath382 . thus there are a total of three distinct primary currents : @xmath381 , @xmath380 , and @xmath383 . the pattern of zeros can be fully specified by specifying the scaling dimensions of these simple current operators , @xmath384 , @xmath385 , and @xmath386 , and the matrix @xmath29 , which specifies the @xmath346 part of the electron operators in the cft . applying @xmath387 for this choice of @xmath28 shows that the simple - current scaling dimensions are all integer or half integer : @xmath388 . one general way of encoding these fusion relations and the associated scaling dimensions is to write the electron operators in the cft using simple - current operators that consist of multiple copies of the @xmath220 parafermion cft . the most general choice for the electron operators can be written in the form @xmath389 where @xmath390 is a simple current from the @xmath376th copy of the @xmath220 parafermion cft and @xmath391 and @xmath392 refer to either @xmath340 , @xmath341 , or @xmath393 . some explicit forms for such correlators of simple - current operators in the @xmath220 parafermion cft were discussed in . computing these correlation functions provides one way not necessarily unique of constructing a wave function with the desired pattern of zeros . these pattern - of - zeros solutions can naturally be grouped into two distinct classes . in the first class , @xmath394 and @xmath395 contain only one kind of simple - current , either @xmath380 , @xmath381 , or @xmath396 , from each copy of the @xmath220 cft . for example , if @xmath397 appears in @xmath394 , then @xmath395 can not contain @xmath398 or @xmath399 . in such a situation , we can think of @xmath400 as being the majorana fermion from the ising cft . this means that these states can be written in terms of multiple copies of the ising cft . this class of multilayer states can be written by choosing @xmath401 where @xmath402 is the majorana fermion from the @xmath376th copy of the ising cft . the wave function for this class of states is therefore : @xmath403 @xmath358 represents the coordinates in both layers : @xmath359 this is the simplest generalization of the single - layer non - abelian states to a class of non - abelian bilayer states : the interlayer pfaffian states . the simplest version of this , with @xmath404 , is closely related to ( and identical to for certain choices of @xmath29 ) the spin - charge separated non - abelian spin singlet wavefunction proposed in . the second class of states can not be written in terms of multiple copies of the ising cft the full @xmath220 parafermion cft is necessary . the first class of states , which can be written only using the ising cft , have the property that their pattern of zeros satisfies @xmath405 for all choices of @xmath64 , @xmath83 , and @xmath142 . the second class of states , for which the full @xmath220 cft is necessary , satisfies @xmath406 for certain choices of @xmath64 , @xmath83 , and @xmath142 . let us compare the simple - current algebra of the @xmath220 parafermion cft to what one would obtain using two copies of the ising cft . if we used two copies of the ising cft , we could have @xmath407 with scaling dimension @xmath408 , @xmath409 with scaling dimension @xmath408 , and @xmath410 with scaling dimension @xmath411 . this satisfies @xmath412 . on the other hand , in the @xmath220 parafermion cft , the only difference is that @xmath396 also has scaling dimension @xmath408 . thus in this latter theory , two fermions combine to give another fermion . this yields @xmath413 . the fact that there are valid single - valued translationally invariant fqh wavefunctions that arise from unitary cfts and that have @xmath406 for certain choices of @xmath64 , @xmath83 , and @xmath142 suggests ( see section [ delta3evensec ] ) that we should impose @xmath414 not in general but only if @xmath64 , @xmath83 , and @xmath142 are collinear through the origin . let us examine the pattern of zeros for a few of the simplest examples of these non - abelian bilayer states . there is a fermionic @xmath1 state with @xmath415 this is the pattern of zeros for the interlayer pfaffian state , which is of the form @xmath416 we use the notation @xmath417 there are also fermionic states at @xmath418 and @xmath419 . these have the following pattern of zeros : @xmath420 @xmath421 the state at @xmath419 is the non - abelian spin singlet state that was proposed in . note once again that the pattern of zeros @xmath29 and @xmath158 refer to the pattern of zeros of the symmetric polynomial , @xmath422 . in single - layer quantum hall samples , a quantum hall plateau is seen at @xmath423 , but not at @xmath424 . the reason is that even though in all of these cases there is a single half - filled landau level , the existence of the two filled extra landau levels modifies the effective interactions between the electrons in the unfilled level . in the @xmath423 case , numerical calculations suggest that the these effective interactions are modified in such a way that a non - abelian quantum hall state may be realized . experiments on multicomponent quantum hall systems should be able to probe an even wider variety of regimes with distinct effective interactions . for example , for a two - component fqh system , we can study systems in which the spin degree of freedom is present , two - dimensional electron systems with two quantum wells , wide single - layer systems in which the electrons spontaneously form a double - layer system due to coulomb repulsion , or systems in which there may be two valleys for the free quasiparticle spectrum ( such as in graphene or sige heterostructures ) , etc . in many of these cases , experimentalists can also tune to some extent the degree of correlation between the two components . for example , in double layer systems , application of a parallel magnetic field can tune the tunneling and correlation between the layers . there may also be some degree of tunability in the relative densities between the two components in addition to being able to probe fqh states with different numbers of filled landau levels . with this greatly increased amount of variability and tunability in the effective interactions between electrons in the unfilled landau levels , it is possible that a non - abelian state can be realized in a two - component quantum hall system . since the pattern of zeros provides a systematic classification and characterization of a wide variety of quantum hall states , it provides us with a general sense of how all of the non - abelian bilayer states are related and which ones are simpler than other ones . just as we know that the single - layer pfaffian quantum hall state is the simplest non - abelian generalization of the laughlin states , we can determine the simplest non - abelian generalization of the halperin bilayer states and therefore single out some of the possibilities that may be experimentally viable . in , we have given an overview of some of the simplest non - abelian bilayer states that we find and that occur at filling fractions at which experiments on two - component fqh systems have already observed incompressible states . here we briefly summarize that discussion and supplement details of the calculations of various topological properties of the candidate states . experiments have so far observed fqh plateaus in two - component systems at @xmath1 , @xmath2 , @xmath3 , @xmath425 , @xmath426 , @xmath427 , @xmath428 , etc . @xcite in some cases , these plateaus have been observed in both bilayer and spin - unpolarized single - layer systems , while in others , the plateau has only been observed in one of them . at all of these filling fractions , there exists also one ( or several ) candidate abelian phase(s ) ; in most cases , it is assumed that these plateaus are described by one of the abelian phases . however , the pattern - of - zeros construction also yields many simple non - abelian states at these filling fractions . in some situations , we expect the non - abelian states to be good candidate states . there are at least two dimensionless quantities that are important determining factors for which fqh state is realized . the first parameter is @xmath429 , where @xmath430 is the potential for interlayer repulsion and @xmath431 is the potential for intralayer repulsion . the second parameter is @xmath432 , where @xmath433 is the interlayer hopping amplitude . in the limit @xmath434 and @xmath435 , the system will be a fqh state that consists of two independent single - layer fqh states in each layer . in the limit @xmath436 and @xmath434 , a single - layer fqh state may be observed . but if we keep @xmath435 and increase @xmath437 from @xmath438 , then the fqh state formed by two independent single - layer fqh states in each layer must undergo a phase transition into either a compressible phase or a new incompressible state . in the latter case , an abelian hierarchy state ( such as a bilayer composite fermion state ) may form , which would in most cases be a state described by a @xmath439 or more complicated @xmath344-matrix and would have four or more edge modes . the other possibility is that a non - abelian two - component state may form . the pattern - of - zeros construction yields non - abelian two - component wave functions that have zeros when particles from different layers approach each other , indicating that they can accomodate situations in which @xmath440 . additionally , these states generally have less than @xmath441 edge modes ; if we use the number of edge modes as a measure of the complexity of the state , then the non - abelian states are simpler and may therefore be realized experimentally . at @xmath1 , experiments on wide single quantum wells have observed a phase transition from a bilayer to single - layer state while experiments on single - layer systems have seen a phase transition from a spin - polarized to a spin - unpolarized state . in the limit @xmath442 and @xmath435 , the system should be in the @xmath443 state . as @xmath437 is increased while @xmath435 , one possibility is the @xmath444 state . this wave function appears unphysical , because it has higher order zeros as particles from different layers approach each other than particles from within the same layer . another wave function , which has the same topological order as @xmath444 , is a spin - singlet composite fermion state.@xcite there are two other plausible non - abelian states in this situation . one is the following interlayer pfaffian state ( see ( [ interpfpoz ] ) and ( [ interpfwf ] ) ): @xmath445 the other is the following intralayer pfaffian state @xmath446 which has even higher order zeros as particles from different layers approach each other . @xmath447 has a cluster structure @xmath448 while @xmath449 has a cluster structure @xmath450 . at @xmath451 there are also two single - layer possibilities that may be realized as @xmath452 is increased . these are the particle - hole conjugate of the @xmath453 laughlin state and the @xmath454 parafermion read - rezayi state . . [ proposed ] proposed explanations for incompressible states at experimentally relevant filling fractions , @xmath1 , @xmath2 , @xmath3 , and @xmath428 , in two - component fqh systems . the bilayer composite fermion state @xmath455 @xcite refers to the state @xmath456 , where @xmath457 is a single layer composite fermion state at filling fraction @xmath458 . for @xmath459 , we have taken the single layer @xmath460 state to be the particle - hole conjugate of the laughlin state . @xmath461 indicates that there are @xmath462 right - moving edge modes and @xmath463 left - moving edge modes . see appendix [ calculations ] for details of how to calculate the number of edge modes and the shift @xmath43 . [ cols="^,<,^,^",options="header " , ] in table [ proposed ] we summarize some of the filling fractions at which incompressible states have been experimentally observed in two - component fqh systems . for each filling fraction we list some of the proposed wave functions that may characterize the topological order of those phases , the number of edge modes , and their respective shifts on the sphere . we list the qasiparticles , their electric charges , and their scaling dimensions for the interlayer pfaffian state at @xmath1 in table [ ipfqpops ] . in table [ qpdata ] , we list the quasiparticles with the minimal electric charge and their scaling dimensions @xmath464 for the non - abelian fqh states discussed in this paper [ see eqns . ( [ interpfwf ] ) , ( [ 45poz ] ) , ( [ 47poz ] ) , ( [ su3states ] ) , ( [ na14 ] ) ] . those minimally charged quasiparticles may dominate interedge tunneling and give rise to the following i - v curve : @xmath465 in the @xmath466 limit . in summary , we find many simple non - abelian bilayer states that occur at experimentally observed filling fractions . for certain effective interactions among the electrons in the unfilled landau levels , these states may be more favorable than their abelian counterparts . in these cases , the non - abelian states have larger interlayer correlations and therefore may be energetically more favorable in situations in which the interlayer repulsion is comparable to the intralayer repulsion . a useful tool for identifying fqh states in numerical studies of exact diagonalization on finite systems on a sphere is to look at what values of the shift , @xmath467 , a ground state with zero total angular momentum is found . this then limits the possibilities of which topological phase is realized in the system to those that have that particular value of the shift . similarly , in such numerical studies of multilayer systems , one can look for the different sets @xmath468 that yield a ground state with zero total angular momentum . each topological phase will have its own list of @xmath469 that let it fill the sphere ; analyzing this can be a useful way of determining which topological phase is obtained numerically . in section [ shiftrisphere ] , we found conditions that @xmath212 and @xmath470 should satisfy for the fqh state to fill the sphere . & & + & & 2 & 4 & 6 & 8 & 10 + & 2 & ( 3,3 ) & ( 7,5 ) & ( 11 , 7 ) & ( 15,9 ) & ( 19,11 ) + & 4 & ( 5,7 ) & ( 9,9 ) & ( 13,11 ) & ( 17,13 ) & ( 21,15 ) + & 6 & ( 7,11 ) & ( 11,13 ) & ( 15,15 ) & ( 19,17 ) & ( 23 , 19 ) + & 8 & ( 9,15 ) & ( 13,17 ) & ( 17,19 ) & ( 21,21 ) & ( 25,23 ) + & 10 & ( 11,19 ) & ( 15,21 ) & ( 19 , 23 ) & ( 23,25 ) & ( 27,27 ) + & & + & & 2 & 4 & 6 & 8 & 10 + & 2 & ( 2,2 ) & ( 6,3 ) & ( 10 , 4 ) & ( 14,5 ) & ( 18,6 ) + & 4 & ( 3,6 ) & ( 7,7 ) & ( 11,8 ) & ( 15,9 ) & ( 19,10 ) + & 6 & ( 4,10 ) & ( 8,11 ) & ( 12,12 ) & ( 6 , 13 ) & ( 20 , 14 ) + & 8 & ( 5,14 ) & ( 9,15 ) & ( 13,6 ) & ( 17,17 ) & ( 21,18 ) + & 10 & ( 6,18 ) & ( 10,19 ) & ( 14 , 20 ) & ( 18,21 ) & ( 22,22 ) + for states that have a cluster structure @xmath471 , we find that the condition ( [ nss ] ) becomes trivial as long as @xmath181 , where @xmath153 is an integer . this means that as long as @xmath472 and @xmath473 are even and @xmath474 , @xmath475 satisfy ( [ snsne ] ) , then these states can fill the sphere . in this case , we find that ( [ snsne ] ) reduces to the form @xmath476 where @xmath477 is a @xmath478 matrix and @xmath43 is the shift , which can be calculated using eqn . ( [ shift ] ) . the states that we have been considering are of the form @xmath479 , for which @xmath480 . tables [ fluxnotable23 ] and [ fluxnotable45 ] lists some examples . in this paper , we generalized the pattern - of - zeros characterization and classification of fqh states to multicomponent cases . we found that the topological orders in a multicomponent fqh state can be characterized by the following data : a matrix @xmath28 that describes the cluster structure , a matrix @xmath29 and a sequence @xmath158 that describes the pattern of zeros . our pattern - of - zeros characterization gives us a general quantitative view on a large class of abelian and non - abelian bilayer fqh states , which allow us to determine which states are simpler than other states . we find some simplest non - abelian generalizations of the laughlin - halperin abelian bilayer states . those simple non - abelian states may describe some of the bilayer / spin - unpolarized fqh states observed in experiments and numerical calculations . this research is partially supported by nsf grant no . in the single - component pattern - of - zeros description , there is an occupation number characterization that is a useful way to understand both the ground states and the quasiparticles in fqh states . the generalization to multilayer states does not appear to be quite as simple or useful , but for the sake of completeness we will analyze it below . a convenient set of single - particle basis states for particles in the lowest landau level are monomials of the form @xmath481 , for integer @xmath29 . thus , a basis for symmetrized wavefunctions of @xmath179 particles of type @xmath27 is given by : @xmath482 where @xmath483 is a permutation of the particles of type @xmath27 , @xmath484 is an integer , and @xmath485 is a vector whose component @xmath486 is the number of particles of type @xmath27 that occupy the @xmath487 orbit . the polynomial @xmath10 that we are interested in can be expanded in terms of these basis states as : @xmath488 now we may ask what kind of boson occupations @xmath489 will be present in the sum ( [ basisexpansion ] ) for a polynomial @xmath10 with a given pattern - of - zeros @xmath158 . to answer this question , let us set @xmath490 in @xmath491 . since @xmath10 is nonzero when @xmath490 due to translation invariance , there must be a boson occupation @xmath492 in the above sum that contains at least one boson occupying the @xmath493 orbital . that is , there is a term in the above sum with @xmath494 . now , suppose that we bring a second particle of the same type , @xmath495 to 0 . the minimal power of @xmath495 in @xmath496 is @xmath497 : @xmath498 thus , among those @xmath489 that have at least one boson of type @xmath411 occupying the @xmath493 orbital , there must also be an @xmath489 that contains a second boson of type @xmath411 occupying the @xmath499 orbital where @xmath500 . next , assume that two bosons occupy the @xmath501 and @xmath499 orbital , and bring a third particle of type @xmath411 to 0 ; the minimal power of @xmath502 is @xmath503 : @xmath504 thus , among those @xmath492 that have two type @xmath411 bosons occupying the @xmath505 and @xmath506 orbitals , there is a third boson of type @xmath411 occupying the @xmath507 orbital . continuing in this way , we see that there must be a type @xmath411 boson occupying the orbitals @xmath508 for @xmath509 . after taking all the type @xmath411 particles to 0 , we may begin to take the type 2 particles to zero , one by one , thus obtaining that there must be a type @xmath510 boson occupying the orbitals @xmath511 . continuing this argument for bosons of every type , we find that there must be a term in the sum ( [ basisexpansion ] ) with occupation number described by the above sequence of @xmath102 s . however , in the above argument , we chose a particular sequence in which to take various particles to zero . we first took all of the type 1 particles to zero one - by - one , and then all of the type 2 ones , and so on . but we could just as well have made the argument with any sequence . suppose that after taking @xmath17 particles to the origin , there is a @xmath512-cluster at the origin . thus @xmath513 is a sequence that describes the order in which we take particles to the origin until all particles are at the origin . for every such sequence , we may make the above argument and argue that if @xmath514 , then there must be a type @xmath27 boson occupying the orbital @xmath515 . if we enumerate all the different sequences @xmath516 by an integer @xmath437 , then by considering each @xmath437 , we see that there must be a term in the sum ( [ basisexpansion ] ) with occupation number @xmath517 . @xmath518 would be the number of @xmath17 , along the sequence @xmath516 , for which @xmath519 . notice that @xmath102 must be non - zero . thus we have the following important condition on @xmath30 : @xmath520 the analysis above can be thought of in the following way . consider an @xmath14-dimensional lattice @xmath521 , where @xmath14 is the number of layers . at every site @xmath64 of this lattice ( @xmath522 ) we can associate the nonnegative integer @xmath30 . on each link ( @xmath64 , @xmath523 ) , of the lattice we may also associate an integer @xmath524 . now consider any directed path from the origin to @xmath212 ( @xmath179 is the number of particles of type @xmath27 ) , in which the sum of the coordinates of every point on the path is one larger than the sum of the coordinates of the point preceding it , and enumerate the set of these paths by @xmath437 . to each such path we associate an occupation number sequence @xmath525 , where @xmath518 is the number of links along the path @xmath437 whose @xmath526 . if @xmath10 has a pattern - of - zeros @xmath158 , then its basis expansion ( [ basisexpansion ] ) must contain a term with occupation number @xmath517 . thus we may rewrite ( [ basisexpansion ] ) as @xmath527 the two sequences @xmath158 and @xmath528 contain the same information and are one - to - one labellings of each other . however , @xmath528 is redundant in the sense that it does not need to be specified for every @xmath437 in order to reconstruct @xmath106 . the @xmath529 that appear in the second sum characterize the subleading terms that appear when coordinates are brought together ; thus those @xmath530 correspond to sequences @xmath531 where @xmath532 . in the single - layer case , @xmath533 naturally defined an occupation number sequence @xmath534 , which also described the fqh state in the thin - cylinder limit . in the multi - component generalization , we have @xmath535 , which seems to admit no simple generalization of the above occupation number sequence . instead , one has such occupation number distributions for a large number of sequences which we enumerated above by @xmath437 . we have not analyzed on general grounds which particular sequences contribute the most weight to the wavefunction in the thin cylinder limit . some of the simplest non - abelian bilayer states are closely related to @xmath220 parafermion cft . this cft has central charge @xmath536 and has three simple currents , @xmath537 , @xmath538 , and @xmath539 , all of which square to the identity and have scaling dimension @xmath408 . there are four other primary fields , which are associated with the fundamental representation of @xmath319 . their scaling dimensions are listed in table [ su3paraf ] . the fusion rules for these fields all follow from the following fusion rule : @xmath540 cft operator & scaling dimension + @xmath541 & @xmath542 + @xmath543 & @xmath542 + @xmath544 & @xmath545 + @xmath546 & @xmath542 + the total number of edge modes is equal to the central charge of the corresponding cft for the states that are described by the pattern of zeros . for the hierarchy states , the number of edge modes is given by the rank of the @xmath344-matrix . furthermore , in the latter case , the number of right ( left ) -moving edge modes is given by the number positive ( negative ) eigenvalues of the @xmath344-matrix . the interlayer pfaffian states are described by a cft that consists of the ising cft , with @xmath547 , and two scalar boson cfts , each with @xmath548 . thus the number of edge modes for the interlayer pfaffian is @xmath549 . the central charge of the @xmath220 parafermion cft is @xmath536 ; the two - component fqh states based on this are described by the @xmath220 theory and two scalar bosons , for a total of @xmath551 edge modes . for the @xmath552 states , the @xmath344-matrix is @xmath553 . these states have 2 edge modes ; if @xmath554 , all edge modes move in the same direction ; if @xmath555 , then there is one right - moving and one left - moving edge mode . for the states @xmath556 , the @xmath344 matrix is @xmath557 where @xmath558 is the @xmath344-matrix in the hierarchical basis of the state @xmath559 . for the @xmath560 state , @xmath558 in the hierarchical basis is @xmath561 for the @xmath562 state , @xmath558 in the hierarchical basis is @xmath563 for the @xmath451 p - h conjugate of the @xmath564 laughlin state , @xmath558 in the hierarchical basis is @xmath565 using this , we find that the p - h conjugate of the @xmath564 laughlin state has edge modes @xmath566 , @xmath567 has edge modes @xmath568 , and @xmath569 has @xmath570 . now consider the bilayer composite fermion state @xmath556 : @xmath574 let @xmath575 be the maximum power of @xmath44 in @xmath576 . it satisfies : @xmath577 where @xmath578 is the shift of the state @xmath559 . the factor @xmath579 increases the power of @xmath44 by @xmath580 . thus the maximum power of @xmath44 in @xmath581 is @xmath474 : @xmath582 in our cases , @xmath583 , and the number of flux quanta is the same in each layer , so @xmath584 @xmath585 is the filling fraction in one layer . thus we see that the shift of @xmath581 is also @xmath578 . for @xmath459 , we take @xmath586 to be the particle - hole conjugate of the @xmath564 laughlin state . @xmath586 has shift @xmath587 . thus @xmath459 also has shift @xmath572 . @xmath588 has shift @xmath441 , because @xmath589 has shift @xmath441 . the electron operators for the @xmath220 fqh states that we discuss are of the form : @xmath593 where @xmath594 , @xmath408 , and @xmath595 for the @xmath418 , @xmath3 and @xmath425 states , respectively . the quasiparticle operators with minimal total charge are of the form : @xmath596 and have scaling dimension @xmath597 . the total charge of the quasiparticle is @xmath598 , @xmath599 , and @xmath600 for the @xmath418 , @xmath3 and @xmath425 states , respectively . @xmath601 , @xmath411 , and @xmath599 , respectively , for these states .

a growing body of evidence suggests that the stellar population of the galactic nucleus is distinct from that of the bulge . surface brightness measurements from the nir to the fir show that in contrast to the bulge , the central few hundred parsecs of the galaxy are dominated by a flat , disk - like distribution of stars , gas , and dust @xcite . this difference is also reflected in the kinematics : oh / ir stars in the central hundred parsecs show higher rotational velocities than expected for an inner single - component , bulge population @xcite . evidence for ongoing star formation also distinguishes the nucleus from the bulge . the galactic bulge is composed primarily of an old starburst - like population formed early , some 710 gyr ago @xcite . in contrast , the galactic nucleus harbors substantial young and intermediate - age stellar populations . intermediate - age populations have been most commonly inferred from broad - band photometry @xcite . star formation tracers have also been used to study intermediate - age populations , including young supergiants and luminous agbs , which trace 10 myr - 1 gyr populations @xcite , and oh / ir stars , which trace 1 - 3 gyr populations @xcite . the young stellar populations in the galactic nucleus have been the subject of much recent work . these populations are predominantly concentrated in three young , massive clusters : the arches and quintuplet clusters ( figer et al . 1999 , 2002 ; stolte et al . 2005 , and references therein ) , approximately 30 - 50 pc in projected distance from the galactic center , and the central cluster , located within the central parsec ( krabbe et al . 1991 , 1995 ; paumard et al . 2006 , and references therein ) . star formation in the nucleus is thought to be ultimately linked to the inward transport of gas induced by the galactic bar @xcite . however , precisely how gas is funneled from the outer galaxy to the nucleus and how this process affects the resultant star formation , remains poorly known . at present , there is a concentration of molecular gas 2 - 8 pc from the galctic center termed the circumnuclear disk @xcite . if such a structure routinely fuels star formation in the central parsec , the stars formed in this region may represent an entirely distinct population from the arches and quintpulet clusters or the intermediate age populations found throughout the @xmath5200 pc nucleus . dynamical effects due to the supermassive black hole ( genzel et al . 2003 , ghez et al . 2005 , and references therein ) and the large stellar and remnant density in the central parsec may also lead to population differences between the central cluster and that in the larger nucleus , either prior to star formation or afterwards . a number of investigators have attempted to address the nature of the central cluster and its relationship to the larger 100 - 300 pc nucleus . most recently , @xcite investigated the properties of the young stellar population in the central parsec , spectroscopically identifying nearly 100 ob and wolf - rayet stars . they find that the majority of the young stars reside in two , inclined and counter - rotating disks , suggesting in situ star formation in dense gas accretion disks . studies of the older stellar population in the central parsec have been primarily limited to broadband photometry . they have suggested that the fraction of low mass stars in the population increases with distance from the center , perhaps due to dynamical mass segregation @xcite . a detailed analysis , however , of the late - type giant population requires spectroscopy , owing to the large scatter in extinction near the galactic center and the intrinsic variations in giant star colors . @xcite pioneered work in this area , using spectroscopic and photometric observations of the most luminous giants and supergiants ( 50% complete at @xmath610 mag . ) to construct an h - r diagram for stars in the inner 5 pc . they report that the gc star formation rate in the central few parsecs is largely similar to that of the bulge . specifically , they find that the majority of stars formed more than 5 gyr ago , though they also find evidence that significant star formation also occurred during the past 100 myr . however , their conclusions are limited by their bright magnitude limit , which samples only short - lived evolutionary stages , for which theoretical models are uncertain . in this paper , we build on the work of @xcite in an effort to better characterize the late - type giant population within the central parsec . we report deep photometric and spectroscopic observations of 329 late - type giants in the gc , complete to 50% at @xmath615.5 mag . our observations for the first time include the helium - burning red clump , as well as the red giant branch and asymptotic giant branch . this improved magnitude limit allows for the most robust picture of the galactic center star formation history to date , as these stars are much better understood than the supergiants and luminous agb stars studied by @xcite . in 2 , we present our observations and the techniques used to construct the hertzsprung - russell ( h - r ) diagram . section 3 discusses the resulting h - r diagram and implications for the gc star - formation history . we discuss our results in 4 and conclude with 5 . we observed eight galactc center fields between march and september 2006 using the integral field spectrometer spiffi @xcite in conjunction with the macao adaptive optics module @xcite mounted on the sinfoni eso vlt facility . we observed these fields for the dual purpose of identifiying main sequence b stars outside the central parsec ( martins et al . 2007b , in prep ) and gathering spectroscopic observations for a large number of late - type giants . we chose fields with a radial distance from sgr a@xmath8 of less than @xmath9 arcsec , selecting regions outside the minispiral to avoid nebular contamination in brackett @xmath10 , and choosing fields north of sgr a@xmath8 to minimize separation from the ao guide star . we also avoided regions with extremely bright stars ( @xmath11 ) . as stars this bright are relatively rare , the selection bias in avoiding these stars is negligible . if the fields had instead been chosen at random , we expect no more than a few @xmath11 stars would be selected . we observed each selected sinfoni field in 50 @xmath12 100 mas pixel mode , resulting in 0.2 arcsec resolution and a @xmath13 field of view . we simultaneously observed all fields in @xmath2 and @xmath14 band , leading to a spectral resolution of r @xmath15 1500 . figure [ mos ] displays the location of the observed fields . the total exposure time for all fields was 4200 s , except for the fields at ( 17@xmath16 , 17@xmath16 ) , ( 5@xmath16 , 11@xmath16 ) , and ( -8@xmath16 , 8@xmath16 ) , which had total exposure times of 6600 s , 7200 s , and 2400 s , respectively . the fields span a range in projected distance from sgr a * of 4 arcsec to 26 arcsec . a histogram of the observed stellar positions is shown in figure [ dist_hist ] . we reduced the raw data using a standard procedure to perform flat - fielding , sky subtraction , and wavelength calibration @xcite . following extraction , we removed the stellar continua by dividing by a second degree polynomial fitted to the line - free regions of the stellar spectra . the normalized spectra allow us to directly compare the vlt spectra to the normalized catalog spectra in @xcite and @xcite . a typical resultant gc @xmath14 band spectrum following normalization is shown in figure [ spectrum ] . three template spectra from the @xcite spectral library are also shown for comparison . to classify the stars in our sample , we used the @xmath17co 2.2935 @xmath18 m , @xmath19 rovibrational band head , the strongest feature in our spectra and a well - known @xmath7 indicator , for a given luminosity class @xcite . we use only this feature , as it is largely free of nebular and telluric lines , making it more reliable than other @xmath20 and @xmath21band signatures in our spectra . we note that in contrast to the work of @xcite , we do not need the h+k band h@xmath22o feature to reliably predict effective temperatures , as our sample consists only of giants with @xmath2310.3 ( see 2.2 ) . due to their rarity , stars sufficiently bright to be supergiant or long period variable candidates are not contained in our sample . m co index ( see text for definition ) as a function of temperature for the comparison stars taken from the @xcite and @xcite catalogs . the solid line is a least - squares fit to both data sets for stars with 17 @xmath24 co @xmath25 3 . ] to define a @xmath7 - @xmath17co index relation , we computed the co index defined by @xcite . although not as well - known as other co indices ( e.g. equivalent width , photometric indices of @xcite ) , @xcite showed that this index correlates well with other co indices . furthermore , it has the advantage of being insensitive to small variations in the nearby continuum because it does not require a fit to the continuum . the index is defined as co % = [ ( 1 @xmath26 f@xmath27 f@xmath28 ) @xmath12 100 ) ] , where both bands are 0.015 @xmath18 m wide and the continuum and co band centers are defined as 2.284 @xmath18 m and 2.302 @xmath18 m , respectively . we note that although we adopt the same definition for the co index as @xcite , our spectra are normalized while theirs are not , and thus , our computed indices are not directly comparable to theirs . we calculated uncertainties in the computed indices , assuming the noise is dominated by photon statistics of the source and background and that the uncertainty in the band is approximately equal to that of the nearby continuum . following @xcite , we used co strength to estimate effective temperatures for the cool giants in our gc sample . to define a co index versus @xmath7 relation , we used a set of archival comparison star spectra taken from @xcite and @xcite with well - determined effective temperatures given in the literature . in an effort to avoid systematic errors in our derived index@xmath26@xmath7 relation , we used a large number of references , assuming temperatures derived from a wide variety of methods . a summary of the comparison star data is given in table [ comp_table ] , and the resulting relationship between co index and @xmath7 is shown in figure [ comp_relation ] . the @xcite and @xcite relationships are in very good agreement with each other ( for stars in common , @xmath29co @xmath30 ) , suggesting that the co index is largely independent of resolution or observing system . the relation is tightest for co @xmath31 . as no stars in our gc sample have computed indicies exceding this value , we did not include comparison stars with co @xmath32 in our fit . the resulting index-@xmath7 relation using both empirical data sets is : @xmath33 @xmath7 = 3.7351 @xmath26 0.0060 @xmath12 co @xmath26 0.00040 @xmath12 co@xmath34 , for all stars with measured indices @xmath35 co @xmath25 3 . the errors in the derived coefficients are 0.0092 , 0.0022 , and 1.2@xmath36 , respectively . llllllll hr 7001 & a0 va & wh & 1.760 & 0.004 & 9420 & 60 & @xcite + hr 4295 & a1 v & wh & 1.434 & 0.003 & 9600 & 25 & @xcite + hr 5054 & a1 vp & wh & 1.823 & 0.004 & 8760 & 260 & @xcite + hr 6378 & a2 v & wh & 1.953 & 0.004 & 8690 & 78 & @xcite + hr 4534 & a3 v & wh & 1.695 & 0.005 & 8857 & 185 & @xcite + hr 4357 & a4 v & wh & 1.516 & 0.007 & 8243 & 150 & @xcite + hr 7557 & a7 v & wh & 1.147 & 0.004 & 7500 & 200 & @xcite + hr 2943 & f5 iv - v & wh & 1.176 & 0.014 & 6532 & 39 & @xcite + hr 6927 & f7 v & wh & 1.219 & 0.014 & 6087 & 22 & @xcite + hr 4375 & f8.5 v & wh & 1.540 & 0.015 & 5676 & 66 & @xcite + hr 21 & f2 iii & wh & 1.563 & 0.011 & 6847 & 137 & @xcite + hr 403 & a5 iii - iv & wh & 1.828 & 0.008 & 8420 & 360 & @xcite + hr 1412 & a7 iii & wh & 1.554 & 0.007 & 7690 & 320 & @xcite + hr 1457 & k5@xmath37 iii & wh & 12.633 & 0.023 & 3866 & 35 & @xcite + hr 2985 & g8 iii & wh & 5.540 & 0.018 & 5001 & 56 & @xcite + hr 3003 & k4 iii & wh & 13.944 & 0.023 & 3961 & 99 & @xcite + hr 3323 & g5 iiia & wh & 3.045 & 0.017 & 5136 & 88 & @xcite + & & kh & 3.039 & 0.007 & & + hr 4031 & f0 iii & wh & 2.145 & 0.011 & 6880 & 150 & @xcite + hr 4069 & m0 iii & wh & 14.143 & 0.021 & 3730 & 149 & @xcite + hr 4517 & m1 iii & wh & 13.894 & 0.019 & 3828 & 53 & @xcite + hr 4883 & g0 iiip & wh & 1.819 & 0.014 & 5747 & 54 & @xcite + & & kh & 1.377 & 0.007 & & + hr 5017 & f3 iii & wh & 1.395 & 0.017 & 7141 & 181 & @xcite + hr 5340 & k1.5 iiip & wh & 10.997 & 0.021 & 4233 & 55 & @xcite + hr 6299 & k2 iii & wh & 8.653 & 0.020 & 4571 & 100 & @xcite + & & kh & 8.530 & 0.009 & & + hr 6703 & g8.5 iii & wh & 4.426 & 0.020 & 5011 & 35 & @xcite + hr 6705 & k5 iii & wh & 12.859 & 0.023 & 3934 & 42 & @xcite + & & kh & 12.569 & 0.010 & & + hr 7635 & m0@xmath26 iii & wh & 13.035 & 0.025 & 3867 & 50 & @xcite + & & kh & 12.869 & 0.010 & & + hr 7806 & k2.5 iii & wh & 9.993 & 0.022 & 4286 & 100 & @xcite + & & kh & 9.785 & 0.010 & & + hr 7886 & m6 iii & wh & 16.722 & 0.025 & 3243 & 79 & @xcite + hr 8317 & k0.5 iii & wh & 7.319 & 0.024 & 4658 & 19 & @xcite + hr 8694 & k0@xmath26 iii & wh & 6.691 & 0.021 & 4830 & 100 & @xcite + & & kh & 6.463 & 0.010 & & + hr 8905 & f8 iii & wh & 1.265 & 0.014 & 5942 & 42 & @xcite + & & kh & 0.797 & 0.007 & & + swvir & m7 iii & wh & 18.350 & 0.024 & 2966 & 36 & @xcite + & & kh & 18.033 & 0.008 & & + bkvir & m7@xmath26 iii & wh & 20.449 & 0.022 & 2944 & 34 & @xcite + & & kh & 20.730 & 0.006 & & + rxboo & m7.5 - 8 iii & wh & 18.262 & 0.024 & 2786 & 46 & @xcite + [ comp_table ] in order to separate cool giants from warmer giants and main sequence stars , warm stars with co @xmath38 3 are removed from the sample . of the 355 detected stars in our spectroscopy , 329 were categorized as co - absorbers , and we assigned effective temperatures to these stars using the above index@xmath26@xmath7 relation . prior to temperature calculation , we cross - correlated all co - absorbing stars with a co - template in order to remove radial velocities . errors in temperatures were estimated based on the noise in each spectrum and the intrinsic dispersion in our index@xmath26@xmath7 relation ( @xmath39 = 0.0050 ) . as noted by @xcite , errors determined in this way are strictly only lower limits . the spectral classification and analysis of the remaining early - type ( non - co - absorbing ) stars in our sample is less straight - forward @xcite , and we defer analysis of these stars to a later paper ( martins et al . 2007b , in prep ) . we made photometric observations using the imaging system naos / conica ( naco ) , consisting of the adaptive optics system naos @xcite and the nir camera conica @xcite at the 8.2-m ut4 ( yepun ) of the eso vlt . in april 2006 , we collected several ao - corrected images in each of @xmath2 and @xmath3 band with a pixel size of @xmath40 27 mas . the total exposure time was 64 s for each band , broken into thirty - two dithered images , in which every fourth image was separated from the previous exposure by 16 arcsec . we used our standard reduction pipeline to perform sky subtraction , bad pixel corrections , flat field corrections , and stacking of images to create final mosaics . in figure [ mos ] , we display a sub - section of the final @xmath3 band mosaic , containing all stars analyzed in this paper . photometry was extracted for only this region to help ensure psf constancy across the image . we used the crowded field photometry package starfinder @xcite to establish relative photometry and source detection . an empirical psf for the central core ( fwhm @xmath41 0.10 arcsec ) was extracted from each image using seven bright , isolated stars . the full radial extent of the extracted psf was r @xmath41 0.19 arcsec . to derive the photometric curve of growth needed to place these results on an absolute scale , we adopted the mtf - fitting technique of @xcite . this technique fits the power spectrum of the image using a combination of the source spatial distribution function determined by starfinder and a parameterized description of the modulation transfer functions ( mtf ) of the atmosphere , telescope , ao system , and science camera . the advantage of this technique is that it derives the psf encircled energy curve of growth , including the extended , seeing - limited halo , and thus provides the aperture correction required for absolute photometric calibration . in deriving the curve of growth from the data themselves , we avoid systematic calibration errors introduced from using a psf standard acquired at a different time and under different observing conditions than the target data . given the rapid variations in ao performance @xcite , this technique is crucial for deriving accurate absolute photometry . the mtf - fitting technique has so far only been applied to data obtained for the keck observatory lgsao system @xcite . to apply this technique to the vlt images , we used the @xcite software with appropriate input parameters for the vlt telescope pupil size and geometry , the camera platescale and pixel size , and the deformable mirror actuator spacing . the software uses the idl procedure mpfit to perform a levenberg - marquardt least - squares fit of the parameterized model mtf to the data . the best fits to the @xmath2 and @xmath3 band image power spectra are shown in figure [ pspec ] . the spatial frequency , @xmath42 , is normalized relative to the telescope cutoff frequency in the image plane ( d@xmath43 ) . band data are aliased , and we therefore , fit only the power spectrum for @xmath44 ( see 2.2 ) . the spatial frequency , @xmath42 , is normalized relative to the cutoff frequency in the image plane ( d@xmath43 ) . the strehl ratios implied by these fits are 10@xmath45% and 19@xmath46% , for @xmath2 and @xmath3 band , respectively . ] all images , even seeing - limited images , contain power up to the spatial frequency cutoff , which is @xmath47 for a circular pupil . to faithfully record all spatial information allowed by the telescope aperture , the pixel sampling frequency must therefore satisfy @xmath48 ( see @xcite for a detailed discussion ) . the @xmath3 band data are nyquist - sampled : @xmath49 = 1 . however , the h - band data are undersampled : @xmath50 . this aliasing causes power at frequencies @xmath51 to corrupt frequencies @xmath52 . the contamination of frequencies between 0.52 and @xmath50.7 is negligible , as the power aliased to these frequencies consists only of detector and background noise . this power is orders of magnitude less than the non - aliased power at these spatial frequencies , and thus , has very little effect on the total power spectrum . on the other hand , the aliased power for contaminated spatial frequencies @xmath53 is of the same order - of - magnitude as the non - aliased power , so the power spectrum at these frequencies is heavily influenced by the aliasing . we , therefore , chose only to fit the power spectrum for all spatial frequencies @xmath54 . as only detector and background noise contribute to the power spectrum for spatial frequencies @xmath55 , we are excluding very little information about the psf . the relatively minor discrepency in figure [ pspec ] between the model and data power spectra in @xmath3 band is likely due to uncertainties in the deformable mirror influence function , as an explicit description of the vlt cilas mirror influence function was not available . we , therefore , used the approximate influence function for the keck xinetics mirror from @xcite . the overall power spectra fits are satisfactory , and we do not expect that the photometric accuracy of our technique is significantly affected by this choice . the strehl ratios implied by the psfs reconstructed from the fits in figure [ pspec ] are 10% and 19% , for @xmath2 and @xmath3 band , respectively . to test the integrity of these fits , we divided the images into nine subimages and fit the power spectrum of each subimage . this process yielded fairly consistent results for all subimages , implying an error in the @xmath2 and @xmath3 band strehl ratios of 0.5% and 1.0% , respectively . this agreement between subimages suggests that psf variations due to variable distance from the guide star and variable exposure time in the final mosaic are only present at a low level . we determined the photometric zero point from observations of the near - ir standard star 9178 @xcite , obtained on the same night . the measured atmospheric extinction coefficients used to correct for the difference in airmass between our standard stars and science targets were 0.06 mag . airmass@xmath56 for @xmath2 and 0.07 mag . airmass@xmath56 for @xmath3 . we derived photometric errors and completeness using the standard technique of inserting and recovering artificial stars of specified magnitude . completeness is not uniform , so we calculated errors and completeness separately for each spectral field . to avoid artificially crowding our images , we inserted only fifty stars at a time to each spectral subimage and repeated this procedure five times for each magnitude bin . we computed the photometric error at each magnitude using a gaussian fit to the difference between the input and recovered magnitudes . figure [ comp_err ] shows the average photometric errors and completeness at each magnitude bin . also shown is the spectroscopic completeness , estimated by comparing the number of stars detected in the @xmath2 and @xmath57 band photometry to the number detected in the sinfoni data . the incompleteness in the sinfoni data at @xmath613 - 15 mag . is caused by crowding and confusion with nearby bright stars in the field of view . and @xmath3 band photometric errors and completeness based on the artificial star tests described in 2 . the spectroscopic completeness is estimated based on the number of co - stars for which we could extract spectra , compared to the total number of stars detected in the photometry . ] using the derived photometry and effective temperatures presented in the preceding section , we are able to place our gc sample on the hertzsprung - russell ( h - r ) diagram . to convert the measured photometric magnitudes to luminosity , we use models that assume the @xmath14 band magnitudes are measured in the johnson - cousins - glass system @xcite . therefore , we first transformed the measured @xmath3 magnitudes to the johnson - cousins - glass system , using the approximate transformations given in @xcite , assuming the naco and 2mass @xmath58 color systems are identical , as in @xcite . we next adopted empirical bolometric corrections of @xcite for m stars and theoretical corrections of girardi 2005 ( http://pleiadi / pd.astro.it ) for hotter stars . for the theoretical corrections , we assumed a solar metallicity and a surface gravity of @xmath59 , though the results are largely independent of these assumptions . for all temperatures in the galactic center sample , we found systematic differences @xmath60 mag . in bc@xmath61 for @xmath62=1.0 $ ] dex and @xmath63 dex , leading to systematic differences in the luminosity of @xmath64 . additionally , we assumed a distance to the galactic center of 8.0 kpc @xcite . there has been much recent debate concerning the precise distance to the galactic center . however , the choice of distance has only a small systematic effect on our derived luminosities ; a change in distance of @xmath65 kpc induces a change in luminosity of @xmath66 . finally , we corrected each star s luminosity individually for extinction using the photometric color , the derived @xmath7 , and the interstellar extinction law of @xcite , derived from nicmos observations of galactic center stars . all results are tabulated in table [ gc_table ] . lccccccccccccc 1 & 6.26 & 6.24 & 15.275 & 0.042 & 17.419 & 0.072 & 8.28 & 3.658 & 0.006 & 2.99 & 0.15 & 1.88 & 0.06 + 2 & 6.32 & 6.05 & 15.259 & 0.042 & 17.316 & 0.072 & 8.30 & 3.658 & 0.007 & 2.86 & 0.15 & 1.83 & 0.07 + 3 & 5.67 & 5.91 & 14.898 & 0.031 & 17.051 & 0.056 & 6.93 & 3.674 & 0.004 & 3.03 & 0.11 & 2.09 & 0.05 + 4 & 5.67 & 6.13 & 14.239 & 0.013 & 16.344 & 0.049 & 8.61 & 3.654 & 0.006 & 2.93 & 0.09 & 2.25 & 0.04 + 5 & 5.51 & 6.24 & 15.304 & 0.042 & 17.438 & 0.072 & 7.01 & 3.673 & 0.005 & 3.00 & 0.15 & 1.91 & 0.06 + 6 & 5.02 & 6.48 & 15.531 & 0.042 & 17.826 & 0.204 & 7.15 & 3.672 & 0.008 & 3.23 & 0.37 & 1.91 & 0.15 + 7 & 4.86 & 6.29 & 15.131 & 0.031 & 17.274 & 0.072 & 7.01 & 3.673 & 0.007 & 3.01 & 0.14 & 1.98 & 0.06 + 8 & 4.99 & 6.16 & 15.327 & 0.042 & 17.406 & 0.072 & 7.34 & 3.669 & 0.005 & 2.91 & 0.15 & 1.86 & 0.06 + 9 & 4.75 & 6.13 & 15.503 & 0.042 & 17.678 & 0.072 & 8.76 & 3.652 & 0.006 & 3.03 & 0.15 & 1.78 & 0.06 + 10 & 4.35 & 6.34 & 15.474 & 0.042 & 17.135 & 0.056 & 6.82 & 3.675 & 0.004 & 2.30 & 0.12 & 1.57 & 0.05 + 11 & 4.35 & 6.51 & 15.240 & 0.031 & 17.387 & 0.072 & 6.69 & 3.677 & 0.004 & 3.02 & 0.14 & 1.96 & 0.06 + 12 & 3.54 & 6.56 & 14.318 & 0.017 & 16.427 & 0.049 & 7.73 & 3.665 & 0.005 & 2.95 & 0.09 & 2.26 & 0.04 + 13 & 3.56 & 6.40 & 13.803 & 0.013 & 15.707 & 0.012 & 9.40 & 3.643 & 0.005 & 2.62 & 0.03 & 2.27 & 0.02 + 14 & 3.48 & 6.13 & 14.969 & 0.031 & 16.899 & 0.056 & 6.90 & 3.675 & 0.005 & 2.70 & 0.11 & 1.93 & 0.05 + 15 & 3.08 & 6.13 & 12.132 & 0.003 & 14.305 & 0.006 & 11.62 & 3.611 & 0.005 & 2.96 & 0.02 & 2.99 & 0.01 + 16 & 3.05 & 5.83 & 14.123 & 0.013 & 16.226 & 0.042 & 8.65 & 3.653 & 0.006 & 2.93 & 0.08 & 2.30 & 0.04 + 17 & 2.81 & 5.86 & 15.867 & 0.078 & 17.802 & 0.204 & 8.77 & 3.652 & 0.005 & 2.67 & 0.39 & 1.49 & 0.16 + 18 & 3.27 & 5.62 & 15.596 & 0.042 & 17.751 & 0.204 & 9.25 & 3.645 & 0.007 & 2.99 & 0.37 & 1.71 & 0.15 + 19 & 2.78 & 5.32 & 15.115 & 0.031 & 17.234 & 0.056 & 9.20 & 3.646 & 0.007 & 2.94 & 0.11 & 1.88 & 0.05 + 20 & 4.35 & 5.18 & 11.399 & 0.002 & 13.519 & 0.003 & 12.10 & 3.604 & 0.005 & 2.87 & 0.01 & 3.23 & 0.01 + 21 & 4.78 & 5.37 & 14.438 & 0.017 & 16.735 & 0.049 & 9.71 & 3.639 & 0.004 & 3.19 & 0.09 & 2.24 & 0.04 + 22 & 5.13 & 5.51 & 15.436 & 0.042 & 17.641 & 0.072 & 8.55 & 3.655 & 0.004 & 3.08 & 0.15 & 1.84 & 0.06 + 23 & 4.97 & 5.10 & 15.341 & 0.042 & 17.335 & 0.072 & 10.09 & 3.634 & 0.005 & 2.74 & 0.15 & 1.68 & 0.06 + 24 & 5.53 & 4.54 & 14.770 & 0.031 & 17.054 & 0.056 & 6.77 & 3.676 & 0.006 & 3.22 & 0.11 & 2.22 & 0.05 + 25 & 5.10 & 3.46 & 14.078 & 0.013 & 15.666 & 0.012 & 8.31 & 3.658 & 0.005 & 2.17 & 0.03 & 2.02 & 0.02 + 26 & 5.16 & 3.67 & 15.087 & 0.031 & 17.025 & 0.056 & 9.92 & 3.636 & 0.011 & 2.66 & 0.12 & 1.76 & 0.06 + 27 & 5.45 & 3.91 & 15.316 & 0.042 & 17.369 & 0.072 & 10.07 & 3.634 & 0.007 & 2.82 & 0.15 & 1.73 & 0.06 + 28 & 5.08 & 4.18 & 13.152 & 0.005 & 15.121 & 0.006 & 10.32 & 3.631 & 0.006 & 2.69 & 0.02 & 2.53 & 0.02 + 29 & 5.13 & 4.51 & 13.238 & 0.005 & 15.345 & 0.012 & 10.91 & 3.622 & 0.005 & 2.88 & 0.02 & 2.55 & 0.02 + 30 & 5.05 & 4.70 & 13.980 & 0.013 & 16.298 & 0.049 & 12.37 & 3.600 & 0.007 & 3.16 & 0.09 & 2.30 & 0.04 + 31 & 4.56 & 4.56 & 13.382 & 0.009 & 15.542 & 0.012 & 8.63 & 3.653 & 0.004 & 3.01 & 0.03 & 2.63 & 0.02 + 32 & 4.29 & 4.29 & 15.536 & 0.042 & 17.606 & 0.072 & 7.99 & 3.662 & 0.004 & 2.89 & 0.15 & 1.74 & 0.06 + 33 & 4.59 & 4.13 & 16.137 & 0.078 & 18.223 & 0.204 & 5.27 & 3.692 & 0.003 & 2.95 & 0.39 & 1.61 & 0.16 + 34 & 4.40 & 3.48 & 13.845 & 0.013 & 15.739 & 0.012 & 8.65 & 3.653 & 0.006 & 2.62 & 0.03 & 2.28 & 0.02 + 35 & 4.24 & 3.70 & 13.676 & 0.009 & 15.619 & 0.012 & 7.61 & 3.666 & 0.005 & 2.71 & 0.03 & 2.42 & 0.02 + 36 & 4.32 & 3.02 & 13.194 & 0.005 & 15.295 & 0.012 & 9.46 & 3.642 & 0.005 & 2.91 & 0.02 & 2.63 & 0.02 + 37 & 3.00 & 3.32 & 12.562 & 0.003 & 14.834 & 0.006 & 3.70 & 3.707 & 0.004 & 3.24 & 0.01 & 3.20 & 0.01 + 38 & 3.11 & 3.83 & 15.976 & 0.078 & 18.227 & 0.204 & 9.31 & 3.645 & 0.008 & 3.13 & 0.39 & 1.61 & 0.16 + 39 & 3.62 & 3.56 & 15.106 & 0.031 & 17.234 & 0.056 & 5.89 & 3.686 & 0.004 & 3.00 & 0.11 & 2.03 & 0.05 + 40 & 4.02 & 3.97 & 15.532 & 0.042 & 17.603 & 0.072 & 7.97 & 3.662 & 0.006 & 2.89 & 0.15 & 1.74 & 0.06 + 41 & 3.62 & 4.05 & 15.160 & 0.031 & 17.431 & 0.072 & 9.04 & 3.648 & 0.005 & 3.17 & 0.14 & 1.96 & 0.06 + 42 & 2.92 & 4.13 & 12.738 & 0.003 & 15.117 & 0.006 & 12.73 & 3.594 & 0.005 & 3.23 & 0.02 & 2.82 & 0.02 + 43 & -6.97 & 10.15 & 15.030 & 0.026 & 16.988 & 0.058 & 8.98 & 3.649 & 0.008 & 2.71 & 0.11 & 1.83 & 0.05 + 44 & -7.24 & 10.31 & 14.751 & 0.026 & 16.765 & 0.058 & 7.28 & 3.670 & 0.006 & 2.82 & 0.11 & 2.05 & 0.05 + 45 & -7.10 & 9.85 & 15.112 & 0.026 & 17.048 & 0.058 & 5.96 & 3.685 & 0.006 & 2.72 & 0.11 & 1.91 & 0.05 + 46 & -7.32 & 9.88 & 15.075 & 0.026 & 17.022 & 0.058 & 6.46 & 3.680 & 0.006 & 2.73 & 0.11 & 1.91 & 0.05 + 47 & -7.94 & 9.88 & 15.494 & 0.040 & 17.295 & 0.060 & 4.70 & 3.698 & 0.006 & 2.53 & 0.13 & 1.72 & 0.06 + 48 & -8.99 & 10.12 & 13.124 & 0.005 & 15.144 & 0.008 & 10.30 & 3.631 & 0.008 & 2.77 & 0.02 & 2.57 & 0.02 + 49 & -8.80 & 9.67 & 15.374 & 0.040 & 17.408 & 0.060 & 7.61 & 3.666 & 0.009 & 2.84 & 0.13 & 1.80 & 0.06 + 50 & -9.29 & 9.37 & 15.199 & 0.026 & 17.210 & 0.058 & 8.72 & 3.652 & 0.007 & 2.79 & 0.11 & 1.81 & 0.05 + 51 & -8.69 & 9.21 & 15.156 & 0.026 & 17.093 & 0.058 & 6.25 & 3.682 & 0.006 & 2.71 & 0.11 & 1.88 & 0.05 + 52 & -7.45 & 9.37 & 11.937 & 0.003 & 13.686 & 0.002 & 5.88 & 3.686 & 0.004 & 2.44 & 0.01 & 3.07 & 0.01 + 53 & -5.59 & 9.21 & 10.319 & 0.002 & 12.222 & 0.002 & 11.08 & 3.619 & 0.008 & 2.58 & 0.02 & 3.59 & 0.02 + 54 & -6.75 & 8.80 & 15.448 & 0.040 & 17.392 & 0.060 & 6.35 & 3.681 & 0.005 & 2.72 & 0.13 & 1.76 & 0.06 + 55 & -6.75 & 8.56 & 15.948 & 0.090 & 18.090 & 0.245 & 6.47 & 3.679 & 0.008 & 3.02 & 0.46 & 1.68 & 0.19 + 56 & -6.18 & 8.18 & 15.273 & 0.040 & 17.416 & 0.060 & 8.85 & 3.651 & 0.009 & 2.98 & 0.13 & 1.85 & 0.06 + 57 & -6.29 & 8.07 & 15.706 & 0.040 & 17.964 & 0.245 & 8.91 & 3.650 & 0.007 & 3.15 & 0.44 & 1.74 & 0.18 + 58 & -6.10 & 7.86 & 15.506 & 0.040 & 17.688 & 0.060 & 4.64 & 3.699 & 0.005 & 3.10 & 0.13 & 1.94 & 0.06 + 59 & -5.97 & 7.29 & 15.037 & 0.026 & 17.188 & 0.058 & 4.32 & 3.702 & 0.008 & 3.05 & 0.11 & 2.12 & 0.05 + 60 & -6.16 & 6.43 & 13.182 & 0.005 & 15.300 & 0.009 & 7.56 & 3.667 & 0.008 & 2.97 & 0.02 & 2.73 & 0.02 + 61 & -6.51 & 6.64 & 15.628 & 0.040 & 17.772 & 0.245 & 9.21 & 3.646 & 0.008 & 2.98 & 0.44 & 1.69 & 0.18 + 62 & -6.80 & 6.86 & 14.604 & 0.022 & 16.766 & 0.058 & 7.46 & 3.668 & 0.006 & 3.03 & 0.11 & 2.19 & 0.05 + 63 & -7.32 & 7.78 & 13.190 & 0.005 & 15.195 & 0.008 & 8.64 & 3.653 & 0.006 & 2.78 & 0.02 & 2.61 & 0.02 + 64 & -7.45 & 8.29 & 15.413 & 0.040 & 17.374 & 0.060 & 6.44 & 3.680 & 0.005 & 2.75 & 0.13 & 1.79 & 0.06 + 65 & -7.64 & 8.56 & 15.467 & 0.040 & 17.477 & 0.060 & 6.86 & 3.675 & 0.006 & 2.82 & 0.13 & 1.78 & 0.06 + 66 & -7.48 & 8.99 & 12.976 & 0.005 & 14.980 & 0.008 & 12.14 & 3.603 & 0.009 & 2.70 & 0.03 & 2.53 & 0.02 + 67 & -8.23 & 8.88 & 15.498 & 0.040 & 17.511 & 0.060 & 5.75 & 3.687 & 0.008 & 2.83 & 0.13 & 1.81 & 0.06 + 68 & -8.91 & 8.64 & 15.444 & 0.040 & 17.651 & 0.060 & 9.59 & 3.641 & 0.008 & 3.06 & 0.13 & 1.79 & 0.06 + 69 & -8.59 & 8.18 & 12.292 & 0.004 & 14.300 & 0.005 & 8.80 & 3.651 & 0.005 & 2.78 & 0.01 & 2.97 & 0.02 + 70 & -8.13 & 8.13 & 14.233 & 0.010 & 16.282 & 0.047 & 8.98 & 3.649 & 0.009 & 2.84 & 0.09 & 2.21 & 0.04 + 71 & -8.18 & 8.32 & 15.497 & 0.040 & 17.501 & 0.060 & 7.83 & 3.664 & 0.007 & 2.79 & 0.13 & 1.72 & 0.06 + 72 & -7.94 & 8.45 & 15.375 & 0.040 & 17.420 & 0.060 & 9.34 & 3.644 & 0.007 & 2.83 & 0.13 & 1.73 & 0.06 + 73 & -7.75 & 8.26 & 15.616 & 0.040 & 17.814 & 0.245 & 6.29 & 3.682 & 0.006 & 3.10 & 0.44 & 1.85 & 0.18 + 74 & -7.97 & 7.51 & 12.569 & 0.004 & 14.867 & 0.008 & 10.02 & 3.635 & 0.009 & 3.19 & 0.02 & 2.97 & 0.02 + 75 & -7.61 & 7.21 & 14.774 & 0.026 & 16.954 & 0.058 & 5.67 & 3.688 & 0.007 & 3.08 & 0.11 & 2.20 & 0.05 + 76 & -8.45 & 6.99 & 10.843 & 0.002 & 13.278 & 0.002 & 12.81 & 3.593 & 0.013 & 3.32 & 0.03 & 3.60 & 0.03 + 77 & -9.21 & 6.53 & 12.543 & 0.004 & 14.852 & 0.008 & 10.86 & 3.623 & 0.008 & 3.18 & 0.02 & 2.95 & 0.02 + 78 & -8.64 & 7.34 & 15.402 & 0.040 & 17.464 & 0.060 & 8.56 & 3.654 & 0.007 & 2.87 & 0.13 & 1.76 & 0.06 + 79 & 2.94 & 12.93 & 15.314 & 0.049 & 16.938 & 0.055 & 5.66 & 3.688 & 0.004 & 2.26 & 0.13 & 1.65 & 0.06 + 80 & 2.67 & 12.74 & 14.953 & 0.027 & 16.828 & 0.055 & 7.95 & 3.662 & 0.005 & 2.60 & 0.11 & 1.86 & 0.05 + 81 & 2.54 & 12.45 & 15.140 & 0.027 & 16.947 & 0.055 & 8.00 & 3.661 & 0.006 & 2.50 & 0.11 & 1.74 & 0.05 + 82 & 2.59 & 12.04 & 15.189 & 0.027 & 16.632 & 0.046 & 4.90 & 3.696 & 0.005 & 2.00 & 0.09 & 1.62 & 0.04 + 83 & 3.21 & 12.69 & 16.003 & 0.075 & 17.874 & 0.210 & 7.56 & 3.667 & 0.007 & 2.60 & 0.40 & 1.45 & 0.16 + 84 & 3.00 & 12.64 & 15.868 & 0.075 & 17.752 & 0.210 & 6.33 & 3.681 & 0.006 & 2.64 & 0.40 & 1.56 & 0.16 + 85 & 5.62 & 11.88 & 15.501 & 0.049 & 17.546 & 0.080 & 9.55 & 3.641 & 0.006 & 2.82 & 0.17 & 1.67 & 0.07 + 86 & 5.83 & 11.42 & 11.834 & 0.003 & 14.160 & 0.003 & 13.30 & 3.585 & 0.007 & 3.14 & 0.02 & 3.11 & 0.02 + 87 & 5.94 & 11.12 & 13.811 & 0.011 & 16.328 & 0.046 & 10.99 & 3.621 & 0.008 & 3.49 & 0.09 & 2.56 & 0.04 + 88 & 5.48 & 11.45 & 13.479 & 0.008 & 16.289 & 0.046 & 11.75 & 3.609 & 0.007 & 3.90 & 0.08 & 2.83 & 0.04 + 89 & 5.18 & 11.53 & 15.624 & 0.049 & 17.880 & 0.210 & 6.90 & 3.675 & 0.005 & 3.18 & 0.38 & 1.86 & 0.16 + 90 & 4.83 & 11.74 & 15.531 & 0.049 & 17.641 & 0.080 & 7.38 & 3.669 & 0.005 & 2.96 & 0.17 & 1.79 & 0.07 + 91 & 4.78 & 11.53 & 15.510 & 0.049 & 17.643 & 0.080 & 4.91 & 3.696 & 0.006 & 3.02 & 0.17 & 1.90 & 0.07 + 92 & 4.35 & 11.69 & 14.863 & 0.027 & 16.784 & 0.055 & 8.03 & 3.661 & 0.005 & 2.67 & 0.11 & 1.92 & 0.05 + 93 & 3.70 & 11.42 & 12.593 & 0.003 & 14.495 & 0.006 & 5.95 & 3.685 & 0.003 & 2.67 & 0.01 & 2.90 & 0.01 + 94 & 3.43 & 11.39 & 13.861 & 0.011 & 15.757 & 0.042 & 8.13 & 3.660 & 0.005 & 2.63 & 0.08 & 2.30 & 0.03 + 95 & 2.38 & 10.96 & 12.391 & 0.003 & 14.350 & 0.006 & 11.31 & 3.616 & 0.004 & 2.66 & 0.01 & 2.78 & 0.01 + 96 & 2.59 & 10.77 & 13.164 & 0.005 & 14.799 & 0.006 & 6.53 & 3.679 & 0.004 & 2.26 & 0.01 & 2.49 & 0.01 + 97 & 3.16 & 10.69 & 14.694 & 0.019 & 16.701 & 0.046 & 5.54 & 3.690 & 0.006 & 2.83 & 0.09 & 2.13 & 0.04 + 98 & 3.73 & 10.53 & 12.842 & 0.005 & 14.355 & 0.006 & 8.50 & 3.655 & 0.005 & 2.06 & 0.01 & 2.47 & 0.01 + 99 & 4.51 & 10.85 & 11.905 & 0.003 & 13.794 & 0.003 & 5.57 & 3.689 & 0.003 & 2.65 & 0.01 & 3.18 & 0.01 + 100 & 5.10 & 10.66 & 13.593 & 0.008 & 15.708 & 0.008 & 9.17 & 3.646 & 0.007 & 2.93 & 0.02 & 2.49 & 0.02 + 101 & 5.67 & 10.80 & 15.151 & 0.027 & 17.216 & 0.055 & 8.89 & 3.650 & 0.005 & 2.86 & 0.11 & 1.85 & 0.05 + 102 & 5.40 & 9.96 & 15.114 & 0.027 & 17.121 & 0.055 & 6.77 & 3.676 & 0.006 & 2.81 & 0.11 & 1.92 & 0.05 + 103 & 5.16 & 9.88 & 15.134 & 0.027 & 17.133 & 0.055 & 8.75 & 3.652 & 0.007 & 2.77 & 0.11 & 1.83 & 0.05 + 104 & 5.91 & 9.45 & 15.730 & 0.049 & 17.717 & 0.080 & 6.20 & 3.682 & 0.009 & 2.79 & 0.17 & 1.68 & 0.07 + 105 & 5.08 & 10.18 & 13.307 & 0.008 & 15.413 & 0.008 & 11.02 & 3.620 & 0.007 & 2.88 & 0.02 & 2.51 & 0.02 + 106 & 3.59 & 9.45 & 13.281 & 0.008 & 15.303 & 0.008 & 11.49 & 3.613 & 0.006 & 2.74 & 0.02 & 2.45 & 0.02 + 107 & 3.29 & 9.40 & 15.212 & 0.027 & 17.153 & 0.055 & 7.64 & 3.666 & 0.009 & 2.70 & 0.11 & 1.81 & 0.05 + 108 & 3.13 & 9.53 & 15.129 & 0.027 & 16.961 & 0.055 & 6.68 & 3.677 & 0.005 & 2.55 & 0.11 & 1.81 & 0.05 + 109 & 3.08 & 9.83 & 13.819 & 0.011 & 15.715 & 0.008 & 9.31 & 3.644 & 0.006 & 2.61 & 0.03 & 2.27 & 0.02 + 110 & 3.46 & 10.12 & 13.302 & 0.008 & 15.255 & 0.008 & 14.05 & 3.572 & 0.007 & 2.56 & 0.03 & 2.27 & 0.02 + 111 & 2.94 & 10.07 & 14.905 & 0.027 & 16.849 & 0.055 & 8.69 & 3.653 & 0.005 & 2.69 & 0.11 & 1.89 & 0.05 + 112 & 2.51 & 10.23 & 14.381 & 0.019 & 15.823 & 0.042 & 5.62 & 3.689 & 0.005 & 1.99 & 0.08 & 1.92 & 0.04 + 113 & 2.67 & 9.53 & 15.518 & 0.049 & 17.613 & 0.080 & 7.98 & 3.662 & 0.007 & 2.92 & 0.17 & 1.76 & 0.07 + 114 & -4.86 & 12.72 & 15.647 & 0.042 & 17.899 & 0.224 & 7.68 & 3.665 & 0.007 & 3.16 & 0.41 & 1.82 & 0.16 + 115 & -5.40 & 12.39 & 14.498 & 0.019 & 16.366 & 0.048 & 7.14 & 3.672 & 0.005 & 2.60 & 0.09 & 2.07 & 0.04 + 116 & -6.18 & 12.69 & 15.679 & 0.042 & 17.406 & 0.072 & 6.76 & 3.676 & 0.005 & 2.40 & 0.15 & 1.53 & 0.06 + 117 & -6.59 & 12.77 & 16.205 & 0.068 & 18.072 & 0.224 & 5.31 & 3.692 & 0.005 & 2.62 & 0.42 & 1.45 & 0.17 + 118 & -6.78 & 12.66 & 16.404 & 0.119 & 18.180 & 0.224 & 6.30 & 3.681 & 0.008 & 2.48 & 0.45 & 1.28 & 0.19 + 119 & -7.48 & 12.26 & 15.019 & 0.033 & 16.737 & 0.048 & 7.69 & 3.665 & 0.005 & 2.37 & 0.10 & 1.75 & 0.05 + 120 & -7.70 & 12.45 & 15.224 & 0.033 & 16.988 & 0.053 & 6.94 & 3.674 & 0.005 & 2.45 & 0.11 & 1.73 & 0.05 + 121 & -7.07 & 11.93 & 15.166 & 0.033 & 16.674 & 0.048 & 5.40 & 3.691 & 0.004 & 2.09 & 0.10 & 1.65 & 0.04 + 122 & -6.86 & 11.85 & 15.556 & 0.042 & 17.281 & 0.072 & 6.11 & 3.683 & 0.004 & 2.40 & 0.15 & 1.60 & 0.06 + 123 & -6.56 & 11.99 & 15.669 & 0.042 & 17.310 & 0.072 & 5.47 & 3.690 & 0.004 & 2.29 & 0.15 & 1.53 & 0.06 + 124 & -6.21 & 12.10 & 15.638 & 0.042 & 17.358 & 0.072 & 6.14 & 3.683 & 0.006 & 2.40 & 0.15 & 1.56 & 0.06 + 125 & -5.16 & 12.20 & 17.135 & 0.131 & 19.005 & 0.252 & 6.87 & 3.675 & 0.005 & 2.61 & 0.51 & 1.03 & 0.21 + 126 & -4.89 & 12.12 & 15.325 & 0.042 & 17.257 & 0.072 & 5.98 & 3.685 & 0.005 & 2.71 & 0.15 & 1.82 & 0.06 + 127 & -5.18 & 11.85 & 15.313 & 0.042 & 17.172 & 0.053 & 5.02 & 3.695 & 0.004 & 2.62 & 0.12 & 1.82 & 0.05 + 128 & -5.83 & 11.56 & 14.756 & 0.033 & 16.508 & 0.048 & 7.64 & 3.666 & 0.004 & 2.42 & 0.10 & 1.88 & 0.05 + 129 & -6.56 & 11.45 & 14.853 & 0.033 & 16.636 & 0.048 & 8.90 & 3.650 & 0.005 & 2.45 & 0.10 & 1.80 & 0.05 + 130 & -7.21 & 11.53 & 13.517 & 0.008 & 15.305 & 0.010 & 9.99 & 3.635 & 0.005 & 2.43 & 0.02 & 2.29 & 0.02 + 131 & -7.51 & 11.74 & 14.319 & 0.019 & 16.096 & 0.045 & 7.41 & 3.669 & 0.004 & 2.46 & 0.09 & 2.08 & 0.04 + 132 & -7.75 & 11.58 & 15.081 & 0.033 & 16.952 & 0.053 & 8.21 & 3.659 & 0.007 & 2.59 & 0.11 & 1.80 & 0.05 + 133 & -8.23 & 11.45 & 15.838 & 0.068 & 17.605 & 0.072 & 7.01 & 3.673 & 0.004 & 2.46 & 0.18 & 1.48 & 0.08 + 134 & -8.29 & 11.12 & 15.889 & 0.068 & 17.677 & 0.072 & 6.84 & 3.675 & 0.004 & 2.49 & 0.18 & 1.48 & 0.08 + 135 & -7.80 & 11.18 & 14.080 & 0.011 & 15.992 & 0.045 & 10.11 & 3.634 & 0.005 & 2.62 & 0.08 & 2.14 & 0.04 + 136 & -7.70 & 10.83 & 15.377 & 0.042 & 17.235 & 0.053 & 6.32 & 3.681 & 0.005 & 2.60 & 0.12 & 1.74 & 0.05 + 137 & -7.32 & 11.29 & 14.858 & 0.033 & 16.667 & 0.048 & 7.69 & 3.665 & 0.004 & 2.51 & 0.10 & 1.87 & 0.05 + 138 & -7.07 & 10.85 & 15.836 & 0.068 & 17.186 & 0.053 & 5.97 & 3.685 & 0.006 & 1.85 & 0.15 & 1.27 & 0.07 + 139 & -6.83 & 10.88 & 14.494 & 0.019 & 16.366 & 0.048 & 10.31 & 3.631 & 0.006 & 2.55 & 0.09 & 1.94 & 0.04 + 140 & -7.02 & 10.53 & 14.002 & 0.011 & 16.085 & 0.045 & 9.61 & 3.641 & 0.005 & 2.88 & 0.08 & 2.29 & 0.04 + 141 & -7.26 & 10.50 & 14.588 & 0.019 & 16.339 & 0.048 & 7.36 & 3.669 & 0.005 & 2.43 & 0.09 & 1.96 & 0.04 + 142 & -5.99 & 11.18 & 11.861 & 0.003 & 13.400 & 0.002 & 10.47 & 3.628 & 0.005 & 2.06 & 0.01 & 2.79 & 0.01 + 143 & -5.89 & 10.66 & 13.946 & 0.011 & 15.939 & 0.045 & 8.36 & 3.657 & 0.005 & 2.77 & 0.08 & 2.31 & 0.04 + 144 & -5.45 & 10.31 & 15.461 & 0.042 & 17.386 & 0.072 & 5.11 & 3.694 & 0.004 & 2.71 & 0.15 & 1.79 & 0.06 + 145 & -5.26 & 10.18 & 14.684 & 0.019 & 16.372 & 0.048 & 7.67 & 3.666 & 0.004 & 2.33 & 0.09 & 1.87 & 0.04 + 146 & -4.86 & 9.85 & 12.872 & 0.004 & 14.806 & 0.007 & 10.37 & 3.630 & 0.005 & 2.64 & 0.02 & 2.62 & 0.01 + 147 & 4.94 & 16.85 & 15.552 & 0.033 & 17.771 & 0.193 & 8.39 & 3.657 & 0.007 & 3.10 & 0.35 & 1.80 & 0.14 + 148 & 4.18 & 16.82 & 14.313 & 0.014 & 16.510 & 0.043 & 9.01 & 3.649 & 0.006 & 3.06 & 0.08 & 2.26 & 0.04 + 149 & 3.51 & 16.69 & 15.608 & 0.033 & 17.420 & 0.058 & 7.22 & 3.671 & 0.005 & 2.52 & 0.12 & 1.59 & 0.05 + 150 & 5.94 & 16.58 & 15.878 & 0.073 & 18.001 & 0.193 & 5.31 & 3.692 & 0.005 & 3.00 & 0.37 & 1.74 & 0.15 + 151 & 5.83 & 16.28 & 15.617 & 0.033 & 17.657 & 0.058 & 7.14 & 3.672 & 0.006 & 2.86 & 0.12 & 1.72 & 0.05 + 152 & 5.59 & 16.09 & 15.837 & 0.073 & 17.933 & 0.193 & 6.11 & 3.683 & 0.005 & 2.95 & 0.37 & 1.71 & 0.15 + 153 & 4.97 & 16.42 & 12.691 & 0.003 & 15.117 & 0.005 & 14.50 & 3.564 & 0.006 & 3.24 & 0.02 & 2.76 & 0.02 + 154 & 4.32 & 16.28 & 16.030 & 0.073 & 18.088 & 0.193 & 6.94 & 3.674 & 0.008 & 2.89 & 0.37 & 1.58 & 0.15 + 155 & 2.73 & 16.60 & 15.914 & 0.073 & 17.716 & 0.058 & 5.83 & 3.686 & 0.008 & 2.52 & 0.17 & 1.51 & 0.08 + 156 & 2.84 & 16.25 & 13.465 & 0.007 & 15.255 & 0.007 & 8.70 & 3.653 & 0.006 & 2.46 & 0.02 & 2.37 & 0.02 + 157 & 3.00 & 15.98 & 15.508 & 0.033 & 17.399 & 0.058 & 4.92 & 3.696 & 0.006 & 2.66 & 0.12 & 1.76 & 0.05 + 158 & 3.35 & 15.90 & 15.638 & 0.033 & 17.629 & 0.058 & 7.04 & 3.673 & 0.005 & 2.79 & 0.12 & 1.69 & 0.05 + 159 & 3.73 & 15.96 & 15.501 & 0.033 & 17.434 & 0.058 & 6.61 & 3.678 & 0.006 & 2.71 & 0.12 & 1.73 & 0.05 + 160 & 4.67 & 15.93 & 15.194 & 0.026 & 17.162 & 0.046 & 6.63 & 3.678 & 0.004 & 2.76 & 0.09 & 1.87 & 0.04 + 161 & 5.32 & 15.77 & 15.439 & 0.033 & 17.512 & 0.058 & 7.90 & 3.663 & 0.005 & 2.89 & 0.12 & 1.78 & 0.05 + 162 & 5.43 & 15.88 & 15.235 & 0.026 & 17.402 & 0.058 & 9.64 & 3.640 & 0.005 & 3.00 & 0.11 & 1.85 & 0.05 + 163 & 5.78 & 15.80 & 15.236 & 0.026 & 17.399 & 0.058 & 8.57 & 3.654 & 0.007 & 3.01 & 0.11 & 1.89 & 0.05 + 164 & 5.83 & 15.52 & 15.698 & 0.033 & 18.004 & 0.193 & 8.79 & 3.651 & 0.006 & 3.22 & 0.35 & 1.78 & 0.14 + 165 & 5.99 & 15.58 & 16.165 & 0.073 & 18.239 & 0.193 & 6.62 & 3.678 & 0.008 & 2.91 & 0.37 & 1.54 & 0.15 + 166 & 5.08 & 15.50 & 16.363 & 0.076 & 18.561 & 0.216 & 8.51 & 3.655 & 0.006 & 3.07 & 0.41 & 1.46 & 0.17 + 167 & 4.99 & 15.15 & 15.308 & 0.033 & 17.408 & 0.058 & 10.85 & 3.623 & 0.007 & 2.88 & 0.12 & 1.72 & 0.05 + 168 & 3.81 & 15.31 & 12.615 & 0.003 & 14.638 & 0.005 & 13.49 & 3.581 & 0.007 & 2.68 & 0.02 & 2.61 & 0.02 + 169 & 3.38 & 15.31 & 15.383 & 0.033 & 17.319 & 0.058 & 8.35 & 3.657 & 0.006 & 2.68 & 0.12 & 1.71 & 0.05 + 170 & 3.16 & 15.15 & 14.040 & 0.008 & 16.061 & 0.043 & 9.34 & 3.644 & 0.006 & 2.79 & 0.08 & 2.25 & 0.04 + 171 & 2.92 & 15.66 & 15.052 & 0.026 & 16.976 & 0.046 & 8.92 & 3.650 & 0.008 & 2.66 & 0.09 & 1.81 & 0.05 + 172 & 2.78 & 15.34 & 14.778 & 0.026 & 16.934 & 0.046 & 9.07 & 3.648 & 0.005 & 3.00 & 0.09 & 2.05 & 0.04 + 173 & 2.54 & 14.80 & 13.166 & 0.004 & 14.934 & 0.005 & 5.00 & 3.695 & 0.005 & 2.48 & 0.01 & 2.62 & 0.01 + 174 & 3.21 & 14.42 & 15.317 & 0.033 & 17.176 & 0.046 & 6.99 & 3.674 & 0.006 & 2.59 & 0.10 & 1.74 & 0.05 + 175 & 3.24 & 14.66 & 15.590 & 0.033 & 17.545 & 0.058 & 7.46 & 3.668 & 0.006 & 2.73 & 0.12 & 1.67 & 0.05 + 176 & 3.62 & 14.80 & 15.449 & 0.033 & 17.362 & 0.058 & 9.52 & 3.642 & 0.005 & 2.63 & 0.12 & 1.61 & 0.05 + 177 & 4.10 & 14.88 & 15.905 & 0.073 & 18.090 & 0.193 & 7.66 & 3.666 & 0.005 & 3.06 & 0.37 & 1.67 & 0.15 + 178 & 6.10 & 15.07 & 15.197 & 0.026 & 17.898 & 0.193 & 10.05 & 3.634 & 0.006 & 3.78 & 0.35 & 2.16 & 0.14 + 179 & 5.75 & 14.50 & 10.803 & 0.002 & 13.327 & 0.002 & 16.04 & 3.536 & 0.007 & 3.31 & 0.02 & 3.49 & 0.02 + 180 & 4.56 & 14.23 & 12.404 & 0.003 & 15.485 & 0.007 & 14.27 & 3.568 & 0.007 & 4.22 & 0.02 & 3.28 & 0.02 + 181 & 4.00 & 14.28 & 15.637 & 0.033 & 17.484 & 0.058 & 4.80 & 3.697 & 0.003 & 2.60 & 0.12 & 1.69 & 0.05 + 182 & 3.89 & 13.77 & 12.332 & 0.003 & 14.286 & 0.005 & 12.22 & 3.602 & 0.007 & 2.62 & 0.02 & 2.75 & 0.02 + 183 & 3.67 & 14.09 & 15.154 & 0.026 & 17.022 & 0.046 & 9.60 & 3.641 & 0.005 & 2.56 & 0.09 & 1.70 & 0.04 + 184 & 3.21 & 13.47 & 13.180 & 0.004 & 15.232 & 0.005 & 11.63 & 3.611 & 0.007 & 2.78 & 0.02 & 2.50 & 0.02 + 185 & 3.00 & 14.09 & 15.071 & 0.026 & 17.010 & 0.046 & 7.33 & 3.670 & 0.005 & 2.70 & 0.09 & 1.88 & 0.04 + 186 & 2.46 & 13.45 & 15.667 & 0.033 & 17.564 & 0.058 & 7.33 & 3.670 & 0.005 & 2.64 & 0.12 & 1.61 & 0.05 + 187 & 4.75 & 13.82 & 15.838 & 0.073 & 17.896 & 0.193 & 7.32 & 3.670 & 0.004 & 2.88 & 0.37 & 1.64 & 0.15 + 188 & 5.35 & 13.53 & 15.493 & 0.033 & 18.000 & 0.193 & 9.48 & 3.642 & 0.009 & 3.51 & 0.35 & 1.95 & 0.14 + 189 & 5.21 & 13.01 & 13.400 & 0.007 & 15.575 & 0.007 & 11.59 & 3.612 & 0.006 & 2.97 & 0.02 & 2.49 & 0.02 + 190 & 4.86 & 13.04 & 15.419 & 0.033 & 17.675 & 0.058 & 4.92 & 3.696 & 0.005 & 3.20 & 0.12 & 2.01 & 0.05 + 191 & 4.62 & 13.28 & 15.571 & 0.033 & 17.613 & 0.058 & 7.31 & 3.670 & 0.005 & 2.86 & 0.12 & 1.74 & 0.05 + 192 & 4.29 & 13.42 & 15.229 & 0.026 & 17.173 & 0.046 & 7.96 & 3.662 & 0.007 & 2.70 & 0.09 & 1.79 & 0.04 + 193 & 4.13 & 13.18 & 15.177 & 0.026 & 17.135 & 0.046 & 5.54 & 3.690 & 0.006 & 2.76 & 0.09 & 1.91 & 0.04 + 194 & 3.81 & 12.96 & 14.855 & 0.026 & 17.095 & 0.046 & 7.21 & 3.671 & 0.005 & 3.15 & 0.09 & 2.14 & 0.04 + 195 & 3.75 & 12.82 & 14.958 & 0.026 & 16.816 & 0.046 & 6.62 & 3.678 & 0.006 & 2.59 & 0.09 & 1.90 & 0.04 + 196 & 3.46 & 12.18 & 14.986 & 0.026 & 16.925 & 0.046 & 7.58 & 3.667 & 0.007 & 2.70 & 0.09 & 1.90 & 0.04 + 197 & 4.62 & 12.31 & 14.028 & 0.008 & 16.714 & 0.043 & 11.82 & 3.608 & 0.007 & 3.72 & 0.08 & 2.53 & 0.04 + 198 & 4.67 & 11.96 & 15.444 & 0.033 & 17.517 & 0.058 & 8.54 & 3.655 & 0.011 & 2.88 & 0.12 & 1.76 & 0.06 + 199 & 5.48 & 12.34 & 14.066 & 0.008 & 16.048 & 0.043 & 9.55 & 3.641 & 0.009 & 2.73 & 0.08 & 2.21 & 0.04 + 200 & 5.75 & 11.96 & 15.084 & 0.026 & 18.396 & 0.216 & 10.30 & 3.631 & 0.010 & 4.68 & 0.39 & 2.55 & 0.16 + 201 & 5.83 & 12.50 & 15.007 & 0.026 & 16.824 & 0.046 & 7.83 & 3.664 & 0.007 & 2.52 & 0.09 & 1.81 & 0.04 + 202 & 5.56 & 12.80 & 15.206 & 0.026 & 17.276 & 0.058 & 7.35 & 3.669 & 0.009 & 2.90 & 0.11 & 1.90 & 0.05 + 203 & 5.32 & 12.64 & 15.004 & 0.026 & 16.793 & 0.046 & 7.93 & 3.662 & 0.006 & 2.47 & 0.09 & 1.79 & 0.04 + 204 & 17.98 & 18.79 & 13.875 & 0.008 & 15.987 & 0.041 & 12.13 & 3.603 & 0.006 & 2.86 & 0.08 & 2.23 & 0.03 + 205 & 17.42 & 18.66 & 12.761 & 0.005 & 14.750 & 0.005 & 7.67 & 3.665 & 0.004 & 2.77 & 0.01 & 2.82 & 0.01 + 206 & 16.96 & 18.36 & 15.226 & 0.021 & 17.257 & 0.058 & 8.52 & 3.655 & 0.005 & 2.82 & 0.11 & 1.82 & 0.05 + 207 & 16.63 & 18.68 & 14.970 & 0.021 & 16.876 & 0.048 & 8.24 & 3.658 & 0.005 & 2.64 & 0.09 & 1.86 & 0.04 + 208 & 16.42 & 18.63 & 15.582 & 0.027 & 17.474 & 0.058 & 6.17 & 3.683 & 0.004 & 2.65 & 0.11 & 1.69 & 0.05 + 209 & 16.42 & 18.39 & 15.578 & 0.027 & 17.470 & 0.058 & 8.39 & 3.657 & 0.007 & 2.62 & 0.11 & 1.60 & 0.05 + 210 & 15.55 & 18.36 & 15.736 & 0.027 & 17.868 & 0.202 & 6.30 & 3.681 & 0.006 & 3.00 & 0.36 & 1.76 & 0.15 + 211 & 15.47 & 18.68 & 16.367 & 0.083 & 18.105 & 0.202 & 5.70 & 3.688 & 0.004 & 2.43 & 0.39 & 1.30 & 0.16 + 212 & 15.39 & 19.06 & 16.023 & 0.061 & 17.959 & 0.202 & 3.62 & 3.708 & 0.004 & 2.74 & 0.38 & 1.62 & 0.15 + 213 & 15.15 & 18.58 & 14.538 & 0.015 & 16.357 & 0.047 & 7.22 & 3.671 & 0.004 & 2.53 & 0.09 & 2.02 & 0.04 + 214 & 14.61 & 18.50 & 15.400 & 0.027 & 17.426 & 0.058 & 7.51 & 3.667 & 0.008 & 2.83 & 0.11 & 1.79 & 0.05 + 215 & 14.61 & 18.04 & 12.841 & 0.005 & 14.809 & 0.006 & 5.63 & 3.689 & 0.003 & 2.77 & 0.01 & 2.85 & 0.01 + 216 & 15.07 & 18.09 & 15.299 & 0.027 & 17.248 & 0.048 & 6.82 & 3.675 & 0.006 & 2.73 & 0.10 & 1.81 & 0.04 + 217 & 15.61 & 17.90 & 15.055 & 0.021 & 16.851 & 0.048 & 4.29 & 3.702 & 0.003 & 2.53 & 0.09 & 1.90 & 0.04 + 218 & 15.82 & 17.93 & 15.329 & 0.027 & 17.293 & 0.058 & 5.46 & 3.690 & 0.004 & 2.76 & 0.11 & 1.86 & 0.05 + 219 & 16.09 & 18.01 & 15.249 & 0.021 & 17.087 & 0.048 & 7.03 & 3.673 & 0.004 & 2.56 & 0.09 & 1.76 & 0.04 + 220 & 16.36 & 17.82 & 16.300 & 0.083 & 18.281 & 0.204 & 8.44 & 3.656 & 0.006 & 2.75 & 0.39 & 1.36 & 0.16 + 221 & 16.96 & 17.98 & 13.942 & 0.008 & 15.841 & 0.041 & 9.37 & 3.644 & 0.004 & 2.61 & 0.07 & 2.22 & 0.03 + 222 & 17.60 & 18.09 & 15.513 & 0.027 & 17.565 & 0.058 & 9.45 & 3.643 & 0.006 & 2.83 & 0.11 & 1.67 & 0.05 + 223 & 17.12 & 17.74 & 14.996 & 0.021 & 16.761 & 0.048 & 8.00 & 3.661 & 0.004 & 2.44 & 0.09 & 1.78 & 0.04 + 224 & 16.98 & 17.58 & 13.668 & 0.005 & 15.518 & 0.007 & 9.04 & 3.648 & 0.004 & 2.54 & 0.02 & 2.31 & 0.01 + 225 & 16.52 & 17.55 & 15.812 & 0.061 & 18.124 & 0.202 & 5.87 & 3.686 & 0.004 & 3.28 & 0.38 & 1.85 & 0.15 + 226 & 16.31 & 17.28 & 14.495 & 0.015 & 16.480 & 0.047 & 9.03 & 3.648 & 0.004 & 2.74 & 0.09 & 2.06 & 0.04 + 227 & 15.74 & 17.39 & 15.275 & 0.027 & 17.219 & 0.048 & 6.77 & 3.676 & 0.006 & 2.72 & 0.10 & 1.82 & 0.04 + 228 & 15.34 & 17.39 & 15.470 & 0.027 & 17.529 & 0.058 & 6.90 & 3.675 & 0.005 & 2.89 & 0.11 & 1.80 & 0.05 + 229 & 15.42 & 16.98 & 16.262 & 0.083 & 18.066 & 0.202 & 6.45 & 3.680 & 0.007 & 2.52 & 0.39 & 1.35 & 0.16 + 230 & 15.34 & 16.71 & 15.123 & 0.021 & 17.461 & 0.058 & 6.69 & 3.677 & 0.005 & 3.30 & 0.11 & 2.12 & 0.05 + 231 & 15.93 & 16.71 & 13.527 & 0.005 & 15.422 & 0.007 & 10.90 & 3.622 & 0.006 & 2.57 & 0.02 & 2.31 & 0.02 + 232 & 16.44 & 17.15 & 13.639 & 0.005 & 15.636 & 0.007 & 10.57 & 3.627 & 0.005 & 2.73 & 0.02 & 2.34 & 0.02 + 233 & 16.98 & 17.17 & 14.927 & 0.021 & 16.440 & 0.047 & 7.26 & 3.670 & 0.004 & 2.07 & 0.09 & 1.68 & 0.04 + 234 & 17.25 & 17.04 & 15.273 & 0.027 & 17.212 & 0.048 & 8.20 & 3.659 & 0.006 & 2.69 & 0.10 & 1.76 & 0.04 + 235 & 17.39 & 17.20 & 14.481 & 0.015 & 16.331 & 0.047 & 8.79 & 3.651 & 0.005 & 2.55 & 0.09 & 2.00 & 0.04 + 236 & 17.77 & 16.69 & 14.440 & 0.015 & 16.309 & 0.047 & 7.90 & 3.663 & 0.007 & 2.59 & 0.09 & 2.06 & 0.04 + 237 & 17.31 & 16.69 & 15.226 & 0.021 & 16.971 & 0.048 & 4.79 & 3.697 & 0.004 & 2.45 & 0.09 & 1.79 & 0.04 + 238 & 16.98 & 16.77 & 16.383 & 0.083 & 18.381 & 0.204 & 6.51 & 3.679 & 0.008 & 2.80 & 0.39 & 1.42 & 0.16 + 239 & 17.52 & 16.07 & 15.026 & 0.021 & 16.916 & 0.048 & 10.57 & 3.627 & 0.009 & 2.57 & 0.09 & 1.72 & 0.05 + 240 & 17.31 & 15.71 & 13.073 & 0.005 & 15.095 & 0.006 & 10.61 & 3.626 & 0.005 & 2.77 & 0.02 & 2.58 & 0.02 + 241 & 16.39 & 15.63 & 15.333 & 0.027 & 17.192 & 0.048 & 6.66 & 3.677 & 0.005 & 2.59 & 0.10 & 1.75 & 0.04 + 242 & 16.12 & 15.42 & 14.285 & 0.015 & 16.163 & 0.041 & 8.51 & 3.655 & 0.008 & 2.59 & 0.08 & 2.10 & 0.04 + 243 & 16.04 & 15.77 & 15.011 & 0.021 & 16.470 & 0.047 & 4.92 & 3.696 & 0.004 & 2.02 & 0.09 & 1.70 & 0.04 + 244 & 15.28 & 15.47 & 15.066 & 0.021 & 16.950 & 0.048 & 6.97 & 3.674 & 0.005 & 2.63 & 0.09 & 1.86 & 0.04 + 245 & 15.31 & 16.09 & 15.352 & 0.027 & 17.243 & 0.048 & 6.55 & 3.679 & 0.007 & 2.64 & 0.10 & 1.76 & 0.05 + 246 & 15.50 & 16.42 & 15.382 & 0.027 & 17.339 & 0.058 & 7.03 & 3.673 & 0.004 & 2.73 & 0.11 & 1.77 & 0.05 + 247 & 15.12 & 16.36 & 14.726 & 0.015 & 16.606 & 0.047 & 10.30 & 3.631 & 0.006 & 2.56 & 0.09 & 1.85 & 0.04 + 248 & 18.71 & 13.90 & 14.621 & 0.016 & 17.520 & 0.065 & 11.38 & 3.615 & 0.005 & 4.04 & 0.12 & 2.44 & 0.05 + 249 & 18.58 & 13.80 & 13.958 & 0.009 & 15.403 & 0.009 & 6.33 & 3.681 & 0.004 & 1.99 & 0.02 & 2.07 & 0.02 + 250 & 17.87 & 13.93 & 14.910 & 0.027 & 16.616 & 0.044 & 4.50 & 3.700 & 0.004 & 2.39 & 0.09 & 1.90 & 0.04 + 251 & 16.98 & 13.61 & 14.856 & 0.027 & 17.116 & 0.055 & 8.54 & 3.655 & 0.004 & 3.16 & 0.11 & 2.10 & 0.05 + 252 & 16.17 & 13.66 & 16.296 & 0.100 & 18.025 & 0.198 & 5.75 & 3.687 & 0.006 & 2.41 & 0.39 & 1.32 & 0.16 + 253 & 15.93 & 13.45 & 15.774 & 0.064 & 17.581 & 0.065 & 5.53 & 3.690 & 0.007 & 2.53 & 0.16 & 1.58 & 0.07 + 254 & 16.33 & 13.23 & 14.934 & 0.027 & 16.765 & 0.055 & 6.30 & 3.681 & 0.004 & 2.56 & 0.11 & 1.90 & 0.05 + 255 & 16.60 & 13.45 & 16.083 & 0.064 & 18.269 & 0.233 & 4.92 & 3.696 & 0.003 & 3.10 & 0.43 & 1.70 & 0.17 + 256 & 16.90 & 13.45 & 15.023 & 0.027 & 16.766 & 0.055 & 6.39 & 3.680 & 0.004 & 2.43 & 0.11 & 1.81 & 0.05 + 257 & 17.09 & 13.18 & 15.868 & 0.064 & 17.650 & 0.065 & 5.28 & 3.692 & 0.006 & 2.50 & 0.16 & 1.54 & 0.07 + 258 & 17.28 & 13.18 & 14.923 & 0.027 & 16.713 & 0.044 & 6.56 & 3.678 & 0.004 & 2.49 & 0.09 & 1.88 & 0.04 + 259 & 17.55 & 13.31 & 15.131 & 0.027 & 16.904 & 0.055 & 7.08 & 3.672 & 0.005 & 2.46 & 0.11 & 1.76 & 0.05 + 260 & 18.14 & 13.58 & 15.066 & 0.027 & 16.834 & 0.055 & 7.71 & 3.665 & 0.004 & 2.45 & 0.11 & 1.76 & 0.05 + 261 & 18.31 & 13.50 & 15.182 & 0.027 & 16.984 & 0.055 & 7.14 & 3.672 & 0.004 & 2.50 & 0.11 & 1.76 & 0.05 + 262 & 18.90 & 13.20 & 11.054 & 0.002 & 12.618 & 0.002 & 11.69 & 3.610 & 0.005 & 2.06 & 0.01 & 3.06 & 0.01 + 263 & 18.50 & 12.58 & 14.975 & 0.027 & 16.649 & 0.044 & 7.40 & 3.669 & 0.005 & 2.31 & 0.09 & 1.75 & 0.04 + 264 & 18.31 & 12.28 & 12.522 & 0.003 & 14.414 & 0.005 & 10.74 & 3.624 & 0.004 & 2.57 & 0.01 & 2.72 & 0.01 + 265 & 18.71 & 12.26 & 15.774 & 0.064 & 17.527 & 0.065 & 5.80 & 3.687 & 0.005 & 2.45 & 0.16 & 1.54 & 0.07 + 266 & 19.17 & 12.18 & 12.940 & 0.005 & 14.683 & 0.005 & 11.04 & 3.620 & 0.007 & 2.34 & 0.02 & 2.45 & 0.02 + 267 & 17.87 & 12.64 & 15.189 & 0.027 & 17.110 & 0.055 & 7.94 & 3.662 & 0.005 & 2.67 & 0.11 & 1.79 & 0.05 + 268 & 17.74 & 12.26 & 13.417 & 0.008 & 15.210 & 0.006 & 9.23 & 3.646 & 0.004 & 2.46 & 0.02 & 2.37 & 0.01 + 269 & 17.15 & 12.18 & 16.087 & 0.064 & 18.519 & 0.233 & 6.83 & 3.675 & 0.006 & 3.44 & 0.43 & 1.78 & 0.17 + 270 & 16.98 & 12.31 & 15.043 & 0.027 & 16.845 & 0.055 & 8.59 & 3.654 & 0.003 & 2.48 & 0.11 & 1.75 & 0.05 + 271 & 16.69 & 12.18 & 12.248 & 0.003 & 14.120 & 0.003 & 12.93 & 3.591 & 0.006 & 2.48 & 0.02 & 2.70 & 0.02 + 272 & 16.77 & 12.99 & 14.951 & 0.027 & 16.704 & 0.044 & 6.96 & 3.674 & 0.003 & 2.43 & 0.09 & 1.83 & 0.04 + 273 & 16.39 & 12.61 & 12.552 & 0.003 & 14.315 & 0.005 & 4.08 & 3.704 & 0.002 & 2.48 & 0.01 & 2.89 & 0.01 + 274 & 15.55 & 13.10 & 15.324 & 0.039 & 17.308 & 0.065 & 4.81 & 3.697 & 0.003 & 2.80 & 0.13 & 1.89 & 0.06 + 275 & 15.42 & 12.50 & 14.718 & 0.016 & 16.613 & 0.044 & 6.34 & 3.681 & 0.003 & 2.65 & 0.08 & 2.03 & 0.03 + 276 & 15.96 & 12.23 & 13.624 & 0.008 & 15.550 & 0.009 & 8.26 & 3.658 & 0.004 & 2.67 & 0.02 & 2.41 & 0.01 + 277 & 15.47 & 11.77 & 15.327 & 0.039 & 17.333 & 0.065 & 4.94 & 3.696 & 0.005 & 2.83 & 0.13 & 1.90 & 0.06 + 278 & 15.50 & 11.56 & 15.479 & 0.039 & 17.545 & 0.065 & 6.01 & 3.684 & 0.008 & 2.91 & 0.14 & 1.84 & 0.06 + 279 & 15.85 & 11.34 & 13.934 & 0.009 & 15.885 & 0.045 & 7.63 & 3.666 & 0.006 & 2.72 & 0.08 & 2.32 & 0.04 + 280 & 16.01 & 11.15 & 15.087 & 0.027 & 17.174 & 0.055 & 9.49 & 3.642 & 0.008 & 2.89 & 0.11 & 1.86 & 0.05 + 281 & 15.82 & 11.10 & 15.398 & 0.039 & 17.480 & 0.065 & 6.14 & 3.683 & 0.005 & 2.93 & 0.13 & 1.87 & 0.06 + 282 & 16.50 & 11.58 & 15.047 & 0.027 & 16.877 & 0.055 & 8.00 & 3.661 & 0.010 & 2.53 & 0.11 & 1.79 & 0.05 + 283 & 16.74 & 11.58 & 14.931 & 0.027 & 16.922 & 0.055 & 8.07 & 3.661 & 0.009 & 2.77 & 0.11 & 1.93 & 0.05 + 284 & 16.71 & 11.31 & 15.040 & 0.027 & 17.024 & 0.055 & 9.80 & 3.638 & 0.005 & 2.73 & 0.11 & 1.81 & 0.05 + 285 & 16.71 & 10.99 & 15.251 & 0.039 & 17.147 & 0.055 & 10.42 & 3.629 & 0.009 & 2.58 & 0.12 & 1.64 & 0.06 + 286 & 16.96 & 10.66 & 16.019 & 0.064 & 17.658 & 0.065 & 5.89 & 3.686 & 0.006 & 2.28 & 0.16 & 1.37 & 0.07 + 287 & 17.17 & 11.20 & 15.337 & 0.039 & 17.459 & 0.065 & 8.96 & 3.649 & 0.007 & 2.95 & 0.14 & 1.81 & 0.06 + 288 & 17.20 & 11.58 & 15.334 & 0.039 & 17.229 & 0.055 & 5.43 & 3.691 & 0.005 & 2.66 & 0.12 & 1.81 & 0.05 + 289 & 17.58 & 11.29 & 15.161 & 0.027 & 17.013 & 0.055 & 4.87 & 3.696 & 0.002 & 2.61 & 0.11 & 1.88 & 0.05 + 290 & 18.28 & 11.50 & 15.082 & 0.027 & 16.835 & 0.055 & 6.09 & 3.684 & 0.005 & 2.44 & 0.11 & 1.81 & 0.05 + 291 & 18.33 & 11.69 & 15.143 & 0.027 & 16.849 & 0.055 & 5.51 & 3.690 & 0.004 & 2.38 & 0.11 & 1.78 & 0.05 + 292 & 18.06 & 10.77 & 14.916 & 0.027 & 16.814 & 0.055 & 10.69 & 3.625 & 0.008 & 2.58 & 0.11 & 1.76 & 0.05 + 293 & 17.77 & 10.75 & 14.598 & 0.016 & 16.558 & 0.044 & 11.05 & 3.620 & 0.005 & 2.66 & 0.08 & 1.91 & 0.04 + 294 & 17.63 & 10.34 & 15.301 & 0.039 & 17.319 & 0.065 & 8.23 & 3.659 & 0.007 & 2.81 & 0.14 & 1.79 & 0.06 + 295 & -6.67 & 18.85 & 14.814 & 0.014 & 16.883 & 0.042 & 10.09 & 3.634 & 0.007 & 2.85 & 0.08 & 1.93 & 0.04 + 296 & -7.51 & 19.31 & 15.600 & 0.029 & 17.609 & 0.051 & 4.45 & 3.700 & 0.004 & 2.84 & 0.11 & 1.81 & 0.05 + 297 & -7.70 & 19.12 & 15.219 & 0.014 & 17.213 & 0.042 & 5.71 & 3.688 & 0.004 & 2.81 & 0.08 & 1.91 & 0.03 + 298 & -8.29 & 19.01 & 14.783 & 0.014 & 16.753 & 0.042 & 8.22 & 3.659 & 0.005 & 2.74 & 0.08 & 1.97 & 0.03 + 299 & -8.37 & 18.82 & 15.067 & 0.014 & 16.961 & 0.042 & 5.94 & 3.685 & 0.003 & 2.66 & 0.08 & 1.90 & 0.03 + 300 & -9.15 & 18.74 & 11.437 & 0.002 & 13.493 & 0.002 & 8.46 & 3.656 & 0.003 & 2.86 & 0.01 & 3.35 & 0.01 + 301 & -8.26 & 18.23 & 14.468 & 0.011 & 16.434 & 0.044 & 9.28 & 3.645 & 0.004 & 2.71 & 0.08 & 2.05 & 0.03 + 302 & -6.72 & 18.14 & 13.651 & 0.005 & 15.524 & 0.007 & 9.17 & 3.646 & 0.004 & 2.58 & 0.02 & 2.33 & 0.01 + 303 & -7.07 & 17.77 & 15.142 & 0.014 & 17.064 & 0.042 & 7.86 & 3.663 & 0.007 & 2.67 & 0.08 & 1.82 & 0.04 + 304 & -8.13 & 17.63 & 15.598 & 0.029 & 17.594 & 0.051 & 5.03 & 3.695 & 0.004 & 2.82 & 0.11 & 1.78 & 0.05 + 305 & -8.48 & 17.74 & 15.391 & 0.029 & 17.322 & 0.051 & 7.87 & 3.663 & 0.005 & 2.68 & 0.11 & 1.72 & 0.05 + 306 & -8.72 & 17.63 & 15.467 & 0.029 & 17.360 & 0.051 & 7.25 & 3.670 & 0.004 & 2.64 & 0.11 & 1.69 & 0.05 + 307 & -9.94 & 17.90 & 15.508 & 0.029 & 17.631 & 0.051 & 10.16 & 3.633 & 0.006 & 2.93 & 0.11 & 1.69 & 0.05 + 308 & -10.07 & 17.28 & 15.566 & 0.029 & 17.617 & 0.051 & 7.43 & 3.668 & 0.006 & 2.87 & 0.11 & 1.74 & 0.05 + 309 & -10.12 & 17.50 & 15.794 & 0.060 & 17.879 & 0.196 & 5.97 & 3.685 & 0.007 & 2.94 & 0.36 & 1.72 & 0.15 + 310 & -8.88 & 17.28 & 15.160 & 0.014 & 17.095 & 0.042 & 8.64 & 3.653 & 0.005 & 2.68 & 0.08 & 1.78 & 0.03 + 311 & -8.80 & 17.15 & 15.482 & 0.029 & 17.506 & 0.051 & 9.00 & 3.649 & 0.007 & 2.80 & 0.11 & 1.69 & 0.05 + 312 & -8.56 & 17.31 & 15.606 & 0.029 & 17.457 & 0.051 & 8.29 & 3.658 & 0.004 & 2.56 & 0.11 & 1.57 & 0.05 + 313 & -8.05 & 16.98 & 15.477 & 0.029 & 17.505 & 0.051 & 7.95 & 3.662 & 0.006 & 2.83 & 0.11 & 1.74 & 0.05 + 314 & -6.94 & 17.12 & 15.234 & 0.014 & 17.284 & 0.051 & 9.04 & 3.648 & 0.007 & 2.84 & 0.10 & 1.80 & 0.04 + 315 & -6.59 & 16.98 & 15.524 & 0.029 & 17.516 & 0.051 & 5.68 & 3.688 & 0.005 & 2.80 & 0.11 & 1.79 & 0.05 + 316 & -6.86 & 16.71 & 15.647 & 0.029 & 17.546 & 0.051 & 6.88 & 3.675 & 0.004 & 2.65 & 0.11 & 1.64 & 0.05 + 317 & -6.94 & 16.58 & 15.254 & 0.029 & 17.181 & 0.042 & 6.51 & 3.679 & 0.003 & 2.70 & 0.09 & 1.83 & 0.04 + 318 & -7.64 & 16.39 & 15.395 & 0.029 & 17.375 & 0.051 & 8.66 & 3.653 & 0.007 & 2.74 & 0.11 & 1.72 & 0.05 + 319 & -6.61 & 15.77 & 14.757 & 0.014 & 16.307 & 0.044 & 10.12 & 3.633 & 0.006 & 2.08 & 0.08 & 1.65 & 0.04 + 320 & -7.48 & 15.36 & 15.453 & 0.029 & 17.392 & 0.051 & 8.26 & 3.658 & 0.005 & 2.69 & 0.11 & 1.68 & 0.05 + 321 & -7.75 & 15.55 & 13.625 & 0.005 & 15.630 & 0.007 & 10.90 & 3.622 & 0.007 & 2.73 & 0.02 & 2.33 & 0.02 + 322 & -8.34 & 16.52 & 13.477 & 0.005 & 15.438 & 0.007 & 8.02 & 3.661 & 0.004 & 2.73 & 0.02 & 2.50 & 0.01 + 323 & -8.96 & 16.96 & 14.418 & 0.011 & 16.562 & 0.044 & 9.48 & 3.642 & 0.005 & 2.97 & 0.08 & 2.17 & 0.04 + 324 & -9.10 & 16.79 & 13.835 & 0.008 & 15.898 & 0.039 & 10.18 & 3.633 & 0.005 & 2.84 & 0.07 & 2.32 & 0.03 + 325 & -8.53 & 15.66 & 12.615 & 0.003 & 14.755 & 0.006 & 9.82 & 3.638 & 0.005 & 2.96 & 0.01 & 2.87 & 0.02 + 326 & -8.59 & 15.34 & 13.698 & 0.005 & 15.683 & 0.007 & 10.55 & 3.627 & 0.005 & 2.71 & 0.02 & 2.31 & 0.01 + 327 & -9.05 & 15.28 & 12.764 & 0.004 & 14.750 & 0.005 & 9.86 & 3.637 & 0.006 & 2.73 & 0.02 & 2.72 & 0.02 + 328 & -9.18 & 15.85 & 15.145 & 0.014 & 17.151 & 0.042 & 6.83 & 3.675 & 0.004 & 2.81 & 0.08 & 1.90 & 0.03 + 329 & -9.99 & 15.98 & 14.581 & 0.011 & 16.522 & 0.044 & 5.01 & 3.695 & 0.004 & 2.73 & 0.08 & 2.16 & 0.03 + [ gc_table ] the gc h - r diagram is shown in figure [ hr ] . our data clearly show the red clump at @xmath67 , as well as the upper red giant branch / early asymptotic giant branch . there is an indication of the agb bump at @xmath68 , and we also detect some lower red giant branch stars at lower luminosities , though the observations are highly incomplete in this region . the data set in figure [ hr ] allows for a more robust analysis of the galactic center star formation history than any other previously published data set . photometric studies are limited to the modeling of luminosities alone @xcite , due to intrinsic variations in late - type giant colors and the large variation in galactic center extinction . this point is illustrated in figure [ cmd ] , which compares the observed h - r diagram to a color - magnitude diagram ( cmd ) for the same gc stars . the rgb / agb and rc populations are more clearly distinguished in the h - r diagram than in the cmd , and there is less scatter in @xmath7 than @xmath69 , due to variations in gc extinction . as a result , only the k - band luminosity function can reliably be modeled with broadband photometry alone . the observed h - r diagram in this study is also an improvement with respect to previous spectroscopic work @xcite , due to our improved magnitude limit ( @xmath55 mag . deeper ) , which allows for the analysis of well - populated regions in the h - r diagram , in which the evolutionary models are fairly well understood . than @xmath70 , due to variations in gc extinction . ] the errors in temperature and luminosity are shown with the data in grey in figure [ hr ] . average errors in temperature and luminosity for stars with luminosities @xmath71 ( the red clump region ) are shown at right . the temperature limit for the minimum co index and the 50% completeness limit for the average extinction correction ( @xmath72 mag . ) are also shown . solar metallicity ( in red , left ) and metal - poor ( z=0.008 , in blue , right ) isochrones from @xcite are overplotted . the isochrones show the wide range of ages spanned by the galactic center population . figure [ hr ] also demonstrates the age - metallicity degeneracy inherent in this part of the diagram . stars in the same part of the h - r diagram may represent a metal - rich , intermediate - age ( @xmath73 a few gyr ) population or a metal - poor , old ( @xmath74 5 gyr ) population . for stars younger than @xmath55 gyr , the galactic center stellar population is known to be approximately solar @xcite . for older stars , the galactic center metallicity distribution is not well - known . we will address this uncertainty later in the paper . to investigate the star formation history implied by our sample , we begin with a qualitative discussion , followed by quantitative analysis in 3.4 - 3.5 . we start by considering three models based on candidate star formation histories presented in the literature . the first scenario consists of an ancient burst of star formation 7.5 - 8.5 gyr ago , similar to the single stellar population of the bulge . @xcite compared this scenario to the k band luminosity function ( klf ) of the inner parsec . the second model consists of constant star formation between 10 myr and 10 gyr ago . this model is based on the best - fit model of @xcite , who considered the galactic center klf within 40 pc of the center . the third model corresponds to the best - fit star formation history of @xcite , who fit the h - r diagram of asymptotic giant branch and cool supergiant stars within the central 5 pc , using four specified age bins . all models are summarized in table [ sf_models ] . llccccccccc 1 & bulge - like & 7.50 - 8.50 & 0.015 & 2.35 & 0.7 & 120 & 3228.8 & 0.00 & 1.00 & - + 2 & continuous & 0.01 - 10.0 & 0.019 & 2.35 & 0.7 & 120 & 404.9 & 3.76 & 1.00 & - + 3 & blum et al . & 0.01 - 0.10 & 0.019 & 2.35 & 0.7 & 120 & 678.8 & 0.01 & 0.65 & - + & & 0.10 - 1.00 & 0.019 & 2.35 & 0.7 & 120 & & & 0.06 & - + & & 1.00 - 5.00 & 0.019 & 2.35 & 0.7 & 120 & & & 0.09 & - + & & 5.00 - 12.0 & 0.019 & 2.35 & 0.7 & 120 & & & 0.21 & - + 4 & two - bin , solar & 0.01 - 5.00 & 0.019 & 2.35 & 0.7 & 120 & 384.5 & 3.03 & 0.82 & 0.04 + & & 5.00 - 12.0 & 0.019 & 2.35 & 0.7 & 120 & & & 0.18 & 0.07 + 5 & two - bin , solar & 0.01 - 7.00 & 0.019 & 2.35 & 0.7 & 120 & 270.0 & 16.66 & 0.90 & 0.04 + & & 7.00 - 12.0 & 0.019 & 2.35 & 0.7 & 120 & & & 0.10 & 0.09 + 6 & two - bin , poor & 0.01 - 5.00 & 0.019 & 2.35 & 0.7 & 120 & 450.0 & 0.68 & 0.91 & 0.05 + & & 5.00 - 12.0 & 0.008 & 2.35 & 0.7 & 120 & & & 0.09 & 0.08 + 7 & closed - box & 0.01 - 12.0 & 0.004 - 0.019 & 2.35 & 0.7 & 120 & 1035.6 & 0.00 & 1.00 & - + 8 & flat imf & 0.01 - 12.0 & 0.019 & 0.85 & 0.7 & 120 & 242.8 & 39.70 & 1.00 & - + 9 & high @xmath75 & 0.01 - 12.0 & 0.019 & 2.35 & 2.5 & 120 & 4201.8 & 0.00 & 1.00 & - + 10 & high @xmath75 & 0.01 - 12.0 & 0.019 & 2.35 & 1.5 & 120 & 551.4 & 0.00 & 1.00 & - + 11 & model 5 + 8 hybrid & 0.01 - 7.00 & 0.019 & 0.85 & 0.7 & 120 & 257.0 & 26.3 & 0.48 & 0.02 + & & 7.00 - 12.0 & 0.019 & 0.85 & 0.7 & 120 & & & 0.51 & 0.20 + [ sf_models ] for each scenario , we generated model h - r diagrams using the synthetic color - magnitude diagram computation algorithm iac - star @xcite . the algorithm uses a monte carlo approach to compute composite stellar populations on a star - by - star basis , using a specified set of evolutionary tracks . the code accommodates several additional inputs , including the initial mass function , star formation rate function , and chemical enrichment law . to compare the generated models to the data , we added gaussian noise based on the average observed errors in luminosity and temperature . we also randomly removed stars from the models according to our estimated spectroscopic completeness ( fig [ comp_err ] ) . the models are shown against our galactic center sample in figure [ sf_models_lit ] . a qualitative comparison suggests that the ancient burst model is insufficient to fully describe the observed data set . the continuous star formation model and the @xcite model span the full range in temperature and luminosity of the observed data set but seem to have relatively too many stars at cool temperatures . it is unlikely this observed effect is caused by systematic errors or a selection bias . a systematic underestimate of the gc co indices would shift the pattern to hotter temperatures . however , the red edge of the hertzsprung gap ( log @xmath7 @xmath76 ) is well matched by the models , making such a systematic effect unlikely . in addition , some cool rgb / agb stars are present in the data , while a systematic shift to warmer temperatures would allow none . finally , the brightness and co strength of cool rgb / agb stars compared to warmer , dimmer red clump stars means that cool rgb / agb stars are relatively easy to detect and classify . there is no obvious way of selectively removing these stars from the sample . motivated by the comparison in figure [ sf_models_lit ] , we consider three alternative scenarios that could potentially explain the relative paucity of stars at low temperatures : 1 . the first and simplest possibility is that the star formation rate was low at early times ( @xmath77 gyr ) . to test this hypothesis , we generated two models consisting of linear combinations of constant star formation in two specified age bins ( models 4 and 5 in table [ sf_models ] ) . model 4 assumes bins of 0.01 - 5 gyr and 5 - 12 gyr . model 5 assumes bins of 0.01 - 7 gyr and 7 - 12 gyr . a second possibility is that the oldest stars are metal poor . the previously described models assume a solar metallicity for all times . this assumption is motivated by the work of @xcite , who found the stellar [ fe / h ] abundance to be approximately solar for stars younger than @xmath55 gyr . for older stars , the galactic center metallicity distribution has not been determined . however , the bulge , which formed 7 - 12 gyr ago , is known to have a nearly solar distribution @xcite . an extremely metal poor ancient population would , therefore , represent a completely separate population from the bulge . given the current observed differences between the bulge and the nucleus discussed in 1 , we can not exclude this possibility . we , therefore , generated two models to represent this scenario . the first model , model 6 , consists of a linear combination of constant star formation in two age bins with solar metallicity for 0.01 - 5 gyr and a metallicity of 0.008 for 5 - 12 gyr . the second model , model 7 , is a simple closed box model assuming constant star formation for 0.01 - 12 gyr , a starting metallicity of 0.004 and an ending metallicity of solar . 3 . a final way to explain the relative lack of stars at cool temperatures invokes a non - standard initial mass function ( imf ) . in all above models , we assume a salpeter imf with an upper mass limit of 120 @xmath78 and a lower mass limit of 0.7 @xmath78 . as stars less massive than @xmath79 have main sequence lifetimes comparable to the age of the universe , we are not sensitive to stars below this mass limit . recent results indicate that the current initial mass function of the galactic center is flatter than a salpeter function or has an unusally high lower mass cut - off @xcite . such a mass function could explain the relative lack of low mass stars in the observed old stellar population . to test this hypothesis , we consider three models , all assuming continuous star formation at solar metallicity for 0.01 - 12 gyr . the first model , model 8 , assumes a flat single power - law slope ( @xmath80 ) , chosen to match the results of @xcite . models 9 and 10 both assume a standard salpeter slope ( @xmath81 ) , but with lower mass limits of 2.5 @xmath78 and 1.5 @xmath78 , respectively . results stemming from these three hypotheses are tested in 3.5 . to quantitatively compare the models described above to the data , we adopted the numerical techniques described in @xcite . we first binned the observed and model h - r diagrams in temperature and luminosity for stars above our 50% completeness threshold , using uniform bins with size three times our average errors ( @xmath82 , @xmath83 ) . the choice of bins in this technique is somewhat subjective . however , tests using different bins sizes showed that while the binning scheme does change the fit quality , it does not significantly affect the derived star formation rates . this finding is in agreement with @xcite . for models with two age bins , we used the numerical recipes procedure amoeba @xcite to search for the linear combination of models that minimized the poisson maximum likelihood parameter : @xmath84 @xcite . here , @xmath85 is the number of observed stars and @xmath86 is the number of model stars in bin @xmath87 . for models with fixed relative star formation rates , we scaled the model distribution to minimize @xmath88 . to estimate the errors in the derived star formation history , we implemented the technique described in @xcite . we built a set of 100 h - r diagrams consisting of a random sampling of 329 stars drawn from the observed h - r diagram , allowing each observed star to be selected any number of times . we then re - derived the star formation history for each h - r diagram and given model . the resulting standard deviation in the derived star formation rate was taken as the 1@xmath39 uncertainty . we measured the fit quality through a second set of monte carlo simulations . in this set , we generated 10,000 synthetic data sets drawn from the fitted models , selecting 329 stars randomly for each trial and re - deriving the star formation history for the selected subset . for a model that is an accurate representation of the true star formation history , @xmath88 derived from the actual data set should be comparable to @xmath88 derived from a typical synthetic data set . therefore , to establish the goodness of fit , we calculated the percentage of trials , @xmath89 , in which @xmath88 was larger when fitting the synthetic data sets to the models than when fitting the true data sample to the models . if the model is a good representation of the true star formation history , @xmath89 should be approximately 50% . the results of all model fits are listed in table [ sf_models ] . hess diagrams showing the difference between the observed data histogram and the best - fit model histograms are shown in figure [ hess ] . white indicates regions where the model produces too many stars relative to the observed data set , and black indicates regions where the model produces too few stars relative to the observed data set . examination of the @xmath89 values in table [ sf_models ] confirms the qualitative result discussed in 3.3 that no literature model ( models 1 - 3 ) is a likely description of the observed data set . this result is also reflected in the first three panels in figure [ hess ] , which show white diagonal streaks , corresponding to the overdensity of cool stars in the models relative to the data . table [ sf_models ] also suggests that the early - history low star formation rate models ( models 4 and 5 ) are unlikely representations of the data , though model 5 returns a considerably better fit , and we are unable to completely exclude this model . however , examination of the fourth and fifth panels in figure [ hess ] show that both models systematically overpredict the number of cool stars , as in models 1 - 3 . the early - history low metallicity models ( models 6 and 7 ) also show systematic deficiencies when compared to the data . both models predict a red clump / horizontal branch morphology that is too blue compared to the observed distribution . similar discrepencies between the models and data are present in the high imf mass limit models ( models 9 and 10 ) . each predicts a hotter and more luminous red clump than is observed . however , while models 9 and 10 predict no old , cool stars , the flat imf model ( model 8) predicts few old , cool stars , in good agreement with the observations . model 8 also gives a satisfactory goodness of fit ( @xmath89 = 40% ) suggesting that this model is a likely representation of the observed distribution . using the models listed in table [ sf_models ] , we tested for population differences as a function of distance from the cluster center . such differences within the central 1 - 2 pc are expected due to mass segregation , and several broadband photometry studies have found indications of such an effect @xcite . to test for this effect in our data , we separated the data into two sets containing the four innermost ( @xmath90 , @xmath91 median ) and outermost ( @xmath92 , @xmath93 median ) spectral regions shown in figure [ mos ] . we then refit the two best - fit models ( models 5 and 8) to each subset , applying the average errors and completeness specific to that subset . this process returned fits for each subset that are consistent with those obtained for the entire data set . in particular , the subset @xmath89 values were within 10% of that obtained for the entire data set , and the relative star formation rates returned by both subsets for model 5 were within 1@xmath39 of the rates returned by the entire data set . we therefore , find no significant evidence in our data for a population gradient . this result is not that surprising , as the broadband photometry results of @xcite suggest that variations in the cool , low - mass stellar population within the central parsec should be most evident 3 - 7 arcsec from sgr a@xmath0 . our sample is outside of the region expected to exhibit the largest population differences ( see figure [ mos ] ) . in the future , the technique we describe could be applied to regions closer to sgr a@xmath0 , although limited spectroscopic completeness due to increased stellar density would complicate such a study . still , a thorough spectroscopic study of the red - clump and rgb / agb populations as a function of distance from sgr a@xmath0 could provide the first definitive test of mass segregation within the central parsec @xcite . with the restriction to simple models , the top - heavy imf model appears to be superior to the remaining three model families we consider ( low star formation rate ( sfr ) at early times , low metallicity at early times , imf with high lower mass limit ) . the low - sfr , low - z , and high @xmath75 families all show consistent systematic deficiencies when compared to the data . it is , therefore , unlikely one could slightly change the chosen parameters ( i.e. precise metallicity , age bin cut - offs , mass ranges , etc . ) within these families to produce a model that is an adequate description of the data . of the models considered in table [ sf_models ] , model 8 ( continuous star formation at solar metallicity with a top - heavy imf ) fits the observations best , and our monte carlo tests show that it is a reasonable description of the data . based on the goodness of fit , we can not completely exclude model 5 . however , as discussed in the previous section , this model shows systematic deviations from the data ; we , therefore , favor model 8 . we note that model 8 is probably not the only possible description of the data with this degree of likelihood . all models so far discussed assume a simplistic description of the gc star formation history , and there are likely more complicated scenarios that also adequately describe the data . for instance , an additional model with age bins identical to model 5 and an imf slope identical to model 8 yields a reasonable fit ( @xmath94 , @xmath95 ) . however , we do not feel that increasing the number of free parameters is justified , given the limits , uncertainties , and size of our data set . in the context of our current knowledge about the galactic center , continuous star formation with a top - heavy imf is reasonable . at present , there is substantial evidence for stars in the galactic nucleus spanning a wide range of ages , based on results from broadband photometry , and several distinct age tracers , including supergiants , agb stars , and oh / ir stars ( see 1 ) . the present distribution and kinematics of gas in the inner galaxy is also consistent with continuous star formation @xcite . in addition , the most recent epoch of star formation in the galactic center is likely represented by a top - heavy initial mass function @xcite . if this recent epoch of star formation is not anomolous and periodic bursts of gc star formation have occurred throughout the history of the galaxy , there is no a priori reason to believe the gc imf would change with time . the present data set appears to support this picture . we note that the mass traced by giants in the central cluster is primarily set by the initial mass formed , rather than dynamical effects , since mass segregation is expected to only strongly affect the central @xmath50.01 pc of the galactic center @xcite . on the 1 - 2 pc scale studied in this paper , the efficiency of mass segregation is thought to be much lower , in agreement with observational photometry results @xcite . though our findings are broadly consistent with several stellar population studies of the galactic center and the nucleus at large , there remains some disagreement . in particular , our results are somewhat different from those presented in @xcite , who probed the central 5 pc and used supergiants and bright agb stars to quantify the star formation history in a method very similar to that employed here . using their observed h - r diagram , @xcite argued for variable star formation over a wide range of ages , with the majority of stars formed more than 5 gyr ago at solar metallicity . we find a much smaller fraction of our sample is represented by solar - metallicity low - mass stars with ages @xmath74 5 gyr . the present work and the work of @xcite study different regions in the h - r diagram . each regime has relative advantages and disadvantages for studying the gc star formation history , and we discuss each in turn : 1 . _ field of view _ : due to the rarity of supergiants and bright agb stars in the galactic center , @xcite probed all late - type stars above their magnitude limit for the entire central cluster ( r @xmath96 pc ) . they , therefore , derived absolute star formation rates and were able to compare the cluster mass implied to estimates from dynamical studies . in contrast , our study probes a relatively small region on the sky ( @xmath50.2 pc@xmath34 within r @xmath97 pc ) . as such , our results are somewhat susceptible to population inhomogeneities in the central cluster , and we are able to derive only relative star formation rates . stellar crowding _ : spectral extraction from our sinfoni fields is complicated by stellar crowding . while we took care to ensure that our derived co indices are largely independent of exact pixel and background selection , contamination by neighboring stellar spectra remains a source of uncertainty for many of the dim stars . the stars studied by @xcite , on the other hand , are well separated from each other . in addition , the gc agbs and supergiants are much brighter than neighboring stars , so there is negligible uncertainty in the @xcite spectra due to crowding . spectral classification _ : the spectral classification method presented by @xcite is very similar to that presented here . our depper study has the advantage that only giants are observed , and thus , a separate luminosity class determination is not required . additionally , the stars we study are warmer than those studied by @xcite , and thus , our co-@xmath7 relationship is tighter than theirs , due to reduced uncertainties in model atmosphere spectra at warmer temperatures @xcite . we further note that the derived temperatures for the red supergiants in the work of @xcite are likely systematically too cool , as @xcite recently showed that red supergiants are @xmath5400 k warmer than previously thought . adequacy of stellar evolution models _ : the evolutionary models in the part of the h - r diagram probed by our study ( red clump , red giant branch , early agb ) are much less uncertain than the part of the h - r diagram studied by @xcite ( supergiants , tp - agbs ) . @xcite review the adequacy of agb stellar evolution models for deriving star formation histories , concluding that the observed bright agb populations are often more sensitive to poorly known modeling parameters than the star formation history . the input physics for the late agb stages is not well determined ( e.g. mass loss , convection , efficiency of the third dredge - up ) , and currently , only the padova libraries include these stages @xcite . while uncertainties in mass loss and convection are also present for the red clump , rgb , and early agb , the input physics for these stages is much better understood , and a number of stellar evolution models including these phases have been compared and tested against observations . for this reason , we believe our star formation history fit is more reliable than that presented in @xcite . finally , our data set also has the benefit that it contains several evolutionary features ( red clump , agb bump ) that are cleraly distinguishable from surrounding regions in the h - r diagram . the diagram morphology assures us that no large systematic effects are present in the data and allows us to construct physically motivated models . the region of the h - r diagram studied by @xcite has no such morphological features . while we believe that our results are more robust than those presented in @xcite , we also note that the findings are not necessarily in conflict . the region probed by @xcite extends to @xmath52.5 pc from the center , whereas we probe only the central parsec . furthermore , we note that there are some similarities in our findings . both studies find evidence for star formation throughout the history of the galaxy , and both studies suggest that purely solar metallicity models are needed to produce the observed data . still , further work is needed to resolve the remaining discrepencies ( i.e. , variable versus continuous star formation , salpeter versus flat imf ) . specifically , a sample tracing a large region on the sky and containing thousands of stars would represent a significant step forward in this field . the planned flamingos-2 gc survey on gemini @xcite will obtain 4000 late - type giant spectra out to one degree in galacto - centric radius ( @xmath5140 pc ) and with a spectral resolution of r@xmath520,000 . it will , therefore , provide unprecedented information on the galactic center chemical enrichment and star formation history . we observed 329 late - type giants @xmath98 north of sgr a@xmath0 with the integral field spectrometer sinfoni on the vlt . combining spectral classifications with naco photometry , we derived luminosities and effective temperatures for these stars . due to the improved magnitude limit of our sample relative to previous work , our derived h - r diagram clearly shows the red clump , as well as the red giant branch and asymptotic giant branch . using a maximum likelihood analysis , we compared the observed distribution to models representing ten possible star formation histories . the best - fit model corresponds to continuous star formation over the last 12 gyr with a top - heavy imf . the similarity of this result to the imf observed for the most recent epoch of star formation is intriguing and perhaps suggests a connection between recent star formation and the stars formed throughout the history of the galactic center . the upcoming flamingos-2 gc survey on gemini will provide important information needed to further understand this suggestive result . the authors thank amiel sternberg for his helpful discussions and insights . is grateful to the graduate research fellowship program at nsf and the graduate opportunity program fellowship at uc berkeley for funding this research . acknowleges support from the alexander von humboldt foundation . t.a . is supported by minerva grant 8563 and a new faculty grant by sir h. djangoly , cbe , of london , uk . this work has been supported by the national science foundation science through ast 0205999 and the center for adaptive optics , managed by the university of california at santa cruz under cooperative agreement no . ast 9876783 . this work has made use of the iac - star synthetic cmd computation code . iac - star is suported and maintained by the computer division of the instituto de astrofisica de canarias . adelman , s. j. , pintado , o. i. , nieva , f. , rayle , k. e. , & sanders , s. e. 2002 , , 392 , 1031 alexander , t. , & sternberg , a. 1999 , , 520 , 137 alexander , t. 2005 , , 419 , 65 alonso , a. , arribas , s. , & martnez - roger , c. 1999 , , 139 , 335 aparicio , a. , & gallart , c. 2004 , , 128 , 1465 ballero , s. k. , matteucci , f. , origlia , l. , & rich , r. m. 2007 , , 467 , 123 bell , r. a. , & gustafsson , b. 1989 , , 236 , 653 bessell , m. s. , & brett , j. m. 1988 , , 100 , 1134 blackwell , d. e. , & lynas - gray , a. e. 1994 , , 282 , 899 blackwell , d. e. , & lynas - gray , a. e. 1998 , , 129 , 505 blum , r. d. , sellgren , k. , & depoy , d. l. 1996a , , 470 , 864 blum , r. d. , sellgren , k. , & depoy , d. l. 1996b , , 112 , 1988 blum , r. d. , ramrez , s. v. , sellgren , k. , & olsen , k. 2003 , , 597 , 323 bonnet , h. , et al . 2003 , , 4839 , 329 carpenter , j. m. 2001 , , 121 , 2851 carr , j. s. , sellgren , k. , & balachandran , s. c. 2000 , , 530 , 307 davidge , t. j. , simons , d. a. , rigaut , f. , doyon , r. , & crampton , d. 1997 , , 114 , 2586 diolaiti , e. , bendinelli , o. , bonaccini , d. , close , l. , currie , d. , & parmeggiani , g. 2000 , , 147 , 335 dolphin , a. 1997 , new astronomy , 2 , 397 dolphin , a. e. 2002 , , 332 , 91 dyck , h. m. , benson , j. a. , van belle , g. t. , & ridgway , s. t. 1996 , , 111 , 1705 eikenberry , s. , et al . 2006 , , 6269 , eisenhauer , f. , et al . 2003a , the messenger , 113 , 17 eisenhauer , f. , et al . 2003b , , 4841 , 1548 eisenhauer , f. , schdel , r. , genzel , r. , ott , t. , tecza , m. , abuter , r. , eckart , a. , & alexander , t. 2003c , , 597 , l121 engelke , c. w. 1992 , , 104 , 1248 feast , m. w. 1996 , , 278 , 11 figer , d. f. , mclean , i. s. , & morris , m. 1999 , , 514 , 202 figer , d. f. , et al . 2002 , , 581 , 258 figer , d. f. , rich , r. m. , kim , s. s. , morris , m. , & serabyn , e. 2004 , , 601 , 319 freitag , m. , amaro - seoane , p. , & kalogera , v. 2006 , , 649 , 91 frogel , j. a. , persson , s. e. , matthews , k. , & aaronson , m. 1978 , , 220 , 75 fitzgerald , m. p. , & graham , j. r. 2006 , , 637 , 541 fluks , m. a. , plez , b. , the , p. s. , de winter , d. , westerlund , b. e. , & steenman , h. c. 1994 , , 105 , 311 gallart , c. , zoccali , m. , & aparicio , a. 2005 , , 43 , 387 genzel , r. , et al . 2003 , , 594 , 812 ghez , a. m. , salim , s. , hornstein , s. d. , tanner , a. , lu , j. r. , morris , m. , becklin , e. e. , & duchne , g. 2005 , , 620 , 744 girardi , l. , bressan , a. , bertelli , g. , & chiosi , c. 2000 , , 141 , 371 gray , d. f. , & brown , k. 2001 , , 113 , 723 guesten , r. , genzel , r. , wright , m. c. h. , jaffe , d. t. , stutzki , j. , & harris , a. i. 1987 , , 318 , 124 hartung , m. , et al.2003 , , 4841 , 425 hopman , c. , & alexander , t. 2006 , , 645 , l133 jackson , j. m. , geis , n. , genzel , r. , harris , a. i. , madden , s. , poglitsch , a. , stacey , g. j. , & townes , c. h. 1993 , , 402 , 173 kent , s. m. 1992 , , 387 , 181 kleinmann , s. g. , & hall , d. n. b. 1986 , , 62 , 501 kormendy , j. , & kennicutt , r. c. , jr . 2004 , , 42 , 603 krabbe , a. , genzel , r. , drapatz , s. , & rotaciuc , v. 1991 , , 382 , l19 krabbe , a. , et al . 1995 , , 447 , l95 launhardt , r. , zylka , r. , & mezger , p. g. 2002 , , 384 , 112 levesque , e. m. , massey , p. , olsen , k. a. g. , plez , b. , josselin , e. , maeder , a. , & meynet , g. 2005 , , 628 , 973 lindqvist , m. , habing , h. j. , & winnberg , a. 1992 , , 259 , 118 malagnini , m. l. , & morossi , c. 1990 , , 85 , 1015 malagnini , m. l. , & morossi , c. 1997 , , 326 , 736 martins , f. , genzel , r. , hillier , d. j. , eisenhauer , f. , paumard , t. , gillessen , s. , ott , t. , & trippe , s. 2007 , , 468 , 233 mccaughrean , m. j. , close , l. m. , scholz , r .- d . , lenzen , r. , biller , b. , brandner , w. , hartung , m. , & lodieu , n. 2004 , , 413 , 1029 morris , m. , & serabyn , e. 1996 , , 34 , 645 narayanan , v. k. , gould , a. , & depoy , d. l. 1996 , , 472 , 183 nayakshin , s. , & sunyaev , r. 2005 , , 364 , l23 paumard , t. , et al . 2006 , , 643 , 1011 perrin , g. , coude du foresto , v. , ridgway , s. t. , mariotti , j .- m . , traub , w. a. , carleton , n. p. , & lacasse , m. g. 1998 , , 331 , 619 persson , s. e. , murphy , d. c. , krzeminski , w. , roth , m. , & rieke , m. j. 1998 , , 116 , 2475 philipp , s. , zylka , r. , mezger , p. g. , duschl , w. j. , herbst , t. , & tuffs , r. j. 1999 , , 348 , 768 press , w. h. , teukolsky , s. a. , vetterling , w. t. , & flannery , b. p. 1992 , cambridge : university press , |c1992 , 2nd ed . , ramirez , s. v. , depoy , d. l. , frogel , j. a. , sellgren , k. , & blum , r. d. 1997 , , 113 , 1411 ramrez , s. v. , sellgren , k. , carr , j. s. , balachandran , s. c. , blum , r. , terndrup , d. m. , & steed , a. 2000 , , 537 , 205 ramrez , i. , & melndez , j. 2005 , , 626 , 446 reid , m. j. 1993 , , 31 , 345 reid , n. & hawley , s. l. 2000 , new light on dark stars : red dwarfs , low mass stars , brown dwarfs / neill reid and suzanne l. hawley . new york : springer , 2000 . ( springer - praxis series in astronomy and astrophysics ) rieke , m. j. 1987 , nearly normal galaxies . from the planck time to the present , 90 rieke , m. j. 1999 , asp conf . ser . 186 : the central parsecs of the galaxy , 186 , 32 rousset , g. , et al . 2003 , , 4839 , 140 sadler , e. m. , rich , r. m. , & terndrup , d. m. 1996 , , 112 , 171 schoedel , r. , et al . 2007 , arxiv astrophysics e - prints , arxiv : astro - ph/0703178 schreiber , j. , thatte , n. , eisenhauer , f. , tecza , m. , abuter , r. , & horrobin , m. 2004 , asp conf . ser . 314 : astronomical data analysis software and systems ( adass ) xiii , 314 , 380 sellgren , k. , hall , d. n. b. , kleinmann , s. g. , & scoville , n. z. 1987 , , 317 , 881 serabyn , e. , keene , j. , lis , d. c. , & phillips , t. g. 1994 , , 424 , l95 sheehy , c. d. , mccrady , n. , & graham , j. r. 2006 , , 647 , 1517 sjouwerman , l. o. , habing , h. j. , lindqvist , m. , van langevelde , h. j. , & winnberg , a. 1999 , asp conf . ser . 186 : the central parsecs of the galaxy , 186 , 379 smalley , b. 1993 , , 265 , 1035 sokolov , n. a. 1998 , , 130 , 215 stolte , a. , brandner , w. , grebel , e. k. , lenzen , r. , & lagrange , a .- m . 2005 , , 628 , l113 taylor , b. j. 2003 , , 398 , 721 theodossiou , e. , & danezis , e. 1991 , , 183 , 91 vacca , w. d. , sheehy , c. d. , & graham , j. r. 2007 , arxiv astrophysics e - prints , arxiv : astro - ph/0701628 van dam , m. a. , le mignant , d. , & maccintosh , b. a. 2004 , appl opt . , 43 , 5458 van loon , j. t. , et al . 2003 , , 338 , 857 wallace , l. , & hinkle , k. 1997 , , 111 , 445 wood , p. r. , habing , h. j. , & mcgregor , p. j. 1998 , , 336 , 925 zoccali , m. , et al . 2003 , , 399 , 931 zoccali , m. , et al . 2006 , , 457 , l1

the manipulation and control of quantum systems is becoming an important task in many fields @xcite-@xcite , such as atomic physics @xcite , molecular chemistry @xcite and quantum information @xcite . it is desirable to develop quantum control theory in a systematic way in order to adapt it to the development of quantum technology @xcite . several useful tools from classical control theory have been introduced to the control analysis and design of quantum systems . for example , optimal control theory has been used to assist in control design for closed and dissipative quantum systems @xcite-@xcite . a learning control method has been presented for guiding the control of chemical reactions @xcite , @xcite . quantum feedback control approaches including measurement - based feedback and coherent feedback have been used to improve performance for several classes of tasks such as preparing quantum states , quantum error correction , controlling quantum entanglement @xcite-@xcite . robust control tools have been introduced to enhance the robustness of quantum feedback networks and linear quantum stochastic systems @xcite , @xcite . although some progress has been made , more research effort is necessary in controlling quantum phenomena . in particular , the robustness of quantum control systems has been recognized as a key issue in developing practical quantum technology @xcite-@xcite . in this paper , we focus on the robustness problem for quantum control systems . in @xcite , james and co - workers have formulated and solved a quantum robust control problem using the @xmath0 method for linear quantum stochastic systems . here , we develop a variable structure control approach with sliding modes to enhance the robustness of quantum systems . the variable structure control strategy is a widely used design method in classical control theory and industrial applications where one can change the controller structure according to a specified switching logic in order to obtain desired closed - loop properties @xcite , @xcite . in @xcite and @xcite , dong and petersen have formulated and solved a variable structure control problem for the control of quantum systems . however , the results in @xcite only involve open - loop control design using an idea of changing controller structures and do not consider the robustness which can be obtained through sliding mode control . @xcite and ref . @xcite have briefly discussed the possible application of sliding mode control to quantum systems . in @xcite , two approaches based on sliding mode design have been proposed for the control of quantum systems and potential applications of sliding mode control to quantum information processing have been presented . following these results , this paper formally presents a sliding mode control method for two - level quantum systems to deal with bounded uncertainties in the system hamiltonian @xcite . in particular , we propose two approaches of designing the measurement period for different situations which are dependent on the bound on the uncertainties and the allowed probability of failure . variable structure control design with sliding modes generally includes two main steps : selecting a sliding surface ( sliding mode ) and controlling the system to and maintaining it in this sliding surface . being in the sliding surface guarantees that the quantum system has the desired dynamics . we will select an eigenstate of the free hamiltonian of the controlled quantum system as a sliding mode . in the second step , direct feedback control is not directly applicable since we generally can not acquire state information without destroying the quantum system s state . hence , we propose a new method to accomplish this task , which is based on the lyapunov methodology and periodic projective measurements . the lyapunov methodology is a powerful tool for designing control laws in classical control theory and has also been applied to quantum control problems @xcite-@xcite . most existing results on lyapunov control of quantum systems focus on designing a control law to ensure that the controlled quantum system s state asymptotically converges to the target state . the existing lyapunov design methods in quantum control rely on perfect knowledge of the initial quantum states and system hamiltonian . in our approach , once the lyapunov control steers the quantum system into a sliding mode domain , we make a projective measurement on the system . hence , the lyapunov design method can tolerate small drifts ( uncertainties ) when carrying out our control tasks , which will be demonstrated by simulation in section ii.c . periodic projective measurements are employed to maintain the system s state in the sliding mode domain when uncertainties exist in the system hamiltonian . if the measurement period is small enough and the initial state is an eigenstate , the frequent measurements make the system collapse back to the initial state . this is related to the quantum zeno effect ( for details , see @xcite , @xcite and @xcite ) . in contrast to the quantum zeno effect , our objective is to design a measurement period which is as large as possible . the framework of the proposed method involves unitary control ( lyapunov control ) and projective measurement . in this sense , it is similar to the discrete - time quantum feedback stabilization problem in @xcite and @xcite . however , these papers do not consider possible uncertainties in the system hamiltonian and use generalized measurements rather than periodic projective measurements . the main feature of the proposed method is that the control law can guarantee control performance when bounded uncertainties exist in the system hamiltonian . this paper is organized as follows . section [ sec2 ] introduces a quantum control model , defines the sliding mode and formulates the control problem . in section [ sec3 ] , we present a sliding mode control method based on the lyapunov methodology and periodic projective measurements for two - level quantum systems with bounded uncertainties . using the known information about uncertainties ( e.g. , the uncertainty bound and type of uncertainties ) , we propose two approaches ( i.e. , eqs . ( [ 1period ] ) and ( [ 2period ] ) ) for designing the measurement period to guarantee the control performance . an illustrative example is presented to demonstrate the proposed method . the detailed proofs of the main theorems are presented in section [ sec4 ] . conclusions are given in section [ sec5 ] . in this section , we first introduce a two - level quantum control model . then a sliding mode is defined using an eigenstate . finally the control problem considered in this paper is formulated . in this paper , we focus on two - level pure - state quantum systems . the quantum state can be represented by a two - dimensional unit vector @xmath1 in a hilbert space @xmath2 . since the global phase of a quantum state has no observable physical effect , we do not consider the effect of global phase . if we denote the pauli matrices @xmath3 as follows : @xmath4 we may select the free hamiltonian of the two - level quantum system as @xmath5 . its two eigenstates are denoted as @xmath6 and @xmath7 . to control a quantum system , we introduce the following control hamiltonian @xmath8 , where @xmath9 and @xmath10 is a set of time - independent hamiltonians . for simplicity , the control hamiltonian for two - level systems can be written as @xmath11 , where @xmath12 the controlled dynamical equation can be described as ( we have assumed @xmath13 by using atomic units in this paper ) @xmath14 this control problem is converted into the following problem : given an initial state and a target state , find a set of controls @xmath15 in ( [ controlled model ] ) to drive the controlled system from the initial state to the target state . in practical applications , we often use the density operator ( or density matrix ) @xmath16 to describe the quantum state of a quantum system . for a pure state @xmath1 , the corresponding density operator is @xmath17 . for a two - level quantum system , the state @xmath16 can be represented in terms of the bloch vector @xmath18 : @xmath19 the evolution equation of @xmath16 can be written as @xmath20\ ] ] where @xmath21=ab - ba$ ] and @xmath22 is the total system hamiltonian . after we represent the state @xmath16 with the bloch vector , the pure states of a two - level quantum system correspond to the surface of the bloch sphere , where @xmath23 , @xmath24 $ ] , @xmath25 $ ] . an arbitrary pure state @xmath1 for a two - level quantum system can be represented as @xmath26 sliding modes play an important role in variable structure control @xcite . usually , the sliding mode is constructed so that the system has desired dynamics in the sliding surface . for a quantum control problem , a sliding mode may be represented as a functional of the state @xmath1 and the hamiltonian @xmath22 ; i.e. , @xmath27 . for example , an eigenstate @xmath28 of the free hamiltonian @xmath29 ( i.e. , @xmath30 where @xmath31 is one eigenvalue of @xmath29 ) can be selected as a sliding mode . we can define @xmath32 . if the initial state @xmath33 is in the sliding mode ; i.e. , @xmath34 , we can easily prove that the quantum system will maintain its state in this surface under only the action of the free hamiltonian @xmath29 . in fact , @xmath35 , and we have @xmath36 that is , an eigenstate of @xmath29 can be identified as a sliding mode . for two - level quantum systems , we may select either @xmath6 or @xmath7 as a sliding mode . without loss of generality , we identify the eigenstate @xmath6 of a two - level quantum system as the sliding mode in this paper . in section ii.b , we have identified an eigenstate @xmath6 as a sliding mode . this means that if a quantum system is driven into the sliding mode , its state will be maintained in the sliding surface under the action of the free hamiltonian . however , in practical applications , it is inevitable that there exist noises and uncertainties . in this paper , we suppose that the uncertainties can be approximately described as perturbations in the hamiltonian . that is , the uncertainties can be denoted as @xmath37 . the unitary errors in @xcite belong to this class of uncertainties and uncertainties in one - qubit ( one quantum bit ) gate also correspond to this class of uncertainties @xcite . for a spin system in solid - state nuclear magnetic resonance ( nmr ) , external noisy magnetic fields and unwanted coupling with other spins may lead to uncertainties in this class . further , we assume the uncertainties are bounded ; i.e. , @xmath38 when @xmath39 , @xmath40 . that is , there exist no uncertainties , which is trivial for our problem . hence , in the following we assume @xmath41 . an important advantage of classical sliding mode control is its robustness . our main motivation for introducing sliding mode control to quantum systems is to deal with uncertainties . we further suppose that the corresponding system without uncertainties is completely controllable and arbitrary unitary control operations can be generated . this assumption can be guaranteed for a two - level quantum system if we can realize arbitrary rotations along the @xmath42-axis and @xmath43-axis ( @xmath44 ) ( e.g. , see @xcite for details ) . the control problem under consideration is stated as follows : design a control law to drive and then maintain the quantum system s state in a sliding mode domain even when bounded uncertainties exist in the system hamiltonian . here a sliding mode domain may be defined as @xmath45 , where @xmath46 is a given constant . here we assume @xmath47 since the case @xmath48 only occurs in the sliding mode surface and the case @xmath49 is always true . hence , the two cases with @xmath48 and @xmath49 are trivial for our problem . the definition of the sliding mode domain implies that the system has a probability of at most @xmath46 ( which we call the probability of failure ) to collapse out of @xmath50 when making a measurement . this behavior is quite different from that which occurs in traditional sliding mode control . hence , we expect that our control laws will guarantee that the system s state remains in @xmath50 except that a measurement operation may take it away from @xmath50 with a small probability ( not greater than @xmath46 ) . the control problem considered in this paper includes three main subtasks : ( i ) for any initial state ( assumed to be known ) , design a control law to drive the system s state into a defined sliding mode domain @xmath50 ; ( ii ) design a control law to maintain the system s state in @xmath50 ; ( iii ) design a control law to drive the system s state back to @xmath50 if a measurement operation takes it away from @xmath50 . for simplicity , we suppose that there exist no uncertainties during the control processes ( i ) and ( iii ) . the first task is to design a control law to drive the controlled system to the chosen sliding mode domain @xmath50 . lyapunov - based methods are widely used to accomplish this task in traditional sliding mode control . if the gradient of a lyapunov function is negative in the neighborhood of the sliding surface , then the controlled system s state will be attracted to and maintained in @xmath50 . the lyapunov methodology has also been used to design control laws for quantum systems @xcite-@xcite . however , these existing results do not consider the issue of robustness against uncertainties . since the measurement of a quantum system will inevitably destroy the measured state , most existing results on lyapunov - based control for quantum systems in fact use a feedback design to construct an open - loop control . that is , lyapunov - based control can be used to first design a feedback law which is then used to find the open - loop control by simulating the closed - loop system . then the control can be applied to the quantum system in an open - loop way . hence , the traditional sliding mode control methods using lyapunov control can not be directly applied to our problem . although quantum measurement often has deleterious effects in quantum control tasks , recent results have shown that it can be combined with unitary transformations to complete some quantum manipulation tasks and enhance the capability of quantum control @xcite , @xcite-@xcite . for example , vilela mendes and manko @xcite showed nonunitarily controllable systems can be made controllable by using measurement plus evolution " . quantum measurement can be used as a control tool as well as a method of information acquisition . it is worth mentioning that the effect of measurement on a quantum system as a control tool can be achieved through the interaction between the system and measurement apparatus . in this paper , we will combine the lyapunov methodology and projective measurements ( with the measurement operator @xmath51 ) to accomplish the sliding mode control task for two - level quantum systems . the projective measurement with @xmath51 on a two - level system makes the system s state collapse into @xmath6 ( corresponding to eigenvalue @xmath52 of @xmath51 ) or @xmath7 ( corresponding to eigenvalue @xmath53 of @xmath51 ) . the steps in the control algorithm are as follows ( see fig . 1 ) : 1 . select an eigenstate @xmath6 of @xmath29 as a sliding mode @xmath54 , and define the sliding mode domain as @xmath55 . 2 . for a known initial state @xmath33 , construct a lyapunov function @xmath56 to find the control law that can drive @xmath33 into the sliding mode @xmath57 . 3 . for a specified probability of failure @xmath46 and @xmath56 , construct the control period @xmath58 so that the control law can drive the system s state into @xmath50 in a time period @xmath58 . 4 . for an initial condition which is another eigenstate @xmath7 , design a lyapunov function @xmath59 and construct the period @xmath60 by using a similar method to that in 3 ) . 5 . according to @xmath46 and @xmath61 , design the period @xmath62 for periodic projective measurements . use the designed control law to drive the system s state into @xmath50 in @xmath58 and make a projective measurement at @xmath63 . then repeat the following operations : make periodic projective measurements with the period @xmath62 to maintain the system s state in @xmath50 ; if the measurement result corresponds to @xmath7 , we use the corresponding control law to drive the state back into @xmath50 . from the above control algorithm , we see that the design of lyapunov functions and the selection of the period @xmath62 for the projective measurements are the two most important tasks . to design a control law for quantum systems , several lyapunov functions have been constructed , such as state distance - based and average value - based approaches @xcite-@xcite . here we select a function based on the hilbert - schmidt distance between a state @xmath1 and the sliding mode state @xmath28 as a lyapunov function @xcite , @xcite ; i.e. , @xmath64 it is clear that @xmath65 . the first - order time derivative of @xmath66 is @xmath67\\ & = -\sum_{k = x , y , z}u_{k}|\langle \psi|\phi_{j}\rangle|\im[e^{i\angle \langle \psi|\phi_{j}\rangle } \langle \phi_{j}|i_{k}|\psi\rangle],\\ \end{array}\ ] ] where @xmath68=b$ ] ( @xmath69 ) and @xmath70 denotes the argument of a complex number @xmath71 . to ensure @xmath72 , we choose the control laws as in @xcite : @xmath73 ) , \ \ \ ( k = x , y , z)\ ] ] where @xmath74 may be used to adjust the control amplitude and @xmath75 satisfies @xmath76 . define @xmath77 when @xmath78 . when one employs a lyapunov methodology to design a control law , lasalle s invariance principle is a useful tool to analyze its convergence . that is , if @xmath79 is an autonomous dynamical system with phase space @xmath80 and @xmath81 is a lyapunov function on @xmath80 satisfying @xmath82 for all @xmath83 and @xmath84 , any bounded solution converges to the invariant set @xmath85 as @xmath86 ( for details , see @xcite ) . for two - level quantum systems , lasalle s invariance principle can guarantee that the quantum state converges to the sliding mode @xmath6 under the control law in ( [ lyapunov_control ] ) ( for details , see @xcite ) . the convergence is asymptotic . hence , we make a projective measurement with the measurement operator @xmath51 when we apply the lyapunov control to the system for @xmath58 ( corresponding to the initial condition @xmath33 ) or @xmath60 ( corresponding to the initial condition @xmath7 ) , which will drive the system into @xmath6 with a probability not less than @xmath87 . another important task is to design the measurement period @xmath62 . we can estimate a bound according to the bound @xmath61 on the uncertainties and the allowed probability of failure @xmath46 . then , we construct a period @xmath62 to guarantee control performance according to the estimated bound . an extreme case is @xmath88 . that is , after the quantum system s state is driven into the sliding mode , we make frequent measurements . this corresponds to the quantum zeno effect @xcite , which is the inhibition of transitions between quantum states by frequent measurement of the state ( see , e.g. , @xcite and @xcite ) . frequent measurements ( i.e. , @xmath88 ) can guarantee that the state is maintained in the sliding mode in spite of the presence of uncertainties . however , it is usually a difficult task to make such frequent measurements . we may conclude that the smaller @xmath62 is , the bigger the cost of accomplishing the periodic measurements becomes . hence , in contrast to the quantum zeno effect , we wish to design a measurement period @xmath62 as large as possible . in the following subsection , we will propose two approaches of designing @xmath62 for different situations . we select the sliding mode as @xmath89 . if there exist no uncertainties and we have driven the system s state to the sliding mode at time @xmath90 , it will be maintained in the sliding mode using only the free hamiltonian @xmath29 ; i.e. , @xmath91 . that is , if the quantum system s state is driven into the sliding mode , it will evolve in the sliding surface . however , in practical applications , some uncertainties are unavoidable , which may drive the system s state away from the sliding mode . we wish to design a control law to ensure the desired robustness in the presence of uncertainties . assume that the state at time @xmath92 is @xmath93 . if we make measurements on this system , the probability @xmath94 that it will collapse into @xmath7 ( the probability of failure ) is @xmath95 where @xmath96 . we have assumed that the possible uncertainties can be described by @xmath37 , where unknown @xmath97 , @xmath98 and @xmath99 satisfy @xmath100 . we now give detailed discussions to design the measurement period @xmath62 for possible uncertainties . first we consider a special case @xmath101 ( @xmath102 ) . this case corresponds to phase - flip type bounded uncertainties . for any @xmath103 ( where @xmath104 ) , if @xmath105 , we have @xmath106 this type of uncertainty does not drive the system s state away from the sliding mode . hence we ignore this type of uncertainty in our method . now we consider the unknown uncertainties @xmath107 ( where @xmath108 ) and have the following theorem . for a two - level quantum system with the initial state @xmath109 at the time @xmath110 , the system evolves to @xmath111 under the action of @xmath112 ( where @xmath113 and @xmath114 ) . if @xmath115 $ ] , where @xmath116 the system s state will remain in @xmath117 ) . when one makes a projective measurement with the measurement operator @xmath51 at the time @xmath92 , the probability of failure @xmath118 is not greater than @xmath46 . using theorem 1 , we may try to maintain the system s state in @xmath50 ( i.e. , the subtask ( ii ) ) by implementing periodic projective measurements with the measurement period @xmath119 . if we have more knowledge about the uncertainties , it is possible to improve the measurement period @xmath120 . now assume that the uncertainty is @xmath121 ( @xmath122 ) and @xmath123 $ ] . we have the following theorem . for a two - level quantum system with the initial state @xmath109 at the time @xmath110 , the system evolves to @xmath111 under the action of @xmath124 ( where @xmath44 , @xmath102 and @xmath41 ) . if @xmath125 $ ] and @xmath126 $ ] , where @xmath127}{\sqrt{1+\epsilon^{2}}},\ ] ] the system s state will remain in @xmath117 ) . when one makes a projective measurement with the measurement operator @xmath51 at the time @xmath92 , the probability of failure @xmath118 is not greater than @xmath46 . the proofs of theorem 1 and theorem 2 will be presented in section [ sec4 ] . the two theorems mean the following fact . for a two - level quantum system with unknown uncertainties @xmath107 ( where @xmath113 ) , if its initial state is in the sliding mode @xmath6 , we can ensure that the probability of failure is not greater than a given constant @xmath46 ( @xmath128 ) through implementing periodic projective measurements with the measurement period @xmath119 using ( [ 1period ] ) . further , if we know that @xmath46 and @xmath61 satisfy the relationship @xmath129 and there exists only one type of uncertainty ( i.e. , @xmath130 or @xmath131 , where @xmath102 ) , we can design a measurement period @xmath132 using ( [ 2period ] ) which is larger than @xmath120 . the proof of theorem 2 also shows that @xmath133 is an optimal measurement period . this measurement period will guarantee the required robustness . it is easy to prove the relationship @xmath134 for arbitrary @xmath123 $ ] . the detailed proof will be presented in the appendix . based on the above analysis , the selection rule for @xmath62 is summarized in table i. moreover , from ( [ 1period ] ) and ( [ 2period ] ) , it is clear that for a constant @xmath61 , @xmath135 and @xmath136 when @xmath137 . that is , for a given bound @xmath61 on the uncertainties , if we expect to guarantee the probability of failure @xmath137 , it requires us to implement frequent measurements such that the measurement period @xmath88 . another special case is @xmath138 , which leads to @xmath135 and @xmath136 . that is , to deal with very large uncertainties , we need to make frequent measurements ( @xmath88 ) to guarantee the desired robustness . from ( [ 1period ] ) , we also know that for a given @xmath46 , @xmath120 monotonically decreases with increasing @xmath61 . this means that we need to employ a smaller measurement period to deal with uncertainties with a larger bound @xmath61 . @xmath107 & @xmath121 ( @xmath139 + allowed probability of failure @xmath46 & @xmath128 & .a summary on the selection rule of the measurement period @xmath62 [ cols="^,^ " , ] + now we present an illustrative example to demonstrate the proposed method . assume @xmath140 . consider two cases : ( a ) @xmath141 ; ( b ) @xmath142 . for simplicity , we assume @xmath143 . hence , @xmath144 . we first design the control and @xmath60 using ( [ lyapunov_control ] ) . here , we consider control only using @xmath145 . using ( [ lyapunov_control ] ) , we select @xmath146)$ ] and @xmath147 . let the time stepsize be given by @xmath148 . we can obtain the probability curve of @xmath6 shown in fig . [ smc_probability ] , the control value shown in fig . [ smc_control ] and @xmath149 . for @xmath141 , we can design the measurement period @xmath150 using ( [ 1period ] ) . for @xmath142 , we can design the measurement period @xmath151 using ( [ 1period ] ) . since @xmath152 when @xmath142 , if the uncertainties take the form of @xmath121 ( @xmath139 , we can improve the measurement period to @xmath153 using ( [ 2period ] ) . it is clear that @xmath154 in these two cases . for some practical quantum systems such as spin systems in nmr , we can use strong control actions ( e.g. , @xmath155 ) to drive the system from @xmath7 into @xmath50 within a short time period @xmath60 @xcite . these facts make the assumption of no uncertainties in the control process reasonable . moreover , the fact that the measurement period @xmath62 is much greater than the control time required to go to @xmath6 from @xmath7 indicates the possibility of realizing such a periodic measurement on a practical quantum system . under lyapunov control.,width=345 ] .,width=345 ] in the process of designing the control law for driving the system s state from @xmath7 to @xmath6 , we employ an approach based on the lyapunov methodology . an advantage of such an approach is that it is relatively easy to find a control law by simulation . it is worth noting that most existing applications of the lyapunov methodology to quantum systems do not involve measurement . here , we combine the lyapunov - based control and projective measurements for controlling quantum systems , which in some applications make our method more useful than the lyapunov - based control for quantum systems proposed in previous papers . in @xcite , an approach based on time - optimal control design has also been proposed to complete this task . the advantage of such an approach is that we take the shortest time to complete the control task . however , it is generally difficult to find a complete time - optimal solution for high - dimensional quantum systems . for the above simple task , it has been proven that the time - optimal control employs a bang - bang control strategy @xcite . using the method in @xcite , we should take @xmath156 in @xmath157 $ ] and then use @xmath158 in @xmath159 $ ] . in this case , the total time required is @xmath160 ( @xmath161 ) . in the process of designing the lyapunov control for driving the system s state from @xmath7 to @xmath6 , we ignore possible uncertainties . by simulation , we find that small uncertainties can also be tolerated in this process . for example , if @xmath141 and the uncertainty @xmath162 is the noise with a uniform distribution on the interval @xmath163 $ ] , the probability curves of @xmath6 are shown in fig . [ figure_with_noise ] when we apply the control obtained from fig . [ smc_control ] to the quantum system . the probabilities of @xmath6 for the cases with uncertainties are very close to the probability of @xmath6 for the case without uncertainties . by more simulation , we find that the final probability of @xmath6 is @xmath164 for @xmath165 ( @xmath102 where @xmath141 or @xmath166 ) , the final probability of @xmath6 is @xmath167 for @xmath168 ( @xmath169 ) and the final probability of @xmath6 is @xmath170 for @xmath168 ( @xmath171 ) . if we use a smaller probability of failure @xmath172 ( e.g. , @xmath173 ) as the terminal condition of the lyapunov control or employ a bigger @xmath174 for the same @xmath60 , these simulations suggest that it is possible to ensure that the lyapunov control will drive the system s state into the sliding mode domain even when there exist small uncertainties . for the case without uncertainties ( without noise ) and the cases with uncertainties ( noise ) when we apply the control in fig . [ smc_control ] to the quantum system . the @xmath175-axis noise ( @xmath176 or @xmath177 ) means the existence of @xmath178 where @xmath162 is the noise with a uniform distribution on the interval @xmath179$].,width=576 ] this section will present the detailed proofs of theorem 1 and theorem 2 . the proof of theorem 1 involves the following steps : ( i ) compare the probabilities of failure for @xmath180 and @xmath181 ( @xmath182 and @xmath183 are constant ) ; ( ii ) compare the probabilities of failure for @xmath184 and @xmath185 ( @xmath102 ) ; ( iii ) use the previous results to compare the probabilities of failure for @xmath184 and @xmath186 ( @xmath187 ) ; ( iv ) use @xmath184 to estimate the measurement period . the basic steps for the proof of theorem 2 include : ( i ) formulate the problem of finding the worst " case as an optimal control problem of @xmath188 ; ( ii ) obtain the optimal control solution for nonsingular cases ; ( iii ) exclude the possibility of singular cases ; ( iv ) use the worst " case to estimate the measurement period . considering that the arguments in the proof of theorem 2 are useful for the proof of theorem 1 , we will first present the proof of theorem 2 and then prove theorem 1 . for @xmath189 , using @xmath190 $ ] and ( [ eq4 ] ) , we have @xmath191 that is , @xmath192 where @xmath193 . we now consider @xmath162 as a control input and select the performance measure as @xmath194 from ( [ fidelity ] ) , we know that the worst " case ( i.e. , the case maximizing the probability of failure ) corresponds to minimizing @xmath195 . also , we introduce the lagrange multiplier vector @xmath196 and obtain the corresponding hamiltonian function as follows : @xmath197 where @xmath198 . that is @xmath199 according to pontryagin s minimum principle @xcite , a necessary condition for @xmath200 to minimize @xmath201 is @xmath202 the necessary condition provides a relationship to determine the optimal control @xmath200 . if there exists a time interval @xmath203 $ ] of finite duration during which the necessary condition ( [ necessary_condition ] ) provides no information about the relationship between @xmath204 , @xmath200 , @xmath205 , we call the interval @xmath203 $ ] a singular interval @xcite . if we do not consider singular cases ( i.e. , @xmath206 ) , the optimal control @xmath200 should be chosen as follows : @xmath207 that is , the optimal control strategy for @xmath162 is bang - bang control ; i.e. , @xmath208 . now we consider @xmath209 which leads to the state equation @xmath210 where @xmath193 . the corresponding solution is @xmath211 where @xmath212 . from ( [ theorem1-bound - solution ] ) , we know that @xmath213 is a monotonically decreasing function in @xmath92 when @xmath214 $ ] . hence , we only consider the case @xmath215 $ ] where @xmath216 $ ] . now consider the optimal control problem with a fixed final time @xmath217 and a free final state @xmath218 . according to pontryagin s minimum principle , @xmath219 . from this , it is straightforward to verify that @xmath220 . now let us consider another necessary condition @xmath221 which leads to the following relationships : @xmath222 where @xmath220 . the corresponding solution is @xmath223 we obtain @xmath224 . % \end{array}\ ] ] it is easy to show that the quantity @xmath225 occurring in ( [ bangbang ] ) does not change its sign when @xmath216 $ ] and @xmath215 $ ] . now we further exclude the possibility that there exists a singular case . suppose that there exists a singular interval @xmath226 $ ] ( where @xmath227 ) such that when @xmath228 $ ] @xmath229 we also have the following relationship @xmath230 where we have used ( [ thm2stateeq ] ) and the following costate equation @xmath231 if @xmath232 , we have @xmath193 . by the principle of optimality @xcite , we may consider the case @xmath233 . using ( [ singular1eq ] ) , ( [ singular2eq ] ) and @xmath234 , we have @xmath235 and @xmath236 . using the relationship of @xmath237 , we obtain @xmath238 or @xmath53 . if @xmath238 , the initial state and the final state are the same state @xmath6 . however , if we use the control @xmath239 , from ( [ theorem1-bound - solution ] ) we have @xmath240 . hence , this contradicts the fact that we are considering the optimal case @xmath241 . if @xmath242 , there exists @xmath243 such that @xmath244 . by the principle of optimality @xcite , we may consider the case @xmath245 . from the two equations ( [ singular1eq ] ) and ( [ singular2eq ] ) , we know that @xmath246 which contradicts @xmath244 . hence , no singular condition can exist if @xmath232 . if @xmath247 , using ( [ bangbang ] ) we must select @xmath239 when @xmath248 $ ] . from ( [ singularcondition ] ) , we know that there exist no @xmath249 satisfying @xmath250 . hence , there exist no singular cases for our problem . from the above analysis , @xmath239 is the optimal control when @xmath251 $ ] . hence @xmath252 . from ( [ fidelity ] ) , it is clear that the probabilities of failure satisfy @xmath253 . that is , the probability of failure @xmath254 is not greater than @xmath255 for @xmath214 $ ] . when @xmath214 $ ] , @xmath256 is monotonically decreasing and @xmath257 is monotonically increasing . when @xmath258 , using ( [ theorem1-bound - solution ] ) we have @xmath259 that is , the probability of failure @xmath260 . hence , we can design the measurement period @xmath62 using the case of @xmath261 when @xmath262 . using ( [ fidelity ] ) and ( [ theorem1-bound - solution ] ) , for @xmath251 $ ] we obtain the probability of failure @xmath263 hence , we can design the maximum measurement period as follows @xmath264}{\sqrt{1+\epsilon^{2}}},\ ] ] for @xmath131 ( where @xmath265 ) , we can obtain the same conclusion as that in the case @xmath130 ( where @xmath102 ) . to prove theorem 1 , we first prove two lemmas ( lemma [ lemmab ] and lemma [ lemma2 ] ) . [ lemmab ] for a two - level quantum system with the initial state @xmath266 ( i.e. , @xmath6 ) , the system evolves to @xmath267 and @xmath268 under the action of @xmath269 ( @xmath182 is a nonzero constant ) and @xmath270 , respectively . for arbitrary @xmath271 $ ] , @xmath272 . for the system with hamiltonian @xmath269 , using @xmath273 $ ] and ( [ eq4 ] ) , we obtain the following state equations @xmath274 where @xmath275 . the corresponding solution is as follows @xmath276 where @xmath277 . for the system with hamiltonian @xmath270 , using @xmath273 $ ] and ( [ eq4 ] ) , we obtain the following state equations @xmath278 where @xmath279 . we can obtain the corresponding solution as @xmath280 since @xmath281 , we may first assume @xmath282 . we define @xmath283 and @xmath284 as follows . @xmath285 @xmath286 now we consider @xmath287 $ ] , and obtain @xmath288 it is clear that @xmath289 only when @xmath110 . hence @xmath284 is a monotonically increasing function and @xmath290 hence , we have @xmath291 from this , it is clear that @xmath283 is a monotonically increasing function and @xmath292 hence @xmath293 when @xmath294 $ ] . moreover , it is clear that @xmath295 and @xmath296 is a monotonically decreasing function when @xmath271 $ ] . it is easy to obtain the following relationship @xmath297}\leq z_{t}^{b}|_{t=\frac{\pi}{\sqrt{1+\epsilon_{0}^{2}}}}<z_{t}^{a}|_{t=\frac{\pi}{\sqrt{1+\epsilon_{0}^{2}}}}.\ ] ] hence we can conclude that @xmath272 for arbitrary @xmath271 $ ] . let @xmath298 , and we have the following corollary . [ corollary ] for a two - level quantum system with the initial state @xmath266 ( i.e. , @xmath6 ) , the system evolves to @xmath267 and @xmath268 under the action of @xmath299 ( where @xmath182 is a nonzero constant ) and @xmath300 , respectively . for arbitrary @xmath271 $ ] , @xmath272 . we now present another lemma . [ lemma2 ] for a two - level quantum system with the initial state @xmath266 ( i.e. , @xmath6 ) , the system evolves to @xmath267 and @xmath268 under the action of @xmath189 ( where @xmath102 ) and @xmath301 , respectively . for arbitrary @xmath302 $ ] , @xmath272 . first , we take an arbitrary evolution state ( except @xmath7 ) starting from @xmath6 as a new initial state . for @xmath303 , the initial state can be represented as @xmath304 , where @xmath305 . we have @xmath306 the corresponding solution is as follows : @xmath307 for @xmath189 , we have @xmath308 where @xmath309 and @xmath310 $ ] . from ( [ state42 ] ) , we have @xmath311 from ( [ equation43 ] ) and ( [ bound ] ) , it is easy to obtain the following relationship : @xmath312+\sin\theta_{0}\epsilon \delta t + o((\delta t)^{2})\\ & = \delta t \sin\theta_{0}(\epsilon+\epsilon(0)\sin\varphi_{0})+o((\delta t)^{2}).\\ \end{array}\ ] ] when @xmath305 , @xmath313 . moreover , it is always true that @xmath314 . if @xmath315 and @xmath316 , we have @xmath317 if @xmath318 , we have @xmath319 . according to corollary [ corollary ] and the proof of theorem 2 , we have @xmath320 for @xmath321 , it corresponds to two cases : ( a ) @xmath322 and @xmath323 ; ( b ) @xmath324 and @xmath325 . in the following , we consider the case ( a ) ( for the case ( b ) we have the same conclusion as the case ( a ) ) . since @xmath323 , @xmath326 . using a similar argument to the proof of theorem 2 , we know that for @xmath327 with @xmath328 , the optimal control for the performance index @xmath329 takes a form of bang - bang control @xmath330 or @xmath331 . so we only need to consider a bang - bang strategy . for such a bang - bang strategy as @xmath332 , we have @xmath333 where @xmath334 and @xmath335 $ ] . we can obtain the corresponding solution as @xmath336 now , we consider the limit as @xmath337 and obtain @xmath338-\frac{\bar{\epsilon}\sin\theta_{0}}{\sqrt{1+\epsilon^{2}}}\sqrt{1+\epsilon^{2}}\delta t\\ & \ \ \ + \frac{\cos\theta_{0}}{1+\epsilon^{2}}-\cos\theta_{0}[1-\frac{\epsilon^{2}(\delta t)^{2}}{2}]+\sin\theta_{0}\epsilon \delta t-\frac{\epsilon^{3}}{6}(\delta t)^{3 } \sin\theta_{0}+o((\delta t)^{4})\\ & = \delta t \sin\theta_{0}(\epsilon-\bar{\epsilon})+\sin\theta_{0 } ( \delta t)^{3}(\bar{\epsilon}+\bar{\epsilon}\epsilon^{2}-\epsilon^{3})+o((\delta t)^{4})\\ & > 0 .\\ \end{array}\ ] ] hence , for arbitrary @xmath339 ( @xmath340 ) , we have @xmath341 for @xmath342 , @xmath343 . hence the relationship @xmath344 is always true . we now define @xmath345 and assume that there exist @xmath346 such that @xmath347 . that is , @xmath348 . since @xmath349 is continuous in @xmath92 and @xmath350 , there exists a time @xmath351 satisfying @xmath352 for @xmath353 $ ] . however , we have proven that for any @xmath354 and @xmath337 , @xmath355 , which contradicts @xmath352 for @xmath356 $ ] . hence , we have the following relationship for @xmath302 $ ] @xmath357 now we can prove theorem 1 using lemma [ lemmab ] and lemma [ lemma2 ] . for @xmath358 , using @xmath273 $ ] and ( [ eq4 ] ) , we can obtain the following state equations @xmath359 where @xmath275 . define @xmath360 and @xmath361 , @xmath362 . this leads to the following equation @xmath363 where @xmath364 and @xmath310 $ ] . from ( [ twouncertainties ] ) , we have @xmath365 from ( [ differential ] ) and ( [ bound ] ) , it is easy to obtain the following relationship : @xmath366+\sin\theta_{0}\epsilon \delta t + o((\delta t)^{2})\\ % & = \epsilon(0)\cos\gamma_{0}\sin\theta_{0}\sin\varphi_{0 } \delta % t\\ % & -\epsilon(0)\sin\gamma_{0 } \sin\theta_{0}\cos\varphi_{0 } \delta t+\sin\theta_{0}\epsilon \delta t + o((\delta t)^2)\\ & = \delta t \sin\theta_{0}(\epsilon+\epsilon(0)\sin(\varphi_{0}-\gamma_{0}))+o((\delta t)^{2}).\\ \end{array}\ ] ] when @xmath305 , @xmath313 . moreover , it is always true that @xmath367 . if @xmath315 and @xmath368 , we have @xmath317 if @xmath318 , we have @xmath319 . according to lemma [ lemmab ] and pontryagin s minimum principle , we have @xmath320 for @xmath369 , it must be true that @xmath322 or @xmath331 . we consider @xmath322 ( for @xmath324 we have the same conclusion as @xmath322 ) . moreover , we have @xmath370 or @xmath371 . for @xmath370 , we first employ the fact that @xmath372 is independent on @xmath373 for @xmath374 and @xmath375 since @xmath376 . then , it is easy to prove the relationship @xmath377 using a similar argument to that in the proof of lemma [ lemmab ] . for @xmath371 , we have the same conclusion as in the case @xmath370 . thus , we obtain @xmath378 now using a similar argument to that in the proof of lemma [ lemma2 ] , for arbitrary @xmath379 , we have @xmath380 from ( [ fidelity ] ) , it is clear that the probabilities of failure satisfy @xmath253 . that is , the probability of failure @xmath254 is not greater than @xmath255 for @xmath381 . hence , we can design the measurement period @xmath62 using the case of @xmath261 . using ( [ fidelity ] ) and ( [ bound ] ) , for @xmath381 , we obtain the probability of failure @xmath382 hence , we can design the maximum measurement period as follows @xmath383 this paper proposes a variable structure control scheme with sliding modes for the robust control of two - level quantum systems where an eigenstate is identified as a sliding mode . we present a design method for the control laws based on a lyapunov methodology and periodic projective measurements to drive and maintain this system s state in the sliding mode domain . the key task of the control problem is converted into a problem of designing the lyapunov functions and the measurement period . the lyapunov function can be constructed to define a control law . by using simulation , we obtain an open - loop control to drive the controlled quantum system s state into the sliding mode domain . for different situations of the uncertainties in the system hamiltonian , we give two approaches to design the measurement period , which guarantees control performance in the presence of the uncertainties . this sliding mode control scheme provides a robust quantum engineering strategy for controller design for quantum systems and has potential applications in state preparation , decoherence control , quantum error correction @xcite , etc . future work which can be carried out in this area is listed as follows . 1 ) the physical implementation of the proposed method on specific quantum systems . for example , spin systems in nmr ( see , e.g. , @xcite , @xcite ) may be a suitable candidate to test the proposed approach . 2 ) the extension from two - level systems to multi - level quantum systems : the basic idea of sliding mode control can be extended in a straightforward way to multi - level quantum systems . however , it is much more difficult to obtain an analytical solution for a multi - level system . in @xcite , a numerical result has been obtained for a three - level quantum system to determine the measurement period and more complex systems are worth exploring by numerical methods . 3 ) the extension to dissipative quantum systems governed by the lindblad equation @xcite or described by a stochastic differential equation : for such cases , it is necessary to develop new methods to drive the system into the sliding mode domain since the lyapunov - based control approach does not usually work @xcite . 4 ) the exploration of practical applications for the proposed approaches : the sliding mode may correspond to an eigenstate or a state subspace and the sliding mode design approach could be used in quantum state preparation and protection of encoded quantum information in a subspace . take @xmath46 as the variable and define @xmath384 for @xmath385 , we have the following relationship @xmath386 it is clear from ( [ f_p0 ] ) and ( [ f_p0 ] ) that @xmath387 for @xmath385 and @xmath388 . hence @xmath389 for @xmath390 . when @xmath391 , @xmath392 @xmath393 let @xmath394 and @xmath395 we have @xmath396 for @xmath397 $ ] , @xmath398 is continuous in @xmath399 and we also know that @xmath400 only when @xmath401 . it is easy to check @xmath402 . hence , we know that for @xmath403 $ ] , @xmath404 . that is , for @xmath114 , @xmath405 . from the relationship @xmath405 , we know @xmath406 for @xmath114 . hence we concluded that for arbitrary @xmath407 $ ] , @xmath134 . the authors would like to thank anonymous reviewers and the associate editor for helpful comments and suggestions . f. ticozzi and s. bolognani , on a canonical qr decomposition and feedback control of discrete - time quantum dynamics " , _ proc . of the 19th international symposium on mathematical theory of networks and systems _ , pp.2065 - 2071 , budapest , hungary , july , 2010 . d. dong , c. chen , t.j . tarn , a. pechen and h. rabitz , incoherent control of quantum systems with wavefunction controllable subspaces via quantum reinforcement learning , " _ ieee transactions on systems man and cybernetics - part b : cybernetics _ , vol . 38 , pp.957 - 962 , 2008 .

quantum communication and computation offer technological advances by performing information processing with quantum mechanics @xcite . in order to perform these tasks over large distances , photons are the obvious choice to carry quantum information . however , transmission over large distances is dramatically hampered by attenuation in optical fibers . unlike classical fiber optic communication , amplifiers placed periodically along the transmission channel can not be used to overcome this loss owing to the no - cloning " theorem @xcite . fortunately , quantum repeaters have been proposed to resolve this issue by using entanglement shared between adjacent nodes and joint measurements to create entanglement between the start node and the terminal node @xcite . a crucial element in this scheme is the ability to efficiently store and retrieve single photons using a quantum memory @xcite . to date , the essential ingredients of a photonic quantum memory have been demonstrated using ensembles of ultracold alkali atoms @xcite , ensembles of ultracold atoms in cavities @xcite , warm vapors of alkali atoms @xcite , ultracold single atoms in cavities @xcite , and solid - state systems composed of rare - earth dopants in crystals @xcite . specifically , retrieval efficiencies as high as 73@xmath0 and storage times as long as 3.2 ms have been simultaneously demonstrated using ultracold atoms @xcite . the combination of large optical depths and long ground state hyperfine coherence make ultracold atomic ensembles an attractive platform for optical quantum memories . while there are quantum memory schemes which use probabilistically - generated spin waves from spontaneous raman scattering in atomic ensembles , it has been shown that schemes based on fast ( rates approaching the ghz scale ) single photon or entangled photon sources can provide better performance @xcite . thus , one would ideally like a source of quantum light states that is on - demand and bright , meaning it can produce these photons on fast timescales ( broadband ) in a triggered fashion . in addition , these broadband photons should be indistinguishable so they bunch perfectly on a beamsplitter @xcite . a promising source of such on - demand , non - classical light states are epitaxially grown , gaas - based semiconductor quantum dots ( qds ) @xcite . the short spontaneous emission lifetime of a qd , more than an order of magnitude shorter than that of an alkali atom , and the fact that it is embedded in a robust optoelectronic material make it an attractive candidate as a quantum light source . as an individual two - level system , the qd naturally emits one photon when it is excited and collection of this photon can be very efficient by proper design of the surrounding dielectric @xcite . single photon count rates can in principle approach 1 ghz , with higher rates possible by taking advantage of a purcell enhancement . in addition , qds have been shown to emit polarization entangled photons by means of a cascaded decay @xcite . finally , qd - generated photons can have a high degree of indistinguishability @xcite . aside from these optical properties , the internal spin states of charged quantum dots have received considerable attention for quantum information processing @xcite . spin coherence times as along as 1 @xmath2s have been measured @xcite and because the spin state can be entangled with the polarization of an emitted photon @xcite , charged quantum dots are a natural candidate for a solid - state qubit that interacts strongly with light . while qds are promising sources of single or entangled photons , they must be coupled to a high quality quantum memory in order to be a viable source for long distance quantum information processing . here , we investigate the storage and retrieval of a broadband , qd - generated single photon with a 1 ns lifetime in an ultracold , dense ensemble of @xmath1rb atoms . taking previously measured experimental parameters , we find that the total efficiency ( @xmath3 , where @xmath4 and @xmath5 are the storage and retrieval efficiencies ) can exceed 28@xmath0 for storage and backwards retrieval of a photon with a 1 ns lifetime in a @xmath1rb ensemble with an on - resonance optical depth of 150 . because the bandwidth of the qd photon can approach the excited - state hyperfine splitting in @xmath1rb , this result was obtained by extending the @xmath6-system theory of gorshkov _ et al _ @xcite to a four - level system . using the gradient ascent approach outlined in ref . @xcite , we determine the maximum efficiency as well as the optimal control pulse for photon storage . we find that the control pulse can be easily generated using a 12 mw laser diode in contrast to the broadband , raman - based scheme in ref . @xcite where a mode - locked laser is required . finally , we consider the effects of excess dephasing and spectral wandering of the qd optical transition and show that the memory efficiency remains fairly robust . these results show that with existing technology , qd - generated photons can be reliably interfaced with ultracold atomic ensembles , paving the way for their future use in quantum information as well as opening interesting avenues in the study of hybrid quantum systems . this paper is arranged as follows : in sec . [ sec:3l ] , we briefly review the physics of photon storage and retrieval in a @xmath6-system following the treatment of gorshkov _ et al _ @xcite . in sec . [ sec:4l ] , we extend this treatment to a four - level atom , which is relevant for broadband photon storage in atomic systems with non - negligible hyperfine structure in the excited state . section [ sec : qd ] implements the model to study storage and retrieval of a broadband single photon emitted by a quantum dot using a @xmath1rb ensemble . in sec . [ sec : dephase ] , deleterious effects resulting from imperfect indistinguishability of the quantum dot photon are discussed . in sec . [ sec : highod ] we investigate how the efficiencies increase for very high optical depth . the work is concluded and summarized in sec . [ sec : conc ] and details of the calculations and numerical implementation are discussed in the appendices a and b. appendix c discusses the role of four - wave mixing ( fwm ) in the storage process . photon storage in a three level , @xmath6-type system has been the subject of many articles and reviews @xcite . here , we briefly touch on the main points and restrict the discussion to those schemes described by the theory of gorshkov _ et al _ , namely those based on electromagnetically induced transparency ( eit ) , off - resonant raman , or photon echo interactions @xcite . in that work , these three schemes were shown to yield similar storage efficiencies for the same optical depth of the @xmath6 medium and are in this sense equivalent . however , there are tradeoffs in the actual physical implementation for these schemes which will be discussed later . the basic picture of photon storage in such a system is shown in fig . [ fig : fig1 ] . each atom in the ensemble is composed of two ground states @xmath7 and @xmath8 that can be optically coupled to the excited state @xmath9 . these two transitions can be individually addressed using a sufficient ground state splitting or by selection rules so as to avoid cross - coupling . the storage procedure starts with all of the atoms initialized into @xmath7 . then a quantum field @xmath10 which is to be stored , addresses the @xmath11 transition while a classical control field @xmath12 addresses the @xmath13 transition ( see fig . [ fig : fig1](a ) ) . the control pulse facilitates the transfer of the quantum field to a spatially - dependent coherence of the two ground states ( often called a spin wave " ) @xmath14 as depicted in fig . [ fig : fig1](b ) . the exact physical mechanism that accomplishes this task depends on the scheme . in eit storage , the control pulse dynamically creates a transparency window in the absorption profile of the @xmath11 transition while adiabatically reducing the group velocity of the quantum field to 0 @xcite . in raman storage , the control pulse enables the quantum field to be absorbed into @xmath8 by a two - photon raman transition @xcite . for photon echo storage , the quantum field promotes an atom to the excited state @xmath9 and the control pulse transfers the excitation to @xmath8 by performing a fast @xmath15 pulse . in all cases , the quantum field transfers one atom from @xmath7 to @xmath8 , however there is no knowledge of which atom . thus the excitation is coherently distributed over the entire ensemble and it is this collective behavior that makes the process efficient . the quantum field is now stored in this collective ground state coherence and is therefore sensitive to decoherence processes . these processes , which can include magnetic field fluctuations or atomic motion , set the limit for the spin wave coherence time , and hence , the duration the photon can be stored . notably , the spin wave is more robust against decoherence than many other multi - particle entangled states @xcite . the spin wave can be re - converted into a photon by using another control pulse ( see fig . [ fig : fig1](c)-(d ) ) . et al _ showed that the optimal retrieval is simply the time - reverse of storage ; so - called backwards retrieval " @xcite . this process creates a photon propagating in the opposite direction to its initial propagation . forwards retrieval is also possible , but will not be treated here . for all of these storage schemes , the dynamics of the interaction between the two optical fields and the atoms is described by treating the signal field on the @xmath11 transition quantum mechanically while treating the @xmath13 control field semiclassically . the atomic level structure with relevant energy scales for photon storage in a three - level @xmath6 system is shown in fig . [ fig : fig2 ] . the ensemble of @xmath6 systems is composed of @xmath16 atoms distributed over a volume of length @xmath17 and cross - sectional area @xmath18 with linear density @xmath19 . in this analysis , we ignore the motion of the atoms and thereby restrict the discussion to ultracold ensembles . as was shown in ref . @xcite , the important parameter governing the efficiency of the storage process is the optical depth , defined as @xmath20 where @xmath21 is the single - photon coupling constant for the transition and @xmath22 is the decay of the coherence of the excited state ( for a purely radiative decay with no additional dephasing @xmath23 , the spontaneous emission rate ) . note that the @xmath24 used here and in ref . @xcite is equal to half of the optical depth as usually defined ( @xmath25 where @xmath26 is the resonant absorption cross - section of a single atom and @xmath27 is the number density ) . the classical control field is described by a rabi frequency envelope @xmath28 centered at frequency @xmath29 . as also noted in app . a , this rabi frequency is defined to be 1/2 of the usually defined rabi frequency . the quantum field is described by a slowly varying envelope operator @xmath30 centered at frequency @xmath31 . in addition to @xmath30 , there are two other operators required to describe the dynamics : the polarization operator @xmath32 and the spin - wave operator @xmath33 where @xmath34 are slowly - varying collective atomic operators defined in app . a. it was shown in ref . @xcite that in order to determine normally - ordered quantities such as efficiencies , one can neglect quantum noise operators and treat all dynamical variables as complex numbers . the equations of motion governing @xmath35 , @xmath36 , and @xmath37 are derived by calculating the heisenberg equation of motion for all dynamical variables using the dipole and rotating wave approximations . then , two further approximations are made related to the fact that the quantum field is weak . first , it is assumed that almost all of the atoms remain in the ground state @xmath7 during the whole process . second , only terms to linear order in @xmath35 are retained . under these approximations ( see ref . @xcite and app . a for details ) , the following equations of motion are obtained : @xmath38 these equations of motion use a coordinate system @xmath39 , where @xmath40 is a dimensionless length parameter ( @xmath41 $ ] ) and @xmath42 is the time in a co - moving reference frame . compared to the results of ref . @xcite , eqs . [ eqn:3leome]-[eqn:3leoms ] also include relative dipole moments @xmath43 . these are defined as dipole moments relative to that of the two - level , cycling transition @xmath44 where a measurement of the optical depth would take place . this sets the natural scale of the atom s dipole strength and allows for easy comparison between different physical implementations of levels @xmath7 , @xmath8 , and @xmath9 ( see app . a for details ) . for storage , we would like to take a quantum field with initial envelope @xmath45 ( non - zero on the interval @xmath46 $ ] ) and map it into a spin wave @xmath47 using a classical pulse @xmath12 . the boundary conditions for the dynamical variables are @xmath48 and @xmath49 . we then want to compute the efficiency of this mapping once @xmath50 at @xmath51 . if the envelope of the initial quantum field is normalized ( @xmath52 ) , the storage efficiency @xmath4 is given by @xcite @xmath53 for a given input field and optical depth of the medium , one would like to determine the optimal classical pulse shape so as to maximize the storage efficiency . in ref . @xcite it was shown that in the adiabatic limit , @xmath54 , an analytic solution for @xmath12 could be found and that the efficiency scaled as @xmath55 . for broadband photon storage , the adiabatic limit is not necessarily met and the optimal @xmath12 must be found numerically . _ used a gradient ascent algorithm in ref . @xcite to numerically optimize @xmath4 and found that the results matched the analytical solution in the adiabatic limit . in addition , it was shown that photons beyond the adiabatic regime ( @xmath56 ) could also be stored efficiently by using this optimization technique . this is the approach we will use in the following analysis and now briefly review . because we take a numerical approach , we now restrict the discussion to near - resonant storage schemes and exclude off - resonant raman based storage which would require a much larger computational domain so as to capture the rapidly varying detuning ( @xmath57 ) . gorshkov _ et al _ mentioned in ref . @xcite how one can obtain the optimal control pulse for raman - based storage from the optimal control pulse found for resonant storage , but we do not pursue this here . we do , however , note that the control pulse required for raman storage can be several orders of magnitude more intense than that required for resonant schemes @xcite for the same optical depth @xmath24 and hence the same efficiency . a gradient ascent algorithm simply starts with a trial solution and proceeds to the optimal solution by moving along the gradient of the quantity to be maximized . at each step in the algorithm , the gradient is determined and the control pulse is updated . mathematically , this replacement rule is @xmath58 where @xmath59 is the step size parameter and @xmath60 is the quantity to be maximized . in the case of photon storage , we want to maximize the storage efficiency subject to the constraint that all of the dynamical variables fulfill the equations of motion ( eq . [ eqn:3leome]-[eqn:3leoms ] ) . thus , @xmath60 takes the form @xmath61 + c.c.\right\ } \nonumber \\ & + \int_0^t d\tau \int_0 ^ 1 d\tilde{z}\ , \big\ { \bar{p}_1^*\big [ \left(-\partial_{\tau}+i \delta_g-\gamma \right ) p_1+i \tilde{\mu}_{1s } \omega(\tau ) s \nonumber \\ & + i \tilde{\mu}_{1 g } \sqrt{d \gamma } \mathcal{e } \big ] + c.c . \big\ } \nonumber \\ & + \int_0^t d\tau \int_0 ^ 1 d\tilde{z}\,\big\ { \bar{s}^*\big[-\partial_{\tau}s+ i \left(\delta_g - \delta_s \right ) s \nonumber \\ & + i \tilde{\mu}_{1s}\omega^*(\tau)p_1 \big ] + c.c . \big\ } , \label{eqn:3lj}\end{aligned}\ ] ] where the first term is @xmath4 and lagrange multipliers @xmath62 , @xmath63 , and @xmath64 have been introduced to include the equations of motion . the maximum storage efficiency is found when @xmath60 is stationary with respect to variations in all dynamical variables and @xmath12 . requiring stationarity with respect to variations in @xmath35 , @xmath36 , and @xmath37 results in equations of motion and boundary conditions for the lagrange multipliers @xmath62 , @xmath63 , and @xmath64 . the equations of motion are @xmath65 with boundary conditions @xmath66 and @xmath67 . as pointed out in ref . @xcite , these are exactly the equations of motion and boundary conditions for backwards retrieval . thus , solving these equations will yield the retrieval efficiency , @xmath5 , and the total efficiency @xmath68 for storage followed by backwards retrieval . explicitly , these quantities can be determined using @xmath69 where @xmath70 is the output quantum field . the variation of @xmath60 with respect to variations in @xmath12 ( the gradient ) can be identified from eq . [ eqn:3lj ] as @xmath71}. \label{eqn:3lgrad}\ ] ] thus , the prescription for obtaining the maximum storage efficiency and optimal control pulse for a given input quantum field and optical depth is as follows . first , take a trial control pulse and solve the equations of motion eq . [ eqn:3leome]-[eqn:3leoms ] with the storage boundary conditions to obtain @xmath72 and @xmath14 . then , solve eq . [ eqn:3leombe]-[eqn:3leombs ] using the boundary conditions for backwards retrieval ( note that these equations run backwards in time and space ) to obtain @xmath73 and @xmath74 . now the gradient can be determined from eq . [ eqn:3lgrad ] and the control pulse can be updated using the replacement rule . this process is then repeated until the desired tolerance of the storage ( or total ) efficiency is obtained . in blue ( red ) . solid curves are guides to the eye . ] as an example we consider the case of an idealized rb gas ( @xmath75 , @xmath76 mhz ) storing a quantum - dot photon with @xmath77 where @xmath78 is the heaviside step function and @xmath79 is the spontaneous emission lifetime ( taken to be 1 ns ) . we take both fields to be exactly on resonance @xmath80 . as a function of @xmath24 , the storage and total efficiencies are determined by performing gradient ascent for each value of @xmath24 . details and limitations of the numerical implementation can be found in appendix [ app : num ] . the results of the numerical calculation are shown in fig . [ fig : fig3 ] where the storage ( blue ) and total ( red ) efficiencies are plotted as a function of @xmath24 . both curves reach an asymptotic value of @xmath8196.0@xmath0 for @xmath82 limited by the finite computational domain and the instantaneous rise of the quantum field , which is non - physical but chosen for calculational simplicity ( see app . [ app : num ] ) . determined using gradient ascent . the rabi frequencies are scaled by @xmath76 mhz . ] in addition to determining the efficiencies , it is instructive to see how the optimal control pulse @xmath12 changes as a function of @xmath24 . figure [ fig : fig4 ] shows the optimal control pulse for a few selected values of @xmath24 and it can be seen that the required peak rabi frequency @xmath83 increases with @xmath24 . for these parameters , we find that @xmath84 . furthermore , @xmath12 starts to become much more heavily weighted at the start of the pulse . both of these observations can be understood as the optimal storage scheme slowly changes from a photon echo type storage towards an eit - based storage as the adiabatic limit is approached . lastly , the widths of these control pulses are not ultrafast , i.e. they do not require a mode - locked laser but rather are achievable with more flexible techniques such as direct intensity modulation . the combination of relatively high total efficiencies and realistic control pulses make it seem that storage of a quantum dot photon in an atomic gas like rb is certainly feasible with current technology . importantly , the peak rabi frequency @xmath83 of the control pulse surpasses 50@xmath22 even for relatively low @xmath24 . for such large values of @xmath83 , the more complicated level structure of the storage medium will begin to play a significant role in the dynamics . for instance , the excited - state splittings of the commonly - used @xmath85 transitions of @xmath1rb are around 100 mhz ( @xmath86 ) . as a result , the three - level @xmath6 treatment is insufficient to describe broadband , on - resonance storage and more levels must be taken into account . we now extend the treatment of sec . [ sec:3l ] to include another excited state that can couple to both the control field and the input quantum field as shown in fig . [ fig : fig5 ] . because the maximum rabi frequency of the control pulse can exceed the excited state level splitting @xmath87 , an atom can go from @xmath7 to @xmath8 via @xmath9 or @xmath88 . depending on the dipole matrix elements @xmath43 , these paths can interfere constructively or destructively . following the same procedure as in sec . [ sec:3l ] and adding another polarization @xmath89 results in the following equations of motion ( see app . a for details ) : @xmath90 \label{eqn:4leome}\\ \partial_{\tau}p_1 & = \left(i \delta_g-\gamma \right ) p_1+i \tilde{\mu}_{1s } \omega(\tau ) s + i \tilde{\mu}_{1 g } \sqrt{d \gamma } \mathcal{e } \label{eqn:4leomp1}\\ \partial_{\tau}p_2 & = \left(i\delta_g -i\delta_e-\gamma\right ) p_2 + i \tilde{\mu}_{2s } \omega(\tau ) s \nonumber \\ & + i \tilde{\mu}_{2 g } \sqrt{d \gamma } \mathcal{e } \label{eqn:4leomp2}\\ \partial_{\tau}s & = i \left(\delta_g - \delta_s \right ) s + i\omega^*(\tau)\left [ \tilde{\mu}_{1s}p_1 + \tilde{\mu}_{2s}p_2 \right ] \label{eqn:4leoms}\end{aligned}\ ] ] where new relative dipole moments @xmath91 and @xmath92 have been introduced to represent the couplings to the additional excited state . as the spin wave @xmath14 and the field @xmath93 are unchanged , the definitions of the storage , retrieval , and total efficiencies remain the same . furthermore , the boundary conditions for storage are the same as before except @xmath94 is added . we can also find the four - level version of @xmath60 @xmath95 + c.c.\big\ } \nonumber \\ & + \int_0^t d\tau \int_0 ^ 1 d\tilde{z}\ , \big\ { \bar{p}_1^*\big[(-\partial_{\tau}+i \delta_g-\gamma ) p_1+i \tilde{\mu}_{1s } \omega(\tau ) s \nonumber \\ & + i \tilde{\mu}_{1 g } \sqrt{d \gamma } \mathcal{e } \big ] + c.c . \big\ } \nonumber \\ & + \int_0^t d\tau \int_0 ^ 1 d\tilde{z}\ , \big\ { \bar{p}_2^*\big [ \left(-\partial_{\tau}+i\delta_g - i\delta_e-\gamma \right ) p_2 \nonumber \\ & + i \tilde{\mu}_{2s } \omega(\tau ) s+ i \tilde{\mu}_{2 g } \sqrt{d \gamma } \mathcal{e } \big ] + c.c . \big\ } \nonumber \\ & + \int_0^t d\tau \int_0 ^ 1 d\tilde{z}\,\big\ { \bar{s}^*\big[-\partial_{\tau}s+ i \left(\delta_g - \delta_s \right ) s \nonumber \\ & + i\omega^*(\tau)\left ( \tilde{\mu}_{1s}p_1+\tilde{\mu}_{2s}p_2 \right ) \big ] + c.c . \big\ } , \label{eqn:4lj}\end{aligned}\ ] ] in order to obtain the optimal control and efficiencies . by requiring that @xmath60 is stationary with respect to variations in @xmath35 , @xmath36 , @xmath96 , and @xmath37 , we find the equations of motion @xmath97 for the lagrange multipliers and their boundary conditions ( @xmath66 and @xmath98 ) . the gradient along @xmath12 is now modified to @xmath99 . \label{eqn:4lgrad}\end{aligned}\ ] ] it is clear that if @xmath100 , all of the expressions from sec . [ sec:3l ] are recovered . in the limit where @xmath101 , then @xmath96 can be adiabatically eliminated and the results of sec . [ sec:3l ] are again recovered . thus , the extension to a four level system reproduces the results of the three level case in the appropriate limits . to gain some insight into the ramifications of including a fourth level in an intermediate regime of @xmath87 , we directly compare the results of the gradient ascent algorithm applied to the three level case to that for two scenarios of the four level case . for all calculations , we again store a resonant quantum dot photon with @xmath77 using a resonant control pulse ( @xmath80 ) . for the parameters of the storage medium we take @xmath76 mhz , @xmath102 , and @xmath103 . the first four level scenario ( 4l+ ) we consider is @xmath104 mhz and @xmath105 while the second scenario ( 4l- ) has the same excited state energy splitting but @xmath106 . the optimal control pulses resulting from the gradient ascent optimization are plotted in fig . [ fig : fig6 ] for each scenario . not only does the inclusion of a fourth level dramatically change the optimal control pulse , but the sign of the relative dipole moment also has a large effect . more importantly , the computed storage ( total ) efficiencies are 73.6@xmath0 , 77.6@xmath0 , and 43.5@xmath0 ( 63.4@xmath0 , 65.7@xmath0 , and 26.3@xmath0 ) for the 3l , 4l+ , and 4l- scenarios respectively . if one naively uses the optimal control for 3l in the 4l+ case , the storage ( total ) efficiency is only 56.5@xmath0 ( 48.4@xmath0 ) , while for the 4l- case 20.8@xmath0 ( 4.8 @xmath0 ) is obtained . from this analysis it is clear that additional excited states must be taken into account and a careful choosing of those levels is required for any implementation of broadband photon storage where the peak rabi frequency of the control pulse is comparable to the excited state splitting . . the slight asymmetry about @xmath107 is due to the finite accuracy of the numerics ( see appendix [ app : num ] ) . ] another question one could ask is if there is an optimal choice of the optical detuning @xmath108 in the four level system . to address this , we again apply the gradient ascent optimization to the 4l+ configuration described above , but we vary the detuning @xmath109 from @xmath110 to @xmath111 in steps of @xmath112 . the storage efficiency is plotted in fig . [ fig : fig7 ] as a function of @xmath109 . surprisingly , the efficiencies are smallest near the mid - point between @xmath9 and @xmath88 at @xmath113 , while the maxima occur near @xmath114 and @xmath115 . nonetheless , the change in storage efficiency over the entire range is within @xmath116 , and there is not much to be gained or lost by changing the detuning at least for this set of @xmath43 . we have also verified that varying the detuning has a small effect ( @xmath117 variation ) when using the dipole moments for the relevant @xmath1rb transitions . with these mathematical and numerical tools in hand , we can address storage of a quantum dot photon using the actual atomic levels of @xmath1rb . because of the long hyperfine ground - state coherence time and the ability to create cold , dense gases , @xmath1rb has proven to be a natural choice for implementation of quantum memory schemes . indeed , both the 5@xmath118p@xmath119 and 5@xmath118p@xmath120 excited state manifolds corresponding to the @xmath121 and @xmath85 optical transitions have been used for quantum memory applications . before assigning @xmath1rb hyperfine states @xmath122 to the various states of the four level model , the problem of frequency matching of a qd transition to the @xmath85/@xmath121 transitions at 780/795 nm must be addressed . the most commonly studied self - assembled qds are composed of in@xmath123ga@xmath124as islands embedded in a gaas matrix , which are made to emit light in the 900 - 1000 nm band . in such qds , the so - called wetting layer " ( a thin quantum well ) defines a barrier near 850 nm , below which no confined states exist in the dot . therefore , the photons produced by these qds can not be stored in @xmath1rb , but cs transitions ( 852/895 nm ) are close to within reach and @xmath125yb@xmath126 ( 935 nm ) has a transition compatible with qds @xcite . nonetheless , efficient quantum frequency conversion techniques have been demonstrated using qds @xcite such that the single photon produced by a qd could be frequency translated to another wavelength without destroying its quantum characteristics . such techniques could bridge the frequency gap between in@xmath123ga@xmath124as qd transitions and those of @xmath1rb . in addition , there are qds composed of gaas embedded in algaas and ingaas qds embedded in algaas which have been shown to emit in the 780 nm - 795 nm region @xcite . the optical properties of these qds are much less well - known compared to standard ingaas / gaas qds and detailed investigations are ongoing . thus , either use of quantum frequency conversion techniques or proper material choices permit the study of broadband qd photon storage and retrieval in @xmath1rb . notably , photons from a qd have been made to interact with @xmath1rb atoms in a recent experiment @xcite . in this case the atoms acted as a passive medium whose dispersion near the @xmath85 transition was used to reduce the group velocity of the qd photons . in this present work , we are interested in actively manipulating the atoms to controllably store and retrieve the qd photon . we proceed by determining which hyperfine states to assign to the four levels of the model . there are two hyperfine ground - state manifolds in @xmath1rb with total angular momentum @xmath127 and @xmath128 separated by 6.835 ghz . the choice of which ground - states within these manifolds to use is determined by several factors . firstly , cross - coupling of the control field and the quantum field can lead to unwanted processes such as four - wave mixing ( fwm ) , which is described in detail in app . [ app : fwm ] . in order to mitigate cross - coupling , it is advantageous to use ground - states that are widely separated in energy and couple to the excited state through perpendicular polarizations . this becomes especially important at high optical depth and large control fields @xcite . one can choose ground states from the same @xmath129 manifold , but then the energy splitting is limited to what can be obtained by zeeman shifting the levels . for the purposes of qd photon storage , we choose one ground state in @xmath127 and one in @xmath128 . furthermore , since both states must be coupled to a common excited state , the difference between the @xmath130 values must be between -2 and 2 . this sets the first condition for choosing the ground states . as discussed in sec . [ sec : intro ] , the long coherence between the hyperfine ground states of @xmath1rb make it very attractive as a quantum memory . this long coherence is usually limited experimentally by magnetic field fluctuations that cause small zeeman shifts of the @xmath130 levels . therefore , choosing @xmath130 levels that are common mode to magnetic field fluctuations for the ground states @xmath7 and @xmath8 is crucial to achieving long storage times . notably , it is not possible to find two states in the same @xmath129 manifold that are common mode to magnetic field fluctuations . however , there are two non - degenerate choices for the ground states from different @xmath129 manifolds that are common mode to magnetic field fluctuations ( to first order ) and can be coupled to the same excited state . these levels are the clock states @xmath131 and @xmath132 and the states @xmath133 and @xmath134 ( @xmath135 and @xmath136 is a degenerate choice , but is not magnetically - trappable ) . for each pair of ground states , there are two optical transitions that are of interest ; the @xmath85 transition to the @xmath137p@xmath120 states near 780 nm and the @xmath121 transition to the @xmath137p@xmath119 states near 795 nm . we will investigate these transitions separately in the following sections . the excited states of the @xmath85 transition of @xmath1rb are composed of four hyperfine manifolds @xmath138 0 , 1 , 2 , and 3 . because the two ground states must share a common excited state , only states in @xmath139 and @xmath140 are eligible . for the pair of ground states @xmath133 and @xmath134 , there are then two possible excited states ; @xmath141 and @xmath142 as shown in blue in fig . [ fig : fig8 ] . rb that are eligible for implementation of a quantum memory . ] in the following , we always work on resonance with one of the excited states ( @xmath143 ) and let the value of @xmath87 change signs . that is to say , @xmath7 and @xmath8 couple resonantly to @xmath9 and are detuned to @xmath88 . two choices to assign @xmath7 and two choices for @xmath9 result in four unique configurations for each set of four levels . to determine which configuration will yield the highest storage efficiency , it is instructive to consider the results of sec . [ sec:3l ] . because the storage efficiency depends critically on the optical depth @xmath24 , and the effective optical depth on the @xmath11 transition is @xmath144 , it would seem optimal to choose the configuration where this relative dipole moment is the largest . a secondary concern is that @xmath145 should be large to avoid extremely intense control pulses . the four relative dipole moments are listed in table [ tab : d2_1_mu ] ( see app . a for details ) , where it can be seen that the largest moment is for the @xmath146 transition . .relative dipole moments @xmath43 for @xmath1rb @xmath85 transitions with ground states @xmath133 and @xmath134 @xcite . [ cols="^,^,^",options="header " , ] [ tab : d1_2_eff ] . ] configuration 2 yields comparable efficiencies to the best configuration for @xmath85 , with only a minor increase . if we perform gradient ascent for this configuration but neglect the additional excited state ( configuration 3 ) , we obtain almost identical efficiencies . comparison of the two optimal control pulses ( fig . [ fig : fig11 ] ) shows that they are almost exactly the same , except the four level control has a small modulation at the frequency of the excited state splitting . if we use the three level control on the four level system ( configuration 4 ) , we obtain almost exactly the same efficiencies although the control pulse is much simpler to generate . this analysis simply implies that the large excited state splitting almost completely nullifies the effects of the additional excited state , as anticipated in sec . [ sec:4l ] . consequently , the efficiencies we find in this case match well with those predicted by fig . [ fig : fig3 ] if @xmath147 is used as the effective optical depth of the storage transition . up to this point , our analysis has assumed that the photons emitted by a qd are completely indistinguishable , which is an important property in photon - based quantum information processing @xcite . however , due to their dynamic solid - state environment the photons produced by excitons in quantum dots may not have fourier transform - limited spectra @xcite . this leads to an imperfect photon indistinguishability between subsequently emitted photons . up to now , the highest measured two photon visibility is @xmath148 @xcite , corresponding to a linewidth that is @xmath149 times larger than the transform - limited linewidth @xmath150 . while it is possible to approach @xmath151 by careful sample selection and resonant excitation , in this section the effect of excess linewidth broadening on the storage efficiency is calculated . the excess linewidth of quantum dot generated photons can be caused by two distinct physical processes . the first , sometimes referred to as spectral wandering , is a slow process that causes changes in the carrier frequency of the photons from shot to shot . this simply means that subsequent photons have slightly different carrier frequencies due to changes in the qd environment on timescales longer than the spontaneous emission lifetime @xmath79 . the second , referred to as pure dephasing , is a perturbation of the qd s energy levels on timescales shorter than @xmath79 . this leads to a time - dependent phase within each shot . in a time - averaged spectral measurement , these effects can both produce a lorentzian lineshape with a linewidth greater than the fourier limit . here , we treat these cases separately but show that they cause exactly the same effect on storage and retrieval efficiencies in a quantum memory for a given amount of added linewidth @xmath152 . in the case of spectral wandering , the waveform of each photon remains @xmath77 but the carrier frequency @xmath153 is drawn from a probability distribution @xmath154 . the time - averaged spectrum of these photons is then an integral over all realizations of @xmath153 @xmath155 where @xmath156 is the transform limited spectrum centered about @xmath153 . if @xmath154 is a lorentzian distribution with linewidth @xmath152 , the time - average spectrum is a lorentzian with linewidth @xmath157 . the average storage and total memory efficiencies can be calculated similarly by determining how the efficiencies depend on @xmath158 and then integrating over all realizations . we numerically calculate the storage and total efficiencies as a function of @xmath159 using the optimal control pulse found for configuration 4 of @xmath85 storage with @xmath160 . all other parameters are kept fixed ( @xmath161 , @xmath102 , @xmath162 ns ) and the results are plotted in fig . [ fig : fig12 ] . storage with ground states @xmath132 and @xmath131 as a function of @xmath159 . all other parameters are kept fixed . ] using @xmath163 and @xmath164 , the efficiencies as a function of @xmath152 can be obtained by @xmath165 where @xmath166 is taken to be a normalized lorentzian distribution of width @xmath152 centered at @xmath167 . performing this integration results in the plot shown in fig . [ fig : fig13 ] . storage with ground states @xmath132 and @xmath131 as a function of @xmath152 resulting from spectral wandering . ] notably , spectral wandering causing a factor of 2 increase in the total linewidth ( @xmath168 ) roughly leads to a factor of 2 reduction in the storage and total efficiencies . while the initial drop is steep , the total efficiency remains above 5@xmath0 even if the total linewidth is more than 7 times the transform limit , showing that the storage process is relatively robust to the effects of spectral wandering . the effects of fast , pure dephasing can also be calculated numerically . in this case , the waveform remains at the carrier frequency @xmath153 but the slowly varying amplitude is modified to @xmath169 which includes a time - dependent phase @xmath170 . in the simplest model of pure dephasing , @xmath170 is driven by a markovian langevin force @xmath171 characterized by @xmath172 and @xmath173 with diffusion constant @xmath174 . this force will cause phase diffusion , resulting in a lorentzian time - averaged spectrum with total linewidth @xmath175 . the effect on the storage and total efficiency can be determined by calculating the efficiencies for several trajectories of @xmath170 and averaging . storage with ground states @xmath132 and @xmath131 as a function of @xmath176 resulting from pure dephasing ( red , open circles ) . uncertainties in @xmath177 are given by the standard deviation of the ensemble . for comparison , results from spectral wandering ( fig . [ fig : fig13 ] ) are shown as blue lines . ] using the same configuration as for the spectral wandering investigation , we performed storage and retrieval calculations for 100 phase trajectories for several values of @xmath178 . the average storage and total efficiencies are plotted as red circles in fig . [ fig : fig14 ] , with the uncertainties given by the standard deviation . for comparison , results from spectral wandering ( fig . [ fig : fig13 ] ) are shown as blue lines . evidently , both spectral wandering and pure dephasing lead to the same reduction in efficiency for the same amount of @xmath152 , indicating that the efficiencies depend only on the time - averaged spectrum and not on the physical mechanism causing an excess linewidth . a recent experiment by sparkes _ et al . _ has demonstrated an optical depth of 1000 in an ultracold @xmath1rb gas @xcite using spatial and temporal dark spots . we therefore investigate how the efficiencies increase for such high optical depth . if the optical depth is increased to @xmath24=500 in the optimization using the clock ground states ( @xmath131 and @xmath132 ) and configuration 4 of @xmath85 storage , the storage ( total ) efficiency is increased to 51 @xmath0 ( 34 @xmath0 ) . if instead configuration 2 of @xmath121 storage with clock ground states is used , the optimization yields a storage ( total ) efficiency of 82 @xmath0 ( 76 @xmath0 ) . notably , the efficiencies are not dramatically increased for @xmath85 storage while @xmath121 storage is much more promising . there are two factors that contribute to the worse performance of @xmath85 ; the small excited - state splitting and the large relative dipole element between the storage state @xmath8 and the unwanted excited state @xmath88 . naively , one would expect an increase in efficiency according to fig . [ fig : fig3 ] , but that is only true if the additional excited state does nt play a large role in the storage process . however , as the optical depth increases so does the required control pulse amplitude and therefore the coupling to unwanted excited states . for the case of @xmath85 storage , the extra excited state can not be neglected and limits the achievable efficiencies even for large optical depth . on the other hand , for @xmath121 storage of a 1 ns photon at @xmath179 we find very high efficiencies , comparable to the efficiency of 80 @xmath0 reported in ref . @xcite for much lower bandwidth photons . in conclusion , we have calculated the efficiency with which a quantum dot generated single photon can be stored and retrieved from an optically - thick @xmath1rb ensemble . our calculations take into account the multi - level structure of @xmath1rb by extending the standard three - level model of an atomic ensemble quantum memory . using an optical depth of 150 , the storage ( total ) efficiency can reach 46@xmath0 ( 28@xmath0 ) for a photon resulting from the 1 ns spontaneous excitonic decay in a quantum dot . importantly , this storage can be performed using control pulses obtained from a diode laser rather than requiring pulse energies only achievable with ultrafast , mode - locked laser sources . increasing the optical depth , for example by using bose - condensed ensembles , an optical cavity , or advanced trapping techniques can increase the efficiencies to near unity for storage on the @xmath121 transition . in addition , we have studied the effects of spectral diffusion and pure dephasing of the quantum dot generated photons on the storage efficiency and shown that a factor of 2 increase in the time - averaged photon linewidth roughly leads to a factor of 2 reduction in the efficiency . thus , storage and retrieval of single photons from a quantum dot in an @xmath1rb ensemble is feasible with demonstrated experimental parameters even in the presence of non - ideal properties of the quantum dot . integration of quantum dot sources with atomic ensemble quantum memories may lead to high - speed quantum networks for communication or distributed computation as well as entanglement between collective atomic degrees of freedom and the spin of an electron or hole confined in the quantum dot . we thank a. v. gorshkov , m. fleischhauer , and k. srinivasan for helpful discussions . m. t. r. acknowledges support from a marie curie international incoming fellowship . this work was also supported by the nccr quantum science and technology . the equations of motion eq . [ eqn:4leome]-[eqn:4leoms ] ( and by extension eq . [ eqn:3leome]-[eqn:3leoms ] ) are derived by considering an ensemble of @xmath16 motionless , four - level atoms interacting with a quantum field and a classical field closely following the treatment of ref . @xcite , whose notation we also adopt for the most part . the hamiltonian of this system can be expressed as @xmath180 where @xmath181 and the interaction between the light fields and the atoms in the dipole approximation is @xmath182 here , the operators @xmath183 change the internal state of the @xmath184th atom from @xmath185 to @xmath186 and @xmath187 is the dipole moment of an atom for the @xmath188 transition . the total electric field @xmath189 is composed of a classical field @xmath190 and a quantum field @xmath191 . the @xmath192-propagating classical field with polarization orientation @xmath193 can be written as @xmath194},\ ] ] where @xmath195 is the envelope function and @xmath196 is the carrier frequency . we have assumed that the classical pulse propagates with a group velocity of @xmath197 , which is valid if almost all of the atomic population remains in state @xmath7 . the quantum field is taken to be a sum of modes @xmath198 centered about frequency @xmath199 with polarization @xmath200 and cross - sectional area @xmath18 @xmath201 where h.c . denotes the hermitian conjugate . in this treatment , the quantum field only drives the @xmath11 and @xmath202 transitions while the classical control field drives the @xmath13 and @xmath203 transitions . as shown in app . [ app : fwm ] , cross - coupling of the control field leads to a four - wave mixing process which can reduce the storage efficiency , but it is safe to neglect for the parameter regime considered here . applying this assumption and making the rotating wave approximation allows eq . [ eq : appa_hint ] to be written as the sum of the interaction with the classical field @xmath204 and the quantum field @xmath205 where @xmath206 are the rabi frequencies associated with the classical field and @xmath207 are the couplings to the quantum field . note that the rabi frequencies @xmath208 are defined differently compared to the standard definition such that @xmath209 . in order to treat the ensemble as a continuous density distribution , we divide the ensemble into thin slices of thickness @xmath210 such that the quantum field can be taken to be constant over this range while also ensuring that the number of atoms in a slice @xmath2111 . then , we define slowly varying operators @xmath212 using these operators , we can rewrite @xmath213 as @xmath214 + \int d\omega\ , \hbar \omega \hat{a}^{\dag}_{\omega } \hat{a}_{\omega}\end{aligned}\ ] ] and the two parts of the interaction as @xmath215 \\ \mathcal{h}_{\textrm{int , q } } & = -\hbar \int_0^l dz \ , n(z ) \big\{g_{1 g } \hat{\mathcal{e}}(z , t ) \hat{\sigma}_{1g}(z , t ) \nonumber \\ & + g_{2 g } \hat{\mathcal{e}}(z , t ) \hat{\sigma}_{2g}(z , t)+ \textrm{h.c . } \big\},\end{aligned}\ ] ] where @xmath19 is the linear number density of atoms along the length of the ensemble and we have assumed the cross - sectional area of the field @xmath18 matches that of the atomic cloud . the dynamics are determined by finding the heisenberg equations of motion for the operators . using the commutation relations @xmath216 & = \delta(\omega-\omega ' ) \\ \left [ \hat{\sigma}_{\alpha \beta}(z , t),\hat{\sigma}_{\lambda \rho}(z',t ) \right]&= \frac{\delta(z - z')}{n(z ) } \big\{\delta_{\beta \lambda}\hat{\sigma}_{\alpha\rho}-\delta_{\alpha \rho}\hat{\sigma}_{\lambda\beta } \big\}\end{aligned}\ ] ] one obtains @xmath217 \\ \partial_t \hat{\sigma}_{g1 } & = i \delta_g \hat{\sigma}_{g1}+i \omega_{1s } \hat{\sigma}_{gs } \nonumber \\ & + i \hat{\mathcal{e } } \big [ g_{1g}(\hat{\sigma}_{gg}-\hat{\sigma}_{11})-g_{2 g } \hat{\sigma}_{21 } \big]\\ \partial_t \hat{\sigma}_{g2 } & = i ( \delta_g-\delta_e ) \hat{\sigma}_{g2}+i \omega_{2s } \hat{\sigma}_{gs } \nonumber \\ & + i \hat{\mathcal{e } } \left[g_{2g}(\hat{\sigma}_{gg}-\hat{\sigma}_{22})-g_{1 g } \hat{\sigma}_{12 } \right]\\ \partial_t \hat{\sigma}_{s1 } & = i \delta_s \hat{\sigma}_{s1}+i\omega_{1s } ( \hat{\sigma}_{ss}-\hat{\sigma}_{11 } ) \nonumber \\ & - i\omega_{2s}\hat{\sigma}_{21}+i g_{1 g } \hat{\mathcal{e } } \hat{\sigma}_{sg}\\ \partial_t \hat{\sigma}_{s2 } & = i ( \delta_s-\delta_e ) \hat{\sigma}_{s2}+i\omega_{2s } ( \hat{\sigma}_{ss}-\hat{\sigma}_{22 } ) \nonumber \\ & - i\omega_{1s}\hat{\sigma}_{12}+i g_{2 g } \hat{\mathcal{e } } \hat{\sigma}_{sg}\\ \partial_t \hat{\sigma}_{12 } & = -i \delta_e \hat{\sigma}_{12}+i(\omega_{2s}\hat{\sigma}_{1s}-\omega_{s1}\hat{\sigma}_{s2 } ) \nonumber\\ & + i(g_{2 g } \hat{\mathcal{e}}\hat{\sigma}_{1g}-g_{g1}\hat{\mathcal{e}}^{\dag}\hat{\sigma}_{g2})\\ \partial_t \hat{\sigma}_{gs } & = i(\delta_g-\delta_s)\hat{\sigma}_{gs}+i(\omega_{s1}\hat{\sigma}_{g1}+\omega_{s2}\hat{\sigma}_{g2 } ) \nonumber\\ & - i\hat{\mathcal{e}}(g_{1g}\hat{\sigma}_{1s}+g_{2g}\hat{\sigma}_{2s})\\ \partial_t \hat{\sigma}_{gg } & = i ( g_{g1}\hat{\mathcal{e}}^{\dag}\hat{\sigma}_{g1}-g_{1g}\hat{\mathcal{e}}\hat{\sigma}_{1 g } \nonumber \\ & + g_{g2}\hat{\mathcal{e}}^{\dag}\hat{\sigma}_{g2}-g_{2g}\hat{\mathcal{e}}\hat{\sigma}_{2 g } ) \\ \partial_t \hat{\sigma}_{ss } & = i ( \omega_{s1 } \hat{\sigma}_{s1}-\omega_{1s } \hat{\sigma}_{1s } \nonumber \\ & + \omega_{s2 } \hat{\sigma}_{s2}-\omega_{2s } \hat{\sigma}_{2s } ) \\ \partial_t \hat{\sigma}_{11 } & = i ( \omega_{1s}\hat{\sigma}_{1s}-\omega_{s1}\hat{\sigma}_{s1 } ) \nonumber \\ & + i(g_{1g}\hat{\mathcal{e}}\hat{\sigma}_{1g}-g_{g1}\hat{\mathcal{e}}^{\dag}\hat{\sigma}_{g1 } ) \\ \partial_t \hat{\sigma}_{22 } & = i ( \omega_{2s}\hat{\sigma}_{2s}-\omega_{s2}\hat{\sigma}_{s2 } ) \nonumber \\ & + i(g_{2g}\hat{\mathcal{e}}\hat{\sigma}_{2g}-g_{g2}\hat{\mathcal{e}}^{\dag}\hat{\sigma}_{g2})\end{aligned}\ ] ] where the time and spatial dependencies have been neglected for brevity . these equations can be reduced considerably by making one simplifying assumption ; the quantum field is weak . the first consequence of this assumption is that almost all atoms remain in @xmath7 for the duration of the dynamics . secondly , we keep only terms that are linear in @xmath218 @xcite . under these assumptions , the equations of motion are reduced to @xmath217 \\ \partial_t \hat{\sigma}_{g1 } & = i \delta_g \hat{\sigma}_{g1}+i\omega_{1s } \hat{\sigma}_{gs}+i g_{1 g } \hat{\mathcal{e } } \\ \partial_t \hat{\sigma}_{g2 } & = i ( \delta_g-\delta_e ) \hat{\sigma}_{g2}+i\omega_{2s } \hat{\sigma}_{gs}+i g_{2 g } \hat{\mathcal{e } } \\ \partial_t \hat{\sigma}_{gs } & = i ( \delta_g-\delta_s)\hat{\sigma}_{gs}+i(\omega_{s1 } \hat{\sigma}_{g1}+\omega_{s2 } \hat{\sigma}_{g2}).\end{aligned}\ ] ] we now introduce the polarization operators @xmath32 and @xmath89 as well as the spin wave operator @xmath33 . in addition , we move to a new coordinate system @xmath39 where @xmath42 is the time in a co - moving reference frame and @xmath40 is a dimensionless length . inserting these definitions into the equations of motion yields @xmath219 where a factor of @xmath220 has been absorbed into @xmath218 and we introduced the optical depths @xmath221 . we have also assumed that both polarizations decay at the same rate @xmath22 and that @xmath222 is real . finally , for notational convenience we compare all transition dipole moments @xmath223 to that of the two - level cycling transition ( @xmath224 ) . in this way , we make the substitutions @xmath225 and @xmath226 in the equations of motion . this enables easy comparison between different state configurations and also sets @xmath227 to where it will be measured experimentally . it also enables @xmath208 to be connected to a light intensity through the two - level relation @xmath228 , where @xmath229 is the saturation intensity . using the value of @xmath229 for the cycling transition @xcite , the peak power can be related to the peak rabi frequency @xmath83 by @xmath230w_o^2 ( \omega_m/\gamma)^2 \label{eq : powerrabi}\end{aligned}\ ] ] where @xmath59 is the wavelength of the transition and @xmath231 is the @xmath232 waist of a gaussian beam . the relevant values for the @xmath1rb cycling transition @xmath233 of the @xmath85 line have been inserted to obtain eq . [ eq : powerrabi ] . similarly , the pulse energy can be determined by @xmath234 wm@xmath235w_o^2 \int_0^t d\tau |\omega(\tau)/\gamma|^2 $ ] . from ref . @xcite , we have @xmath236 where @xmath237 is the reduced dipole moment for the @xmath85 transition . to put the @xmath121 relative dipole moments @xmath43 in units of @xmath44 , the values in tables [ tab : d1_1_mu ] and [ tab : d1_2_mu ] have been multiplied by the factor @xcite @xmath238 further , as discussed in sec . [ sec:3l ] of the main text , the operators can be treated as complex numbers and their associated quantum noise can be neglected because we are interested in computing the expectation values of normally - ordered operators . making the substitutions and dropping the operator notation yields the equations of motion as found in sec . [ sec:3l]-[sec:4l ] , @xmath90 \\ \partial_{\tau}p_1 & = \left(i\delta_g-\gamma \right ) p_1+i \tilde{\mu}_{1s } \omega(\tau ) s + i \tilde{\mu}_{1 g } \sqrt{d \gamma } \mathcal{e}\\ \partial_{\tau}p_2 & = \left(i\delta_g -i\delta_e-\gamma\right ) p_2 + i \tilde{\mu}_{2s } \omega(\tau ) s \nonumber \\ & + i \tilde{\mu}_{2 g } \sqrt{d \gamma } \mathcal{e } \\ \partial_{\tau}s & = i \left(\delta_g - \delta_s \right ) s + i\omega^*(\tau)\left [ \tilde{\mu}_{1s}p_1 + \tilde{\mu}_{2s}p_2 \right].\end{aligned}\ ] ] numerical solutions of the equations of motion are obtained using the method of finite differences . the time - space grid is composed of 9@xmath239 ( 3000 by 3000 ) points and the domain of @xmath240 is chosen such that the optimized control pulses tend toward 0 at @xmath241 . this ensures that the drop to @xmath242 outside of the domain is smooth . the gradient ascent algorithm is implemented using a dynamic step size @xmath59 to guarantee quick convergence . the step size is determined using an inexact line search such that @xmath59 is initialized to a large value ( 1000@xmath22 ) at each step of the ascent and the increase of @xmath243 and its gradient are calculated at the next step . if this step does not meet the wolfe conditions @xcite , the step size is reduced geometrically until they are satisfied . the optimization proceeds until @xmath243 has not been increased by more than 0.001@xmath243 compared to the average of the three previous values . this tolerance was estimated by considering the errors resulting from the numerical integration . errors in the reported efficiencies are @xmath2441 @xmath0 which was determined by examining the variation in efficiencies for perturbations of the time - space grid . for the parameter set @xmath80 , @xmath102 , @xmath245 mhz , the gradient ascent optimization took approximately 20 minutes on a standard computer . the pure dephasing process was simulated by first obtaining @xmath171 on a finite grid using ( pseudo-)random numbers drawn from a normal distribution . then , @xmath170 was calculated using @xmath246 . averaging over many trajectories yielded the correct diffusion behavior @xmath247 ^ 2 \rangle = d_{\phi } \tau$ ] . the extracted @xmath4 and @xmath243 shown in fig . [ fig : fig14 ] were obtained by averaging the results of 100 different phase trajectories for each value of @xmath174 . for the heaviside model and for the three - level ( 3l ) model ( @xmath248 ps ) of the qd photon . ] the slowly varying photon waveform @xmath77 is not physical due to the infinitely sharp rise of @xmath78 . a more realistic model might include the non - zero temporal width of the excitation pulse or fast loading of the qd from another excited state . for the latter , @xmath249 for a three - level model with instantaneous excitation of an ancillary excited state that loads the qd with rate @xmath250 . for comparison , we have implemented this form of @xmath251 with @xmath252 ps @xcite using configuration 4 of @xmath85 storage with @xmath253 . we obtain roughly the same efficiencies ( @xmath254 , @xmath255 ) and peak control rabi frequency ( @xmath256 ) as found in table [ tab : d2_2_eff ] . both input photons and control pulses are plotted in fig . [ fig : figb_1 ] . because the curves and efficiencies are extremely similar , we conclude that the infinitely sharp rise of @xmath78 in the simple photon waveform does not dramatically influence the results and is therefore a sufficiently representative choice . in the preceding analysis , cross - coupling of the control field @xmath12 to the @xmath11 transition was neglected due to the large ground - state hyperfine splitting of @xmath2576.835 ghz . because the control field can be quite strong and the optical depth quite large , this cross - coupling can lead to detrimental effects that reduce the storage and retrieval efficiencies in practice @xcite . of course , a proper choice of ground states and optical polarizations can eliminate cross - coupling completely . for example , choosing ground states whose @xmath130 values differ like @xmath258 and @xmath259 and using circularly - polarized light allows cross - coupling to be neglected . on the other hand , configurations such as @xmath260 and @xmath261 are coupled to the excited state by light of the same polarization . in this case , the control beam can off - resonantly drive the @xmath11 transition as shown in fig . [ fig : figc_1 ] and coherently generate a stokes field @xmath262 in a four - wave mixing ( fwm ) type of process @xcite . the stokes field can interfere with the spin - wave created from storage of the quantum field @xmath35 and lead to reduced storage and retrieval efficiencies . in ref . @xcite , the stokes field was taken into account theoretically and was shown to match experimental results quite well . following the same approach here , we obtain the stokes - modified equations of motion for a three - level system @xmath263 where @xmath262 is the stokes field and the off - resonant interaction @xmath264 has been adiabatically eliminated , leaving an effective coupling between @xmath262 and @xmath37 . in addition , this interaction induces time - dependent light shifts of + @xmath265 and -@xmath266 for states @xmath9 and @xmath7 respectively . using these equations of motion with the optimized control field for on - resonance storage of a qd - generated photon with configuration 3 of @xmath121 storage with ground states @xmath131 and @xmath132 ( fig . [ fig : fig11 ] ) , one obtains the same storage and total efficiencies as in table [ tab : d1_2_eff ] . this result indicates that the effect of fwm in this storage scheme is negligible . another way to determine the relative effect of fwm is to consider the ratio of the last two terms in eq . [ eqn:3lfwms ] . the ratio of the fwm term to the normal @xmath267 term should roughly scale as @xmath268 , which is approximately 2@xmath269 for typical parameters considered here . compared to typical parameters found in ref . @xcite where the onset of fwm effects was measured , this ratio is about three orders of magnitude smaller so fwm can be safely neglected . 60ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop _ ( , , ) link:\doibase 10.1038/nature08812 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevlett.81.5932 [ * * , ( ) ] link:\doibase 10.1038/nature07127 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.83.33 [ * * , ( ) ] link:\doibase 10.1038/35106500 [ * * , ( ) ] link:\doibase 10.1038/nphoton.2009.231 [ * * , ( ) ] link:\doibase 10.1038/nature04315 [ * * , ( ) ] link:\doibase 10.1038/nature06670 [ * * , ( ) ] link:\doibase 10.1038/nphys1153 [ * * , ( ) ] link:\doibase 10.1038/nphys1152 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.98.183601 [ * * , ( ) ] link:\doibase 10.1038/nature04327 [ * * , ( ) ] link:\doibase 10.1038/nature05136 [ * * , ( ) ] link:\doibase 10.1038/nphoton.2010.30 [ * * , ( ) ] link:\doibase 10.1038/nature09997 [ * * , ( ) ] link:\doibase 10.1038/nature09662 [ * * , ( ) ] link:\doibase 10.1038/nature09719 [ * * , ( ) ] link:\doibase 10.1038/nphys2324 [ * * , ( ) ] link:\doibase 10.1103/physreva.76.050301 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1038/nphoton.2007.46 [ * * , ( ) ] link:\doibase 10.1038/nphoton.2007.227 [ * * , ( ) ] link:\doibase 10.1038/nphoton.2009.287 [ * * , ( ) ] link:\doibase 10.1063/1.3615051 [ * * , ( ) ] link:\doibase 10.1038/nature04446 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.103.217402 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.103.167402 [ * * , ( ) ] link:\doibase 10.1038/nnano.2012.262 [ * * , ( ) ] link:\doibase 10.1126/science.1126074 [ * * , ( ) ] link:\doibase 10.1038/nature06472 [ * * , ( ) ] link:\doibase 10.1038/nphys2078 [ * * , ( ) ] link:\doibase 10.1126/science.1173684 [ * * , ( ) ] link:\doibase 10.1103/physrevb.75.081302 [ * * , ( ) ] link:\doibase 10.1038/nature11577 [ * * , ( ) ] link:\doibase 10.1038/nature11573 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physreva.76.033805 [ * * , ( ) ] link:\doibase 10.1103/physreva.77.043806 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.75.457 [ * * , ( ) ] link:\doibase 10.1103/physreva.75.011401 [ * * , ( ) ] link:\doibase 10.1103/physreva.76.033804 [ * * , ( ) ] link:\doibase 10.1103/physreva.80.062330 [ * * , ( ) ] link:\doibase 10.1038/nphoton.2010.221 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.107.083602 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.109.147405 [ * * , ( ) ] link:\doibase 10.1103/physrevb.66.153316 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1063/1.3478232 [ * * , ( ) ] link:\doibase 10.1038/nphoton.2011.16 [ * * , ( ) ] link:\doibase 10.1103/physreva.83.063823 [ * * , ( ) ] @noop link:\doibase 10.1103/physrevlett.105.260502 [ * * , ( ) ] @noop * * , ( ) @noop ( ) , @noop _ _ ( , , ) @noop * * , ( ) link:\doibase 10.1103/physreva.79.013806 [ * * , ( ) ] link:\doibase 10.1080/09500340903159511 [ * * , ( ) ]

ground - based doppler radial velocity surveys and space - based photometric surveys have established that planets are very common , and perhaps ubiquitous , around sun - like stars . in particular , analysis of detections by the nasa _ kepler _ mission show that planets with radii of 1 - 2 earth radii ( r@xmath2 ) are far more common than larger planets @xcite . comparisons between estimates or upper limits on mass from doppler radial velocity measurements and radii from _ kepler _ also suggest that the densities of planets ( at least planets on short - period orbts ) smaller than 1.5m@xmath2 are consistent with a `` rocky '' ( silicates + metals ) composition , while larger planets have an additional low - molecular weight ( e.g. hydrogen ) envelope @xcite . however , the data are insufficiently precise to inform as to the exact composition of these planets and regardless there are degeneracies in mass and radius with different bulk compositions . the photospheres of dwarf stars on the main sequence are broadly representative of the material from which the star and any circumstellar disk formed . additional information about planet composition might therefore be inferred by determining elemental abundances in the host - star photosphere . this connection is supported by the observation that the abundances of refractory elements in the solar photosphere are , to a large extent , mirrored by that in primitive chondritic meteorites , the latter widely used as an analog of the primordial building blocks of planets ( e.g. , * ? ? ? to a lesser extent , these abundances are reflected in the bulk composition of the planets such as earth , but with some important exceptions . the two most abundant heavy elements in the galaxy and sun ( and presumably planet - forming disks ) are carbon ( c ) and oxygen ( o ) . in the cool interstellar medium ( ism ) , star - forming regions , and molecular cloud cores , c and o are present primarily as the co molecule ( e.g. , * ? ? ? because of the unit stochiometry of the co molecule , whether c or o is in excess controls which element is available to form other compounds and thus the chemistry of the gas and solids that condense or equilibrate with the gas . the solar ratio of these two elements is presently estimated at @xmath3 ( * ? ? ? * see also asplund et al . 2009 ) , however this value is not necessarily universal . the two elements differ in their predicted pathways of stellar nucleosynthesis and c / o has a strong positive correlation with metallicity @xcite . motivated by condensation scenarios for the solar nebula @xcite and observations of widely varying c / o among solar - type stars @xcite , @xcite proposed that c / o controls the abundance of water in planetary systems . during condensation from a hot gas with c / o @xmath4 , excess oxygen forms refractory silicate minerals and eventually h@xmath5o . however , in systems with c / o near a critical value ( about 0.88 ) , no o for water is available after silicate condensation and planets accrete from `` dry '' material . at still higher c / o , the deficiency in o causes carbides and graphite to become more stable and replace silicates in the condensation sequence ; thus `` carbide planets '' with very un - earth - like mineralogies , interior structures , and atmospheres might form around stars with c / o @xmath0 . many analyses of the spectra of solar - type stars obtained for galactic chemical evolution ( gce ) studies @xcite and doppler radial velocity surveys found that a significant fraction have c / o @xmath0 @xcite . a series of subsequent works have predicted the final bulk composition of planets by modeling ( i ) the sequential condensation of elements and formation of planetesimals from a cooling gas disk with an initial composition set by host star abundances ; and ( ii ) the accretion of those planetesimals into planets via integration of the dynamical equations of motion . these scenarios have been applied to specific systems where the photospheric abundances of elements have been estimated , variously predicting water - rich `` ocean planets '' and `` carbon planets '' lacking any water @xcite . one question at the foundation of these works is whether c / o really varies widely among galactic disk stars in the solar neighborhood and ever approaches unity . abundances of c and o are more uncertain than for many other elements because of the limited number of usable absorption lines , confusion with lines of other elements , and non - lte corrections . recent studies have revised the c / o of nearby solar - type stars , including those that host known planets , downwards @xcite . these newer studies find no cases where c / o @xmath6 , the threshold where carbide minerals are expected to first form . @xcite has pointed out that the occurrence of c - rich systems in surveys of solar - type stars conflicts with the lack of known c - rich ultracool t dwarfs , as the latter are easily distinguished by their spectra @xcite . a second issue is the assumption that the solids in planetary systems condense from a cooling gas of stellar composition in a circumstellar disk . equilibrium condensation can explain some solids in primitive chondrites , widely considered analogs to the building blocks of planets , but these solids constitute only a small fraction of those meteorites . calcium aluminum inclusions ( cais ) and amoeboid olivene agregates , the most abundant refractory condensates in chondritic meteorites , have volume abundances @xmath7% and typically @xmath8% @xcite . the dominant constituents of chondritic meteorites are chondrules and fine - grained matrix , and there is compelling evidence that these are the product of incomplete sublimation and chemical alteration of pre - existing solids , e.g. older generations of solids and even pre - solar dust from the parent molecular cloud @xcite . this processing occurred under conditions that were very different from a gas of solar composition , i.e. an oxygen - rich environment and dust - to - gas ratios @xmath9 @xcite . this environment could have been created by gravitational settling to the mid - plane of the protoplanetary disk . because of this , the composition of planeteismals reflects the composition of dust grains in the molecular clouds cores that collapse to form stars and planets , plus subsequent thermal processing at mid - plane conditions . ism dust was , in turn , the product of processing in the interstellar medium involving cycling of order @xmath10 times between denser , cooler clouds , where it accreted mantles of atoms , and the hotter intercloud phase , where shock - heated ions and high - energy radiation ( uv , x - rays , cosmic rays ) sputtered atoms from the grains @xcite . an equally intriguing possibility is that the compositions of planetesimals and/or planets can be _ inferred _ from differential measurements of abundances in the photospheres of stars . this approach is motivated by the detection of small differences in stellar abundances of elements that are more refractory ( higher condensation temperature @xmath11 ) and thus more likely to be incorporated into planets ( e.g. , * ? ? ? * ) . these differences could arise either from the accretion of planetesimals or planets onto the star , after the dissipation of disk gas , or the sequestration of solids into planets , and accretion of the dust - poor gas onto the star . this leads to a prediction that these differences correlate with the presence or properties of planetary systems , and provides a potential `` short - cut '' to discovering planets as well as estimating their chemical composition . however , subsequent analyses of larger samples of stars does not support such a correlation @xcite , and alternative explanations for these differences should be considered . in this work , i critically examine the hypothesis that c / o varies significantly among neighboring stars in the galactic disk ( and hence the host stars of doppler - detected exoplanets ) from both theoretical ( sec . [ sec.gce ] ) , observational ( sec . [ sec.mdwarfs ] ) , and cosmochemical ( sec . [ sec.grains ] ) perspectives . for the first i develop a gce model to predict the abundance of c and o over the history of the galactic disk in the vicinity of the sun . for the second , i use the spectra obtained in m dwarf surveys to place strict upper limits on the occurrence of c - rich single m dwarfs among recent large surveys . for the third , i combine a simple model of interestellar dust evolution with uv measurements of the depletion of heavy elements in the ism to determine what controls the c / o ratio and by how much it may vary . finally , i develop an alternative explanation that can explain the observed elemental relative abundance differences between solar - type stars , without resorting to possible signatures of planet formation ( sec . [ sec.gasdust ] ) . galactic chemical evolution ( gce ) models predict changes in the abundances of elements and isotopes with time in both stars and the ism @xcite . they combine yields of nucleosynthetic products in stellar winds and ejecta with the stellar initial mass function ( imf ) and a prescription for star formation rate to estimate the production of elements and isotopes . these are integrated with descriptions of the flow of heavy elements back into the ism and subsequent incorporation in long - lived low - mass stars and remnants from high - mass stars . production varies with both stellar progenitor mass and metallicity , thus the changing metallicity of new generations of stars must be tracked in a gce . carbon and oxygen are synthesized in stars more massive than 1.5m@xmath12 and injected into the interstellar medium through winds and type ii supernovae ( sn ) : type i sn are predicted to contribute @xmath13% @xcite . gce models generally predict a positive trend of c / o with metallicity ( and hence time ) as well as variation between stellar populations with different chemical histories ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? these trends are broadly observed within disk stars and between the thin , thick - disk and bulge populations , and even between massive and metal - poor irregular dwarf galaxies @xcite . at least two effects are thought to contribute to this correlation : ( i ) the short main - sequence lifetime of massive stars , which are major contributors to @xmath14-elements including c and o , compared to intermediate mass asymptotic giant branch ( agb ) progenitors , which contribute comparatively more to the c budget ; ( ii ) increased mass loss and c yield from younger , more metal - rich agb stars at the expense of o. mass - loss rates depends on metallicity through its effect on the opacity of outer stellar atmospheres and the impact on their structure . the implications of the abundance dependence of c / o yields for the chemical evolution of galaxies has been previously pointed out ( e.g. , * ? ? ? * ; * ? ? ? the c / o of bulge and disk populations diverge due to a combination of differences in the stellar imf , wind - driven mass loss in the former and the accretion of primordial metal - poor gas in the former . the flatter imf of bulge stars implies a higher relative production of o from more abundant massive stars . addition ( loss ) of mass from a star - forming system will increase ( decrease ) the amount of star formation required to arrive at a given metal abundance and therefore the c / o at that abundance . one feature of the statistics of stellar metallicities of the galactic disk in the neighborhood of the sun is the relatively narrow dispersion and paucity of metal - poor stars . this is often called the `` g dwarf problem '' and translates into a flat age - metallicity relation for the most of the history of the galactic disk , at least in the solar neighborhood . it is actually not a problem for gce models if addition of metal - poor gas to the disk is admitted to balance stellar nucleosynthesis ( e.g. , * ? ? ? if c / o is strongly correlated with metallicity then a flat age - metallicity relation might mean that c / o evolves very little ( e.g. , * ? ? ? * ) . to re - visit the question of the evolution of c / o in the solar neighborhood , the abundances of the two elements were calculated on an isotope - by - isotope basis using the model described in detail in the appendix . the model simulates the production of these isotopes and release into the ism in sn explosions , winds from massive stars and agb stars , sequestration of isotopes into long - lived low - mass stars , and addition of metal - poor gas by infall onto the galactic disk . the ism is described by two components ; an inter - cloud medium which produces molecular clouds but no stars , and a star - forming giant molecular cloud component . yields for intermediate- and massive stars are taken from a variety of recent sources . best - fit parameters describing the timescale of exponentially - declining gas infall on the disk , the initial metallicity of the infalling gas , the age of the stellar population at the solar galactocentric radius , the index of the power - law describing the imf for massive stars , and the power - law relationship between star formation rate and gas surface density were found by a monte - carlo markov chain analysis : the observational constraints were the age of the sun , the present mass surface density of stars , stellar remanants , and gas at the solar galactocentric radius , solar metallicity ( here taken to be c+o ) , the current metallicity of the ism , and its intrinsic standard deviation . the prediction evolution of c / o is compared to the solar value from @xcite in fig . [ fig.co ] . after initial transients that die away after a few hundred myr , there is only moderate evolution in c / o during most of the history of the galactic disk . the c / o ratio rises from 0.5 to about 0.65 over the first 6 gyr in response to increasing input from stars of lower mass and higher metallicity , as described above , and declines slightly thereafter . the predicted value at the time of the sun s formation is 0.64 , withinin @xmath15 of the @xcite value . the formal error in the predicted value , based on the standard deviation of the mcmc chain after removal of the `` burn in '' , is only 0.003 , but this value does not reflect the dominant source of uncertainty in these calculations the nucleosynthetic yields . although the model is unable to exactly reproduce the solar values of c and o abundances , correctly reproducing these using gce been always been challenging . @xcite also found that satisfactorily reproducing a `` flat '' c / o places constraints on the remnant or `` cut - off '' mass at the center of a sn progenitor which is not injected into the ism . figure [ fig.co2 ] plots the predicted c / o evolution vs. predicted [ o / h ] , again compared to the sun . also plotted are the values found by @xcite for galactic disk stars ( filled points ) and bulge or halo stars ( open points ) and converted to absolute values assuming a solar c / o of 0.55 . the predicted trend is perfectly consistent with the disk values if an offset of about 16% with the predicted vs. measured solar value is artificially removed . the absence of data for disk stars at low [ o / h ] is a consequence of the age - metallicity relationship : comparatively little time elapsed ( @xmath8 gyr according to the model ) and few stars formed in this interval . the predicted constancy of c / o contrasts with previous predictions for an increase in c / o with metallicity / time . for example , both @xcite and @xcite predict a positive slope of @xmath16 dex / dex in [ c / o ] vs. [ o / h ] , and thus a c / o of about 0.06 at [ o / h ] = -1 . since the model and the best - fit parameter values presented here are similar to those of previous works the most likely cause of this difference is in the particular nucleosynthetic yields that were used . the yields from agb stars have been significantly revised @xcite but in the model these are minority contributors to the c and o budgets . indeed , if there contribution is completely removed , c / o still changes little and the predicted solar c / o is slightly closer to the @xcite value . instead , the predictions of this model probably stand out from previous results because of the dominant contribution of massive stars to both c and o. since these stars have very short lives compared to the chemical evolution of the disk , there is no effect from delayed contribution , and the c / o rapidly approaches a steady - state value . previous works have pointed to favorable comparisons of model predictions with observations , but the latter are _ combined _ data on bulge , halo , and thick- and thin - disk populations , therefore assuming a common origin for these populations that may not exist . for example , the [ c / o ] of thin disk stars presented by @xcite and @xcite show no correlation with [ o / h ] and it is only when bulge and thick disk stars are added that such a correlation appears . the apparent consistency between the predicted and observed trends of [ c / o ] vs. [ o / h ] @xcite belies the fact that the models were tuned for the ( thin ) disk population in the solar neighborhood and should not be compared with other stellar populations . it is indisputable that metal - poor bulge / halo stars have low c / o compared to the solar neighborhood @xcite . the model presented here can not , and was not designed to , explain these populations , but their abundances might be a result of a flatter imf , early loss of gas and/or truncation of star formation . but bulge / halo stars are rare in the solar neighborhood and because their lines are very weak they are usually avoided by planet searches using the doppler radial velocity method . c / o may also vary at some level within the solar galactocentric annulus because the sources of nuclides ( stars and stellar clusters ) are discrete and mixing by rotational shear and epicyclic dynamics is not completely efficient . however , gce is occurring on a timescale much longer ( few gyr ) than the rotational time of the galaxy at 8 kpc ( @xmath17250 myr ) which governs the rate of mixing , and thus this limits the magnitude of such heterogeneities . . the @xcite solar value is plotted as well as values for galactic disk stars ( filled points ) and bulge / halo stars ( open points ) from @xcite and tied to the solar value.,width=317 ] in solar - type stars , the absorption lines of c and o are weak and c / o has little effect on the overall spectrum . @xcite pointed out that cooler stars , i.e. brown dwarfs , offer more obvious constraints on the occurrence of high c / o systems . the spectra of m dwarfs , with effective temperatures @xmath18 @xmath19k , also include absorption bands of o - containing molecules , e.g. tio , vo , and caoh , which are sensitive to the available o abundance @xcite , and hence c / o . at c / o @xmath9 , tio is absent and the schwan bands of c@xmath5 , cn , and ch should appear . such spectra are characteristic of carbon stars , evolved stars in which dredge - up of carbon - righ interiors has occurred ( agb stars ) . if the initial mass function of star formation is chemically invariant , a limit on the occurrence of c - rich m dwarfs is also a limit on c - rich g dwarfs . dwarf carbon ( dc ) star have been identified @xcite but these probably accreted carbon - rich gas from a present or former evolved companion @xcite . overall , dc stars are uncommon @xcite . carbon enhanced metal - poor ( cemp ) stars in the galactic halo appear are also rare overall but more prevalent among ( initially ) metal - poor systems because less accreted c is required to increase c / o @xcite . any tally of _ intrinsically _ carbon - rich dwarfs must remove these interlopers or be considered an upper limit . i considered three spectral indices which are continuum - normalized measures of the emission in specific bands @xcite : cah , which is the mean of the cah2 ( 6814 - 6846 ) and cah3 ( 6960 - 6990 ) indices ; tio5 ( 7126 - 7135 ) ; and caoh ( 6230 - 6240 ) . values of these temperature- and gravity - sensitive indices are highly and positively correlated among solar - metallicity m dwarfs . carbon - rich stars can be identified by relatively weak bands of oxygen - containing tio and caoh , i.e. high tio5 or caoh indices , for a given strength of the non - oxygen - containing cah . however , indices of metal - poor subdwarfs or `` extreme '' subdwarfs exhibit similar behavior @xcite . to guide discrimination between c - rich m dwarfs and metal - poor stars based on indices , synthetic spectra were generated using the bt - settl version of the phoenix models @xcite . four cases were considered : a solar - metallicity star , a star with [ m / h ] = 0 but c and o adjusted so that c / o=1 , a metal - poor subdwarf ( sd ) with [ m / h ] = -1 , and an extreme subdwarf ( esd ) with [ fe / h ] = -2 . the last two cases also had an alpha - element enhancement [ @xmath14/fe ] = + 0.4 . effective temperatures over the entire m dwarf and late - k dwarf range in steps of 100 k were considered . all cases had @xmath20 , except for the esd which had @xmath21 . the visible and far - red portions of the spectra are plotted and compared in fig . [ fig.spectra ] . the bands used to compute the cah , tio5 , and caoh indices are indicated . the current bt - settl line lists do not actually include caoh , but the bandpass also includes the tio @xmath22 line which is also sensitive to c / o . k , @xmath20 , and solar abundances ( * ? ? ? * red line ) , metal - poor subdwarf ( [ fe / h ] = -1 , [ @xmath14/fe ] = 0 , green line ) , extreme subdwarf ( [ fe / h ] = -2 , [ @xmath14/fe ] = + 0.4 , @xmath21 , blue line ) , and solar - metallicity star except some o as c such that c / o=1 ( red line ) . bottom : ratios of the three non solar - metallicity spectra described above to the solar metallicity spectrum.,width=317 ] as expected , in the c / o=1 case ( black lines in fig . [ fig.spectra ] ) the tio bands at around 6235 and 6700 are dramatically weakened relative to the solar case ( red line ) while those of cah are essentially unchanged . this is also true for the esd case ( but not the subdwarf case ) . the differential response of tio5 and cah is the basis for the `` @xmath23 '' parameter developed by @xcite . however , the c / o=1 and esd cases differ markedly at 4700 - 5300 and below 4500 because fe lines are very weak in the esd case . there are also difference at @xmath24 ( fig . [ fig.spectra ] ) . absorption by h@xmath5o over a broad wavelength range centered at 1.9@xmath25 m is also weaker in the c / o=1 case , but this would be difficult to ascertain from the ground . index values were calculated for the four cases at each @xmath18 . i searched for carbon - rich m dwarfs in three spectroscopic catalogs : the conch - shell spectroscopic catalog of nearby , bright m dwarfs @xcite , the sloan digital sky survey ( sdss ) catalog of m dwarfs constructed by @xcite , and the lamost spectroscopic catalog of m dwarfs in @xcite . stars in the conch - shell catalog were selected based on brightness ( @xmath26 ) , parallaxes or proper motions consistent with main sequence status , and colors . although some sets of stars in the catalog were selected based on optical and infrared colors consistent with `` normal '' m dwarfs , this criterion was relaxed for another set , thus admitting any c - rich dwarfs with peculiar colors . the selection for `` red '' ( @xmath27 ) stars is not relaxed , and this inevitably eliminates some metal - poor early m - type dwarfs , although not necessarily c - rich examples . figure [ fig.indices ] plots the tio5 vs. cah indices for conchshell stars . the red line is the best - fit quadratic to the locus and the black points and lines are the values computed from phoenix bt - settl spectra . i identified 18 conch - shell stars with a tio5 band that is significantly ( @xmath28 ) weaker ( larger index ) compared to a best - fit second - order fit of tio5 as a function of cah . when calculating the significance of a deviation , an intrinsic dispersion of 0.032 in tio5 was added in quadrature to the formal measurement errors . none of these 18 stars had indices that deviate to the extent predicted for a c / o=1 star . the most deviant star is pm i20050 + 5426/gj 781/wolf 1130 , with a tio5 index that is 0.17 above the best - fit locus value . this was previously identified as an active m1.5 subdwarf @xcite with a metallicity based on an infrared spectrum of @xmath29 @xcite . interestingly , @xcite identified wolf 1130 as a single - lined spectroscopic binary ( @xmath30 d ) and proposed that the unseen companion was a 0.3m@xmath12 helium white dwarf . a hubble space telescope imaging spectrograph ( stis ) spectrum of the star ( fig . [ fig.stis ] ) exhibits some tio absorption and no diagnostic carbon - star features . there is also no indication of a wd companion , although this would be consistent with an advanced age and hence low uv luminosity of any such object . spectra of the other candidates indicate they are metal - poor stars , or have systematic errors . ) . some atomic lines and molecular bands are labeled ; the red dashed lines mark the locations of some cn and c@xmath5 schwan bands observed in carbon stars , but not in this star.,width=317 ] i examined the molecular indices for two much larger samples of m dwarfs from the sdss and lamost @xcite . figure [ fig.dr7 ] shows the tio5 and caoh indices vs. cah index for 70,841 m dwarfs with spectra in data release 7 of the sloan digital sky survey @xcite . the grey scale is linearly related to the density of stars in the tio5- or caoh - vs . cah plots . the black dashed line is the best - fit to the stellar locus , and the red , aquamarine , dashed purple , and black solid lines are the predictions from the bt - settl models for the solar metallicity , subdwarf , extreme subdwarf , and c / o=1 cases , respectively . stars for which either index value is significantly ( @xmath28 ) higher than the best - fit polynomial to the stellar locus , and within @xmath15 of or above the predicted c / o=1 locus and are plotted as open points . the 30 stars where both indices satisfy these criteria are plotted as filled points . inspection of the sdss spectra of these 30 stars found that all are consistent with template spectra of solar - metallicity or metal - poor stars and do not have the featuers expected of c / o=1 stellar atmospheres . figure [ fig.lamost ] is the analogous set of plots for 67,082 candidate m dwarfs from a pilot survey of the large area multi - object fiber spectroscopic telescope ( lamost , * ? ? ? * ) . the intrinsic scatter in index values , after correcting for measurement errors , is much larger for this sample , perhaps due to systematic errors . only two stars have both tio5 and caoh values significantly above the locus , and within 1@xmath31 of the predicted c / o line . however , the index values of these stars are all @xmath32 and probably spurious . the paucity of high c / o candidates in the lamost survey compared to the sdss dr7 sample is undoubtedly due to the larger uncertainties in the indices . based on the conch - shell sample alone , high c / o @xmath16 stars constitute less than @xmath33 of m dwarfs with 95% confidence , a stricter constraint than the limits placed by @xcite based on spectroscopy of 66 solar - type stars . the null result from the dr7 sample places an an even more stringent upper limit of @xmath34 at 99% confidence . it is possible that manual screening performed by @xcite to remove corrupted spectra could have excluded carbon - rich stars , but at the resolution and signal - to - noise of the spectra , the differences that distinguish c / o=1 spectra from the solar case would not make them overwhelmingly abberant . theoretical studies linking the abundances in the photospheres of host stars to the composition of rocky planets generally assume that the building blocks of planetesimals condensed directly from a cooling gaseous disk of the same composition , but there is accumulating meteoritic evidence that primitive solids in the solar system were by and large _ not _ produced by de nova condensation from a gas of solar composition . instead , this material was the outcome of partial thermal processing of pre - existing solids under high dust / gas ratios and oxygen abundance ( fugacity , @xmath35 ) brought about by growth and settling of grains to the nebular mid - plane and/or inward migration of water - rich ( and hence o - rich ) planetesimals @xcite . this evidence includes ( i ) the survival of pre - solar grains and isotopic anomalies inherited from the molecular cloud and older generations of stars @xcite ; ( ii ) the gradual variation in the abundance of elements with condensation temperature in primitive meteorites , which contradicts the sharp cutoff predicted by equilibrium condensation and which can not be explained by radial transport in the disk @xcite ; ( iii ) the retention of volatile sodium and sulfur and lack of isotopic mass fractionation for potassium and silicon in chondrules @xcite ; and ( iv ) molybdenum and tungsten depletions in refractory inclusions @xcite . the equilibrium oxygen fugacities of minerals in many meteorites also suggest high ambient @xmath35 , but some could also be explained by alteration on the parent bodies of the meteorites @xcite . dust growth and settling to the midplane of a disk is predicted to occur in @xmath36-@xmath37 yr , depending on the intensity of turbulence in the disk @xcite . dust settling , as well as growth , might be observed via its effect on the spectral energy distribution of a disk @xcite . tentative evidence for significant dust settling and depletion from the upper layers of the disks of t tauri stars ( ages of @xmath38 yr ) has been presented @xcite , but unambiguous detection of settling is challenging @xcite . if settling occurs faster than the viscous accretion time of a disk ( @xmath39 yr , * ? ? ? * ) then grains will experience high temperatures only in a dust - rich environment . if the precursor material of planetesimals is dust , rather than gas , then to a large extent the composition of interstellar grains governs the composition of rocky exoplanets . interstellar dust begins its existence as condensates in the cooling envelopes and winds of agb and red giant branch stars but these grains are subsequently and completely altered by many cycles of erosion and formation of mantles in the ism . erosion takes place in the lower - density , higher - temperature inter - cloud phase of the ism , principally by sputtering by ions heated by the passage of supernova shocks as well as uv photons . condensation takes place onto surviving grains that are incorporated into the denser , cooler cloud phase of the ism . the cycling time between these two phases ( @xmath40 yr , set by the cloud lifetime ) is short compared to the mean time since formation in a circumstellar wind ( @xmath41 yr ) , hence the bulk elemental composition of ism grains is set by the balance between condensation and sputtering @xcite . the bulk composition of dust can be inferred by measuring the depletion of elements from the gas phase , i.e. using the strength of uv absorption lines along different lines of sight through the ism to some suitable background source . following @xcite , the equations of motion of the depletions @xmath42 in the cloud ( @xmath43 ) and inter - cloud ( @xmath44 ) medium of any particular element can be written as : @xmath45 and @xmath46 respectively . in these equations , @xmath47 , @xmath48 , @xmath49 , and @xmath50 are the rate of mixing from the intercloud to the cloud medium , the rate of mixing from the cloud to the intercloud medium , the rate of grain destruction in the intercloud medium , and the rate of grain growth from molecular cloud gas , respectively . the nucleosynthetic rate of production of an element is much slower than any of the rates of the formation of clouds , dissipation of clouds , destruction by sputtering in the inter - cloud phase , and growth condensation in the cloud phase , respectively , and is ignored here . the steady state solutions are : @xmath51 and @xmath52 thus , the steady - state abundances are governed by only two parameters , the amount of growth in the clouds @xmath53 , and the amount of erosion between clouds : @xmath54 . i estimated parameter values for some elements using the data compiled by @xcite . i set @xmath55 and @xmath56 to @xmath57_1 $ ] ( maximum depletion ) and @xmath58_0 $ ] ( minimum depletion ) , respectively , and solved for @xmath54 and @xmath53 . figure [ fig.rates ] plots these two parameters . three elements with data in @xcite are not shown : to explain the abundances of p , cl , and zn in the context of this model requires _ negative _ destruction in the intercloud medium . this may be an artifact of photoionization since the abundance of these elements is estimated from their singly ionized forms . the comparatively low condensation rates of kr , c , and o reflect the volatility of these elements . the behavior of o deviates strongly from that of the refractory elements , perhaps because the primary carrier of o is not silicates but a much more volatile substance such as water ice . likewise c does not behave as a refractory such as graphite @xcite . this suggests that c is present as relatively volatile organic matter @xcite . kr is not expected to condense and the non - zero value of @xmath53 may be a consequence of measurement errors or departures from solar relative abundances @xcite . the differences in the parameters for c and o manifest themselves as a modified c / o ratio in interstellar dust with respect to the total ( gas+dust ) abundance . in the case of the nominal rates inferred from the data of @xcite , the degree of depletion from cold molecular cloud gas shows that the c / o of the dust is 0.92 times that of the bulk ism , i.e. slightly more oxygen - rich than the current bulk ism value . although o is depleted more rapidly from interstellar grains , it is also accreted more rapidly . one caveat of this estimate for the dust c / o is that the gas - phase depletion of c is more uncertain than other elements because there are few suitable absorption lines @xcite . the c / o could also vary between locations as a result of varying @xmath54 and/or @xmath53 . @xcite estimated the residence time @xmath59 in the warm inter - cloud phase of the ism as @xmath60 yr assuming that it is set by the timescale for shocking by sn and subsequent cooling and collapse into clouds . this was consistent with an ism mass fraction in the warm phase of @xmath1710% . however , the mass in the warm phase is probably comparable to the dense molecular h@xmath5 ( cloud ) phase @xcite and thus the residence time in the inter - cloud phase is similar to the cloud lifetime @xmath61 yr @xcite . over this time , dust grains may experience @xmath62 sn shocks before becoming incorporated into molecular clouds . in contrast , @xmath53 would be expected to vary only to the extent that the lifetimes of molecular clouds varies . i estimated the sensitivity to variations in these rates by multiplying each of the parameter ratios by varying factors @xmath63 $ ] . figure [ fig.contour ] plots the predicted variation of c / o with contours intervals of 10% . the unadjusted parameter values , which predict [ c / o]@xmath64 ( heavy contour ) , are at unit value abcissa and ordinate ( circle ) . variation in the efficiency of grain destruction in the intercloud medium ( @xmath54 ) have more effect on dust c / o than variation in the efficiency of grain growth in molecular clouds ( @xmath53 ) . this difference is a consequence of the larger dispersion ( a ratio of @xmath65 ) in the removal of o vs. c during grain destruction compared to the dispersion in the incorporation of o vs. c during grain growth ( a ratio of @xmath66 , fig . [ fig.rates ] ) . the very volatile behavior of o compared to the other elements is presumably because some of it is incorporated as water ice mantles around dust grains . this model predicts that large ( factors of two ) variation in dust lifetime produced by different shock and uv conditions in the intercloud medium will produce modest ( @xmath1720% ) variation in bulk dust c / o , with dust in the vicinity of massive sn progenitors more carbon - rich . by the arguments presented above , variation in dust c / o could generate diversity in the composition of planeteismals , but this will be limited by the extent that water ice is retained on dust grains ( see below ) . a pronounced correlation between the metallicity of the host star and the occurrence of giant planets was discovered early in the history of exoplanet science @xcite . increasingly precise measurements of the relative abundances of elements in stellar photospheres have permitted more nuanced investigations of correlations with planet occurrence ( e.g. , * ? ? ? @xcite found that the solar composition was anomalous compared to 9 out of a sample of 11 solar `` twins '' and that the sun is deficient ( @xmath67 dex ) in refractory elements ( condensation temperature @xmath68 k ) compared to more volatile elements . they proposed that this was a signature of the sequestration of those elements in rocky planets and cores of gas giants . @xcite also found a statistical correlation between planets and a deficit of refractory elements among larger samples of solar - type stars in doppler surveys . if differences in relative photosphere abundances are a signature of rocky planets or giant planet cores , this would provide a short - cut to planet detection and even a means to probe the composition of the planets themselves @xcite . on the other hand , @xcite find no clear correlation between the slope of the relative abundance vs. @xmath69 and the presence or absence of planets . one factor that may contribute to these contrasting results is the removal of galactic chemical evolution trends , e.g. decreasing [ @xmath14/fe ] with increasing metallicity , by @xcite . moroever , @xcite found no trend with @xmath69 in the relative abundances of members of the solar - age , solar - metallicity cluster m67 . observations of physical pairs of stars with and without ( detected ) planets provide the clearest test of the planet - abundance relation , as both stars should have formed from the same molecular cloud . but the available observations do unambiguously support such a correlation @xcite . in fact , the trend of increasing differential abundance with increasing condensation temperature approximately mirrors the pattern of depletion of elements in ism gas @xcite . this simply reflects the preferential condensation of refractory elements onto interstellar grains and the universality of chemistry . if interstellar dust does not perfectly follow the gas , this will produce variation in element abundances that are proportional to the extent of gas - dust segregation and the dust composition . all else being equal , the abundances of the more refractory elements ( which are mostly in the dust ) would be expected to vary more than volatile elements ( which are mostly in the gas ) , in a manner that roughly correlates with @xmath69 . there is evidence for , and theoretical predictions of , dust - gas segregation in the ism . @xcite found that the power - law describing the spatial distribution of extinction ( a tracer of the column density of dust ) on the sky is shallower than that of @xmath70co ( a tracer of the gas column density ) , such that in the densest regions of the ism , extinction by dust is less than that predicted by perfect correspondence with the gas . this could be explained by grain growth in excess of that predicted by models . ( co condensation would presumably act in the opposite sense * ? ? ? * ) . but @xcite offer spatial variation in dust - to - gas ratio as an alternative explanation . dust - gas segregation could occur at three different scales ; ( i ) over the extent of a giant molecular cloud or star - forming region ; ( ii ) in cloud cores that collapse to form individual stellar systems ; and ( iii ) in the accretion disks around young stellar objects . @xcite estimated that dust drift due to radiation pressure from o stars can remove dust from the centers of h ii regions in @xmath71 myr , but since such low - density regions are not themselves the site of star formation , this effect should not manifest itself in the relative abundances of stellar photospheres . @xcite showed that the different dynamics of dust and gas during bondi - type accretion flow can enhance dust - to - gas ratios by an order of magnitude in cloud cores . the mechanism modeled by @xcite relies on dust velocities of several km sec@xmath72 relative to the gas around the growing cloud core . the size distribution of interstellar dust grains peaks sharply near 0.2 - 0.3 @xmath25 m @xcite . these grains may be dynamically decoupled from the diffuse interstellar medium ( number density @xmath73 @xmath74 , stopping distance @xmath75 pc ) , and there is tentative evidence for this in the trajectories of interstellar meteors @xcite and _ ulysses _ spacecraft measurements @xcite . however , these grains will be tightly coupled to the molecular gas ( @xmath76 @xmath74 ) surrounding a cloud core , i.e. over length scales @xmath4 pc and much smaller than a typical cloud size . under such conditions it seems unlikely that dust acceleration mechanisms ( * ? ? ? * ; * ? ? ? * e.g. , ) can achieve equipartion of energy between ism and dust and speeds of @xmath1710 km s@xmath72 . the second scenario which could produce variation in the dust - to - gas ratio in cloud cores is drift induced by radiation pressure . @xcite showed that the inward radial drift of 0.1-@xmath25 m grains under the influence of a typical radiation field can treble the dust concentration in a static gas sphere of a few solar masses at 10 k in 10 myr . the effect scales linearly with the intensity of the external radiation field and the inverse of the characteristic cloud column density . grains reach terminal velocity on a timescale of @xmath10 yr , i.e. much shorter than the cloud lifetime . thus any density enhancement can be expected to grow linearly and the total dust enhancement can be expected to scale with the total external radiation experienced by the cloud core over its lifetime . @xcite numerically simulated this effect , including gas dynamics as well as coupling between the gas and dust , and found that the dust concentration is enhanced by about an order of magnitude in a narrow , inward - propagating shell , and depleted exterior to that shell . while the mean dust - to - gas ratio of a cloud is not changed by migration , truncation of the cloud by photoevaporation or internal collapse ( to the exclusion of outer regions ) would produce a metal - enhanced object . both @xcite and @xcite point out the relevance of this process to the metallicities of stars and the formation of their planets . a third scenario for dust - gas segregation involves accreting protostars and their disks , or concomitant bondi - hoyle accretion from the molecular cloud @xcite . photoevaporation of gas , but not dust , from a disk produces an enhancement in the dust - to - gas ratio which is inherited by the star as disk accretion continues . photoevaporation of disks driven by x - rays from the central star has been proposed to explain the final , rapid stage of circumstellar disk clearing @xcite . observations of blue - shifted lines of ne ii @xcite and o i @xcite indicate that heavy volatile elements as well as h and he are lost in these winds . the temperature of the x - ray - heated `` surface '' of the disk from which the winds flow is heated to a few thousand k @xcite , but gravitational settling ( and the formation of planetesimals ) keeps dust near the cooler mid - plane . disk masses are typically 1% of the central star , with a large scatter @xcite ; assuming the convection zone of a young solar - mass star contains 0.1m@xmath12 , accretion of an entire disk which has been severaly gas - depleted would increase relative elemental abundances by @xmath77 dex , or @xmath78 dex if the convection zone contains 0.4m@xmath12 . this may fall short of explaining some of the observations . two important clues to the mechanism(s ) of the observed variation are the spatial scale on which the variation occurs , and the @xmath69 of the `` elbow '' in the relative abundance variation below which abundances do not vary in any systematic manner . surveys of nearby , solar - age field stars offer little information on the spatial scale of the segregation as the stars are far from their birthplace . @xcite found no significant difference between m67 cluster stars and the sun , and proposed that the sun also formed in a dense cluster analogous to m67 . they suggested that the common pattern of relative abundances was established by the removal of dust by radiation drift . but there are several problems with this explanation . star formation does not take place in hii regions but in surrounding neutral gas , and , as they point out , the expansion of an hii region should outpace dust drift . this explanation also requires most stars to form after an earlier generation of massive stars that move the dust . it also requires that the vast majority of nearby solar - type stars have _ not _ formed in dense stellar clusters , something not supported by cluster statistics @xcite . finally , abundance differences between stars in physical pairs requires gas - dust segregation on scales smaller than the cloud core . the other clue is the value @xmath69 below which relative abundance variation disappears , as this suggests the temperature and hence location where the dust and gas segregation occurs . if this `` knee '' in the abundance variation vs. @xmath69 is 1000 k , as many data sets suggest @xcite , this would seem to rule out a molecular cloud setting with @xmath79 k where even some c and o condense as organics and ices . all of these observations point to gas - dust segregation during the formation or subsequent evolution of an accretion disk as a plausible explanation of the abundance trends . neither observations of m dwarfs nor models of galactic chemical evolution support the premise that there are a significant number of stellar systems with primordial c / o @xmath16 in the solar neighborhood , i.e. within the samples of doppler radial velocity surveys . moreover , if the process of planet formation is universal , then studies of primitive meteorites show that the chemical composition of small , rocky planets is controlled largely by the composition and thermal processing of pre - existing dust from the parent molecular cloud , rather than condensation from gas with the stellar composition , and that this dust is likely to have a c / o reflecting that of the bulk ism . finally , small differences in the relative patterns between the sun and solar `` twins '' can best be explained by dust - gas segregation at stellar scales and temperatures of up to 1000 k , i.e. in accreting protostars , rather than planet formation per se . when comparing the sun to the solar neighborhood it is usually assumed that they share a common history . however , it is also possible that the two are not related . @xcite estimated `` cosmic '' standard abundances ( rather the present - day abundances in the solar neighborhood ) using bright , slowly - rotating , early - type b stars . due to their short main sequence lives , b stars contain abundance information that is both contemparaneous and local , and their purely radiative atmospheres are comparatively simple to model , although uv photoionization rates must be correctly modeled @xcite . @xcite report metallicities that are very close to solar , depsite 4.5 gyr of intervening gce , and a mean c / o of 0.37 , lower than the solar value of 0.55 . they explain both discrepancies by appealing to outward migration of the sun in the galactic disk . these findings warrant further investigation . depletion patterns in the ism suggest that the c / o of dust largely reflects the ism ; o is incorporated more rapidly in grains than c in molecular clouds but it is depleted from grains more quickly in the intercloud medium , and these two effects approximately balance . variation in the efficiency of grain erosion by sn shocks and uv radiation could produce modest variations in the c / o , however . while the phases of o in interstellar grains are clearly water ice and silicates , the phases of c are more controversial . volatility patterns suggest c is _ not _ in refractory phase(s ) like graphite and carbides but instead more volatile organic molecules , in particular aromatic hydrocarbons @xcite . aliphatic hydrocarbons have been suggested as the source of the ubiquitious ism absorbtion feature at 3.4@xmath25 m but the absence of polarization in this line is difficult to reconcile with a scenario of condensation of carbonaceous mantles onto silicate cores @xcite . surface reactions could explain why growth rate of c in grains is very slow with respect to other elements . while these considerations plus a bulk ism c / o@xmath8 are sufficient to explain the oxidized nature of dust in the solar system , it remains to be explained why primitive meteorites in the solar system are very depleted in both c and o , especially c. based on the meteorite abundances and associated uncertainties compiled by @xcite and solar abundances and uncertainties from @xcite , the depletion has a stochiometric ratio of @xmath80 , i.e. consistent with unity and depletion by the formation and removal of co. in this scenario the removal of c or o is not controlled by volatility , otherwise either c or o would be far more depleted than the other , depending on the effective @xmath69 . this can occur if the reaction occurs in situ , i.e. at temperatures lower than the @xmath69s of the c and o phases ( @xmath81k ) where both elements are retained in grains , and/or where vapor is not lost by turbulent mixing , i.e. in `` dead zone '' where turbulence is suppressed . primitive meteorites , considered analogs to the now - lost building blocks of the solar system , also include the highly - reduced enstatite chondrites . enstatite chondrites contain carbide minerals but virtually no water and have been explained by equilibration of solids with a gas having c / o = 0.83 @xcite . interestingly , removal of an amount of c and o equal to that incorporated in ci chondrites produces a residual gas with a c / o of 0.93 . @xcite have proposed that removal of refractories from a gas of solar composition and equilibration of solids with the remaining gas could produce the precursors for enstatite chondrites . if enstatite chondrites formed from recycling of the precursors of terrestrial planets in an oxygen - poor gas it could explain why the enstatite chondrites and earth lie along the same mass - dependent fractionation line in a three oxygen - isotope plot . this equilibration would necessarily have occurred at a comparatively low dust - to - gas ratio and lower temperatures where moderately volatile elements such as fe would not be lost . al - mg isotopes of cais in unequilibrated enstatite chondrites indicate that these refractory inclusions formed in the same region as cais in other chondrites , but were subsequently exposed to reducing conditions @xcite . reducing conditions could have been established sequentially or in a different part of the protoplanetary disk . the formation times of enstatite chondrites is an area of active research . dating of sulfides in unequilibrated e chondrites using the fe - ni and mn - cr short - lived radionuclide chronometers gives ages of 12 - 13 myr after cai formation @xcite . re - analysis of the mn - cr isotope data for a single sulfide in the macalpine hills 88136 el3 chondrite @xcite plus an initial @xmath82mn/@xmath83mn of @xmath84 @xcite gives an age of 10 myr . in summary , i propose the following scenario for the chemistry of rocky planets : * planet formation proceeds from a two - component ism ( gas and dust ) which are never fully equilibraxted during the formation process . * the composition of the precursor material of solar system material is interstellar dust with refractory abundances and a c / o ratio approximately equal to the bulk ism ; this was set by grain growth in molecular clouds and destruction between clouds . * planetesimals , as represented by carbonaceous ( ci ) chondrites , formed under the oxidizing , high dust - to - gas ratio conditions established by settling of interstellar grains to the disk mid - plane and subsequent depletion of c and o by formation and removal of co. * enstatite chondrites represent reduced material that equilibrated with gas after removal of solids in approximately ci proportions , possibly during a second generation of planetesimal formation . * the chemical composition of rocky exoplanets could be set by mixing of oxidized and reduced generations of planetesimals . the mixing ratio could be determined by the effiency with which the first , oxidizing generation incorporated disk solids , as well as the dynamics of accretion in the disk . giant planets , unlike the small , rocky planets considered in this work , accrete massive gas envelopes that would include volatile species such as co , and the considerations described above do not preclude the possibility of c - rich atmospheres in giant planets . detections of such objects have been claimed , but are controversial ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? if the core - first model of giant planet formation is correct , removal of o as silicates and sequestration into a core increases the c / o ratio of the gas that is subsequently captured into the planet s envelope . beyond the ice - line , condensation of water and removal of more o would drive the c / o of the gas even closer to unity @xcite . dissociation of co in the disk and removal of the o as water ice would further enhance this ratio , and indeed the enrichment of planetary water in @xmath85o and @xmath86o relative to the sun is thought to be a signature of this process @xcite . these processes may facilitate the formation of reduced , enstatite chondrite - like planetesimals as discussed above . the author thanks derek homeier and france allard for generating the c / o=1 cases with the phoenix atmosphere model , gary huss for discussions which stimulated this work , and larry nittler for comments . this research was supported by nasa grants nnx10aq36 g and nnx11ac33 g to eg . 139 natexlab#1#1 , c. m. o. , & ebel , d. s. 2012 , meteoritics and planetary science , 47 , 1157 , f. , homeier , d. , & freytag , b. 2011 , in astronomical society of the pacific conference series , vol . 448 , 16th cambridge workshop on cool stars , stellar systems , and the sun , ed . c. johns - krull , m. k. browning , & a. a. west , 91 , s. m. , rosenfeld , k. a. , kraus , a. l. , & wilner , d. j. 2013 , , 771 , 129 , s. s. , & lewis , j. s. 1976 , , 14 , 81 , t. c. , preston , g. w. , & shectman , s. a. 1992 , , 103 , 1987 , n. t. , bonifacio , p. , ludwig , h .- g . , et al . 2010 , , 513 , a72 , p. m. 2008 , , 678 , 1099 , t. , & feltzing , s. 2008 , in nuclei in the cosmos ( nic x ) , 197 , a. d. , wolfire , m. , & leroy , a. k. 2013 , , 51 , 207 , j. c. , obrien , d. p. , & lauretta , d. s. 2010 , , 715 , 1050 , c. , kleine , t. , dauphas , n. , & wieler , r. 2012 , earth and planetary science letters , 357 , 298 , e. , ludwig , h .- g . , steffen , m. , freytag , b. , & bonifacio , p. 2011 , , 268 , 255 , j. c. , obrien , d. p. , & raymond , s. n. 2012 , , 760 , 44 , s. i. b. , meyer , d. m. , & lauroesch , j. t. 2003 , , 597 , 408 , g. , matteucci , f. , mcwilliam , a. , & chiappini , c. 2009 , , 505 , 605 , j. e. , tielens , a. g. g. m. , adamson , a. j. , & ricca , a. 2013 , , 770 , 78 , l. , & povich , m. s. 2011 , , 142 , 197 , f. j. 2007 , , 654 , l159 ciesla , f. j. 2008 , meteoritics & planetary science , 43 , 639 , r. n. 2002 , , 415 , 860 , i. j. m. , barman , t. , hansen , b. m. s. , tanaka , i. , & kodama , t. 2012 , , 760 , 140 , c. c. , liebert , j. , kron , r. g. , spinrad , h. , & hintzen , p. m. 1977 , , 216 , 757 , a. m. 2011 , proceedings of the national academy of science , 108 , 19142 , e. , israelian , g. , gonzlez hernndez , j. i. , et al . 2010 , , 725 , 2349 , c. l. , gil - pons , p. , lau , h. h. b. , lattanzio , j. c. , & siess , l. 2014 , , 437 , 195 , r. a. , margon , b. , anderson , s. f. , et al . 2004 , , 127 , 2838 , b. t. 2003 , , 41 , 241 . 2011 , physics of the interstellar and intergalactic medium , a. , israelian , g. , santos , n. c. , et al . 2004 , , 426 , 619 . 2006 , , 445 , 633 , b. , andersen , j. , gustafsson , b. , et al . 1993 , , 275 , 101 , s. , meyer , m. r. , & moore , b. 2012 , icarus , 221 , 859 , c. , garca - rojas , j. , carigi , l. , et al . 2014 , , 443 , 624 , d. , nissen , p. e. , asplund , m. , pettini , m. , & akerman , c. 2009 , , 500 , 1143 , b. , & palme , h. 1985 , earth and planetary science letters , 72 , 311 , d. a. , & valenti , j. 2005 , , 622 , 1102 , j. j. 2012 , , 747 , l27 , d. t. , & brown , r. l. 1997 , , 113 , 221 , e. , hartmann , l. , calvet , n. , et al . 2006 , , 165 , 568 , e. , mann , a. w. , lepine , s. , et al . 2014 , arxiv e - prints , arxiv:1406.7353 gaidos , e. j. 2000 , icarus , 145 , 637 , d. r. , skillman , e. d. , dufour , r. j. , et al . 1995 , , 443 , 64 , r. d. , truran , j. w. , williams , r. e. , & starrfield , s. 1998 , , 110 , 3 , j. e. 1997 , , 113 , 806 . 1998 , , 115 , 2053 , g. 1998 , , 334 , 221 , j. i. , delgado - mena , e. , sousa , s. g. , et al . 2013 , , 552 , a6 , p. 2013 , , 765 , 12 grossman , l. , beckett , j. r. , fedkin , a. v. , simon , s. b. , & ciesla , f. j. 2008 , reviews in mineralogy and geochemistry , 68 , 93 , y. , huss , g. r. , & leshin , l. a. 2007 , , 71 , 4082 , y. , huss , g. r. , macpherson , g. j. , & wasserburg , g. j. 2000 , science , 289 , 1330 , b. , karlsson , t. , olsson , e. , edvardsson , b. , & ryde , n. 1999 , , 342 , 426 , c. j. , schwartz , j. c. , & cowan , n. b. 2014 , , 444 , 3632 , p. , edwards , s. , & ghandour , l. 1995 , , 452 , 736 , l. , calvet , n. , gullbring , e. , & dalessio , p. 1998 , , 495 , 385 , d. c. , russell , s. s. , ross , a. j. , & kearsley , a. t. 2008 , meteoritics and planetary science , 43 , 1879 , t. , lazarian , a. , & schlickeiser , r. 2012 , , 747 , 54 , j. , nordstrm , b. , & andersen , j. 2007 , , 475 , 519 , g. r. , meshik , a. p. , smith , j. b. , & hohenberg , c. m. 2003 , , 67 , 4823 , m. , & ruzicka , a. 2000 , meteoritics and planetary science , 35 , 601 , w .- c . , henry , t. j. , beaulieu , t. d. , & subasavage , j. p. 2008 , , 136 , 840 , e. b. 2009 , , 700 , 1299 , a. p. 2009 , in astronomical society of the pacific conference series , vol . 414 , cosmic dust - near and far , ed . t. henning , e. grn , & j. steinacker , 473 , a. p. , fanciullo , l. , khler , m. , et al . 2013 , , 558 , a62 , a. i. 2010 , , 403 , 1413 , a. i. , & lattanzio , j. c. 2014 , arxiv e - prints , arxiv:1405.0062 , c. , karakas , a. i. , & umeda , h. 2011 , , 414 , 3231 , c. , umeda , h. , nomoto , k. , tominaga , n. , & ohkubo , t. 2006 , , 653 , 1145 , h. , & grn , e. 2009 , interstellar dust inside and outside the heliosphere , ed . j. l. linsky , v. v. izmodenov , e. mbius , & r. von steiger , 347 , s. , rich , r. m. , & shara , m. m. 2007 , , 669 , 1235 , q. , liang , s. l. , & li , a. 2014 , , 440 , l56 , m. r. , knutson , h. , wolf , a. s. , & yung , y. l. 2014 , , 783 , 70 , f. , asplund , m. , ramrez , i. , yong , d. , & melndez , j. 2014 , , 442 , l51 lodders , k. 2003 , sciences - new york , 1220 , l. s. 2013 , astrophysics , 56 , 472 , n. , harrington , j. , stevenson , k. b. , et al . 2011 , , 469 , 64 , a. 1992 , , 264 , 105 , g. w. , isaacson , h. , howard , a. w. , et al . 2014 , , 210 , 20 , l. 2010 , , 515 , a68 , j. , asplund , m. , gustafsson , b. , & yong , d. 2009 , , 704 , l66 , j. , madhusudhan , n. , & fischer , d. 2014 , , 787 , 81 , k. 2014 , , 100 , 51 , n. 2011 , , 729 , 133 , t. , & ostriker , j. p. 2006 , , 366 , 899 , m .- f . , & przybilla , n. 2012 , , 539 , a143 , p. e. 2013 , , 552 , a73 , p. e. , chen , y. q. , carigi , l. , schuster , w. j. , & zhao , g. 2014 , arxiv e - prints , arxiv:1406.5218 nomura , h. , & nakagawa , y. 2006 , the astrophysical journal , 640 , 1099 , k. i. , murray - clay , r. , & bergin , e. a. 2011 , , 743 , l16 , a. , gustafsson , b. , & korn , a. 2014 , , 562 , a102 , j. e. , clarke , c. j. , & ercolano , b. 2012 , , 422 , 1880 , p. , cambrsy , l. , juvela , m. , et al . 2006 , , 649 , 807 , i. , & sterzik , m. 2009 , , 702 , 724 , e. a. , howard , a. w. , & marcy , g. w. 2013 , proceedings of the national academy of science , 110 , 19273 , e. a. , & marcy , g. w. 2011 , , 735 , 41 , l. , chiosi , c. , & bressan , a. 1998 , , 334 , 505 prantzos , n. 2008 , sciences - new york , 1 , a. s. , reyl , c. , allard , f. , et al . 2014 , , 564 , a90 , i. , asplund , m. , baumann , p. , melndez , j. , & bensby , t. 2010 , , 521 , a33 , i. n. , hawley , s. l. , & gizis , j. e. 1995 , , 110 , 1838 , a. , & voli , m. 1981 , , 94 , 175 robinson , s. e. , laughlin , g. , bodenheimer , p. , & fischer , d. 2006 , astrophysical journal , 643 , 484 , b. , covey , k. r. , muirhead , p. s. , & lloyd , j. p. 2012 , , 748 , 93 , n. c. , israelian , g. , & mayor , m. 2004 , , 415 , 1153 , s. j. , wallerstein , g. , woolf , v. m. , & bean , j. l. 2009 , , 121 , 1083 , e. r. d. 2007 , annual review of earth and planetary sciences , 35 , 577 , y. m. , kim , j. , & hong , s. s. 2011 , , 728 , 140 , l. 2010 , , 512 , a10 , a. , gaidos , e. , & wu , y. 2014 , arxiv e - prints , arxiv:1406.6048 , u. j. , & parvathi , v. s. 2009 , in astronomical society of the pacific conference series , vol . 414 , cosmic dust - near and far , ed . t. henning , e. grn , & j. steinacker , 236 , k. b. , bean , j. l. , madhusudhan , n. , & harrington , j. 2014 , , 791 , 36 , m. , deroo , p. , tinetti , g. , et al . 2013 , icarus , 225 , 432 , h. , himeno , y. , & ida , s. 2005 , , 625 , 414 , a. d. , baggaley , w. j. , & steel , d. i. 1996 , , 380 , 323 , m. , huss , g. r. , ogliore , r. c. , nagashima , k. , & tachibana , s. 2012 , meteoritics and planetary science , 47 , 2013 , j. k. , cunha , k. , schuler , s. c. , griffith , c. a. , & smith , v. v. 2013 , , 778 , 132 , j. k. , cunha , k. , smith , v. v. , schuler , s. c. , & griffith , c. a. 2014 , , 788 , 39 , j. k. , schuler , s. c. , cunha , k. , smith , v. v. , & griffith , c. a. 2013 , , 768 , l12 , h. b. , & bally , j. 2008 , , 135 , 2380 , a. g. g. m. 1998 , , 499 , 267 , a. g. g. m. 2012 , in iau symposium , vol . 284 , iau symposium , ed . r. j. tuffs & c. c. popescu , 7281 , a. g. g. m. 2012 , in iau symposium , vol . 284 , iau symposium , ed . r. j. tuffs & c. c. popescu , 7281 , f. x. , woosley , s. e. , & weaver , t. a. 1995 , , 98 , 617 , b. m. 1980 , , 5 , 287 , a. , elliott , t. , ulfbeck , d. , et al . 2009 , science , 324 , 374 , m. , melendez , j. , & ramirez , i. 2014 , arxiv e - prints , arxiv:1407.4132 , m. , zinner , e. k. , & crozaz , g. 1997 , meteoritics and planetary science , 32 , 281 , a. a. , morgan , d. p. , bochanski , j. j. , et al . 2011 , , 141 , 97 , j. c. , sneden , c. , & truran , jr . , j. w. 1989 , , 27 , 279 , d. c. b. , goldsmith , p. f. , & pineda , j. l. 2010 , , 720 , 259 , a. p. , & bate , m. r. 2002 , , 333 , 679 williams , j. p. , & gaidos , e. 2007 , the astrophysical journal , 663 , l33 , m. g. , mckee , c. f. , hollenbach , d. , & tielens , a. g. g. m. 2003 , , 587 , 278 , v. m. , lpine , s. , & wallerstein , g. 2009 , , 121 , 117 , s. e. , & weaver , t. a. 1995 , , 101 , 181 , h. 2009 , , 397 , 1093 , z. , luo , a. , song , y. , et al . 2014 , , 147 , 33 , q. 2005 , in astronomical society of the pacific conference series , vol . 341 , chondrites and the protoplanetary disk , ed . a. n. krot , e. r. d. scott , & b. reipurth , 632 , q .- z . , jacobsen , b. , moynier , f. , & hutcheon , i. d. 2007 , , 662 , l43 i computed changes in the abundance of five isotopes in the vicinity of the solar galactocetnric radius : @xmath87c , @xmath70c , @xmath88o , @xmath85o , and @xmath86o . the model accounts for the production of these isotopes and their release into the ism in sn explosions , winds from massive stars , and agb winds , incorporation of isotopes into long - lived low - mass stars , and dilution of the ism by the infall of metal - poor gas . the ism is described by a two - box model with a lower - density , warmer intercloud medium ( icm ) which spawns molecular clouds , and a giant molecular cloud ( gmc ) component that can form stars . the mean lifetime of molecular clouds is @xmath89 and the residence time of gas in the icm before condensing into a cloud is @xmath90 . the equations of motion for the mass surface density @xmath91 of the two components are : @xmath92 @xmath93 where @xmath94 is the infall rate , @xmath95 is the rate of formation rate of stars with @xmath96m@xmath12 , @xmath97 is the ratio of the total stellar mass to the mass in stars @xmath98m@xmath12 , @xmath99 is the flux of sn ejecta into the icm from progenitors that explode in time @xmath100 , @xmath101 is the flux of sn ejecta into the parent molecular cloud from progenitors that contribute to gmcs in time @xmath102 , and @xmath103 is the wind from agb stars , all contributing to the icm . the fluxes from agb winds and sn ejecta are given by : @xmath104 where @xmath105 is the mean mass of stars with @xmath106m@xmath12 , @xmath107 is the ejected mass in winds and/or explosions , @xmath108 is the fractional number of stars per unit mass in the imf , and @xmath109 and @xmath110 are the minimum and maximum masses contributing mass to the three different cases , and @xmath111 is the mass in the wind or ejecta from a progenitor of mass @xmath112 . the equations for the mass surface densities of stars ( @xmath113 ) and stellar remnants ( @xmath114 ) are @xmath115 and @xmath116 f(m _ * ) dm_*,\ ] ] where the limits of integation are over any stars that are moving off the main sequence . the star formation rate is related to the total mass surface density of gas using a schmidt - kennicut law : @xmath117 the equations governing the the mass fraction of the @xmath44th isotope is : @xmath118 @xmath119 @xmath120 are the isotopic abundances of the infalling gas , and @xmath121 where @xmath122 is the production ( equal to the nucleosynthetic yield plus the original mass of the isotope ) . the c / o ratio is calculated by summing the appropriate mass fractions divided by the atomic weights . an approximate metallicity is also calcated from the sum of all the isotopes relative to the abundances in the sun . i considered nucleosynthesis in the progenitors of agb stars ( 1 - 6.5m@xmath12 ) , `` super - agb '' stars ( 6.5 - 11m@xmath12 ) , and sn ( 11 - 120m@xmath12 ) . i used the agb yields from @xcite for agb progenitors up to 6.5m@xmath12 . for super - agb progenitors with masses between 7 and 11m@xmath12 i used the productions from @xcite , or @xcite otherwise . for sn yields i adopted the values in @xcite and @xcite for regular sn with energies of @xmath123 erg and their re - run calcalations for the cases of @xmath124m@xmath12 and @xmath125 , and @xmath126 and @xmath127 . these were supplemented with values from @xcite for progenitor masses of 12 , 60 , 100 , and 120m@xmath12 . in order to arrive at a solar c / o with reasonable choices of parameters i find it necessary to include yields from the _ winds _ of massive stars with @xmath128m@xmath12 estimated by @xcite ( table 3 in that work ) . the literature for sn and agb nucleosynthesis is very heterogeneous , with calculations performed for different ranges / values of progenitor masses and metallicities , and it was necessary to interpolate or extrapolate for some values . specifically , to estimate the production of 6.5m@xmath12 agb stars with sub - solar progenitor metallicities i scaled the @xcite values for sub - solar metallicity 6m@xmath12 progenitors by the ratio of the 6.5m@xmath12 to 6m@xmath12 production for solar - metallicity . for productions from super - agb stars with progenitor masses outside the caclulated range , i either used the production from the most massive progenitor with the same metallicity , or linearly interpolated between the values for the most massive agb and least massive super - agb progenitors with the same metallicity . for a review of the many parameters and uncertainties that enter these calculations see @xcite . i linearly interpolated the production onto a grid of 1000 masses over 1 - 120m@xmath12 with intervals chosen such that the imf has equal total mass in each bin . figure [ fig.yield ] shows the calculated yields ( production - initial incorporation ) for solar metallicity . these values were then used to evaluate eqns . [ eqn.production ] and [ eqn.iso_production ] . i searched for the parameter values that best reproduce the observational constraints using monte - carlo markov chain analysis . the constraints are the age of the sun , the present mass surface densities of stars , stellar remnants , and total gas at the solar galactocentric radius , solar c / o , the current metallicity of the ism ( taken to be + 0.14 based on the metallicities of m44 and the hyades ) , and the mean ( -0.05 ) and intrinsic standard - deviation ( 0.18 dex ) of the metallicity distribution in the solar neighborhood @xcite . i adopted standard errors of 0.12 in c / o @xcite , 1.5 , 1 , and 1 m@xmath12 pc@xmath129 for stars , remnants , and gas , 50 myr for the absolute age of the sun , and 0.01 and 0.02 dex in the mean and standard deviation of the present metallicity . i varied the exponential infall timescale , the metallicity of the infalling material , the age of the galactic disk in the solar annulus , the high - mass imf index @xmath14 , and the schmidt - kennicut index @xmath130 . the simulation is able to adequately reproduce the observed properties of the galactic disk in the solar neighborhood ( @xmath131 with @xmath132 ) . the most significant deviation is the over - prediction of the present ism gas mass and underprediction of the current metallicity . the adopted or fit values for the simulation are given in table [ tab.params ] . the predicted evolution of the sfr , mass surface densities , and the metallicity of the ism are plotted in fig . [ fig.gce ] . estimates of the current total star formation rate of the milky way cluster around @xmath133 m@xmath12 yr@xmath72 @xcite . presuming that star formation has an exponential radial distribution with scale length of 3.5 kpc @xcite then the rate at the solar radius will be about @xmath134 m@xmath12 myr@xmath72 pc@xmath129 . this compares favorably with the predicted value at the present ( fig . [ fig.gce ] ) . @lllll@ & & & & + + @xmath135 & solar oxygen & @xmath136 & & @xcite + @xmath137 & solar carbon & @xmath138 & & @xcite + + @xmath14 & imf index & 2.61 & & fit + @xmath97 & low mass / high - mass ratio & 2.00 & & fit + + @xmath139 & present stellar mass density & @xmath140 & m@xmath12 pc@xmath129 & @xcite + @xmath141 & present gas mass density & @xmath142 & m@xmath12 pc@xmath129 & @xcite + @xmath143 & present remnant mass density & @xmath144 & m@xmath12 pc@xmath129 & @xcite + @xmath89 & gmc gas residence time & 30 & myr & + @xmath145 & imc gas residence time & 150 & myr & + @xmath146 & age of disk & 9554 & myr & fit + @xmath147 & infall e - folding time & 2630 & myr & fit + @xmath148_0 $ ] & metallicity of infalling gas & -1.3 & & fit + @xmath130 & sfr rate index & 1.71 & & fit +

investigation of an uncontrollable or inaccessible quantum system or quantum systems with a large number of degrees of freedom via some controllable quantum system is known as quantum simulation @xcite . to this end , a wide variety of systems such as atoms in optical lattices , trapped ions , nuclear spins , superconducting circuits and photonic systems have been proposed as quantum simulators in different branches of physics such as condensed matter physics , high - energy physics , and cosmology @xcite . the general theory of relativity ( gtr ) is , without doubt , one of the fascinating theories of the history of science describing the most important gravitational phenomena . the light deflection in gravitational field , gravitational waves , and also appearance of black holes , are some famous and spectacular predictions of gtr . one of the most important goals of recent investigations in theoretical physics is the presentation of a unified theory for gravitation and other quantum fields by their quantization on a curved background , i.e. , a quantum field theory ( qft ) in curved spacetime . this theory has some experimental consequences , such as particle creation , hawking radiation , etc . ( see e.g. , @xcite and references therein ) . even in the case of small curvature , various aspects of qft in curved spacetime can not be examined directly with current technologies . nevertheless , they can be investigated via quantum simulation . the geometric formulation of quantum mechanics is a very important step toward the presentation of the analog models of gtr in quantum realm . hence , the investigation of novel models and a better understanding of previous analog models is still open - ended @xcite . various analog models of gtr have been proposed , such as the acoustic fluid @xcite , liquid helium @xcite , fermi gases @xcite , slow light @xcite , nonlinear electromagnetic waveguides @xcite , graphene @xcite , ion rings @xcite and bose - einstein condensate ( bec)@xcite . many of the notions of atomic and molecular physics can be realized on an entirely different scale , namely , in optomechanical systems . the field of optomechanics is currently undergoing rapid experimental and theoretical progress . typically , electromagnetic radiation and macroscopic mechanical oscillators can interact via radiation pressure in a fabry - perot cavity with a movable end mirror . this interaction is the basis of various optomechanical phenomena which can occur in a wide range of system sizes and parameters ( for review , see @xcite ) . many of the rudiments of quantum optomechanics date back to the early attempts of the investigation of one of the most prominent predictions of gtr , i.e. , the gravitational waves . however , during the past decade , significant efforts have been devoted to developing and implementing optomechanical interaction for position or force sensing @xcite , backaction cooling @xcite , quantum state transfer @xcite , optomechanical entanglement generation @xcite , optomechanically induced transparency realization @xcite , and generating self - sustained mechanical oscillations @xcite . another interesting aspect of the optomechanical systems , which has not been investigated much yet , is that they can be considered as promising candidate systems for quantum simulation applications . for instance , it has very recently been shown @xcite how to realize the so - called sphere - coherent motional states of a mechanical oscillator in an optomechanical cavity in the presence of a two - level atom and how to use the optomechanical parameters to control the curvature of the sphere . in the present paper , we investigate the possibility of simulating the curved spacetimes by photonic fields interacting with a mechanical oscillator operating in the reversed dissipation regime ( rdr ) of optomechanics @xcite . due to the radiation pressure force , an effective kerr - type photon - photon interaction arises which is mediated by the mirror motion . the emergence of fluid - like behavior for photons is the consequence of this photon - photon interaction so that we can use the hydrodynamic equations for the photonic fluid to study the corresponding effective metric for the propagation of fluctuations in the photon fluid . in this manner , the optomechanical system can be regarded as a quantum simulator for the propagation of the collective excitations of photon fluid in a curved spacetime . we also show that the photonic fluid as well as the corresponding curved spacetime can be controlled by manipulating the system parameters . the rest of the paper is structured as follows . in section [ sec : sec2 ] , we introduce the quantum field description of our first proposed system , i.e. , a planar optomchanical microcavity . we discuss the equations of motions in the rdr of optomechanics . in section [ sec : sec3 ] , we present our second proposed setup which is a two - dimensional optomechanical array . in section [ sec : sec4 ] , we discuss the so - called hydrodynamic version of the nonlinear schrdinger equation ( nlse ) , which can be linearized to describe an effective metric for the two - dimensional photon fluid . finally , in section [ sec : sec5 ] , we present our concluding remarks as well as some possible outlooks of our work .. ) is used to reach the rdr of cavity optomechanics . ( c ) rdr regime of optomechanics where @xmath0 ( see appendix [ app1 ] for details of calculations).,title="fig:",width=302 ] ) is used to reach the rdr of cavity optomechanics . ( c ) rdr regime of optomechanics where @xmath0 ( see appendix [ app1 ] for details of calculations).,title="fig:",width=302 ] as the first theoretical proposal , we consider a planar optomechanical microcavity . as depicted in fig ( [ fig : fig1]-a ) , we consider a typical optomechanical system i.e. , a fabry - prot cavity with a movable end mirror , where the displacement of the mirror shifts the cavity resonances and results in the dependence of cavity resonance frequency to the position of the mechanical oscillator . we separate the total wave vector @xmath1 , into the transversal in - plane , @xmath2 , and the longitudinal , @xmath3 , components . the photon energy dispersion in the limit of paraxial approximation as a function of in - plane and longitudinal wave vectors reads , @xmath4 where @xmath5 denotes the speed of light . imposing the boundary conditions on the electromagnetic field inside the optomechanical cavity with curved mirrors of radius @xmath6 leads to a discrete wave vector in the @xmath7 direction , namely , @xmath8 $ ] , where the mirrors separation at the distance @xmath9 from the cavity axis is given by @xmath10 and the displacement of the mirror from its equilibrium position is indicated by @xmath7 . in the limit of paraxial approximation @xmath11 the photon dispersion relation becomes @xmath12 where @xmath13 is the trapping potential and @xmath14 is the single - photon optomechanical coupling rate . therefore , in a fabry - prot cavity with a movable end mirror , spatial confinement of the photons in the @xmath7 direction breaks the symmetry and eventually it provides a finite effective mass ( @xmath15 ) for the photons in the plane perpendicular to the cavity axis @xcite . this makes the photon gas in the @xmath16-plane to be effectively two - dimensional . moreover , the mirror curvature provides a trapping potential with frequency @xmath17 . it should be noted that in the case @xmath18 ( cavity with fixed mirrors ) we recover the results of ref . we now consider the case in which the moving mirror is interacting with a continuum of optical modes , schematically shown in fig . ( [ fig : fig1]-a ) . the moving mirror is treated as a single - mode quantum mechanical harmonic oscillator with effective mass @xmath19 , frequency @xmath20 , and the position operator @xmath21 where @xmath22 and @xmath23 satisfy the commutation relation @xmath24=1 $ ] and @xmath25 is the zero - point fluctuation of the mirror s position . the single - mode consideration is valid if the detection bandwidth is chosen such that it includes only a single , isolated , mechanical resonance and mode - mode coupling is negligible @xcite . the continuum of optical modes , characterized by transverse wave vector @xmath26 and the annihilation ( creation ) operator @xmath27 ( @xmath28 ) , is coupled to the movable mirror via the radiation pressure coupling . the optical field operators satisfy the commutation relations @xmath29 = { ( 2\pi ) ^2}\delta ( { \bf{k } } - { \bf{k'}})$ ] and @xmath30 = 0 = [ { \hat a_{\bf{k}}^\dag , \hat a_{{\bf{k'}}}^\dag } ] $ ] . in other words , the cavity free spectral range is much larger than the mechanical frequency ( the single - longitudinal - mode assumption ) . we restrict our considerations for the primary mode to the case of a single - longitudinal - cavity mode . using the dispersion relation of eq . ( [ eq : dispersion_relation ] ) , the total hamiltonian of the system is expressed as @xmath31\hat a_{\bf{k}}^\dag { { \hat a}_{\bf{k } } } + { \omega _ m}{{\hat b}^\dag } \hat b , \label{eq : hamiltonian1}\ ] ] where we have defined @xmath32 . this hamiltonian consists of the free energy of the single mechanical mode and the optical modes , the trapping potential due to the mirror curvature , and the optomechanical interaction energy . we have also defined @xmath33 , with @xmath34 being the single - photon optomechanical strength , as the effective optomechanical coupling strength . we investigate the dynamics of the system by the heisenberg - langevin equations associated with the hamiltonian of eq . ( [ eq : hamiltonian1 ] ) @xmath35{{\hat a}_{\bf{k } } } , \label{eq : heisenberglangevin1}\ ] ] @xmath36 where the zero - mean operator @xmath37 denotes the mechanical noise operator . here , we assume that the fluctuation - dissipation process affects only the mechanical mode . this assumption is justified by using a cavity with a high - quality factor ( small damping rate @xmath38 ) and a mechanical mode with a large damping rate . of course , it is not the normal dissipation regime of optomechanics where the mechanical dissipation rate @xmath39 is much smaller than the cavity linewidth @xmath40 . therefore , one has to engineer a mechanical mode with an effective large damping rate . the mechanism of realizing rdr of optomechanics is depicted in figs . ( [ fig : fig1]-b ) and ( [ fig : fig1]-c ) . the detailed considerations for the realization of the rdr in the system under study are descibed in appendix [ app1 ] . briefly , rdr with respect to the optical mode @xmath41 can be achieved by optomechanical sideband cooling of a high - q mechanical mode with an ancillary optical mode possessing a large damping rate @xmath42 . in principle , for time scales longer than @xmath43 and much shorter than @xmath44 we conclude that the photonic subsystem becomes isolated from the environment . based on the calculations in appendix [ app1 ] , for a typical optomchanical system with a mirror with intrinsic frequency @xmath45 interacting with an ancillary optical mode with damping rate @xmath46 , it is possible to increase the mechanical damping rate to @xmath47 . since the mechanical damping rate is much greater than the cavity decay rate ( @xmath48 ) , the mechanical mode can be adiabatically eliminated on time scales greater than @xmath43 . for this purpose , one can formally integrate eq . ( [ eq : heisenberglangevin2 ] ) to obtain @xmath49 for time scales long compared to @xmath43 , the first term becomes zero . substituting this equation into eq . ( [ eq : heisenberglangevin1 ] ) the following equation is obtained for the optical mode @xmath50{{\hat a}_{\bf{k } } } + \hat f_{\bf{k}}^{in}\left ( t \right ) \nonumber\\ & & \qquad\quad-\ ! 2i{g^2}\left [ { \!\int\limits_0^t { \!dt'{e^{\gamma ( { t ' \!-\ ! t})/2}}\sin { \omega _ m } ( { t ' \!-\ ! t})\!\int \!{\frac{{{d^2}{\bf{k}}}}{{{{(2\pi ) } ^2}}}{{\hat n}_{\bf{k } } } ( { t ' } ) } } } \right]{{\hat a}_{\bf{k}}}\ , , \qquad \label{eq : heisenberglangevin3 } \end{aligned}\ ] ] where we have introduced the generalized noise operator @xmath51 } \,.\ ] ] for time scales much longer than the mechanical characteristic time @xmath43 this generalized noise operator can be approximated by @xmath52 . the net effect of the mirror is a redistribution of the photons between the various transversal modes . now , in the rdr the mirror motion can be adiabatically eliminated , resulting in a mechanical field that is affected too much by the optical field operators . using this approximation and defining @xmath53 which for time scales much longer than @xmath54 can be approximated as @xmath55 eq . ( [ eq : heisenberglangevin3 ] ) becomes @xmath56{{\hat a}_{\bf{k } } } \label{eq : heisenberglangevin4}\ ] ] altogether , the adiabatic elimination of the mirror motion has two consequences on the dynamics of the system . first , it provides a reservoir for the photons inside the cavity that can exchange excitations . second , it gives rise to a nonlinear kerr - type photon - photon interaction with a modified interaction rate , @xmath57 . the fourier transform of @xmath58 ( @xmath59 ) is defined to be the two - dimensional cavity photon field operator @xmath60 ( @xmath61 ) @xmath62 where the field operators satisfy the following equal - time commutation relations @xmath63 = \delta ( { \bf{r } } - { \bf{r'}}),\\ & & [ { \hat \psi ( t,{\bf{r}}),\hat \psi ( t,{\bf{r ' } } ) } ] = 0 = [ { { { \hat \psi } ^\dag } ( t,{\bf{r}}),{{\hat \psi } ^\dag } ( t,{\bf{r ' } } ) } ] . \end{aligned}\ ] ] using the fourier transform the equation of motion ( [ eq : heisenberglangevin4 ] ) becomes @xmath64\hat \psi \left ( { t,{\bf{r } } } \right ) \nonumber\\ & & \qquad\quad\quad\quad-2\hbar { g^2 } \mathcal{t } \left [ { \int { { d^2}{\bf{r}}{{\hat \psi } ^\dag } \left ( { t,{\bf{r } } } \right)\hat \psi ( { t,{\bf{r } } } ) } } \right]\hat \psi ( { t,{\bf{r } } } ) . \end{aligned}\ ] ] we now assume that the volume of integration is small and the integrand is slowly varying and smooth in the region of integration so that we can write approximately @xmath65 where @xmath66 is the volume of the integration . in this manner , we arrive at a nlse describing the two - dimensional photon field in the optomechanical microcavity @xmath67{\mkern 1mu } \hat \psi \left ( { t,{\bf{r } } } \right ) , \ ] ] where we have defined @xmath68 , and the photon - photon coupling strength @xmath69 @xmath70 is determined by the optomechanical parameters . as the second theoretical proposal , we consider a two - dimensional array of coupled optomechanical systems schematically depicted in fig ( [ fig : fig5 ] ) . this system could be realized experimentally in optomechanical crystals . one- and two - dimensional optomechanical crystals have been recently realized experimentally @xcite . in this case , optical ( mechanical ) modes on each site interact locally via radiation pressure while photons can tunnel between adjacent site with hopping rate @xmath71.since in such experiments phonon tunneling rate is much smaller ( four orders of magnitude ) than the photon tunneling rate , we can neglect phonon hopping in our considerations . the optomechanical array hamiltonian is given by @xcite @xmath72 } \nonumber \\ & & \qquad\qquad\qquad\qquad \qquad\quad + \sum\limits_{\left\langle { i , j , k } \right\rangle } { j\left ( { \hat a_{ij}^\dag { { \hat a}_{kj } } + \hat a_{ij}^\dag { { \hat a}_{ik } } } \right ) } \ , , \label{eq : optomechanical_array_hamiltonian}\end{aligned}\ ] ] where @xmath73 and @xmath74 , respectively , stands for the bosonic annihilation operators for phonon and photons on the lattice site denoted by @xmath75 . in the last term @xmath76 denotes the summation over all adjacent lattice sites . similar to the case of planar optomechanical microcavity , we consider the system in the rdr and write down the heisenberg - langevin equations for the quantum fields @xmath77 @xmath78 for a slowly varying optical field which is justified when the optical wave vector @xmath79 , we can take the field operators to be continuous functions of @xmath80 and @xmath81 where @xmath82 is the cells spacing , i.e. , @xmath83 and @xmath84 . this assumption allows us to rewrite the equations of motion as follows @xmath85\hat \psi ( t,{\bf{r}})\nonumber\\ & & \quad\ - ij\left[{\hat \psi ( t,{\bf{r } } \!-\ ! h\hat x ) \!+\ ! \hat \psi ( t,{\bf{r } } \!+\ ! h\hat x ) \!+\ ! \hat \psi ( t,{\bf{r } } \!-\ ! h\hat y ) \!+\ ! \hat \psi ( t,{\bf{r } } \!+\ ! h\hat y ) } \right ] \label{eq : array_equation_of_motion1}\end{aligned}\ ] ] @xmath86 we now expand the field operators @xmath87 and @xmath88 as taylor series up to second order in @xmath89 @xmath90 under this approximation , eq . ( [ eq : array_equation_of_motion1 ] ) takes the form @xmath91\hat \psi ( t,{\bf{r } } ) \label{eq : array_equation_of_motion3 } \end{aligned}\ ] ] again , applying the adiabatic approximation for the mechanical modes , we arrive at an equation analogous to eq . ( [ eq : gpe ] ) with @xmath92 , @xmath93 and @xmath94 . we should notice that in contrast to the former case , here photons gain mass due to the hopping between lattice sites . moreover , the potential @xmath95 could be generally space dependent by introducing a space dependent optical frequency @xmath96 or hopping rate @xmath97 . using the mean - field approximation , the field operator can be written as the sum of its mean value and a fluctuation operator with zero mean value , that is , @xmath98 . the classical mean field and the fluctuation operator satisfy , respectively @xmath99{\mkern 1mu } { \psi _ 0 } , \label{eq : gpe0 } \\ & & i\hbar { \partial _ t}\hat \phi = - \left [ { \frac{{{\hbar ^2}}}{{2m}}{\nabla ^2 } + \frac{{{\hbar ^2}}}{m}\frac{{\nabla { \psi _ 0}}}{{{\psi _ 0}}}\nabla } \right]\hat \phi + n\mathcal{g}(\hat \phi + { { \hat \phi } ^\dag } ) , \qquad \label{eq : bdg } \end{aligned}\ ] ] where @xmath100 is defined to be the two - dimensional photon density . in order to interpret clearly eqs . ( [ eq : gpe0 ] ) and ( [ eq : bdg ] ) , we write the quantum field operator , @xmath101 , in the so - called madulang representation , namely @xmath102 where @xmath103 and @xmath104 denote , respectively , the amplitude of the photon number density fluctuation and the phase fluctuation . the equations of motion for the density fluctuation and the phase fluctuation read @xmath105 , \label{eq : hydrodynamics1 } \end{aligned}\ ] ] where @xmath106 and @xmath107 are the local velocity of fluid and the local speed of excitations , respectively . the so - called healing length is defined to be @xmath108 . within the hydrodynamic approximation , i.e. , over the length scales much larger than @xmath109 the last term in eq.([eq : hydrodynamics1 ] ) can be safely ignored and thus @xmath110 . \label{eq : hydrodynamics3}\ ] ] combining eq . ( [ eq : hydrodynamics3 ] ) and eq . ( [ eq : hydrodynamics2 ] ) results in @xmath111 these fluctuations , within the hydrodynamic approximation , are analogous to the collective equantum field on a curved metric . in fact , the photon phase fluctuation obeys the covariant klein - gordon equation for a massless scalar field propagating in a curved spacetime @xmath112 where the dalembertian operator depends on an effective metric , @xmath113 , and is given by @xmath114 with @xmath115 . \label{eq : metric}\ ] ] here , the greek indices range from @xmath116 to @xmath117 , @xmath118 range from @xmath119 to @xmath117 , and @xmath120 is the determinant of @xmath121 . the line element of this spacetime is given by @xmath122.\ ] ] this derivation is similar to the derivation of analog spacetime given in @xcite . a singularity may occur if the flow velocity and the local excitation velocity have equal values . as depicted in fig . ( [ fig : fig4 ] ) , this configuration corresponds to a black hole since excitation waves traveling with @xmath123 are trapped inside the _ superexcitonic _ region and they are not able to propagate backward . although analog models of gtr have a limited ability to simulate all aspects of gr , they provide accessible experimental models for quantum field theory in curved spacetime . some properties ( mainly kinematic ) of gtr can be investigated by the analogies with accessible physical systems . according to eq . ( [ eq : frequencyshift ] ) if @xmath124 then @xmath69 will be a positive parameter but it is physically unacceptable solution because the system enters an unstable region . therefore , the parameter @xmath69 can only have a negative value for a bare optomechanical system . this is equivalent to the imaginary velocity for excitation waves and hence they feel a euclidean metric . positive nonlinearities can also be introduced into the system by placing an extra kerr medium inside the optomechanical microcavity @xcite . this is equivalent to the real excitonic speed and , consequently , correponds to a lorentzian metric for the analog spacetime which is suitable for the investigation of analog black holes . in both cases , the velocity of cavity excitations can be controlled by the ancillary optical mode . in summary , we have introduced a theoretical scheme for the quantum simulation of the curved spacetimes in two types of optomechanical systems operating in the rdr , that is a planar optomechanical microcavity and a two - dimensional array of coupled optomechanical systems . such optomechanical systems are realizable with the state - of - the - art technologies . key features of our proposal can be categorized as follows . first , the optical mode with higher damping rate renormalizes both the damping rate and the frequency of the mechanical element ( see appendix [ app1 ] ) . second , the mechanical oscillator which operates in the rdr of optomechanics can be adiabatically eliminated to achieve a kerr - like photon - photon interaction . the description of the system with the nlse allows us to connect the photonic fluid with the notion of the analog spacetime . we have shown that the phase fluctuation in the photonic fluid obeys the klein - gordon equation for a massless scalar field propagating in a curved spacetime with a metric given by eq . ( [ eq : metric ] ) which can be regarded as an analog of the curved spacetime . the corresponding metric can be controlled by the system parameters . the system introduced here is rather a simple system . as an outlook for future works , it can be extended to more complex situations which open up several possibilities . for example , by introducing two primary modes with different polarizations , it should be possible to produce a two - component photonic fluid inside an optical cavity . in principle , it could be possible to introduce an effective mass for the photons in two dimensions by adding an optical parametric amplifier in the cavity . studying the quantum nature of the photon fluid through various quantum optical measurements on the leakage photon field is another outlook for future works . for two different couplings of the ancillary mode : @xmath125 ( solid line ) , @xmath126 ( dashed line ) . parameter values are @xmath45 , @xmath127 , @xmath128 , @xmath129 . , title="fig:",width=158 ] for two different couplings of the ancillary mode : @xmath125 ( solid line ) , @xmath126 ( dashed line ) . parameter values are @xmath45 , @xmath127 , @xmath128 , @xmath129 . , title="fig:",width=158 ] in this appendix , we discuss how to prepare the mechanical mode in the rdr of optomechanics . assume that the mechanical mode with intrinsic frequency @xmath130 is coupled to another ancillary optical mode denoted by @xmath131 via the radiation pressure . the hamiltonian of the system in the frame rotating with the laser frequency @xmath132 is given by @xmath133 where @xmath134 , @xmath135 and @xmath136 are the single - photon optomechanical coupling , laser pumping rate and the detuning , respectively . considering a strongly driven field and weak optomechanical coupling allows us to linearize the quantum dynamics of fluctuations around the semiclassical amplitudes , @xmath137 and @xmath138 . the steady - state solutions for the classical mean fields are given by @xmath139 $ ] and @xmath140 . here , the cavity damping rate for the ancillary mode is denoted by @xmath141 and the intrinsic damping rate of the mechanical mode is denoted by @xmath142 . the linearized quantum langevin equations for the fluctuations are obtained as @xmath143 where we have defined @xmath144 and @xmath145 . the zero - mean operators @xmath146 and @xmath147 that denote , respectively , the vacuum optical input noise and the mechanical noise operator satisfy commutation relations @xmath148 = [ { { { \hat d}^{in } } ( t),{{\hat d}^{in,\dag } } ( t ' ) } ] = \delta ( t - t')$ ] and the second order correlations @xmath149 and @xmath150 in which we have assumed that the cavity is at zero temperature and @xmath151 is the mean number of thermal phonons of the mechanical oscillator at heat bath temperature @xmath152 with @xmath153 being the boltzmann constant . equations ( [ eq : linearizedheisenberglangevin1 ] ) and ( [ eq : linearizedheisenberglangevin2 ] ) together with noise correlations fully describe the dynamics of the system under consideration . it is convenient to rewrite the equations of motion in the fourier space @xmath154 \!=\ ! ( i\bar \delta \!-\ ! \kappa ' /2)\hat c[\omega ] \!-\!i g ( { \hat d[\omega ] \!+\ ! { { \hat d}^\dag } [ \omega ] } ) \!-\ ! \sqrt \kappa ' { { \hat a}^{in}}[\omega ] , \\ & & - i\omega \hat d[\omega ] \!=\ ! - \left ( i{{\omega _ i } \!+\ ! { \gamma _ i}/2 } \right)\hat d[\omega ] \!-\ ! i ( { g{{\hat c}^\dag } [ \omega ] \!+\ ! g^*\hat c[\omega ] } ) \!-\ ! \sqrt \gamma_i { { \hat b}^{in}}[\omega ] .\end{aligned}\ ] ] combining these two equations results in @xmath155 = \frac { { - i\sqrt \kappa ' \left\ { { { g^*}\chi [ \omega ] { { \hat a}_{{\rm{in}}}}[\omega ] + g{\chi ^ * } [ - \omega ] \hat a_{{\rm{in}}}^\dag [ \omega ] } \right\ } + \sqrt \gamma_i { { \hat b}_{{\rm{in}}}}[\omega ] } } { { i\left ( { \omega - { \omega _ i } + \sigma [ \omega ] } \right ) - { \gamma _ i}/2}}\,,\ ] ] where @xmath156 = [ - i(\omega + \bar \delta ) + \kappa ' /2]^{-1 } $ ] and @xmath157 = - i|g|^2\left ( { \chi [ \omega ] - { \chi ^ * } [ - \omega ] } \right ) $ ] are the optical susceptibility and the self - energy , respectively . therefore , the effect of the ancillary mode will be a renormalization in the mechanical damping rate ( @xmath158 ) and the mechanical frequency ( @xmath159 ) . the finite cavity lifetime leads to the retarded essence of the radiation - pressure force and , consequently , it introduces the frequency - dependent mechanical frequency shift @xmath160 and optomechanical damping rate @xmath161 given by @xcite @xmath162,\ ] ] @xmath163.\ ] ] the intrinsic mechanical resonator damping rate and frequency modify due to the radiation pressure as follows @xmath164 one can thus increase or decrease the damping rate and frequency of the moving mirror , depending on the sign of the detuning . in particular , for a red - detuned ancillary pump , @xmath165 , there is an increase in the mechanical damping rate of the mechanical oscillator and consequently the cooling of the mechanical oscillator is provided . moreover , the mechanical oscillator will be spring softened . in fig . ( [ fig : fig3]-a ) and fig . ( [ fig : fig3]-b ) we have plotted , respectively , the modified mechanical frequency and the modified optomechanical damping rate versus the normalized frequency @xmath166 for two different values of the optomechanical coupling of the ancillary optical mode for experimentally feasible parameters of a typical optomechanical system @xcite . with @xmath167 , and in the resolved side - band limit ( @xmath168 ) the induced damping rate and the frequency shift of the oscillator are , respectively , given by @xmath169 it is evident that @xmath20 is negative for @xmath170 . under such a circumstance , the system can enter unstable region and the routh - hurwitz stability conditions are violated @xcite . driving the system with an ancillary optical mode with a large damping rate ( @xmath171 and @xmath172 ) tuned to the red side of the optical cavity @xcite , as the consequence of the optomechanical interaction , the ancillary mode induces an optical damping to the mechanical oscillator . working in the resolved - sideband regime and using a red detuned driving laser together with a weak optomechanical coupling the net effect of the ancillary optical mode is to renormalize the mechanical damping rate according to eq . ( [ eq : dampingrate ] ) . in this manner , @xmath173 can be adjusted by the strength of the driving ancillary field so that the total damping rate of the mechanical oscillator , @xmath174 becomes very large . therefore , coupling a high - q mechanical oscillator to an auxiliary cavity mode allows us to realize the rdr of the cavity optomechanics , @xmath48 @xcite ( see figs . ( [ fig : fig1]-b ) and ( [ fig : fig1]-c ) ) . the wide separation between the time scales of dissipation mechanisms allows us to use the mechanical oscillator as an extra dissipative reservoir for the optical mode . m. novello , matt visser , and g. e. volovik , `` _ _ artificial black holes , _ _ '' ( world scientific , 2002 ) . c. barcel , s. liberati and m. visser , `` _ _ analogue gravity , _ _ '' http://www.livingreviews.org/lrr-2005-12[living rev . relativity * 8 * , 12 ( 2005 ) . ] w. g. unruh , `` _ _ quantum analogues : from phase transitions to black holes and cosmology _ _ '' ( springer , 2007 ) . d. faccio , f. belgiorno , s. cacciatori , v. gorini , s. liberati , and u. moschella , `` _ _ analogue gravity phenomenology : analogue spacetimes and horizons , from theory to experiment _ _ '' ( springer , 2012 ) . l. sindoni , `` _ emergent models for gravity : an overview of microscopic models , _ '' http://dx.doi.org/10.3842/sigma.2012.027[sigma * 8 * , 027 ( 2012 ) ] . w. g. unruh , `` _ _ experimental black - hole evaporation ? , _ _ '' https://doi.org/10.1103/physrevlett.46.1351[phys . lett . * 46 * , 1351 ( 1981 ) ] . g. e. volovik , `` _ _ superfluid analogies of cosmological phenomena , _ _ '' http://dx.doi.org/10.1016/s0370-1573(00)00139-3[phys . rep . * 351 * , 195 ( 2001 ) ] . t. a. jacobson and g. e. volovik , `` _ _ event horizons and ergoregions in @xmath175 , _ _ '' https://doi.org/10.1103/physrevd.58.064021[phys . d * 58 * , 064021 ( 1998 ) ] . s. giovanazzi , `` _ _ hawking radiation in sonic black holes , _ _ '' https://doi.org/10.1103/physrevlett.94.061302[phys . . lett . * 94 * , 061302 ( 2005 ) ] . s. giovanazzi , `` _ _ entanglement entropy and mutual information production rates in acoustic black holes , _ _ '' https://doi.org/10.1103/physrevlett.106.011302[phys . 106 * , 011302 ( 2011 ) ] . u. leonhardt and p. piwnicki , `` _ _ relativistic effects of light in moving media with extremely low group velocity , _ _ '' https://doi.org/10.1103/physrevlett.84.822[phys . rev . lett . * 84 * , 822 ( 2000 ) ] . u. leonhardt , `` _ _ space - time geometry of quantum dielectrics , _ _ '' https:/doi.org/10.1103/physreva.62.012111[phys . a * 62 * , 012111 , ( 2000 ) ] . b. reznik , `` _ _ origin of the thermal radiation in a solid - state analogue of a black hole , _ _ '' https:/doi.org/10.1103/physrevd.62.044044[phys . d * 62 * , 044044 ( 2000 ) ] . w. g. unruh and r. schtzhold , `` _ _ on slow light as a black hole analogue , _ _ '' https:/doi.org/10.1103/physrevd.68.024008[phys . d * 68 * 024008 , ( 2003 ) ] . r. schtzhold and w. g. unruh , `` _ _ hawking radiation in an electromagnetic waveguide ? , _ _ '' https:/doi.org/10.1103/physrevlett.95.031301[phys . * 95 * , 031301 ( 2005 ) ] . a. cortijo and m. a. h. vozmediano , `` _ _ effects of topological defects and local curvature on the electronic properties of planar graphene , _ _ '' b * 763 * , 293 ( 2007 ) ] . m. a. h. vozmediano , m. i. katsnelson , and f. guinea , `` _ _ gauge fields in graphene , _ _ '' https:/doi.org/10.1016/j.physrep.2010.07.003[phys . rep . * 496 * , 109 ( 2010 ) ] . b. horstmann , b. reznik , s. fagnocchi , and j. i. cirac , `` _ _ hawking radiation from an acoustic black hole on an ion ring , _ _ '' https:/doi.org/10.1103/physrevlett.104.250403[phys . * 104 * , 250403 ( 2010 ) ] . l. j. garay , j. r. anglin , j. i. cirac , and p. zoller , `` _ _ sonic analog of gravitational black holes in bose - einstein condensates , _ _ '' https://doi.org/10.1103/physrevlett.85.4643[phys . lett . * 85 * , 4643 ( 2000 ) ] . l. j. garay , j. r. anglin , j. i. cirac , and p. zoller , `` _ _ sonic black holes in dilute bose - einstein condensates , _ _ '' https://doi.org/10.1103/physreva.63.023611[phys . a * 63 * , 023611 ( 2001 ) ] . c. barcel , s. liberati , and m. visser , `` _ _ analogue gravity from bose - einstein condensates , _ _ '' http://stacks.iop.org/0264-9381/18/i=6/a=312[class . quantum grav . * 18 * , 1137 ( 2001 ) ] . c. mayoral , a. fabbri , and m. rinaldi , `` _ _ steplike discontinuities in bose - einstein condensates and hawking radiation : dispersion effects , _ _ '' https://doi.org/10.1103/physrevd.83.124047[phys . d * 83 * , 124047 ( 2011 ) ] . f. girelli , s. liberati , and l. sindoni , `` _ _ gravitational dynamics in bose - einstein condensates , _ _ '' https://doi.org/10.1103/physrevd.78.084013[phys . d * 78 * , 084013 ( 2008 ) ] . s. finazzi , s. liberati , and l. sindoni , `` _ _ cosmological constant : a lesson from bose - einstein condensates , _ _ '' https://doi.org/10.1103/physrevlett.108.071101[phys . lett . * 108 * , 071101 ( 2012 ) ] . p. r. anderson , r. balbinot , a. fabbri , and r. parentani , `` _ _ hawking radiation correlations in bose - einstein condensates using quantum field theory in curved space , _ _ '' https://doi.org/10.1103/physrevd.87.124018[phys . d * 87 * , 124018 ( 2013 ) ] . t. j. kippenberg and k. j. vahala , `` _ _ cavity opto - mechanics , _ _ '' https:/doi.org/10.1364/oe.15.017172[optics express * 15 * , 17172 ( 2007 ) ] . p. meystre , `` _ _ a short walk through quantum optomechanics , _ _ '' https:/doi.org/10.1002/andp.201200226[ann . ( berlin ) * 525 * , 215 ( 2013 ) ] . m. aspelmeyer , t. j. kippenberg , and f. marquardt , `` _ _ cavity optomechanics , _ _ '' https:/doi.org/10.1103/revmodphys.86.1391[rev . 86 * , 1391 ( 2014 ) ] . w. p. bowen and g. j. milburn , `` _ _ quantum optomechanics , _ _ '' ( crc press , 2015 ) . k. w. murch , k. l. moore , s. gupta , and d. m. stamper - kurn , `` _ _ observation of quantum - measurement backaction with an ultracold atomic gas , _ _ '' https:/doi.org/10.1038/nphys965[nature physics * 4 * , 561 ( 2008 ) ] . j. d. thompson , b. m. zwickl , a. m. jayich , f. marquardt , s. m. girvin , and j. g. e. harris , `` _ _ strong dispersive coupling of a high - finesse cavity to a micromechanical membrane , _ _ '' https:/doi.org/10.1038/nature06715[nature ( london ) * 452 * , 72 ( 2008 ) ] . t. p. purdy , r. w. peterson , and c. a. regal , `` _ _ observation of radiation pressure shot noise on a macroscopic object , _ _ '' https:/doi.org/10.1126/science.1231282[science * 339 * , 801 ( 2013 ) ] . a. motazedifard , f. bemani , m. h. naderi , r. roknizadeh , and d. vitali , `` _ _ force sensing based on coherent quantum noise cancellation in a hybrid optomechanical cavity with squeezed - vacuum injection , _ _ '' https:/doi.org/10.1088/1367 - 2630/18/7/073040[new j. phys . * 18 * , 073040 ( 2016 ) ] . j. d. teufel , t. donner , d. li , j.w . harlow , m. s. allman , k. cicak , a. j. sirois , j. d. whittaker , k.w . lehnert , and r.w . simmonds , `` _ _ sideband cooling of micromechanical motion to the quantum ground state , _ _ '' https:/doi.org/10.1038/nature10261[nature ( london ) * 475 * , 359 ( 2011 ) ] . j. chan , t. p. m. alegre , a. h. safavi - naeini , j. t. hill , a. krause , s. grblacher , m. aspelmeyer , and o. painter , `` _ _ laser cooling of a nanomechanical oscillator into its quantum ground state , _ _ '' https:/doi.org/10.1038/nature10461[nature ( london ) * 478 * , 89 ( 2011 ) ] . t. a. palomaki , j.w . harlow , j. d. teufel , r.w . simmonds , and k.w . lehnert , `` _ _ coherent state transfer between itinerant microwave fields and a mechanical oscillator , _ _ '' https:/doi.org/10.1038/nature11915[nature ( london ) * 495 * , 210 ( 2013 ) ] . t. a. palomaki , j. d. teufel , r.w . simmonds , and k.w . lehnert , `` _ _ entangling mechanical motion with microwave fields , _ _ '' https:/doi.org/10.1126/science.1244563[science * 342 * , 710 ( 2013 ) ] . m. paternostro , d. vitali , s. gigan , m. s. kim , c. brukner , j. eisert , and m. aspelmeyer , `` _ _ creating and probing multipartite macroscopic entanglement with light , _ _ '' https:/doi.org/10.1103/physrevlett.99.250401[phys . * 99 * , 250401 ( 2007 ) ] . d. vitali , s. gigan , a. ferreira , h. r. bhm , p. tombesi , a. guerreiro , v. verdal , a. zeilinger , and m. aspelmeyer , `` _ _ optomechanical entanglement between a movable mirror and a cavity field , _ _ '' https:/doi.org/10.1103/physrevlett.98.030405[phys . * 98 * , 030405 ( 2007 ) ] . j. zhang , k. peng , and s. l. braunstein , `` _ _ quantum - state transfer from light to macroscopic oscillators , _ _ '' https:/doi.org/10.1103/physreva.68.013808[phys . a * 68 * , 013808 ( 2003 ) ] . s. weis , r. rivire , s. delglise , e. gavartin , o. arcizet , a. schliesser , and t. j. kippenberg , `` _ _ optomechanically induced transparency , _ _ '' https:/doi.org/10.1126/science.1195596[science * 330 * , 1520 ( 2010 ) ] . t. hong , h. yang , h. miao , and y. chen , `` _ _ open quantum dynamics of single - photon optomechanical devices , _ _ '' https:/doi.org/10.1103/physreva.88.023812[phys . a * 88 * , 023812 ( 2013 ) ] . f. marquardt , j. g.e . harris , and s. m. girvin , `` _ _ dynamical multistability induced by radiation pressure in high - finesse micromechanical optical cavities , _ _ '' https:/doi.org/10.1103/physrevlett.96.103901[phys . lett . * 96 * , 103901 ( 2006 ) ] . f. bemani , r. roknizadeh , and m. h. naderi , `` _ _ theoretical scheme for the realization of the sphere - coherent motional states in an atom - assisted optomechanical cavity , _ _ '' https:/doi.org/10.1364/josab.32.001360[j . b * 32 * , 1361 ( 2015 ) ] . a. nunnenkamp , v. sudhir , a. feofanov , a. roulet , and t. j. kippenberg , `` _ _ quantum - limited amplification and parametric instability in the reversed dissipation regime of cavity optomechanics , _ _ '' https:/doi.org/10.1103/physrevlett.113.023604[phys . lett . * 113 * , 023604 ( 2014 ) ] . l. d. toth , n. r. bernier , a. nunnenkamp , e. glushkov , a. k. feofanov , and t. j. kippenberg , `` _ _ a dissipative quantum reservoir for microwave light using a mechanical oscillator , _ _ '' https://arxiv.org/abs/1602.05180[arxiv:1602.05180 ( 2016 ) ] . j. klaers , f. vewinger , and m. weitz , `` _ _ thermalization of a two - dimensional photonic gas in a ` white wall ' photon box , _ _ '' https:/doi.org/10.1038/nphys1680[nat . phys . 6 , 512 ( 2010 ) ] . c. genes , d. vitali , and p. tombesi , `` _ _ simultaneous cooling and entanglement of mechanical modes of a micromirror in an optical cavity , _ _ '' https://doi.org/10.1088/1367-2630/10/9/095009[new j. phys . * 10 * , 095009 ( 2008 ) ] . a. h. safavi - naeini , j. t. hill , s. meenehan , j. chan , s. grblacher , and o. painter , `` _ _ two - dimensional phononic - photonic band gap optomechanical crystal cavity , _ _ '' https://doi.org/10.1103/physrevlett.112.153603[phys . lett . * 112 * , 153603 ( 2014 ) ] . j. chan , a. h. safavi - naeini , j. t. hill , s. meenehan , and o. painter , `` _ _ optimized optomechanical crystal cavity with acoustic radiation shield , _ _ '' http://dx.doi.org/10.1063/1.4747726[appl . * 101 * , 081115 ( 2016 ) ] . m. ludwig and f. marquardt , `` _ _ quantum many - body dynamics in optomechanical arrays _ _ , '' https://doi.org/10.1103/physrevlett.111.073603[phys . lett . * 111 * , 073603 ( 2013 ) ] . w. chen and a. a. clerk , `` _ _ photon propagation in a one- dimensional optomechanical lattice _ _ , '' https://doi.org/10.1103/physreva.89.033854[phys . rev . a * 89 * , 033854 ( 2014 ) ] . v. peano , c. brendel , m. schmidt , and f. marquardt , `` _ _ topological phases of sound and light , _ _ '' https://doi.org/10.1103/physrevx.5.031011[phys . x * 5 * , 031011 ( 2015 ) ] . s. shahidani , m. h. naderi , and m. soltanolkotabi , and s. barzanjeh , `` _ _ steady - state entanglement , cooling , and tristability in a nonlinear optomechanical cavity , _ _ '' https:/doi.org/10.1364/josab.31.001087[j . b * 31 * , 1087 ( 2014 ) ] . s. shahidani , m. h. naderi , and m. soltanolkotabi , `` _ _ normal - mode splitting and output - field squeezing in a kerr - down conversion optomechanical system , _ _ '' https:/doi.org/10.1080/09500340.2014.958111[j . mod . opt . * 62*,116 ( 2015 ) ] . c. genes , d. vitali , p. tombesi , s. gigan , and m. aspelmeyer , `` _ _ ground - state cooling of a micromechanical oscillator : comparing cold damping and cavity - assisted cooling schemes , _ _ '' https:/doi.org/10.1103/physreva.77.033804[phys . a * 77 * , 033804 ( 2008 ) ] . a. hurwitz , `` _ _ selected papers on mathematical trends in control theory _ _ , '' edited by r. bellman and r. kabala ( dover , new york , 1964 ) . f. marquardt , j. p. chen , a. a. clerk , and s. m. girvin , `` _ _ quantum theory of cavity - assisted sideband cooling of mechanical motion , _ _ '' https:/doi.org/10.1103/physrevlett.99.093902[phys . lett . * 99 * , 093902 ( 2007 ) ] . i. wilson - rae , n. nooshi , w. zwerger , and t. j. kippenberg , `` _ _ theory of ground state cooling of a mechanical oscillator using dynamical backaction , _ _ '' https:/doi.org/10.1103/physrevlett.99.093901[phys . rev . lett . * 99 * , 093901 ( 2007 ) ] .

a large number of the ly@xmath0 emitters at high redshift have been detected in sensitive search using the narrow - band filters ( e.g. , hu & mcmahon 1996 ; steidel et al . 2000 ; rhoads et al . 2001 ; 2003 ; kodaira et al . 2003 ; palunas et al . 2004 ; hayashino et al . 2004 ; matsuda et al . 2004 ; taniguchi et al . 2005 ; gronwall et al . 2006 ; iye et al . 2006 , ouchi et al . 2008 ) . from the photometric study , they are generally considered to be young star - forming objects and found to be widely distributed in their luminosity , size , and equivalent width ( e.g. , gawiser et al . 2007 ; finkelstein et al . 2007 ) . while the ly@xmath0 emission is one of the fundamental tools to study the galaxy formation at high redshift , it is , however , still difficult to understand what the dominant origins of the ly@xmath0 emission are , and how the ly@xmath0 photons escape from the galaxy . resonance scattering and extinction by the neutral hydrogen or dust in galaxies may easily modify the observed flux and line profile . there are at least three different physical origins of the ly@xmath0 lines from high - redshift star - forming galaxies ; photo - ionization by hot massive stars , cooling radiation from the gas heated either by the shock during to the gravitational collapse or by the shock due to the gas outflow driven by the thermal energy of frequent supernovae or active galactic nuclei . the ly@xmath0 photons produced by any of these process are affected by the neutral hydrogen gas mixed with or surrounding the emission - line regions . recent results suggest that statistical average of the ly@xmath0 escape fraction of a sample of the star - forming galaxies at redshift @xmath8 is only @xmath9@xmath7 ( hayes et al . 2009 ) . it is important to study their spectroscopic properties , especially their line profiles to study how the lines are produced and modified . if the ly@xmath0 emission is largely reprocessed by the neutral gas with large bulk motion , we may observe the signatures in their line profiles . for example , dijikstra et al . ( 2006 ) studied the expected ly@xmath0 profile of optically thick , spherically symmetric collapsing gas clouds . the expected profiles are generally blueshifted and show double peaked profile possibly dominated by the blue peak due to the combination of the infalling gas motion and line radiation transfer effects . on the other hand , verhamme et al . ( 2006 , 2008 ) modeled the line profiles expected for the expanding neutral hydrogen shell . although the expected profiles are also double - peaked , in many cases they are redshifted and the red peak appears stronger due to the scattered components behind the original ly@xmath0 emission - line region in our line of sight . it is , however , not easy to discriminate these two cases from the observation of the individual ly@xmath0 line itself if we do not know the systemic velocity . it is known that high - redshift star - forming galaxies typically show gas outflow from the sources . pettini et al . ( 2001 ) and shapley et al . ( 2003 ) compared the redshifts of the interstellar absorption line with the rest - frame optical emission lines which are nearly at the systemic velocity of the photo - ionized region and found that the inter - stellar absorption gas is generally blueshifted for typically 200 km s@xmath5 indicating that the outflow gas motion is common among the lyman break galaxies . ly@xmath0 lines in these objects also show p cyg - type profiles , namely , a narrow emission line with a blueshifted absorption line . verhamme et al . ( 2008 ) applied their models to the ly@xmath0 line profile of the lyman break galaxies and found that they can be explained by the scattering in the simple expanding shell models . steidel et al . ( 2010 ) also showed that the cool gas around @xmath10 - 3 lyman break galaxies typically exhibit outward motion . on the other hand , the models of cold accretion ( dekel et al . 2009 ; goerdt et al . 2009 ) also suggest that the observed luminosity , surface brightness , and size distributions of the ly@xmath0 emitters , especially for the extended ly@xmath0 blobs ( steidel et al . 2000 ; matsuda et al . 2004 ) can be explained by the gas heated by the collisional excitation in the cold gas stream . spectroscopic observations to study the line profiles of the ly@xmath0 emitters are essential to discriminate these cases although only a small sample of spectroscopy with enough high spectral resolution is available so far . matsuda et al . ( 2006 ) studied the line profiles of the extended ly@xmath0 emitters , or ly@xmath0 blobs as well as more ordinary ly@xmath0 emitters , and found that the ly@xmath0 lines of the blobs have more structure and the whole velocity width is positively correlated with their size . in this paper , we present the results of our new spectroscopic observations for the 91 ly@xmath0 emitters to show their ly@xmath0 line profiles . the photometric data for the sample selection is similar as in matsuda et al . ( 2006 ) but the number of the spectra for normal ly@xmath0 emitters are significantly increased . we describe our observations and data reduction in section 2 , the observed line profiles in section 3 , and the discussion in section 4 . the sample of the ly@xmath0 emitters are selected based on the same data presented in hayashino et al . ( 2004 ) at the field of ssa22-sb1 where one of the largest overdensity of lyman break galaxies as well as ly@xmath0 emitters at @xmath11=3.1 is known to exist ( steidel et al . 1998 ; 2000 ) . the selection criteria for the emitters are , however , slightly modified to include the objects with relatively low equivalent width . we also did not use the @xmath12 color criteria , which was needed to make the very robust imaging sample of ly@xmath0 emitters to exclude the contamination by the foreground objects . we here applied only the following criteria . @xmath13 or @xmath14 where @xmath15 is the effectively averaged @xmath16 magnitude of the @xmath17 and the @xmath18-band and @xmath19 represents the @xmath16 magnitude in the nb497 narrow - band filter ( hayashino et al . 2004 ) . the observations of the ly@xmath0 profiles of the emitters are done by the faint object camera and spectrograph ( focas ) equipped with the subaru 8.2 m telescope in jul 2004- aug 2005 . the vph grating 600_450 was used to obtain the spectral dispersion of 0.37 per pixel . using slits with a width of 1 arcsec and spectral resolution , @xmath2/@xmath3 @xmath4 was achieved as measured from the widths of the arc lines . in order to observe as many targets as we can in one exposure , we used the custom - made intermediate band filter with @xmath3 @xmath20 to limit the spectral length on the detectors . by this method , we obtain the range of the spectra only around the emission lines , but the spectral resolution is high enough to clearly discriminate the [ oii ] 3727 doublet , which is the only significant contaminants , from the @xmath1 ly@xmath0 lines . the data was reduced in the standard manner using iraf . the accuracy of the wavelength calibration is higher than 0.2 . the one - dimensional spectra are then extracted by averaging the three spatial pixels , each 0.@xmath212 . we also smoothed the spectra by three pixels in the dispersion direction . as a result , we obtained the ly@xmath0 emission - line spectra for the 91 objects . the sample contains the 12 objects classified as ly@xmath0 blobs by matsuda et al . ( 2004 ) whose isophotal area above the surface brightness of @xmath22 erg s@xmath5 @xmath23 arcsec@xmath24 is larger than 16 arcsec@xmath25 or 900 kpc@xmath25 at @xmath1 . matsuda et al . ( 2004 ; 2006 ) showed that the photometric or the spectroscopic properties of the ly@xmath0 blobs are not quite discriminated from other objects but rather continuously distributed in the whole sample of the emitters . as described in matsuda et al . ( 2005 ) and matsuda et al . ( 2006 ) , the sample selected by the criteria in hayashino et al . ( 2004 ) is almost free for contamination ( @xmath26 ) . for the objects with lower equivalent width ( i.e. , @xmath27 ) , the fraction of the [ oii ] emitter at @xmath28 increases , as expected , but still less than @xmath9% . we identified in total of 91 spectroscopic ly@xmath0 emitters in the 11 masks located in the ssa22-sb1 field . fig.1a-1f show the observed ly@xmath0 spectrum for all the 91 objects . two of them ( lae j221709.6 + 001801 and lae j221720.2 + 002019 ) clearly show a very large line width ( @xmath29 1000 km s@xmath5 ) and possibly host active galactic nuclei ( agn ) . table 1 summarizes their observed properties . in fig.1 , we first notice that many of the objects show the characteristic profiles , namely the double - peaked spectra with a strong , asymmetrical red peak and a much weaker blue peak . while the most conspicuous cases are the objects lae j221720.3 + 002438 , lae j221745.3 + 002006 , and lae j221759.2 + 002254 ( lab28 in matsuda et al . 2004 ) , many other cases can be recognized in fig.1 . as a large fraction of the observed ly@xmath0 spectra shows the similar profiles , it may represent the common physical properties of the emitters . we therefore tried to characterize the profile and to evaluate the fraction of the objects with the similar profile among those observed in more objective way . we first fitted the visually isolated peaks in each spectrum with the gaussian profiles . these procedures are rather formal ones but useful in the following discussions . fig.2 and fig.3 show the example of the fitting for the two objects with high and moderate signal - to - noise ratio , respectively . fig.2a and 3a are for the multiple gaussian fitting . the green lines show the each component and the blue ones the total . the red dashed lines in the figures show the noise level measured at the same wavelength by using the blank part of the slits . bottom panels show the residual . on the other hand , fig.2b and 3b show the examples of the single gaussian fitting similar with those tried in matsuda et al . for the objects with multiple peaks , the data points between the peaks are not used for the fitting ( thin crosses in fig.2b and 3b ) . the center of the wavelength in each panel in fig.1 is the central wavelength of the single - component gaussian fitting . the central wavelengths of the primary red and secondary blue peaks ( for the objects with the feature ) as well as that of the single - component fitting ( for all the objects ) are also presented in table 1 . of the 89 objects except for the two broad - line agn , 50 objects have the visually identified strong blue and weak red peaks . 4 other objects ( 221728.3 + 001212 , 221740.3 + 001129 , 221740.9 + 001125 , and 221812.5 + 001433 ) do not show the weak blue peaks but have notable components to the red of the strongest peak . three of these four objects , 221728.3 + 001212 , and closely located objects 221740.3 + 001129 and 221740.9 + 001125 are associated with the ly@xmath0 blobs lab33 and lab7 in matsuda et al . ( 2004 ) , respectively . for these 50 candidates of the strong red and weak blue " profiles , we then evaluated the significance of the weak peaks relative to the noise level . we evaluated the r.m.s . noise in the blank - sky spectra ( the red dashed lines in fig.2 and fig.3 for the example ) at the line core ( @xmath30@xmath31 of the gaussian ) and then selected only the objects with the weak peaks which are more significant than the 3@xmath32 level as the final sample of the strong red and weak blue " profiles . for the cases between 3@xmath32 and 4@xmath32 , we further visually checked the spectra and rejected a few cases . after all , 39 objects among the 89 ( 44@xmath7 ) show the significant characteristic profile . the wavelength of the primary and the secondary peaks are shown by the vertical dotted lines in fig.1 . this is a large fraction , which implies that the strong red and weak blue " profile is a common characteristic property of the observed ly@xmath0 emitters . the fraction should be considered even as the lower limit because it is more difficult to detect the secondary weaker peaks significantly if the objects are faint or their emission - lines are weak . indeed , if we limit the objects with @xmath33 and further with @xmath34 , the fraction is even larger , 53 and 55@xmath7 , respectively . table 2 summarizes the number of the objects with the characteristic profile . we also listed the wavelength of the secondary component for five marginal objects whose blue peaks were visually identified but not found to be significant in table 1 with :. while it is difficult , only from such formal fitting procedure as shown in fig.2 and fig.3 , to tell whether the peaks are the multiple emission lines with different central velocities or the single line modified by the absorption and/or scattering by neutral gas in vicinity , the latter interpretation is plausible as such large fraction of the sample always show much stronger peak at the long wavelength . we confirm this further by the following two different methods . first , we investigated the asymmetry of the strongest peaks of the sample . we identify the pixel with the largest flux and obtained the ratio of the flux integrated over the 4.5 ( about 1.5 times the spectral resolution ) toward longer wavelength to those toward shorter wavelength . fig.4 shows the result for the objects with @xmath35 and @xmath36 , avoiding the objects with relatively low signal to noise ratio . the symmetry index ( the ratio ) of those with the characteristic strong red and weak blue " profile is strongly concentrated around 0.5 while those of other objects are distributed around the unity ( the median is 0.83 ) with the r.m.s . scatter of @xmath37 . 12 objects with characteristic profile has the value below 0.6 and they indeed show the conspicuous strong red and weak blue " profiles . we applied the kolmogorov - smirnov test to the symmetry distribution of the objects with and without the characteristic profile and the probability that they are drawn from the same population is 7% . we also changed the magnitude or color range as well as the wavelength region for the index and found that the trend is always seen . the objects with the characteristic profile with strong red and weak blue components are shown by the blue histogram which strongly peaked at 0.5 . the objects with strong blue and weak red peaks are shown by the green line . other objects are shown by the red line . , width=302 ] and @xmath38 are marked by the large open circles . the dotted line is the best fit regression line.,width=302 ] next , for those objects with the characteristic profile , we compared the observed width of the stronger red peak and the separation of the peaks obtained in the formal gaussian fitting . fig.5 shows the result ; the clear correlation between these two quantities is seen . this strongly support that the two peaks are indeed the parts of a single feature . if they are multiple different components , such trend as seen in fig.5 is not necessary to be observed . fig.6 shows the distribution of the observed fwhm of the most prominent peak of the each object . those with the characteristic profile tend to show the smaller values . and @xmath38 are used . same as in fig.4 , the blue and red line show the objects with multiple characteristic peaks and those with a single peak , respectively . , width=302 ] we found that a large fraction , @xmath3940@xmath7 of the observed ly@xmath0 emitters in ssa22-sb1 field have a characteristic double - peaked line profile with a strong red peak and a much weaker blue peak . what does this characteristic ly@xmath0 profile means ? the double peaked profiles can be a single source absorbed or scattered by the neutral hydrogen gas or multiple components of the ionized gas with different velocities . the multiple component cases are not favored , however , as the observed lines show asymmetry and there is a correlation between the peak separation and the line width as described in the previous section . ly@xmath0 line profile emission is affected by intergalactic medium ( igm ) in the line of sight toward the object ( dijikstra et al . 2007 ; laursen et al . 2011 ) . at @xmath40 , the mean transmission of the photon at the wavelength just shorter than ly@xmath0 due to the absorption by the foreground intergalactic ly@xmath0 clouds is @xmath41 ( madau et al . 1996 ) . due to the fluctuation by the large - scale structure , the absorption varies by the line of sight toward the sources . laursen et al . ( 2011 ) recently examined how the line profile of ly@xmath0 emitter at z=2.5 - 6.5 is affected by igm absorption . the case for @xmath42 ( their figure 7 ) indeed predicts the double peaked ly@xmath0 line profile with the stronger red component . the red and blue peak - strength difference is more conspicuous for the smaller objects . as the size of the emitter becomes larger , the two peaks show comparable strength on average although the distribution of the igm optical depth can produce the variation of the profiles . the prediction may be compared with our observed results . for those objects identified to have the characteristic profile , most of them show the large flux difference between the red and blue peaks and there are few objects with comparable strength ( fig.1 ) . the formal single - component gaussian fitting shows that more than 50% of the line flux is absorbed for the typical cases ( e.g. , fig.2b , fig.3b ) . it seems difficult to explain the current results only by igm absorption in this sense . note that , however , the emitters observed here are located in the dense environment and the igm absorption is much larger than average . we revisit this possibility in sec.4.4 . we then consider the cases that the profiles show the intrinsic properties of the ly@xmath0 emitters , namely the effects by the gas associated with them . scattering in the infalling partially ionized emission - line gas ( dijikstra et al . 2006 ) as well as scattering in the expanding shell - like out - flowing gas ( verhamme et al . 2006 ; 2008 ) can modify the line profile . for the case of spherically symmetric infalling gas model , dijikstra et al . ( 2006 ) argued that the entire line profile shows strong blue symmetry and the blue peaks are observed stronger than the red peaks . as the red peaks are much stronger than the blue peak for the most of the cases studied here , the other interpretation , namely the expanding shell model is more favored . for the spherically symmetrical expanding shell model ( verhamme et al . 2006 ) surrounding the ly@xmath0 source , the ly@xmath0 line is shifted red ward from the systemic velocity and also show strong red asymmetry . indeed , some model profiles shown in verhamme et al . ( 2006 ) ( their fig.15 ) as well as in verhamme et al . ( 2008 ) seems quite similar with the characteristic strong red and weak blue " profiles discussed here . recently , mclinden et al . ( 2011 ) reported that a strong [ oiii ] 5007 emission line is detected for a ly@xmath0 emitter at @xmath1 which show the similar double peaked ly@xmath0 profile . interestingly , the [ oiii ] line locates at the middle of the two ly@xmath0 peaks , which implies that the systemic velocity of the galaxy is indeed at the middle of the two ly@xmath0 peaks . thus the expanding shell by verhamme et al . ( 2008 ) is favored to understand the observed line profiles . in this case , the star - formation in these ly@xmath0 emitters must be dominated by the burst - like events since the continuous gas infall is needed to maintain continuous star - formation activities . the population of the ly@xmath0 emitters are generally considered to be less massive objects in stellar mass as the many of them are not detected even in the deepest ground - based near - infrared images . the early active starburst events may result in the galactic wind activity producing the expanding shells . it is important to note that the ly@xmath0 emission from @xmath43@xmath7 , and possibly a larger fraction of the observed sample , show such evidence . to prove the expanding shell or gas outflow models , another important observational constraint is the surface - brightness distribution of the ly@xmath0 emission . rauch et al ( 2008 ) , having obtained 2-dimensional spectra of lyman alpha emitters and found the surface - brightness to be mostly strongly peaked in the spatial direction , even though the emission could be traced out to several arcseconds . barnes & haehnelt ( 2010 ) argued that this rapid drop of surface brightness with radius is inconsistent with the much flatter radial profile predicted by a simple expanding shell model . in fact , there is a weak trend , though not very significant , that the objects with the characteristic profile are rather small ( see sec.4.3 and fig.8 and fig.9 below ) . this is against the interpretation by the simple expanding - shell models . for much extended objects such as ly@xmath0 blobs , on the other hand , mori and umemura ( 2006 ) successfully reproduce the observed ly@xmath0 morphology by the models of expanding gas heated by supernovae . future improvements both in observations and models ( e.g. , barnes et al . 2011 ) will allow more detailed study on the surface - brightness distributions . it is also interesting to discuss the relation between characteristic line profile and the ly@xmath0 blobs . matsuda et al . ( 2006 ) found that the galaxies with larger size tend to have more structures in the line profiles and total line width larger than fwhm @xmath44 km s@xmath5 if fitted by a single gaussian profile ( as in fig.2b or 3b in this paper ) . for the velocity width , a similar trend is confirmed with the current sample . in table 3 , we listed the velocity fwhm obtained by the single gaussian fitting , namely the same procedure as in matsuda et al . ( 2006 ) , for the 12 objects associated with the ly@xmath0 blobs in matsuda et al . ( 2004 ) . of the 12 objects , 9 have fwhm @xmath45 km s@xmath5 . while the rest three objects have the smaller velocity width , lab32 and lab33 are among the smallest in the sample of the 35 ly@xmath0 blobs in matsuda et al . lab7 has three , or possibly four distinguished ly@xmath0 peaks ( matsuda et al . 2004 ) and the slit position in this study just covers a single south - most peak and the value measured in this observation may not represent the whole nature of the system . of the 12 objects associated with the ly@xmath0 blobs , on the other hand , only 5 are classified as the strong red and weak blue " profile objects . at a glance , this appears somewhat strange , since the line profiles of the ly@xmath0 blobs shown in matsuda et al . ( 2006 ) have more structures with possible absorption than the other smaller emitters . if we closely see fig.1 , however , there are hints of the multiple components which are not significantly detected in the current data . although it is difficult to conclude , as the exposure time ( 7 hours in matsuda et al . 2006 ) , spectral resolution ( @xmath46 ) , instrument ( @xmath47 deimos ) are different from those in this paper , the deeper data may probe significant structures of these lines . other possible interpretation is as follows . if ly@xmath0 blobs are more massive objects , they may have more thick absorbing material and the weak blue line is too much absorbed to be detected . this is likely to be the case for those with the single but asymmetrical component , such as lae j221808.3 + 001022 ( lab15 ) , lae j221812.5 + 001433 ( lab33 ) , and lae j221817.3 + 001209 ( lab21 ) . matsuda et al . ( 2006 ) argued that the large velocity width implies that ly@xmath0 blobs are more massive objects than other normal ly@xmath48 emitters . while the large size of the ly@xmath0 emission of the blobs can be explained by either the expanding gas ionized and excited by the supernova feedback ( mori and umemura 2006 ) or by the infalling cooling gas ( haiman et al . 2000 ; fardal et al . on the other hand , recent models of the cold flows can reproduce the size and luminosity distribution of the ly@xmath0 blobs ( dekel et al . 2009 ; goerdt et al . if the ly@xmath0 blobs are the scaled - up version of the normal emitters whose line profiles are dominated by the outflowing gas , the gas - outflow interpretation may be favored . we also investigated whether the presence of the characteristic profile is related with any other properties of the ly@xmath0 emitters . first , we see the distribution of the observed excess in @xmath49 color , or equivalent width , in fig.7a and 7b . fig.7a shows the result for the color measured in the fixed 2@xmath50-diameter aperture and the fig.7b shows those for the semi - total flux measured in the kron aperture ( i.e. , sextracter mag_auto ) for those laes with @xmath51 mag . the emitters with the characteristic profile seem to dominate at @xmath52 , where 13/16 objects show strong red and weak blue " peaks , although the @xmath49 colors excess of the objects with the characteristic profile is not always large . the trend is also notable in fig.7b . the probabilities for kolmogorov - smirnov test that they follow the same distribution are 17@xmath7 and @xmath53 , respectively . the large excess in the narrow band , or ly@xmath0 equivalent width can be related to the gas outflow such as additional ly@xmath0 emission due to the collisional excitation or ionization by the shock . if the source of the ly@xmath0 emission is the photoionization by massive stars , the large equivalent width suggests the presence of a very young or metal - poor stellar population , which is also consistent with the gas outflow by the supernovae or strong stellar wind . we also studied the distributions of the size , observed fwhm and the isophotal area above the detection threshold and the results are shown in fig.8 and fig.9 . to omit the objects with low s / n ratio , we again limited the sample to those with @xmath35 and @xmath34 . while the distributions for the objects with and without the characteristic profile are not quite distinguishable , there is some concentration of the strong red and blue weak " object for smaller sizes ( kolmogorov - smirnov test probability is @xmath54 ) . this is disadvantageous for the simple expanding shell model which predicts a rather flat surface brightness distribution . if the starburst is so young , however , the area of the outflowing gas can be still very limited within small radius from the galactic center to show nearly unresolved morphology . finally , we revisit the possibility that the characteristic profile is caused by the intergalactic absorption systems . as the field is known to have a high density of star - forming galaxies ( steidel et al . 1998 ; 2000 , hayashino et al . 2004 ) , the neutral gas density may also be enhanced along the filamentary large - scale structure . the characteristic profile can be caused by the absorption by the common gas in such structure in front of the observed emitters . in such a case , absorption redshift is expected to be spatially correlated . fig.10 shows the sky distribution of the ly@xmath0 emitters with the characteristic profile . the redshift of the absorption " that was formally obtained by the multiple component gaussian fitting allowing the absorption features are color coded , from the redshift 3.047 to 3.129 with the interval of 0.008 ( 600 km s@xmath5 ) . no large - scale correlation of the absorption redshift is seen . we thank the staff of the subaru telescope for their assistance . this research is supported in part by the grant - in - aid 20450224 for scientific research of the ministry of education , science , culture , and sports in japan . data analysis was in part carried out on common use data analysis computer system at the astronomy data center , adc , of the national astronomical observatory of japan . ccccccccc name & z@xmath55 & @xmath56 & @xmath57 & @xmath58 & @xmath59 & nb@xmath60 & nb@xmath61 & bv@xmath61 + & & km s@xmath5 & & & & & & + lae j221650.2 + 002750 & 3.099 & 171.1 & 4984.9 & 0.0 & 4984.9 & 25.24 & 25.68 & 26.88 + lae j221650.9 + 002327 & 3.087 & 191.1 & 4970.8 & 0.0 & 4970.8 & 25.00 & 25.14 & 26.31 + lae j221652.7 + 002423 & 3.067 & 217.3 & 4944.7 & 0.0 & 4944.7 & 24.73 & 24.89 & 25.89 + lae j221653.9 + 002137 & 3.081 & 402.5 & 4964.6 & 4950.9 & 4966.6 & 24.57 & 24.67 & 26.27 + lae j221658.4 + 002430 & 3.102 & 392.0 & 4987.1 & 0.0 & 4987.1 & 24.07 & 24.61 & 25.90 + lae j221659.3 + 002304 & 3.069 & 306.0 & 4947.2 & 0.0 & 4947.2 & 23.95 & 24.92 & 25.69 + lae j221659.3 + 002501 & 3.088 & 141.5 & 4970.7 & 4960.9 : & 4969.3 & 24.30 & 24.97 & 26.08 + lae j221700.3 + 002404 & 3.098 & 96.7 & 4981.8 & 0.0 & 4981.8 & 24.61 & 24.78 & 26.35 + lae j221700.7 + 002006 & 3.075 & 42.5 & 4955.3 & 0.0 & 4956.4 & 25.17 & 25.22 & 26.66 + lae j221701.6 + 002135 & 3.070 & 265.8 & 4947.6 & 4932.9 & 4942.2 & 24.59 & 24.96 & 25.63 + lae j221701.8 + 002003 & 3.077 & 161.9 & 4956.1 & 4947.0 & 4953.1 & 25.21 & 25.25 & 28.03 + lae j221702.7 + 002433 & 3.067 & 385.4 & 4944.7 & 0.0 & 4944.7 & 23.58 & 23.88 & 24.91 + lae j221702.8 + 002121 & 3.091 & 0.0 & 4973.5 & 4968.7 & 4971.7 & 25.27 & 25.34 & 27.70 + lae j221704.9 + 002226 & 3.108 & 279.8 & 4995.5 & 0.0 & 4995.5 & 24.80 & 25.26 & 26.63 + lae j221706.7 + 002134 & 3.067 & 330.4 & 4944.2 & 4935.8 & 4941.9 & 23.60 & 24.11 & 24.97 + lae j221709.6 + 001801 & 3.105 & 924.1 & 4994.2 & 0.0 & 4987.8 & 23.07 & 23.08 & 24.75 + lae j221713.3 + 002034 & 3.101 & 278.0 & 4987.2 & 0.0 & 4987.2 & 24.73 & 25.06 & 27.55 + lae j221713.6 + 000640 & 3.113 & 302.7 & 5002.0 & 4991.5 & 5000.9 & 23.23 & 23.31 & 24.82 + lae j221713.7 + 001656 & 3.057 & 249.4 & 4934.2 & 0.0 & 4934.2 & 24.75 & 25.13 & 26.45 + lae j221715.7 + 001906 & 3.101 & 233.2 & 4986.7 & 4972.3 & 4986.0 & 24.55 & 24.60 & 27.03 + lae j221715.8 + 002431 & 3.106 & 276.5 & 4992.4 & 0.0 & 4992.4 & 24.47 & 24.68 & 26.40 + lae j221716.7 + 002309 & 3.066 & 388.7 & 4945.7 & 0.0 & 4945.7 & 24.53 & 25.36 & 26.61 + lae j221717.3 + 000728 & 3.090 & 149.3 & 4972.9 & 0.0 & 4972.9 & 24.47 & 25.18 & 26.53 + lae j221717.5 + 002010 & 3.075 & 563.1 & 4954.4 & 0.0 & 4954.7 & 24.65 & 25.36 & 28.47 + lae j221718.8 + 001518 & 3.067 & 91.1 & 4944.8 & 0.0 & 4943.4 & 24.44 & 24.90 & 26.98 + lae j221719.0 + 001200 & 3.090 & 265.7 & 4972.1 & 4965.6 : & 4970.3 & 25.03 & 25.13 & 26.64 + lae j221719.3 + 001450 & 3.065 & 288.8 & 4942.8 & 4935.2 : & 4943.8 & 24.40 & 0.00 & 0.00 + lae j221719.7 + 001149 & 3.067 & 150.1 & 4945.4 & 4938.1 & 4943.4 & 24.83 & 0.00 & 0.00 + lae j221720.1 + 002226 & 3.104 & 226.7 & 4991.1 & 0.0 & 4991.1 & 24.87 & 25.40 & 26.73 + lae j221720.2 + 002019 & 3.108 & 1997.2 & 5005.8 & 0.0 & 4995.0 & 22.44 & 22.40 & 23.43 + lae j221720.3 + 002438 & 3.070 & 201.4 & 4948.0 & 4935.9 & 4943.9 & 23.59 & 23.87 & 25.18 + ccccccccc name & z@xmath55 & @xmath56 & @xmath57 & @xmath58 & @xmath59 & nb@xmath60 & nb@xmath61 & bv@xmath61 + & & km s@xmath5 & & & & & & + lae j221721.7 + 001223 & 3.068 & 355.1 & 4947.0 & 0.0 & 4947.0 & 25.06 & 25.32 & 26.90 + lae j221722.3 + 000804 & 3.085 & 88.4 & 4967.3 & 4957.4 & 4962.0 & 23.81 & 24.07 & 25.55 + lae j221722.6 + 002145 & 3.104 & 88.1 & 4988.4 & 4978.1 & 4989.0 & 25.00 & 25.05 & 27.06 + lae j221722.9 + 001441 & 3.054 & 0.0 & 4929.2 & 4924.1 & 4928.7 & 24.99 & 25.08 & 26.90 + lae j221723.5 + 002040 & 3.071 & 368.3 & 4950.8 & 4935.7 & 4950.9 & 24.76 & 24.88 & 26.51 + lae j221723.8 + 002155 & 3.102 & 0.0 & 4987.2 & 0.0 & 4987.2 & 24.71 & 25.49 & 27.09 + lae j221724.6 + 001557 & 3.079 & 172.0 & 4960.1 & 4953.5 : & 4958.8 & 25.34 & 25.42 & 28.47 + lae j221724.7 + 002227 & 3.091 & 349.8 & 4974.4 & 0.0 & 4975.2 & 25.12 & 25.37 & 26.75 + lae j221724.8 + 001717 & 3.096 & 358.3 & 4981.2 & 0.0 & 4980.8 & 23.62 & 0.00 & 0.00 + lae j221727.2 + 001622 & 3.096 & 339.5 & 4978.1 & 0.0 & 4978.1 & 24.67 & 24.81 & 26.33 + lae j221727.3 + 001046 & 3.069 & 307.2 & 4946.9 & 4938.1 & 4945.3 & 24.64 & 24.77 & 27.02 + lae j221727.8 + 001737 & 3.092 & 172.9 & 4975.6 & 4967.6 & 4973.2 & 24.31 & 24.40 & 26.97 + lae j221728.3 + 001212 & 3.067 & 162.3 & 4945.3 & 0.0 & 4947.5 & 24.31 & 24.51 & 26.27 + lae j221733.9 + 001215 & 3.106 & 0.0 & 4991.5 & 4982.5 & 4990.7 & 24.67 & 25.04 & 26.93 + lae j221735.9 + 001559 & 3.094 & 673.8 & 4978.0 & 0.0 & 4978.0 & 24.08 & 24.68 & 26.92 + lae j221736.4 + 001251 & 3.058 & 150.5 & 4934.1 & 0.0 & 4934.1 & 25.20 & 25.17 & 27.31 + lae j221736.8 + 002614 & 3.069 & 179.5 & 4948.0 & 4942.2 & 4947.3 & 24.98 & 25.06 & 27.40 + lae j221738.5 + 002215 & 3.096 & 144.4 & 4980.3 & 0.0 & 4980.3 & 24.85 & 24.96 & 26.74 + lae j221738.9 + 001102 & 3.064 & 116.9 & 4940.9 & 4929.9 & 4937.3 & 23.56 & 24.06 & 25.36 + lae j221739.0 + 001726 & 3.072 & 165.0 & 4951.5 & 4943.3 & 4949.0 & 24.66 & 0.00 & 0.00 + lae j221739.2 + 002242 & 3.098 & 189.3 & 4981.4 & 0.0 & 4983.8 & 25.25 & 25.52 & 27.64 + lae j221739.3 + 001610 & 3.093 & 66.0 & 4976.5 & 0.0 & 4976.4 & 24.84 & 25.40 & 27.51 + lae j221739.9 + 002142 & 3.089 & 361.1 & 4972.8 & 0.0 & 4972.8 & 25.24 & 25.21 & 27.66 + lae j221740.3 + 001129 & 3.065 & 71.9 & 4941.4 & 0.0 & 4943.2 & 25.01 & 24.58 & 25.43 + lae j221740.9 + 001125 & 3.089 & 288.4 & 4968.8 & 0.0 & 4968.8 & 23.49 & 24.53 & 25.31 + lae j221741.4 + 002227 & 3.096 & 208.0 & 4980.9 & 0.0 & 4980.9 & 24.31 & 24.55 & 25.35 + lae j221743.3 + 002149 & 3.097 & 88.2 & 4981.9 & 4967.8 & 4975.4 & 24.68 & 0.00 & 0.00 + lae j221743.4 + 001348 & 3.098 & 142.8 & 4982.1 & 4975.9 & 4980.9 & 23.95 & 24.01 & 25.97 + lae j221743.7 + 002451 & 3.101 & 350.0 & 4987.4 & 4977.5 & 4984.3 & 25.04 & 25.18 & 26.91 + lae j221745.3 + 002006 & 3.074 & 195.8 & 4953.0 & 4943.9 & 4950.3 & 24.00 & 24.13 & 26.53 + ccccccccc name & z@xmath55 & @xmath56 & @xmath57 & @xmath58 & @xmath59 & nb@xmath60 & nb@xmath61 & bv@xmath61 + & & km s@xmath5 & & & & & & + lae j221745.6 + 001544 & 3.065 & 241.6 & 4942.3 & 4933.6 & 4939.4 & 24.58 & 24.59 & 26.25 + lae j221745.9 + 002319 & 3.098 & 142.8 & 4981.9 & 4976.0 & 4979.9 & 24.81 & 0.00 & 0.00 + lae j221748.6 + 001334 & 3.056 & 93.5 & 4932.7 & 4924.5 & 4930.5 & 23.90 & 0.00 & 0.00 + lae j221750.5 + 001310 & 3.089 & 57.0 & 4970.6 & 4965.9 & 4967.5 & 25.52 & 25.35 & 27.81 + lae j221751.0 + 001858 & 3.081 & 346.1 & 4963.7 & 4954.0 & 4961.6 & 24.17 & 24.42 & 27.00 + lae j221753.2 + 001238 & 3.095 & 112.4 & 4977.9 & 4969.3 : & 4975.1 & 24.60 & 0.00 & 0.00 + lae j221754.3 + 001224 & 3.093 & 184.0 & 4976.4 & 0.0 & 4973.9 & 24.49 & 24.50 & 26.00 + lae j221755.1 + 002460 & 3.058 & 402.7 & 4935.7 & 4923.4 & 4932.3 & 23.73 & 24.03 & 25.33 + lae j221757.6 + 001204 & 3.093 & 282.1 & 4976.6 & 4966.1 : & 4973.9 & 24.38 & 24.80 & 26.15 + lae j221759.2 + 002254 & 3.087 & 333.3 & 4968.2 & 4958.8 & 4965.1 & 23.00 & 23.34 & 25.43 + lae j221759.3 + 001148 & 3.055 & 19.4 & 4928.9 & 0.0 & 4929.0 & 25.41 & 25.01 & 25.86 + lae j221802.2 + 002556 & 3.081 & 0.0 & 4961.5 & 0.0 & 4960.3 & 23.03 & 23.88 & 25.97 + lae j221803.6 + 002247 & 3.079 & 133.8 & 4958.9 & 0.0 & 4958.9 & 25.44 & 25.42 & 27.50 + lae j221805.9 + 002407 & 3.095 & 0.0 & 4979.0 & 4973.7 & 4977.0 & 25.24 & 25.38 & 27.25 + lae j221807.9 + 002317 & 3.088 & 131.8 & 4970.1 & 4955.8 & 4964.9 & 23.97 & 24.58 & 26.12 + lae j221808.0 + 001151 & 3.096 & 159.7 & 4980.1 & 0.0 & 4980.1 & 24.82 & 25.19 & 26.33 + lae j221808.3 + 001022 & 3.097 & 421.7 & 4982.6 & 4969.6 & 4980.0 & 23.02 & 23.47 & 24.72 + lae j221809.2 + 001242 & 3.096 & 60.0 & 4980.1 & 4972.2 & 4976.9 & 24.82 & 25.05 & 27.06 + lae j221809.4 + 001358 & 3.105 & 124.5 & 4989.9 & 0.0 & 4989.9 & 24.96 & 25.49 & 26.77 + lae j221811.0 + 002508 & 3.095 & 294.8 & 4978.9 & 0.0 & 4978.9 & 25.04 & 25.07 & 26.59 + lae j221812.5 + 001433 & 3.095 & 164.1 & 4978.8 & 0.0 & 4983.9 & 23.57 & 24.15 & 25.25 + lae j221813.2 + 002157 & 3.115 & 176.1 & 5003.6 & 4990.3 & 4999.2 & 25.22 & 25.32 & 26.37 + lae j221813.8 + 000925 & 3.104 & 110.3 & 4989.8 & 4983.0 & 4987.9 & 24.83 & 24.83 & 27.93 + lae j221813.9 + 002222 & 3.089 & 0.0 & 4971.3 & 0.0 & 4971.3 & 24.69 & 0.00 & 0.00 + lae j221816.0 + 001412 & 3.086 & 74.2 & 4968.0 & 4957.4 & 4964.7 & 24.24 & 24.66 & 26.64 + lae j221817.3 + 001209 & 3.087 & 330.9 & 4970.8 & 0.0 & 4966.6 & 24.45 & 24.78 & 25.77 + lae j221818.2 + 001143 & 3.091 & 315.8 & 4975.6 & 0.0 & 4975.6 & 25.00 & 25.17 & 26.11 + lae j221820.8 + 001257 & 3.109 & 119.2 & 4996.0 & 4986.8 & 4991.7 & 25.40 & 25.28 & 27.52 + lae j221820.8 + 001241 & 3.103 & 342.0 & 4989.6 & 4978.0 & 4986.0 & 23.78 & 24.60 & 26.46 + lae j221820.9 + 001031 & 3.097 & 178.3 & 4981.3 & 0.0 & 4981.3 & 24.99 & 25.50 & 28.16 + \(a ) redshift measured by the peak of the strongest emission - line component . ( b ) fwhm of the strongest component after corrected for the instrumental profile . the zero value means that the line is not resolved . ( c ) the central wavelength of the red peak in the multiple - component gaussian fitting . ( d ) the central wavelength of the blue peak in the multiple - component gaussian fitting . ( e ) the central wavelength of peak in the single - component gaussian fitting . ( f ) the aperture magnitude values are measured in the 2-arcsec diameter apertures . lae j221658.4 + 002430 & 428.8 & lab34 & no + lae j221706.7 + 002134 & 562.9 & lab27 & yes + lae j221723.8 + 002155 & 474.3 & lab32 & no + lae j221724.8 + 001717 & 474.4 & lab35 & no + lae j221735.9 + 001559 & 696.0 & lab14 & no + lae j221738.9 + 001102 & 641.2 & lab31 & yes + lae j221740.9 + 001125 & 338.1 & lab7 & no + lae j221759.2 + 002254 & 625.5 & lab28 & yes + lae j221807.9 + 002317 & 582.4 & lab23 & yes + lae j221808.3 + 001022 & 738.8 & lab15 & yes + lae j221812.5 + 001433 & 222.1 & lab33 & no + lae j221817.3 + 001209 & 661.5 & lab21 & no + \(a ) name of the associated ly@xmath0 blobs listed in matsuda et al . ( 2004 ) ( b ) the column indicates whether the ly@xmath0 line profile shows the characteristic ` strong red and weak blue ' one ( yes " ) or not ( no " ) .

for observation windows of @xmath2 and @xmath3 ( smaller observation windows of @xmath4 are also outlined ) . , width=321 ] the hallmarks of the glass transition @xcite are a sudden increase of relaxation times , and dynamic heterogeneity @xcite . a recent paper @xcite described the idea that these phenomena are manifestations of phase coexistence in space - time . the two competing phases are an active state , containing many relaxation events , and an inactive one , where these events are scarce . the purpose of the current paper is to develop these ideas by building on the results of ref . @xcite . in the active phase of the dynamics , the existence of the inactive phase can be inferred by measuring the distribution of any observable that measures the quantity of relaxation events within a finite space - time window @xcite . because of the coexisting inactive phase , these distributions display non - gaussian tails . this behaviour is analogous to the situation in standard phase transitions : the distribution of cavity sizes in a fluid near liquid vapour coexistence has non - gaussian tails @xcite , as does the distribution of box magnetization for the ising model in the vicinity of its phase transition @xcite . the concept of space - time thermodynamics that we use @xcite is illustrated in fig . [ fig : bubble_sketch ] . it employs a simple picture of glassy dynamics at low temperatures , in which diffusing excitations coalesce and branch @xcite . this dynamics is reducible @xcite . that is , there are two steady states , an active one with a finite density of excitations , and an inactive one with strictly zero density . working in the active state , the dominant fluctuations on large length scales are space - time regions in which there are no excitations , as illustrated in fig . [ fig : bubble_sketch ] . we refer to these regions as `` bubbles '' of the inactive phase . we characterise the bubbles by the probability density function @xmath5 for their spatial and temporal extents ( denoted @xmath6 and @xmath7 , respectively ) . to the extent that these rare fluctuations are dynamical analogues of those near phase coexistence , and for bubbles which are large compared with the bulk correlation length of the active state , we can write : @xmath8\right\ } , \label{equ : bubble_dist}\ ] ] where @xmath9 and @xmath10 are , in effect , surface tensions in the spatial and temporal directions respectively , @xmath11 is the difference between the bulk free energy densities of active and inactive phases , and @xmath12 controls the aspect ratio of the bubbles , in conjunction with the velocity parameter @xmath13 . this distribution of ( [ equ : bubble_dist ] ) can be investigated by considering the probability distribution functions for observables that are averaged over finite regions of space - time , such as the boxes in fig . [ fig : bubble_sketch ] @xcite . large rare bubbles dominate the tails of these distributions . the terminology `` free energy '' , as used in the previous paragraph , refers to space - time ; the relevant ensemble is a the space of histories ( or trajectories ) of the system , which is observed for a specified period of time , @xmath14 . this concept of a free energy is central to ref . @xcite , and it is found also in the thermodynamic formulation of dynamics due to ruelle @xcite . in particular , for a @xmath0-dimensional system , one considers statistical mechanics in the corresponding @xmath15-dimensional space . the extra dimension is that of time , so the resulting space is necessarily anisotropic . thus , to appreciate the analogy with phase coexistence for @xmath16 , consider an anisotropic two - dimensional ising model below its critical temperature , with a small field stabilising the spin - up phase . the spins are predominately up , but there will be a distribution of spin - down domains with various sizes . in the case where the interactions are different in the @xmath17 and @xmath18 directions , the distribution of sizes for domains of the spin - down phase is of the form @xmath19\right\ } , \label{equ : ising_dist}\ ] ] where @xmath17 and @xmath18 are the extents of the bubbles in the @xmath17 and @xmath18 directions ; @xmath20 and @xmath21 are surface tensions ; @xmath22 is the free energy difference between up and down phases ( proportional to the small field ) ; and @xmath23 and @xmath24 set the shape of the bubbles . as phase coexistence is approached in the ising system , the free energy difference @xmath22 approaches zero . in the dynamical system , the corresponding limit is @xmath25 , in which case active and inactive states coexist . a central result of this article is that @xmath26 for systems in which the dynamics is reducible into active and inactive partitions . within the thermodynamic formalism of ruelle the free energy difference @xmath11 is the difference in the so - called ` topological pressure ' @xcite between the active and inactive phases of the dynamics . since the phases are at coexistence in these models , a boundary condition suffices to make the system choose one phase . for example , consider a system with periodic spatial boundary conditions , as sketched in fig . [ fig : bc ] . if we choose our initial condition to be completely inactive , then this state persists forever ; all other initial conditions lead to the active state . in this article , we use the the fredrickson - andersen ( fa ) facilitated spin model @xcite as an illustrative system . in the following section , we probe the distribution of ( [ equ : bubble_dist ] ) by analysing the distributions of space - time observables , such as the total amount of dynamical activity inside a given spatial region , integrated over a finite observation time . the forms of these distributions also imply that @xmath26 , in accordance with our analytic arguments . we also compare the fa model with two other systems . in ref . @xcite , a model with appearing and annihilating excitations ( the so - called aa model ) was shown to have the same two point correlations as the fa model ( up to a multiplicative factor ) . however , we show that this model has mesoscopic fluctuations that are very different from those of the fa model . conversely , breaking microscopic reversibility in the fa model moves moves the system into the directed percolation ( dp ) universality class @xcite . in that case , the the two point functions and critical scaling change qualitatively . however , the active and inactive states still coexist , so the fluctuations are still well - described by the phase coexistence analogy . the paper is organized as follows . in section [ sec : model ] we define the model , discuss its trajectories , and introduce the observables of interest . we also show that reducibility of the dynamics is essential to see the phase coexistence effect . in section [ sec : action ] we formalise our discussion of space - time thermodynamics by considering the distribution of the dynamical action that is analogous to the thermodynamic free energy . we consider the effect of breaking microscopic reversibility in section [ sec : dp ] . kinetically constrained models ( see ref . @xcite for a comprehensive review ) are defined in such a way that their non - trivial behaviour is purely dynamical in origin . thus , they are the natural framework for investigating dynamical heterogeneity and its consequences @xcite . a very simple kinetically constrained model is the single - spin facilitated fa model @xcite in dimension @xmath16 . it is defined for a chain of @xmath27 binary variables @xmath28 , with trivial hamiltonian @xmath29 . the model evolves under the following monte - carlo ( mc ) dynamics . at each mc iteration a randomly chosen site @xmath30 changes state according to : @xmath31 where we have set @xmath32 and @xmath33 . the non - trivial part of the dynamics is due to the facilitation function @xmath34 that is , a spin flip on site @xmath30 can take place only if at least one if its nearest neighbours is in the excited state . since @xmath35 does not depend on @xmath36 then the time evolution of the model obeys detailed balance with respect to @xmath37 at temperature @xmath38 . the mean density of up spins in the active state is @xmath39 the unit of time is an mc sweep ( @xmath27 attempted spin flips ) . in the following , we consider an fa model in which @xmath27 is to be taken to infinity in the thermodynamic limit . we will take a subsystem of this model , containing @xmath40 spins , @xmath41 . the length @xmath42 is to remain finite : it is a mesoscopic quantity . a trajectory for the subsystem specifies the state of the @xmath43 spins at @xmath44 mc steps . the trajectory is defined within a space - time observation box of size @xmath45 . we consider observables such as the density of excitations within the box , for a given trajectory : @xmath46 where @xmath47 is the state of the @xmath30th spin in the @xmath7th state of the trajectory . note that the boundaries @xmath48 and @xmath49 do not appear in the sum . the quantity @xmath50 is analogous to the net magnetization of a @xmath15-dimensional ising model , and thus we often refer to it as the `` magnetization '' . however , its physical meaning is that of mobility or activity , i.e. , the net number of cells in the observed space - time volume that exhibit an appreciable degree of molecular motion . this meaning is that of garrahan and chandler s coarse grained view of structural glasses @xcite . in an atomistic treatment , a corresponding quantity can be expressed in terms of the observable @xmath51\ } $ ] where @xmath52 is the position of the @xmath53th particle at time @xmath54 , and @xmath55 is a microscopic wave vector ( on the order of the inverse particle size ) . if @xmath56 is small , then particle @xmath53 experienced a relaxation event between times @xmath54 and @xmath57 . for a subsystem of particles , @xmath58 , we define the density of relaxation events , or net mobility , in space and time to be @xmath59\ ] ] where @xmath60 is the number of particles in the subsystem @xmath58 . the quantity @xmath61 is parametrised by microscopic length and time scales ( @xmath62 and @xmath63 ) , and is summed over a mesoscopic region of space - time . the parameters @xmath62 and @xmath63 coincide with coarse graining lengths and time , respectively . the fluctuations in @xmath61 are related to four point observables @xcite ; the distribution of @xmath61 is also related to the distributions of correlation functions measured in @xcite . results for distributions of the density of relaxation events in atomistic glass - formers show non - gaussian tails similar to those that we present here @xcite . however , the purpose of this article is to discuss generic properties of dynamically heterogeneous systems , so we restrict ourselves to very simple models . we consider the distribution of the trajectory activity , or box magnetisation @xmath50 : @xmath64 the sum is over all possible trajectories of the system , with their associated probabilities @xmath65 . in the limit of @xmath66 or @xmath67 , the central limit theorem states that @xmath68 will be sharply peaked around @xmath69 , with a variance that scales as @xmath70 . the observation of ref . @xcite was that while this is the case in the limit of large @xmath71 , the distribution of @xmath72 remains non - trivial even for @xmath73 and @xmath74 , where @xmath75 and @xmath76 are the correlation length and time associated with the zero temperature dynamical fixed point of this model . at @xmath77 , @xmath4 and various observation times . we use @xmath78 which is large enough so that @xmath68 does not depend on @xmath27 . the exponential tails of @xmath68 all have similar gradients : the dotted lines are @xmath79 with @xmath80 . , width=321 ] we show various @xmath68 in fig . [ fig : pm_simple ] . the peak of @xmath68 is gaussian , and this corresponds to trajectories like the one on the top - right of fig . [ fig : bubble_sketch ] . the tails of @xmath68 are exponential for small @xmath72 . this data was obtained using a combination of transition path sampling ( tps ) @xcite and umbrella sampling @xcite . for details see the appendix . the trajectories in the tail are dominated by rare large regions which have no excitations , like the the one on the bottom - right of fig . [ fig : bubble_sketch ] . to explain the presence of the exponential tail , suppose that we have a trajectory with an initial condition containing a large empty region of size @xmath17 , and that this empty region persists throughout @xmath81 , leading to a box magnetisation that is small . now consider a trajectory whose initial condition has an empty region of size @xmath82 , but is otherwise identical to the original one . then the probabilities of the two trajectories are in the ratio @xmath83 , and their magnetisations differ by approximately @xmath84 . thus we predict @xmath85 . \label{equ : pm_naive}\ ] ] the significance of this result is that the right hand side is independent of @xmath81 : increasing the observation time changes the factor multiplying the tail in @xmath68 , but the gradient of the tail remains constant . [ fig : pm_simple ] shows that the gradient is indeed independent of @xmath81 . however , the quantitative prediction for its value is accurate only to within 20 - 30% . this inaccuracy arises because the large inactive region may extend beyond the edge of the box thus , adding extra down sites to the initial condition may decrease the magnetisation by an amount less than @xmath86 . at @xmath77 for varying @xmath42 and @xmath87 . we use @xmath88 which is large enough that the results do not depend on @xmath27 . ( top ) increasing observation time at at fixed @xmath89 . ( middle ) increasing box size @xmath42 at @xmath90 . as @xmath42 or @xmath81 is increased , we move from a regime in which the tail gradient is independent of the increasing parameter to a regime in which the gradient is proportional to that parameter . ( bottom ) we show a typical trajectory for large @xmath14 and @xmath89 where the observation box is outlined : the size of the total spatial region shown is @xmath91 . for large @xmath42 the trajectories are of the form shown in fig . [ fig : bubble_sketch ] . , width=321 ] at @xmath77 for varying @xmath42 and @xmath87 . we use @xmath88 which is large enough that the results do not depend on @xmath27 . ( top ) increasing observation time at at fixed @xmath89 . ( middle ) increasing box size @xmath42 at @xmath90 . as @xmath42 or @xmath81 is increased , we move from a regime in which the tail gradient is independent of the increasing parameter to a regime in which the gradient is proportional to that parameter . ( bottom ) we show a typical trajectory for large @xmath14 and @xmath89 where the observation box is outlined : the size of the total spatial region shown is @xmath91 . for large @xmath42 the trajectories are of the form shown in fig . [ fig : bubble_sketch ] . , width=321 ] at @xmath77 for varying @xmath42 and @xmath87 . we use @xmath88 which is large enough that the results do not depend on @xmath27 . ( top ) increasing observation time at at fixed @xmath89 . ( middle ) increasing box size @xmath42 at @xmath90 . as @xmath42 or @xmath81 is increased , we move from a regime in which the tail gradient is independent of the increasing parameter to a regime in which the gradient is proportional to that parameter . ( bottom ) we show a typical trajectory for large @xmath14 and @xmath89 where the observation box is outlined : the size of the total spatial region shown is @xmath91 . for large @xmath42 the trajectories are of the form shown in fig . [ fig : bubble_sketch ] . , width=166,height=113 ] in fig . [ fig : pm_simple ] , the exponential tail of @xmath68 depends only on @xmath42 and not on @xmath81 , but this is a result of the shape of the observation box : the behaviour of @xmath68 is symmetric with respect to @xmath42 and @xmath81 . in fig . [ fig : pm_l20 ] we show how @xmath68 changes as we change the aspect ratio of the observation box . there is a crossover at @xmath92 , where @xmath93 at @xmath77 . for @xmath94 the gradient of the exponential tail is independent of @xmath81 as described above , but for @xmath95 then the argument must be modified . typical trajectories at small @xmath72 change their form to that shown in fig.[fig : pm_l20 ] ( right ) . in that case , increasing the width of the bubble does not change @xmath72 ; we must instead increase the time for which the inactive state persists . repeating the argument leads to a situation in which the gradient of the exponential tail is proportional to @xmath81 and independent of @xmath42 , see fig.[fig : pm_l20 ] . in other words , the gradient of the exponential tail is proportional to @xmath96 . the spatial and temporal extent of the box enter on equal footing . the data of fig.[fig : pm_l20 ] is consistent with the crossover from large @xmath42 to large @xmath81 occuring when @xmath81 is of the order of the first passage time across a bubble of size @xmath42 . this time scales as @xmath97 since the spreading velocity @xcite in the fa model is @xmath98 . consistent with the known scaling of the fa model in one dimension @xcite , we find that the temperature dependent of @xmath68 can be accounted for by using rescaled variables . the function @xmath99 depends only weakly on @xmath100 , as shown in fig . [ fig : fa_scaling ] . while subleading corrections to the scaling do appear to be significant , there is no qualitative change on lowering the temperature . this justifies our working at inverse temperatures around @xmath77 , where the data is easier to obtain than in the more glassy regime of the fa model ( @xmath101 ) . as a final observation in this subsection , we note that the behaviour of @xmath68 for small observation boxes , @xmath102 or @xmath103 , is different from that for large boxes . this is shown in fig.[fig : pmfa_small ] . the distribution acquires a secondary peak at @xmath104 : there is a finite probability of observing @xmath105 , so that @xmath68 diverges as @xmath106 . we will return to this feature below . in the previous subsection , the non - trivial structure in @xmath68 occurs in the tails of the distribution . it is associated with rare collective fluctuations in the system . we will argue below that these rare fluctuations do contain significant information about the model . we first compare the magnetisation distributions of the fa model and the aa model of @xcite . the aa model has the same two - point correlation functions as the fa model in equilibrium , but other properties of the two systems are qualitatively different . we define the aa model by specifying local moves for a chain of binary variables : @xmath107 where @xmath108 is an arbitrary temperature dependent factor that rescales the time , chosen for later convenience . these rates again respect detailed balance with respect to @xmath109 , and the inverse temperature is @xmath110 . correlation functions in the steady state of the aa model at a given temperature are related to those in the fa model at a different temperature @xcite ( strictly this holds when the facilitation function is proportional to the number of up neighbours , not when it is equal to zero or one as here , but there is no qualitative change to the physics ) . the relationship between correlators depends only on simple multiplicative factors , for example @xcite @xmath111 where @xmath112 is the inverse temperature in the fa model and @xmath110 the inverse temperature in the aa model . equation ( [ equ : aa_fa_prop ] ) holds when @xmath113 in the fa model at various temperatures , scaled according to ( [ equ : temp_scaling ] ) . we plot @xmath114 : the box sizes are @xmath115 ; the observation times are @xmath116 ; and we use @xmath117 . these temperatures are not very small , so there are subleading corrections to scaling , but there is no qualitative change to the scaled distribution on lowering the temperature . further , the computational time required at @xmath118 is quite significant , so we can not rule out small systematic errors arising from non - convergence of our tps procedure ( see appendix).,width=321 ] it is clear from ( [ equ : aa_fa_prop ] ) that the scaling behaviour and critical properties of the fa and aa models are the same . however , their activity or magnetisation distributions are different , as shown in fig . [ fig : aa_fa ] . the exponential tail is absent in the aa case . moreover , the persistence functions of the two models are different [ the persistence function is the probability that a site does not flip at all in an interval of length @xmath54 ] . the fluctuations responsible for the tails in @xmath68 can be linked to the decoupling of exchange and persistence times in the fa model @xcite , a feature which is absent in the aa case . we argue that there are important differences between the dynamics of the fa and aa models : subsystem observables like @xmath68 make these differences clear where two - point functions like @xmath120 do not . at @xmath77 , @xmath121 , @xmath122 , showing secondary maximum at small @xmath72 . ( inset ) enlargement of the secondary peak , shown in a linear scale for @xmath68.,width=321 ] we have shown that the distribution of subsystem magnetisations in the fa model has an exponential tail at small magnetisation . we also showed that this tail comes from rare mesoscopic regions of space - time in which there are no relaxation events . in this section we study the statistical mechanics of the space - time configurations of the subsystem . we draw an analogy between these space - time statistics and the thermodynamic statistics of a system near a phase coexistence boundary . where @xmath123 is the variance of the instantaneous magnetisation . parameters are @xmath4 , @xmath90 , @xmath78 ; in the fa model @xmath77 ; in the aa model @xmath110 is given by ( [ equ : aa_temp ] ) with @xmath77 . for the aa model , @xmath68 is close to gaussian . the standard deviation @xmath124 is _ not _ trivially related to the variance of the box magnetisation , so the fact that the gaussian parts of the two distributions are very similar is a non - trivial consequence of the exact mapping between the two models.,width=321 ] to make this analogy concrete , we define the probability of a trajectory being generated by some ( markovian ) dynamical rules : @xmath125 where @xmath126 represents the state of the system at time @xmath7 ; the transition probabilities are denoted by @xmath127 ; @xmath128 is the probability of the trajectory of the two boundary spins ; and @xmath129 is the probability of the initial condition ( @xmath130 ) . we note that the transition probabilities into a state @xmath131 depend on the state of the boundary spins as well as the state @xmath132 . it is convenient to define the dynamical action , @xmath133 and the dynamical partition sum , @xmath134 which is equal to unity when @xmath135 . the extensivity properties of the action @xmath136 will be discussed below . to make the correspondence with a thermodynamic partition sum , we simply define @xmath137 where the sum is over configurations of a system ; @xmath138 is the energy of a configuration ; and @xmath112 is the inverse temperature . the application of thermodynamic formalism to dynamical objects such as @xmath139 is found in the work of ruelle @xcite . a very recent and detailed discussion of its application to markov chains is given in @xcite ; the parameter @xmath140 of that paper is @xmath141 in our notation ; the kolmogorov - sinai entropy @xmath142 of the ruelle formalism is the mean action , @xmath143 in our notation . to treat phase coexistence in ordinary thermodynamics , one may decompose the partition function into its pure states @xcite @xmath144 where @xmath145 is the contribution of pure state @xmath146 to the partition sum . for example , in the ising system then the two pure states have positive and negative magnetisations , and statistical weights @xmath147 and @xmath148 . the difference in free energy density between the two states is @xmath149 at phase coexistence , we have @xmath150 . the result is that subextensive contributions to @xmath151 from boundary conditions are sufficient to select one phase over the other . by analogy , for dynamically heterogeneous systems , we make an decomposition of @xmath152 into its reducible partitions @xcite . if there are two such partitions then we write @xmath153 where the two terms are the probabilities that an initial condition lies in one partition or the other . we then define @xmath154 , \label{equ : mu_ic}\ ] ] by analogy with the static free energy difference @xmath22 . the probabilities @xmath155 and @xmath156 are independent of @xmath14 , so @xmath11 vanishes as @xmath14 is taken to @xmath157 , and an initial condition on the system is sufficient to select the active or inactive partition of the dynamics : see figure [ fig : bc ] . in a similar way , a boundary condition is sufficient to select the up or down pure state in an ising system below the critical temperature . recalling ( [ equ : bubble_dist ] ) and ( [ equ : ising_dist ] ) , we observe that phase coexistence and reducibility lead to distributions with @xmath150 and @xmath26 respectively . the free energy cost associated with large bubbles ( or down domains ) comes from their boundaries and not from the bulk . [ fig : bubble_sketch ] illustrates how fa model trajectories with low densities of relaxation events typically contain a large compact `` bubble '' in which there are no relaxation events . in the fa model then these large bubbles are more common than homogeneous reductions in the density . conversely , in the aa model , the large bubbles are absent and the homogeneous regions of low density determine the tail of the magnetisation distribution ( recall figure [ fig : aa_fa ] ) . the fa model is reducible , so the distribution of large bubbles will be that of ( [ equ : bubble_dist ] ) , with @xmath26 [ recall ( [ equ : mu_ic ] ) ] . the derivation of the magnetisation distribution from the bubble distribution depends on the ratio of @xmath42 and @xmath14 . for observation windows such that @xmath158 , trajectories with small @xmath50 contain large bubbles that span the temporal extent of the observation window , as in the lower right panel of figure [ fig : bubble_sketch ] . in that case , we have @xmath159 .\ ] ] where @xmath6 is the spatial extent of the bubble . the probability of the box containing a bubble of this size and shape is @xmath160 assuming that @xmath12 is not too small , so that @xmath161 is sharply peaked around @xmath162 , then @xmath163 , \ ] ] for @xmath164 $ ] . unless @xmath11 is very small , the term proportional to @xmath11 will dominate at large @xmath42 , leading to a gaussian distribution . however , in the case of small @xmath11 , the distribution is exponential in @xmath72 , and the gradient of this exponential tail is proportional to @xmath42 and independent of @xmath81 , as observed in fig . [ fig : pm_simple ] . a similar argument holds in the opposite regime of @xmath165 and @xmath166 $ ] , which explains the exponential tails of fig . [ fig : pm_l20 ] . pursuing the analogy with phase coexistence , we identify the active partition of the dynamics with a dense phase and the inactive partition with a sparse phase . the density of dynamical activity in the active phase is @xmath167 . the typical length scale associated with bubbles of the sparse phase is @xmath168 which scales as @xmath169 . thus , the typical bubble size and the typical spacing between excitations scale in the same way as the temperature is reduced . this scaling @xcite is consequence of detailed balance in the fa model , which implies that excitations are uncorrelated at equal times : @xmath170 . in section [ sec : dp ] , we will show how generalising the fa model to a system which does not obey detailed balance leads to a situation in which the typical bubble sizes are larger than the excitation spacing . this situation is the usual one in systems at phase coexistence : the fa model is a special case because of the extra symmetry of detailed balance . in the analogy between static and dynamic partition sums , the dynamical action of eq . ( [ equ : def_action ] ) corresponds to the energy of the static system . for the fa model , the transition probabilities are @xmath171 where the projector @xmath172 takes the value of unity if @xmath140 and @xmath173 differ in the state of exactly one spin , otherwise it is equal to zero ; @xmath174\ ] ] accounts for transitions into state @xmath140 ; and @xmath175 \right\}\ ] ] accounts for transitions from state @xmath140 . the symbol @xmath176 is equal to unity if and only if states @xmath140 and @xmath173 are identical , and the function @xmath177 was defined in ( [ equ : def_f ] ) : its value is unity if spin @xmath30 is free to flip ; otherwise it is zero . these operators enforce the constraint that only one spin flips on each time step . taking the limit of large @xmath27 , and using ( [ equ : def_action ] ) , we arrive at the action for an allowed trajectory : @xmath178 \nonumber \\ & & + ( \hbox{boundary terms } ) , \label{equ : def_sn}\end{aligned}\ ] ] where @xmath179 is the total number of flips from state @xmath180 to state @xmath181 inside the observation box and , similarly , @xmath182 is the number of flips from @xmath181 to @xmath180 . of course , trajectories containing transitions that are not allowed by the dynamical rules have @xmath183 and their action is formally infinite ; we ignore them in what follows . we now discuss briefly the extensivity properties of the dynamical action . the probability @xmath65 is associated with the states of @xmath42 spins over a time period @xmath14 . however , the positions of the spin flip events in time are specified with a resolution that depends on the system size , @xmath27 . thus , the probability of any trajectory vanishes as @xmath184 and the action diverges logarithmically with @xmath27 as well as being extensive in @xmath42 and @xmath14 . we therefore introduce a coarse - graining timescale @xmath63 . that is , we define @xmath185 where the sum is over trajectories with the same spin flips as the original trajectory , but the flips may happen within an interval @xmath186 of their original times . the second , approximate , equality indicates that as long as @xmath186 is not too large then the probability of all trajectories in the sum will be approximately equal , and the number of such trajectories is @xmath187 . the result is that @xmath188 \nonumber \\ & & + ( \hbox{boundary terms } ) , \label{equ : def_sdt}\end{aligned}\ ] ] where we have converted the sum to an integral , which is valid in the limit of large @xmath27 . this definition of the action is parametrised by the coarse - grained time scale @xmath186 . for fixed @xmath63 , distributions of @xmath189 have a peak whose position is independent of @xmath27 and extensive in @xmath42 and @xmath81 . our results do not depend qualitatively on @xmath63 ; the numerical results below are at @xmath190 . finally , we have no prescription for calculating the boundary terms in the action ; these terms are not extensive in @xmath191 , so we neglect them in what follows . , @xmath90 , obtained with @xmath78 . ( top ) contour plot of the joint probability distribution for action density and magnetisation @xmath192 ( obtained from @xmath193 independent trajectories ) . the contours are at @xmath194 . the dotted line is the prediction ( [ equ : sm_dist ] ) . ( bottom ) distribution of the action density @xmath195 [ where @xmath196.,width=287 ] , @xmath90 , obtained with @xmath78 . ( top ) contour plot of the joint probability distribution for action density and magnetisation @xmath192 ( obtained from @xmath193 independent trajectories ) . the contours are at @xmath194 . the dotted line is the prediction ( [ equ : sm_dist ] ) . ( bottom ) distribution of the action density @xmath195 [ where @xmath196.,width=287 ] results for the distribution of the dynamical action density , @xmath197 , and its joint distribution with the box magnetisation , are shown in figure [ fig : fa_action ] . we see that the distribution of the action density is tightly correlated with that of the magnetisation . to understand this , we define the density of facilitated spins in a trajectory to be @xmath198 now , facilitated down spins flip with rate @xmath199 and facilitated up spins flip with rate unity . since @xmath200 , there are many such independent events within the observation window , so that fluctuations in the number of these events are small . therefore , @xmath201 where the approximate equalities indicate that the joint distribution of each pair of observables are sharply peaked around these values ( we have verified this by measuring joint distributions of the form shown in fig . [ fig : fa_action ] ) . further , each up - spin is associated with two facilitated spins . taking into account spins that are facilitated by two separate up spins , we have @xmath202 where the approximate equality holds in a similar sense , and @xmath50 is the magnetisation or activity of the trajectory of interest , defined in ( [ equ : def_mag ] ) . hence , taking these results together , we predict that the joint distribution of magnetisation and action density is sharply peaked around @xmath203 , \label{equ : sm_dist}\ ] ] in accordance with the results of figure [ fig : fa_action ] . we conclude that the exponential tail in the magnetisation distribution is intrinsically linked with the exponential tail in the distribution of the dynamical action . numerical evidence for this result was given in @xcite , but without concrete explanation . the definition ( [ equ : zb ] ) motivates us to define a new ensemble for the dynamics , in which the probabilities of trajectories are @xmath204 this ensemble has the action distribution @xmath205 . \label{equ : pbs}\ ] ] hence , if @xmath206 has an exponential tail with gradient @xmath146 then @xmath207 has a similar exponential tail with gradient @xmath208 . increasing @xmath209 reduces the gradient , so large inactive regions become more common . observables within the @xmath209-ensemble can be measured by averaging over the dynamics at @xmath135 and reweighting the resulting trajectories according to their action . see the left panel of figure [ fig : dp_simple ] . since the magnetisation and action are tightly correlated , increasing @xmath209 leads to a reduction in the gradient of the exponential tail of @xmath68 . we associate the vanishing of this gradient with the proliferation of trajectories with large bubbles . in @xcite , it was argued that this proliferation would appear as a first order transition to a state with large inactive bubbles , if an appropriate dynamics could be found to generate the @xmath209-ensemble . for a given @xmath42 and @xmath14 , a stochastic process exists whose statistics are those of the @xmath209-ensemble ( the probability of each trajectory is known , and an appropriate dynamics can be constructed ) . this construction is extremely laborious , since the transition probabilities depend explicitly on the size of the space - time box , and on the position ( and time ) within it . alternatively , one can simulate master equations such as those of @xcite , which do not conserve probability . such a simulation requires a dynamics in which ` clones ' of the system are created and destroyed , according to certain rules @xcite . the connection to experimentally accessible dynamics for either alternative is not yet apparent , which limits analysis of the predicted first order transition . on the other hand , the effects of phase coexistence can be probed experimentally by moving towards the second order critical point at the end of the phase coexistence boundary . as discussed in section [ sec : discuss_dist ] above , this critical point is approached as @xmath210 in the fa model . in that limit , the order parameter @xmath211 and the surface tension @xmath212 vanish with the same power of @xmath100 . the function @xmath68 scales according to ( [ equ : temp_scaling ] ) , and it is hard to discern the two coexisting phases . however , we can modify the fa model by introducing an extra process whereby sites can flip from state @xmath181 to state @xmath180 , even when unfacilitated . this new process breaks detailed balance , so the resulting model is not appropriate for an equilibrium physical system , but such models are widely studied in a range of physical contexts @xcite . the rate for spins to flip from down to up is now finite at the critical point , and the new transition is in the directed percolation ( dp ) universality class , where the effects of phase coexistence are more apparent . to be precise , the generalised fa model has dynamical rules @xmath213 the generalised model has @xmath214 since all probabilities must be positive ; @xmath215 is the fa model . for @xmath216 then the system no longer obeys detailed balance , and @xmath112 can no longer be interpreted as an inverse temperature . , for different values of @xmath112 with @xmath217 . the axis @xmath215 is the fa model and the axis @xmath105 is a line of critical points . the dotted line separates the region in which the scaling of directed percolation ( dp ) will apply from those in which @xmath218 can be treated perturbatively [ so the scaling will be that of the coagulation - diffusion ( cd ) fixed point ] . the fa model ( @xmath215 ) is the unique case for which the critical scaling is coagulation - diffusion ; for finite @xmath218 the relevant critical point is dp . , width=321 ] the qualitative behaviour of the model with finite @xmath218 is shown in fig . [ fig : dp_sketch ] . the `` dimensionless '' parameter that determines the effect of @xmath218 is the product of the death rate and the relaxation time of the fa model : @xmath219 which defines @xmath220 . at any finite @xmath112 there is a directed percolation transition to an active state , which occurs when @xmath220 is of order unity . this transition is accompanied by a diverging static correlation length : @xmath221 where @xmath222 is the length scale that diverges at the transition and @xmath223 and @xmath224 are dp exponents @xcite ( the dimensionality is @xmath16 ) . on the other hand , if @xmath225 then there is a transition at @xmath226 that is controlled by the coagulation - diffusion fixed point . in that case we have @xmath227 for all @xmath100 . we show in the figure how the crossover into the dp critical region is always relevant if @xmath220 is finite , but does not affect the behaviour if @xmath225 . we note that between two and four dimensions a similar picture holds , but the scaling at @xmath225 will be gaussian @xcite . turning to the dynamical action , we define @xmath228 + n_{i\tau } r \right\ } \nonumber\\ \label{equ : def_sr}\end{aligned}\ ] ] by analogy with ( [ equ : def_sdt ] ) , where we have set the coarse - graining time scale @xmath190 for ease of writing . equations ( [ equ : fac_rel1]-[equ : nf_rel ] ) remain true since facilitated spins are still equilibrated at a density close to @xmath229 ( but note that the mean density in the system will be less than @xmath100 : the death process reduces the mean density ) . in fig . [ fig : dp_simple ] we plot the distribution of the action density at finite @xmath218 and compare it with the behaviour of the @xmath209-ensemble . since the system does not obey detailed balance the tps procedure becomes inefficient : data at finite @xmath218 is obtained by simple binning and histogramming . in the @xmath209-ensemble , the mean action is reduced as the ( finite ) system crosses over from an active to an inactive state ( this crossover would be a first order phase transition in the thermodynamic limit ) . on the other hand , increasing @xmath218 reduces the gradient of the exponential tail and the mean action , as the second order phase transition is approached ( at @xmath230 ) . with conditions otherwise similar to fig . [ fig : pm_simple ] ( @xmath77 , @xmath4 @xmath231 ) . ( left ) sample from centre of distribution . ( right ) sample with @xmath232 . clearly there are more large inactive regions in these trajectories than in those of fig . [ fig : bubble_sketch ] ; increasing @xmath220 from zero leads to proliferation of large `` bubbles''.,title="fig:",width=151,height=75 ] with conditions otherwise similar to fig . [ fig : pm_simple ] ( @xmath77 , @xmath4 @xmath231 ) . ( left ) sample from centre of distribution . ( right ) sample with @xmath232 . clearly there are more large inactive regions in these trajectories than in those of fig . [ fig : bubble_sketch ] ; increasing @xmath220 from zero leads to proliferation of large `` bubbles''.,title="fig:",width=151,height=75 ] the key difference between the fa model and its generalisation to finite @xmath218 is that the surface tension and mean magnetisation decouple at finite @xmath218 . on reducing the temperature in the fa model ( @xmath215 , @xmath233 ) we have @xmath234 , so the typical bubble size @xmath168 scales in the same way as the inverse density of up spins . in contrast , as we approach the dp fixed point ( @xmath112 finite , @xmath235 ) then we have @xmath236 where @xmath237 are the exponents of the dp fixed point @xcite ( @xmath238 should not be confused with the parameter @xmath112 that enters the transition probabilities ) . as criticality is approached , the mean bubble size increases much faster than the inverse density since @xmath239 . the up spins cluster together , leaving large inactive regions consisting only of down spins . introducing the death process drives the system towards a phase separation into active and inactive regions of space - time . this can be seen in the trajectories of fig . [ fig : dp_traj ] . finally , we note that the critical dp scaling leads to a crossover at small @xmath220 , when the typical size of inactive regions exceeds the observation box ( @xmath240 ) . in that case , @xmath195 acquires a second peak at zero magnetisation ( see the largest value of @xmath220 in fig.[fig : dp_simple ] ) . this is analogous to the situation for small boxes ( @xmath241 ) in the fa case ( recall fig . [ fig : pmfa_small ] ) . in the thermodynamic language , it corresponds to a second minimum in the free energy , as would be observed near a first order phase transition . here , the phase transition is second order , but the finite size of the observation box means that the divergence of the correlation length is cut off at the size of the observation box . thus , we have shown that adding a death process to the fa model strengthens the analogy between trajectory statistics and the statistical mechanics of phase coexistence . in the fa model , the only length scale is the inverse density of up spins ; if @xmath218 is finite then a new length scale appears : the dp correlation length . this second length scale allows the density of active regions to decouple from their separation , leading to richer behaviour than that of the fa model ( which corresponds to the particular case of the phase coexistence in which the two length scales are equal ) . throughout this article , we have emphasised the broad applicability of the idea of space - time phase coexistence by focussing on general features of the models , such as microscopic reversibility and reducibility of the dynamics . we expect the behaviour described here to be generic in kinetically constrained models with reducible dynamics @xcite . more generally , the extent to which this behaviour can be observed in real structural glass - formers remains an open question . we thank fred van wijland for important comments on links with the ruelle formalism . we also benefited from discussions with mauro merolle and tommy miller . rlj was supported in part by nsf grant no . che-0543158 ; jpg by epsrc grants no . gr / r83712/01 and no . gr / s65074/01 , and by university of nottingham grant no . fef 3024 ; and dc by the u. s. department of energy grant no . de - fg03 - 87er13793 . we calculated probability distributions for various observables in the fa model : these were obtained by a combination of transition path sampling ( tps ) @xcite and umbrella sampling @xcite . here we give a brief description of the procedure used . transition path sampling is a monte carlo procedure applied to trajectories of a dynamical system ( a trajectory is a particular realisation of the dynamics ) . we use it to efficiently sample restricted ensembles of trajectories . the procedure is well - established @xcite , but we present some information regarding its application to the fa model for completeness . suppose that we wish to sample trajectories with magnetisation @xmath72 smaller than some reference value @xmath242 . the simplest way to do this is to generate statistically independent trajectories , accepting only those that satisfy the restriction @xmath243 . however , if @xmath242 is much smaller than the mean of @xmath72 then this is inefficient . instead , we can generate unbiased trajectories satisfying the restriction by deforming a set of initial trajectories , as long as the deformations satisfy detailed balance with respect to the probability distribution within the constrained ensemble of trajectories . such deformations are called tps moves . one possibility is to take a trajectory of length @xmath81 and keep only the part with @xmath244 with @xmath245 ; we the generate the rest of the trajectory using the dynamical rules prescribed for the system ( we use a continuous time monte carlo algorithm @xcite ) . if the new trajectory has @xmath243 then it replaces the old one ; otherwise the move is rejected and we retain the old trajectory . this is a `` shooting '' move @xcite . we couple this the move with the reverse procedure in which we discard the part of the trajectory with @xmath246 and regenerate the rest of the trajectory by propagating backwards in time ( since we have detailed balance then the steady state of the fa model is invariant under time - reversal , so this is a valid way to generate unbiased trajectories ) . in addition to these moves , we also use `` shifting '' moves @xcite in which we shift the trajectory in time , discarding the parts of the new trajectory with @xmath247 or @xmath248 . we then regenerate the remaining parts of the new trajectory . we find that this combination of moves is quite efficient for exploring the restricted ensembles of interest . we use umbrella sampling @xcite to calculate probability distributions by measuring ratios such as @xmath249 where @xmath250 and @xmath251 are two cutoffs for the variable @xmath72 with @xmath252 . consider an ensemble of all allowed trajectories for the system , with their statistical weights . in order to measure @xmath253 we sample a restricted ensemble which contains only trajectories with @xmath254 ; these trajectories have the same weights as they would have in the original ensemble . we then measure the probability that @xmath255 within the restricted ensemble : this probability is equal to @xmath253 . 1 . start with a representative set of trajectories from the unrestricted ensemble . 2 . choose an ordered set of cutoff magnetisations @xmath256 for which we will calculate the probabilities @xmath253 . 3 . explore the unrestricted ensemble , measuring the fraction of trajectories with @xmath243 . ( this is done by sampling independent trajectories . ) once we have a good enough estimate for @xmath257 , start a restricted ensemble with trajectories satisfying @xmath243 . typically we store @xmath258 such trajectories . 5 . explore the restricted ensemble using tps , measuring the fraction of trajectories with @xmath259 . typically this takes @xmath260 tps moves per ensemble member . once we have a good enough estimate for @xmath261 , we discard all trajectories with @xmath262 and replace them by trajectories with @xmath259 . these trajectories are generated by continuing the tps procedure and accepting all trajectories that satisfy the new constraint . the resulting set of trajectories are not statistically independent so we equilibrate the new ensemble by allowing it to evolve from the biased initial condition . typically we use around @xmath263 tps moves per ensemble member . we test for equilibration by tracking the fraction of trajectories with @xmath264 , since this will be the quantity that we will measure on the next step . we then repeat steps 5 and 6 for increasingly restricted ensembles . at each step , we measure @xmath253 .

psr b105552 is a middle - aged ( 0.5 myr ) energetic ( @xmath2 erg s@xmath3 ) pulsar which has been detected as a source of pulsed @xmath0-rays . in their study of energetic pulsars , @xcite singled out psr b105552 for comment since it is exceptional in having both an interpulse and wide radio components , with the interpulse ( ip ) well separated from the main pulse ( mp ) . in general the presence of an ip is a strong clue to a pulsar s geometry , and suggests that a study of the radio emission of psr b105552 may give insight into the structure of the magnetosphere and the origin of its @xmath0-rays . psr b105552 was discovered with the molongolo telescope by @xcite , who noted a possible ip in one of their records . the ip was confirmed by @xcite , who commented that the pulse profile resembles that of the crab pulsar . @xcite noted that the mp - ip separation shows no variation with frequency , which is additional evidence that the pulsar s magnetic axis is highly inclined to the rotation axis . @xcite concluded that the observed polarisation position angle ( pa ) swings of the mp and ip suggest an angle of inclination of the magnetic axis of @xmath1 , without ruling out higher values around @xmath4 . this general view was supported in a complex study published by @xcite . psr b105552 is relatively close , which makes it possible to observe its emission over a wide range of wavelengths . the pulsar s @xmath5 distance is only @xmath6 kpc @xcite , roughly consistent with the x - ray data ( e.g. @xcite ) and the non - thermal radio source around psr b105552 ( @xcite ) . the egret detector ( energetic gamma - ray experiment telescope ) on board of the _ compton gamma ray observatory _ ( cgro ) revealed that psr b105552 is a pulsed @xmath0-ray source ( @xcite ) . unlike most @xmath0-ray profiles ( but in common with psr b170644 ) , that of psr b105552 does not show two narrow peaks separated by 0.40.5 in phase , but has a broad @xmath0-ray profile with possible sub - peaks at the leading and trailing edge of the profile which peak 0.25 and 0.05 in phase before the radio mp @xcite . the leading side of the profile has also been seen with the comptel detector ( 0.7530 mev ) , while only the trailing side of the profile is seen above 2 gev by egret , suggesting a longitude dependent spectral index . another feature which makes psr b105552 an interesting source at high energies is its spin - down age of 0.5 myr , making psr b105552 ( at the time of writing ) the oldest of the known @xmath0-ray pulsars . its period of 0.197 seconds makes it also the slowest ( excluding geminga ) . this status as being both an ` old ' young pulsar and , at the same time a ` young ' old pulsar ( most radio pulsars are over 1 myr old ) suggests that it may possess characteristics of both and give insight into the link between high - energy production in the outer magnetosphere and radio emission above the polar cap . it has famously been dubbed as one of the ` three musketeers ' @xcite , along with psr b0656 + 14 and geminga , because these pulsars were found to emit solely thermal x - rays which are interpreted as the product of initial neutron star cooling . the geometry linking the high - energy and radio profiles was recently studied by @xcite , who made use of the pa analysis of @xcite . they concluded that that the radio mp and ip are emitted from separate poles with the @xmath0-ray emission coming from the same pole as the radio mp . they claim that the ip and mp come from different heights , but do not consider the effect of retardation and aberration on the pa - curve . in their model the radio emission occurs on fieldlines which are confined to a polar cap bounded by the last closed fieldlines which , in a dipole geometry , touch the light cylinder . here we use new data to re - analyse the pa - swings and to create a geometry which is fully consistent with retardation and aberrational effects . a full understanding of the geometry is an important prelude to undertaking an analysis of the pulsar s complex single pulse behaviour , which will appear in a subsequent paper . the organisation of this paper is as follows : in sect . [ sectobs ] details about the radio data used are provided , in sect . [ sectprofile ] the properties of the pulse profile of psr b105552 are described and its most likely geometry is derived in sect . [ sectgeometry ] . in sect . [ sectbeamillumination ] we identify the active fieldlines of the mp and ip and the implications of our results are discussed in sect . [ sectdiscussion ] . the pulse profile of psr b105552 was obtained at three different frequencies by the parkes 64-m radio telescope located in australia . during an observing session carried out from 2006 august 24 to 27 data was recorded using the h - oh receiver ( 1369 mhz centre frequency , 256 mhz bandwidth split into 1024 frequency channels with an equivalent system flux density of 43 jy ; @xcite ) and the 10/50 cm receiver ( 3094/653 mhz centre frequency , 1024/64 mhz bandwidth split into 1024/512 frequency channels with an equivalent system flux density of 49/57 jy on a cold sky ) . the signals from the two linear feeds of the receiver were converted into stokes parameters , resampled and folded at the pulse period by a digital filterbank using 512 bins across the profile . in addition we also used the pulse profile with a slightly higher time resolution ( 1024 bins across the profile ) obtained using the 20 cm multibeam receiver ( 1369 mhz centre frequency , 256 mhz bandwidth split into 1024 frequency channels with an equivalent system flux density of 35 jy on a cold sky ) for the fermi timing program . this program started in april 2007 with the aim to provide timing solutions which can be used to fold @xmath0-ray data obtained by the fermi satellite @xcite . the profile is the sum of 33 individual observations with a total length of 6.6 hours . for details about the data - processing we refer to @xcite . the shape of the pulse profile of psr b105552 is quite complex , as one can see in fig . [ fig1055profiles20 ] . in a similar way to @xcite ( who did not detect the first component of the mp ) , we call the four most prominent peaks of the mp peaks i to iv and the three peaks in the ip v to vii . long integrations are required to obtain stable profiles , with evidence for persistent intensity changes in component iii ( also noted by @xcite ) . the peak - to - peak separation between the mp and ip is @xmath7 , well below @xmath8 , so it is not immediately obvious that we are dealing with emission from both poles of the neutron star . as one can see in fig . [ fig1055profiles20 ] , the profile of the mp is highly linearly polarized . this is also the case for the leading edge of the ip , but at later pulse longitudes it becomes almost completely depolarized ( e.g. @xcite ) . notice that all three components of the ip have a distinct linear polarization peak , although they are much weaker for components vi and vii ( fig . [ fig1055profiles20 ] ) . the profiles of the degree of linear polarization at 10 cm and 50 cm ( not shown ) look very similar to that at 20 cm . component ii of the mp has a significant amount of circular polarization , although it is weak . there is a hint that the trailing part ip also has a significant amount of circular polarization , but with an opposite sign . _ top panels : _ the pulse profiles of psr b105552 at three observing frequencies drawn with different vertical offsets . from top to bottom the wavelength are 10 cm , 20 cm and 50 cm . _ bottom panels : _ the pa of the linear polarization and an rvm fit . the offsets are the same as in the top panels . these are the august 2006 observations . , title="fig:"]_top panels : _ the pulse profiles of psr b105552 at three observing frequencies drawn with different vertical offsets . from top to bottom the wavelength are 10 cm , 20 cm and 50 cm . _ bottom panels : _ the pa of the linear polarization and an rvm fit . the offsets are the same as in the top panels . these are the august 2006 observations . , title="fig : " ] the pulse profiles of psr b105552 at different frequencies are shown in fig . [ fig1055profiles ] . the profiles are aligned and normalized using the peak of component ii which is put at @xmath9 pulse longitude . one can see that there is not much evidence for frequency evolution in the relative position of the components , but the peak amplitudes do evolve with frequency . in the mp the first component disappears at 50 cm while components iii and iv become relatively stronger . the ip shows a similar frequency evolution with the leading half of the profile becoming weaker relative to the trailing component . in relative terms , the ip itself is strongest at high frequencies . while the components of the ip are sharper at 10 cm this is less clear in the mp . notice also that component vii is the only component which changes its centroid position with respect to the other components . there is also no evidence of the full widths of the mp and ip ( or their separation ) being frequency - dependent , potentially an important hint as to the pulsar s geometry . a profile at 170 mhz , ascribed to mcculloch et al . and published in @xcite , shows a merging of components iii and iv in the mp , but little overall widening of the profiles . the pa - swing is plotted in the bottom panel of fig . [ fig1055profiles ] for the three frequencies . we found a significant offset of the 50 cm data with respect to the other wavelengths when we use the rotation measure ( rm ) of @xmath10 rad m@xmath11 @xcite . by applying an rm of 46 rad m@xmath11 the pa - swings at the different frequencies line up , suggesting that the shape of the pa - swing , like that of the profile , is independent of frequency . this value of the rm is consistent with the value of @xmath12 rad m@xmath11 measured by @xcite . the shape of the pa - swing can be used to constrain the angles between the magnetic and rotation axis ( @xmath13 ) and the angle between the line of sight and the rotation axis ( @xmath14 , where @xmath15 is the impact parameter ) . according to the rotating vector model ( rvm ; @xcite ) the pa ( @xmath16 ) can be described as a function of the pulse longitude @xmath17 by @xmath18 where @xmath19 and @xmath20 are the pa and the pulse longitude corresponding to the inflection point of the pa - swing . using this definition implies that we follow the observer s sign convention for the pa , which differs from that introduced by @xcite and @xcite . if the emission height @xmath21 is zero , or at least negligible compared with the light cylinder radius @xmath22 , the inflection point of the pa - swing coincides with the fiducial point ( defined as the pulse longitude where the magnetic pole at the surface is directed towards us ) . however , for non - zero emission heights this is no longer the case . as the emission height increases the total intensity appears to shift towards earlier pulse longitudes due to aberration ( by @xmath23 ) and the decreasing light travel distance to the observer ( also by @xmath23 ) , while the pa - swing shifts to later pulse longitudes with respect to the fiducial point ( by @xmath24 ) . thus the total relative shift between the steepest gradient of the pa - swing with respect to the emission beam s centroid is given by @xmath25 ( @xcite ) , where @xmath21 is assumed to be small compared with @xmath22 . in psr b105552 the degree of linear polarization is low in the trailing half of the ip , but fortunately it is enough to obtain the pa . this means we have a full pa coverage at virtually every pulse longitude where emission is detected at a far better time resolution than was available to @xcite . this enables us to re - examine and place constraints on the geometric solution of this pulsar . in this section we formally confirm that a nearly - orthogonal solution is preferable to a nearly - aligned solution and exploit the @xmath0-ray profile to establish an integrated geometry . we first consider the possibility that psr b105552 might be a nearly - aligned rotator . in such a geometry the mp and ip originate from the two edges of a hollow cone - like beam from a single pole ( e.g. @xcite ) . the separation between the two components can be large , but , as observed , will not necessarily be @xmath26 in pulse longitude . the large widths of the components together with the steep trailing edge of the mp and the leading edge of ip are also typical for such geometry , making this model a promising possibility to consider . however , for a nearly - aligned rotator one expects the pa - swing to be flatter than is observed . in the rvm model the steepest gradient is associated with the line of sight passing the fiducial plane . in the case of an aligned rotator this plane should be at a pulse longitude between the mp and ip , hence the observed pa - swing will be relatively flat . as has been pointed out ( by @xcite for example ) , the steepness of the pa - swing observed for both the mp and ip rules out the one pole geometry . it should be added that this conclusion is also valid even when @xmath9 jumps in the pa - swing caused by different plasma modes dominating in different parts of the pulse profile ( e.g. @xcite ) are considered . the pa - swing inflection point can in principle be shifted from its unshifted pulse longitude between the mp and ip ( as would be expected for an aligned rotator ) to appear at the pulse longitude of one of the pulse components ( mp or ip ) when the aberrational shift of eq . [ eqbcw ] is considered . this effect could therefore in principle make the pa - swing steep in one of the two components . however , it can not make them steep in both since the rvm model predicts a steep gradient at only one point in the pa - curve , thereby failing to explain the observed pa - swing . in searching for a self - consistent solution , we follow @xcite and consider a model where the profile of psr b105552 is generated by emission from an orthogonal rotator . although @xcite claimed that the pa - swing can not be extrapolated between the mp and ip , we ( like @xcite with inferred data from @xcite ) had no problem in fitting the pa - curve with the rvm model ( eq . [ eqrvm ] ) . in fact , the reduced @xmath27 of 2.7 is quite good , although not perfect ( fig . [ fig1055chi2 ] ) . the lowest @xmath27 is found for @xmath28 and @xmath29 ( where @xmath30 and @xmath31 are the mp values of @xmath13 and @xmath15 ) , suggesting that the orientation of the magnetic axis is indeed close to orthogonal and in agreement with @xcite . this implies @xmath32 and an impact parameter at the ip of @xmath33 . note that these fits require that all the emission at each pole comes from the same height , a point to which we will return in sect . [ sectemissionheight ] . the best fit is shown in fig . [ fig1055profiles ] as the solid curve . the biggest discrepancy between the fit and the data occurs in the trailing component of the ip , where the fit is too flat compared with the model . the steepest gradient in the mp ( @xmath20 of eq . [ eqrvm ] ) is found exactly at the location of the highest peak in the pulse profile ( @xmath34 in the plots ) . the fact that the pa - swing can be extrapolated between the mp and ip further implies that the emission heights at both poles can not be too large . if there were a large differential emission height , the pa - swings of the mp and ip would be shifted by different amounts ( @xcite ) , making it impossible to fit the pa - swing with a single rvm curve . by modelling these relative shifts we find that the lowest @xmath27 is found when the emission height of the ip is higher than that of the mp by @xmath35 ( @xmath36 km ) with an uncertainty of @xmath37 ( @xmath38 km ) , which is consistent with the emission height being identical at both poles . the uncertainty is derived by finding the differential emission height at which the reduced @xmath27 of the pa fit has doubled . although the solution found by @xcite ( who find an ip / mp emission height differential of between @xmath37 and @xmath39 ) is in the allowed range , we will argue later that this solution is very unlikely . we therefore seek a solution with identical emission heights for the mp and ip . as @xcite pointed out , and as is evident in fig . [ fig1055chi2 ] , there are rvm solutions over a wide range of @xmath30 values which have a comparable @xmath27 . however , we can utilize additional constraints to confirm that we have the right solution . by assuming that @xmath40 must be smaller than @xmath9 ( or else the full opening angle of the radio beam would be larger than @xmath26 ) and that @xmath40 must be positive ( or else the pa - swing of the ip would decrease ) only solutions between the two diagonal solid lines in fig . [ fig1055chi2 ] are allowed . a further and critical constraint can come from the observed phase and width of the @xmath0-ray profile ( @xcite ) . different magnetospheric locations are proposed for the @xmath0-ray emission of pulsars . early models put the acceleration zones at low altitudes above the magnetic poles ( polar cap models ; e.g. @xcite ) . more recently , it has been argued that the emission height could be considerably larger for fieldlines close to the last open fieldlines giving rise to the `` slot gap '' model ( e.g. @xcite ) and the `` two - pole caustic '' model @xcite . a separate class of models are the so - called `` outergap '' models @xcite , which place the acceleration zones above the `` null lines '' @xcite . these are the lines on which the charge density of a pulsar magnetosphere changes sign ( @xcite ) . numerical simulations show that only an outer gap model can reproduce the observed shape of the @xmath0-profile of psr b105552 to a good approximation @xcite . the two - pole caustic model , which has @xmath0-ray production below the null line , predicts in general profiles with observed emission from both magnetic poles , which makes it difficult to reproduce the observed @xmath0-ray pulse width . in order to reproduce the observed single - peaked @xmath0-ray profile the two - pole caustic model predicts either @xmath13 or @xmath41 values smaller than allowed by the pa - swing parameters between the two solid diagonal lines of fig . [ fig1055chi2 ] . polar cap models , which have low emission heights , also require too small @xmath13 and @xmath41 values in order to get a broad enough @xmath0-ray profile . on the other hand outer gap models predict , by definition , that only emission from one pole can be seen for any given observer , thereby qualitatively reproducing the observed @xmath0-ray profile for a wide range of geometries . in order to reproduce the observed width of the @xmath0-ray profile the gap thickness @xmath42 ( the fraction of the angle from the last closed fieldline to the magnetic axis ) should be of the order of @xmath43 ( assuming that the @xmath0-ray efficiency is equal to @xmath42 ) , showing that the gap is very large in comparison with most @xmath0-ray pulsars @xcite . large outer gaps are typical for highly efficient old @xmath0-ray pulsars ( e.g. @xcite ) . the outer gap model further predicts that two caustics are formed at the edges of the @xmath0-ray beam , which could account for the sub - structure visible in the egret data . if we accept that the @xmath0-rays are produced above the null surface of the mp ( i.e. they are produced by the outer gap ) large parts of the @xmath44 plane can be ruled out . together with the @xmath45 restriction we can eliminate the entire hashed regions in fig . [ fig1055chi2 ] from consideration . as one can see in fig . [ fig1055chi2_2 ] , using these constrains the geometry of psr b105552 is already constrained even without fitting the shape of the pa - swing of the ip as there is only a relatively small wedge shaped region of allowed solutions in the plot . this is important , because it shows that our conclusions do not depend strongly on the shape of the pa - swing of the ip which deviates slightly from the rvm prediction . we will therefore adopt the rvm solution with the lowest @xmath27 in this paper which has @xmath28 and @xmath29 . the geometry of the dipole field of psr b105552 assuming @xmath28 . the rotation axis points upwards and the light cylinder is indicated by the vertical lines . the line of sight ( dotted line `` los '' ) makes a large angle ( @xmath46 ) with the magnetic axis of the mp ( right hand side of the plot ) , while this angle is much smaller ( @xmath47 ) for the ip ( left hand side of the plot ) . the dashed lines indicate the location of the null surface , and the last closed fieldlines which conventionally define a corotating region are shown . at the mp we must be viewing fieldlines which lie above the null surface ( in the shaded region ) and which will be well within the corotating region if the radio emission is , as usually assumed , close to the surface . the @xmath0-ray profile , which is observed to overlap the mp radio emission , is most likely formed in the shaded area , which indicates the volume above the null surface where an outer gap could potentially be formed . ] the emission geometry in the fiducial plane is illustrated in fig . [ fig1055field ] . one can see that the line of sight passes the magnetic axis of the ip much more closely than to the pole of the mp , consistent with the steeper slope of the pa - swing of the ip . thus , the ip emission must arise from open fieldlines as is generally thought to be the case for radio pulsars . however , the incidence of the line of sight with the magnetic axis of the mp is very large , which immediately leads to one of two possibilities . if we require the emission to arise from open fieldlines , then it must originate at a considerable fraction of the light cylinder . otherwise we are forced to accept emission from presumed closed fieldlines at low altitudes . [ fig1055field ] also shows the null lines ( dashed ) . these are the @xmath48 lines on which the charge density of the magnetosphere changes sign . the tangent of the fieldlines are perpendicular to the rotation axis at the point where they intersect the null line . it immediately follows that for the assumed geometry we ` see ' fieldlines above the null line for the mp ( fig . [ fig1055field ] ) , which suggests that the radio mp is produced in a region where an outergap is likely to exist . it can be shown that this is a general rule . in any pulsar where emission from both poles are detected , at one of the poles our line of sight must see emission from the region above the null line , i.e. the likely site of any `` outergap '' . assuming that this outergap is the site of the observed @xmath0-ray emission , it is no surprise that the phase range of the @xmath0-ray profile overlaps that of the mp rather than the ip . we note that were we to choose an rvm solution with @xmath49 , then the ip would be the pole which views the outergap region , and the @xmath0-ray profile would be at the wrong phase . most other known @xmath0-ray pulsars are , unlike psr b105552 , typically found to have a single peaked radio profile with a large offset from the @xmath0-ray profile . the lines of sight of these pulsars are also believed to have a large @xmath15 value for one of the poles . this causes the line of sight to miss the radio beam of the magnetic pole for which the @xmath0-rays are observed , hence a large offset is seen between the radio and the @xmath0-rays . psr b105552 is different because , for some reason , it has a large enough radio beam to be seen by a line of sight with a large angle @xmath15 . one further conclusion can also be drawn . with @xmath31 being so large at the mp ( @xmath50 ) , the corresponding @xmath40 at the ip will be _ positive _ ( since @xmath33 ) . this implies that at the ip we view the pole on the side _ away _ from the null line , and therefore do not see fieldlines which thread the outergap region . we now use the widths of the mp and ip pulse profile to place additional constraints on the magnetosphere s geometry . the first point that needs to be considered is the deviation of the mp - ip separation from @xmath26 in pulse longitude . this could in principle be caused by a differential emission height , but , as we argue in sect . [ sectemissionheight ] , this solution is very unlikely . this leaves us with the less satisfactory conclusion that at least one of the radio beams is not fully active ( e.g. @xcite ) . our next aim will be to investigate which parts of the mp and ip beam should be active in order to explain the observations . this is done by mapping the observed pulse longitudes onto the fieldlines which are pointing in our direction . the pulse longitudes of the edges of the mp and ip are therefore an important factor in determining which fieldlines are active . we take the positions of the steepest gradient of the rvm fit ( @xmath20 and @xmath51 , see eq . [ eqrvm ] ) as the reference . these are located in component ii of the mp and at the trailing edge of the ip . the emission of the mp is located at a pulse longitude range of @xmath52 to @xmath53 with respect to the first reference point and the emission of the ip confined to @xmath54 to @xmath55 using the second . this can be used to identify the fieldlines which are active after correcting for the relative shift between the pa - swing ( which is used as reference ) and the total intensity for a non - zero emission height ( eq . [ eqbcw ] ) . this means the true longitude of the emission point with respect to the fiducial point in the corotating frame is @xmath56 . if @xmath57 is the opening angle of the fieldline at the emission height and @xmath58 is the azimuthal angle of the fieldline with respect to the magnetic axis such that zero corresponds to the fiducial plane , these are given by @xmath59 where @xmath60 is the observed pulse longitude difference with respect to the reference point as discussed above ( @xcite see also appendix [ sctapp ] for a derivation ) . note that @xmath57 is only identical for the leading and trailing edge of a beam when the beam is symmetric about the fiducial plane . in that case @xmath61 , where @xmath62 is the full pulse width . assuming a dipolar field , the opening angle of the fieldline @xmath57 can be expressed in terms of the polar angle @xmath63 ( colatitude ) of the point of emission @xcite @xmath64 which then can be expressed in terms of the footprint parameter @xmath65 on the polar cap @xmath66 the magnetic axis corresponds to @xmath67 , and @xmath68 corresponds to last open fieldlines . as will be discussed in sect . [ sectpolarmap ] , the polar cap is not necessarily strictly circular if the magnetic axis makes an angle with the rotation axis . the geometrical solution for the leading and trailing edges of the mp and ip are shown in fig . [ fig1055illumination ] as a function of the emission height ( because @xmath69 is a function of the emission height ) . these plots identify which part of the polar cap should be active in order to explain the component widths and separations in combination with the shape and shift of the pa - swing . one can see that @xmath65 becomes very large when the emission height is very low . this is because there is a trade - off between the emission height and the @xmath65 value in order to get the required observed component width . at large emission heights there is a turnover and @xmath65 increases with emission height . this can be understood in terms of the shift of the pa - swing with respect to the emission profile . in order to prevent the pa - swing from having its steepest gradient far outside the beam , @xmath65 has to increase . at the same time the fraction of the beam which is active has to decrease in order to avoid generating profile components which are larger than observed . this can be seen in the @xmath58 curves , which approach each other for large emission heights such that only the trailing edges of the beams are active . it is not possible to determine the emission height based just on the shape of the pulse profile in combination with the shape of the pa - swing . additional assumptions are required to constrain the emission height further . @xcite make two assumptions in order to derive their solution . firstly , both the mp and ip beam are assumed to be filled ( i.e. that @xmath58 of the leading and trailing edge have the same magnitude but opposite sign ) . this choice is not available to us because we have insisted on a solution fully consistent with retardation and aberrational effects . the second assumption made by @xcite is that the @xmath65 parameter parameter corresponds to @xmath70 in @xcite . ] of the mp is assumed to be 1 . one can see in fig . [ fig1055illumination ] that the trailing edge ( dashed line ) of the ip and especially the mp have to be emitted from fieldlines further away than @xmath68 from the magnetic axis ( which is true for all possible emission heights ) . this shows again that the solution presented by @xcite is not self - consistent . [ fig1055illumination ] shows that the solid and dashed @xmath65 curves of the ip cross at @xmath71 . this corresponds to a beam which is completely filled such that the distances of the footprints of the active fieldlines to the magnetic axis are the same for the leading and trailing edge . this is consistent with the @xmath58 values at this height which have the same magnitude but an opposite sign . this happens at an @xmath65 value slightly larger than 1 , indicating that the beam is slightly bigger than the conventional polar cap size . the same emission height corresponds to a special geometry for the mp as the solid @xmath58 line crosses zero at that height . this means that the mp beam is exactly half filled for the emission height at which the ip is exactly completely filled . this arises because at this emission height @xmath72 for the ip . for the same emission height at the mp @xmath73 . there is no obvious reason why an exactly half filled mp should have physical significance , so it is presumably a coincidence this solution occurs at the same emission height as the fully filled ip solution . moreover , we can not know if the emission height is exactly @xmath71 . the trailing edge of the radio mp and the leading edge of the radio ip are relatively steep ( fig . [ fig1055profiles20 ] ) . this suggests that we only see the trailing side of the radio mp and only the leading side of the ip ( e.g. @xcite ) . if we accept this idea , then the emission height must be smaller than @xmath74 km , because for higher emission heights the ip s trailing edge ( rather than leading ) would correspond to an edge of the beam ( see fig . [ fig1055illumination ] ) . this would imply that the radio emission height of psr b105552 is not abnormal compared with other pulsars ( e.g. @xcite ) . if there is a differential emission height between the two poles , then that of the ip is larger ( see sect . [ sctorthogonal ] ) . however , the ip s emission height can not be larger than @xmath35 , or else its leading edge would no longer be the edge of the beam . the emission height of the mp could be slightly lower , although not too much as it would make its @xmath65 value unrealistically large . therefore roughly similar emission heights at @xmath35 ( or slightly lower ) is the most likely solution . the polar cap map of psr b105552 based on the assumption that both poles have identical structure so that we can `` reflect '' the observed ip mapping in the horizontal axis . the conventional polar cap size of an aligned rotator is indicated by the circle with radius @xmath68 . the dashed ellipse shows the polar cap including meridional compression . the fieldlines inside the kidney - shaped area are open and do not pass through any null line . the top and bottom solid curves are the lines of sight for the mp and ip respectively , where the active portions ( assuming an emission height of @xmath35 ) are the thicker parts of the curves . the location of the peaks are indicated as well ( see fig . [ fig1055profiles20 ] ) . the dotted semi - circle indicates approximately the footprints of the fieldlines on which @xmath0-rays are thought to be produced . ] having chosen a preferred ( maximum ) height , we may use the geometric solution of fig . [ fig1055illumination ] to generate a map of the polar cap of psr b105552 . the result is shown in fig . [ figpolarmap ] and combines the ip and mp on a single pole , thereby assuming that the emission structure at both poles is identical . obviously , because we only have information about the line of sight which passes over different parts of the polar cap of the mp and ip , we can not know whether this assumption is true . it must be borne in mind that both the ip and mp are observed on the same meridional planes of @xmath58 ( i.e. they both are viewed below the magnetic axis in fig . [ fig1055field ] ) , and in fig . [ figpolarmap ] the ip line of sight traverse has been `` reflected '' in the horizontal axis . this reflection ensures that a point in the polar map is associated with fieldlines which have identical geometric properties for both the mp and ip . the circle @xmath68 corresponds to the footprints of the last closed fieldlines of an aligned dipole touching the light cylinder . for non - aligned dipoles the polar cap shape gets distorted by meridional compression ( e.g. @xcite ) . the predicted shape for a dipole field inclined at @xmath1 is shown as the dashed oval in fig . [ figpolarmap ] . however , it must be noted that even models which take into account the distortion of the magnetic field lines close to the light cylinder predict roughly circular beams ( e.g. @xcite ) . in fig . [ figpolarmap ] we also indicate the kidney - shaped region within which fieldlines of a dipole inclined at @xmath1 do not cross the null surface inside the light cylinder . this shape is numerically derived by determining for each dipole fieldline whether the null surface @xmath48 is penetrated within the light cylinder . it can also be determined analytically ( @xcite ) . in the upper part of the polar map the null surface will lie at a great distance from the polar region ( see fig . [ fig1055field ] ) , but in the lower section it may be close enough to the star to set up an active outergap . we also indicate the approximation of the footprints of the @xmath75 fieldlines , argued by @xcite to be the fieldlines on which the @xmath0-ray emission is produced ( dotted semi - circle in fig . [ figpolarmap ] ) . note that the locus of active fieldlines on which the ip is generated is symmetric about @xmath76 and barely extends outside @xmath68 . furthermore , the ip fieldlines , apart from their outer edges , are contained within the kidney - shaped locus of the open fieldlines which do not cross a null surface . it was suggested by @xcite that such fieldlines carry the outflowing current from the pulsar . by contrast , the footprints of the active mp fieldlines lie well outside @xmath68 , extending to radii more that double this value . strangely , they cover only the latter half of a symmetrically - defined full beam . however , as pointed out above , the mp s fieldlines will thread the null surface close to the star s surface ( at about 160 km ) and the emission will occur in a plasma region usually considered to be inactive and corotating , so we can be sure that the mp is not a traditional polar cap profile . the solution shown is for @xmath71 , which is argued to be the maximum allowed emission height ( sect . [ sectemissionheight ] ) . for smaller emission heights both the line of sight of the mp and the ip will move to larger values of @xmath65 ( see fig . [ fig1055illumination ] ) and the active parts of the lines of sight will start leftwards of their current location ( more negative values of @xmath58 ) . in our study of psr b105552 we have focused on investigating the geometry of its magnetosphere . its wide mp and ip , together with a clearly - positioned single gamma - ray peak , place considerable constraints on any potential model . in an earlier attempt to locate the emission regions , @xcite fitted an rvm model to published figures ( @xcite ) for the pulsar s pa - swing and did not consider the constraints that aberration and retardation might place on the choice of emission heights . we have been able to exploit new data , gathered at parkes in recent years , to verify the rvm fit and to attempt a self - consistent model . one of our principal conclusions is that both the mp and ip are emitted at roughly the same height . the optimum fit ( @xmath28 , @xmath29 ) is in line with earlier results . this pa fit has a number of interesting implications for psr b105552 which are independent of assumptions about the emission heights . since @xmath77 ( i.e. @xmath78 , at @xmath79 is considerably greater than @xmath4 , our line of sight at the mp must see emission on fieldlines at locations which are above the `` null '' line ( fig . [ fig1055field ] ) , in or beyond territory normally ascribed to an outergap ( as was also noted by @xcite ) . in fact , on elementary geometric grounds , it must be true for every pulsar with a detected ip that at one pole we are viewing above the null line and at the other below it a point not widely realised before . however , in the case of psr b105552 ( but not necessarily for all pulsars with ip s ) , @xmath80 and so our line of sight at the ip must cross the magnetic axis on fieldlines which do not thread the outergap at the opposite pole ( fig . [ figpolarmap ] ) . on these grounds alone , we can expect that emission at the mp and ip will reflect the different conditions of these two regions . this can be contrasted with the emission from psr b170219 , where at the mp , in one of its two possible solutions , @xmath81 and @xmath82 @xcite . then , at one pole ( ip ) we must be viewing fieldlines above the null line and at the other ( mp ) equivalent fieldlines before they cross the null line , with a greater likelihood of coordinated emission . although the radio emission at both poles must come from about the same height , the precise height is not easy to determine . to first - order , aberration at a fixed height simply shifts the entire pa curve to later pulse longitudes by an amount proportional to the height . thus , as long as the emission height is small compared to the light cylinder radius and unchanged across the profile , we have no way of knowing from the shape of the pa curve by how much it has been shifted . the exceptionally wide mp and ip lead us to believe that , whatever the assumed height , it is impossible to confine the footprints of the emitting regions to a polar cap of conventional dimensions ( i.e. defined by last closed dipole fieldlines touching the light cylinder ) . this requirement is particularly acute for the mp , whose trailing components must lie on fieldlines located about twice the conventional polar radius from the pole . furthermore , since the fiducial point of the mp profile must lie ahead of the steepest gradient of the pa - swing , the observed mp width will never occupy much more than half its notional beam . choosing a height less than about 500 km would force the polar cap size to escalate rapidly ( see fig . [ fig1055illumination ] ) and therefore a balance has to be struck . here a height 7 per - cent of the light cylinder radius is assumed , so that the ip emission fills its beam and has footprints at around @xmath68 on the opposite side of the pole ( assuming identical polar caps at both poles ) . this choice also has the advantage that the steep edges of the mp ( trailing ) and ip ( leading ) correspond to beam edges ( see figs . [ fig1055illumination ] and [ figpolarmap ] ) . we therefore conclude that the emission height is unlikely to be higher or much lower than 700 km , a typical value for pulsars . a schematic picture of a physical model for psr b105552 , deduced from its radio and @xmath0-ray profiles . the corotation region , bounded by the fieldline @xmath83 , is shaded and lies well within region defined by the last closed fieldline to touch the light cylinder . it is surrounded by an `` auroral '' zone , which may or may not be entirely closed , but which must support electric fields parallel to its magnetic fieldlines . it makes two different intersections with the null surface , one very close to the star and a second no longer as distant from the surface as in a conventionally - defined corotation region . this may give rise to a complex outergap structure , with pair - created particles continually reflected between the null surfaces . the @xmath0-ray emission , appropriate to a `` middle - aged '' pulsar , is located above the outergap , stretching from the null line to the light cylinder.,width=302,height=302 ] the kidney shaped central area of fig . [ figpolarmap ] , which almost completely encloses the footprints of the ip s line of sight , highlights a significant consequence of the inferred geometry : the active fieldlines of the ip are open fieldlines which not only do not thread the outergap region beyond the null line close to the surface , but they also do not thread the null surface at any point ( assuming a pure dipole geometry as in fig . [ figmodel ] ) before they leave the light cylinder . this suggests that along these fieldlines the pulsar may expel or accrete charged particles without having to encounter any charge sign boundary . such fieldlines appear to bunch around the magnetic pole bounded by the radial parameter @xmath68 , the conventional limit for a polar cap defined by the last closed fieldlines touching the light cylinder . thus the ip , if observed on a line of sight which missed the mp , would not suggest an unusual geometry . by contrast , the fieldlines generating the mp have polar footprints well beyond @xmath68 ( fig . [ figpolarmap ] ) , although the observed emission region is itself well above the null line ( fig . [ fig1055field ] ) . we are therefore led to the conclusion that in psr b105552 s mp we are seeing emission from regions of the magnetosphere normally regarded as a closed and corotating . a polar cap radius of @xmath83 is the footprint of a fieldline which closes at a distance of only a quarter of the light cylinder radius and which intersects the null line at a height of only 160 km above the polar cap ( fig . [ figmodel ] ) . in fig . [ figpolarmap ] we see that radio emission occurs on fieldlines with parameters @xmath84 ( or smaller ) to @xmath83 , indicating emission on fieldlines for a significant region between the @xmath83 surface shaded in fig . [ figmodel ] and the last closed fieldlines ( @xmath85 ) . radio emission from these fieldlines implies particle acceleration and hence the presence of electric fields parallel to the fieldlines . so it is conceivable that such fieldlines are open and cross the light cylinder ( as they do , for example , in the models of aligned magnetospheres with reduced corotation regions given by @xcite ) . the closeness of the outergap to the neutron star surface may lead to an active outergap accelerator merging with the polar gap ( see @xcite ) and the much reduced corotating region ( shaded region in fig . [ figmodel ] ) . alternatively , the quasi - closed region between fieldlines @xmath68 and @xmath83 can be seen as an auroral zone , analogous to those known to be present in the planets , in which accretion from a debris disk may occur and modulate the radio signal ( @xcite ) . in fact , accretion could modulate the entire outergap region through an injection of charge ( @xcite ) . since all the surface @xmath83 is now so close to the neutron star , it is not impossible to envisage interaction between the _ two _ null surfaces which intersect it in the meridional plane ( fig . [ figmodel ] ) . possibly `` gaps '' arise at both at which pair - creation may be stimulated by thermal x - rays from the star s surface leading to secondary particles `` bouncing '' between the gaps and stimulating the observed mp emission . whatever conditions are present in the envelope between @xmath83 and @xmath68 , psr b105552 s @xmath0-rays are almost certainly generated relatively close to the light cylinder above an outer gap which stretches from the last closed fieldline ( whether @xmath68 or @xmath83 ) towards the light cylinder . unlike most other @xmath0-ray pulsars , b105552 has a single broad peak which is shown by @xcite to match closely to the geometric predictions of the outergap model of @xcite , lying on a fieldline with @xmath86 . outergaps with @xmath0-ray emission regions lying significantly above the last closed fieldline are thought to be a feature of relatively `` old '' pulsars , as psr b105552 indeed is by the standards of known @xmath0-ray pulsars . the fact that the @xmath0-ray pulse overlaps the phase of the mp indicates that both regions point to us at about the same time , which lends support to our model ( fig . [ figmodel ] ) . our conclusions about the geometry of psr b105552 result from a rigid application of the rvm model to high - quality data , taking into account the expected effects from aberration . we are convinced that this pulsar emits radiation from fieldlines lying well outside the conventionally - defined polar cap boundary . consequently , we would question the common assumption that all pulsars have closed , quiescent , corotating regions stretching to the light cylinder . possibly this only occurs after the outergap has become inactive , so that energetic ( high @xmath87 ) pulsars can often have wide radio profiles , as observed @xcite . large polar caps might also help explaining the inconsistency found in calculating emission heights of energetic pulsars using last - closed - fieldline methods @xcite . an unsatisfactory aspect of our model is the one - sidedness of the mp emission region . this might arise from the very different azimuthal trajectories which particles on the trailing and leading halves must take . on the trailing half they flow counter to the sense of the pulsar s rotation and can smoothly cross the light cylinder to the wave zone . however , those leaving the leading half of the polar cap will increase speed and hence lorentz factor as they gain height on corotating fieldlines . it has been suggested that this may lead to additional pair - creation and high - energy emission as the particles free themselves from fieldlines near the light cylinder ( @xcite ) , but subsequent screening and its consequences for radio emission have not been considered . the authors would like to thank simon johnston , don melrose , qinghuan luo , matthew verdon , kyle watters , mike keith and wim hermsen for useful discussions , as well as the referee jarek dyks . gw thanks the university of sussex for an ongoing visiting research fellowship , and is grateful to the csiro for support for a stay at atnf , during which most of this work was carried out . the australia telescope is funded by the commonwealth of australia for operation as a national facility managed by the csiro . becker w. , trmper j. , 1997 , a&a , 326 , 682 biggs j. d. , 1990a , mnras , 246 , 341 biggs j. d. , 1990b , mnras , 245 , 514 blaskiewicz m. , cordes j. m. , wasserman i. , 1991 , apj , 370 , 643 cheng k. s. , ho c. , ruderman m. , 1986 , apj , 300 , 522 combi j. a. , romero g. e. , azcrate i. n. , 1997 , ap&ss , 250 , 1 j. m. , lazio t. j. w. , 2002 , astro - ph/0207156 j. m. , shannon r. m. , 2008 , apj , 682 , 1152 a. a. , kahn f. d. , 1982 , mnras , 199 , 211 damour t. , taylor j. h. , 1992 , phys . d , 45 , 1840 daugherty j. k. , harding a. k. , 1996 , apj , 458 , 278 j. , 2008 , mnras , 391 , 859 j. , rudak b. , 2003 , apj , 598 , 1201 j. e. , weisberg j. m. , 2001 , apj , 553 , 341 r. t. , gupta y. , 2001 , apj , 555 , 31 gil j. a. , gronkowski p. , rudnicki w. , 1984 , a&a , 132 , 312 goldreich p. , julian w. h. , 1969 , apj , 157 , 869 s. p. , mestel j. , mestel l. , wright g. a. e. , 2004 , mnras , 349 , 213 y. , gil j. , kijak j. , sendyk m. , 2004 , a&a , 426 , 229 j. a. , arons j. , 2001 , apj , 546 , 382 k. , 2006 , apj , 652 , 1475 k. , shibata s. , 2002 , apj , 564 , 369 n. j. , 1973 , nature , 246 , 6 johnston s. , 2002 , pub . of the asa , 19 , 277 a. , johnston s. , 2007 , mnras , 380 , 1678 kijak j. , gil j. , 1997 , mnras , 288 , 631 q. , melrose d. , 2007 , mnras , 378 , 1481 lyne a. g. , manchester r. n. , 1988 , mnras , 234 , 477 mcculloch p. m. , hamilton p. a. , ables j. g. , komesaroff m. m. , 1976 , mnras , 175 , 71p mcculloch p. m. , hamilton p. a. , manchester r. n. , ables j. g. , 1978 , mnras , 183 , 645 manchester r. n. , lyne a. g. , 1977 , mnras , 181 , 761 manchester r. n. , taylor j. h. , huguenin g. r. , 1975 , apj , 196 , 83 l. , wright g. a. e. , westfold k. c. , 1976 , mnras , 175 , 257 mitra d. , rankin j. m. , 2002 , apj , 577 , 322 a. g. , harding a. k. , 2004a , apj , 606 , 1143 a. g. , harding a. k. , 2004b , apj , 606 , 1143 a. , johnston s. , kramer m. , karastergiou a. , 2008 , mnras , 386 , 1881 gelman h. , finley j. p. , 1993 , apj , 413 , l31 g. j. , lee k. j. , wang h. g. , xu r. x. , han j. l. , 2004 , apj , 606 , l49 radhakrishnan v. , cooke d. j. , 1969 , astrophys . lett . , 3 , 225 rankin j. m. , 1983 , apj , 274 , 333 romani r. w. , 1996 , apj , 470 , 469 romani r. w. , yadigaroglu i .- a . , 1995 , apj , 438 , 314 smart w. m. , 1960 , text - book on spherical astronomy , 6th ed .. cambridge university press , cambridge , england d. a. , guillemot l. , camilo f. , cognard i. , dumora d. , espinoza c. , freire p. c. c. , gotthelf e. v. , harding a. k. , hobbs g. b. , johnston s. , kaspi v. m. , kramer m. , livingstone m. a. , lyne a. g. , manchester r. n. , marshall f. e. , et al . 2008 , a&a , 492 , 923 taylor j. h. , manchester r. n. , lyne a. g. , 1993 , apjs , 88 , 529 thompson d. j. , bailes m. , bertsch d. l. , cordes j. , damico n. , esposito j. a. , finley j. , hartman r. c. , hermsen w. , kanbach g. , kaspi v. m. , kniffen d. a. , kuiper l. , lin y. c. , manchester r. , matz s. m. , mayer - hasselwander h. a. , michelson p. f. , et al . 1999 , apj , 516 , 297 vaughan a. e. , large m. i. , 1972 , mnras , 156 , 27p h. g. , qiao g. j. , xu r. x. , liu y. , 2006 , mnras , 366 , 945 watters k. , romani r. w. , weltevrede p. , johnston s. , 2008 , apj in press ( astro - ph/0812.3931 ) p. , johnston s. , 2008 , mnras , 391 , 1210 p. , wright g. a. e. , stappers b. w. , 2007 , a&a , 467 , 1163 [ lastpage ] we are interested to find the relation between the observed pulse longitude and the magnetic coordinates @xmath57 ( the angular distance from the magnetic pole ) and @xmath58 ( the azimuthal angle around the magnetic pole ) . this coordinate transformation can be derived using spherical geometry on the celestial sphere of the pulsar . two trigonometric relations which can be used are the sine and cosine rules for sides ( e.g. @xcite ) @xmath88 see the top panel of fig . [ figtrigonometry1 ] for the definitions of the angles in these equations . the openings angle of the beam can be derived by applying the cosine rule for sides ( eq . [ eqcos ] ) on the triangle plotted in the bottom panel of fig . [ figtrigonometry1 ] : @xmath89 ( @xcite ) . here @xmath90 is the pulse longitude measured from the fiducial plane . in the case of a profile which is symmetric around the fiducial plane @xmath91 . the azimuthal angle @xmath58 can be derived by using eq . [ eqsin ] on the same triangle plotted in the bottom panel of fig . [ figtrigonometry1 ] and one directly obtains @xmath92 @xcite . it can be shown that this equation is equivalent to the following expression which is used by @xcite : @xmath93 where @xmath94 . _ top plot : _ a spherical triangle . the angular lengths of the sides of the triangle are the angles @xmath95 , @xmath96 and @xmath97 , while @xmath98 , @xmath99 and @xmath100 are the angles between the sides of the triangle measured on the surface of the sphere . _ bottom plot : _ a triangle on the celestial sphere used to derive the half openings angle of the beam @xmath57 and the azimuthal angle of a fieldline @xmath58 . the line of sight makes an angle @xmath41 with the rotation axis and and the magnetic axis makes an angle @xmath13 with the rotation axis.,title="fig : " ] + _ top plot : _ a spherical triangle . the angular lengths of the sides of the triangle are the angles @xmath95 , @xmath96 and @xmath97 , while @xmath98 , @xmath99 and @xmath100 are the angles between the sides of the triangle measured on the surface of the sphere . _ bottom plot : _ a triangle on the celestial sphere used to derive the half openings angle of the beam @xmath57 and the azimuthal angle of a fieldline @xmath58 . the line of sight makes an angle @xmath41 with the rotation axis and and the magnetic axis makes an angle @xmath13 with the rotation axis.,title="fig : " ]

the experiments on resistivity of high mobility two - dimensional electron gas ( 2deg ) in presence of a relatively weak magnetic field and microwave radiation led to a discovery of striking zero - resistance states ( zrs ) induced by a microwave field by mani _ _ @xcite and zudov _ et al . other experimental groups also found the microwave induced zrs in various 2deg samples ( see e.g. @xcite ) . a similar behavior of resistivity is also observed for electrons on a surface of liquid helium in presence of magnetic and microwave fields @xcite . these experimental results obtained with different systems stress the generic nature of zrs . various theoretical explications for this striking phenomenon have been proposed during the decade after the first experiments @xcite . an overview of experimental and theoretical results is give in the recent review @xcite . in our opinion the most intriguing feature of zrs is their almost periodic structure as a function of the ratio @xmath0 between the microwave frequency @xmath1 and cyclotron frequency @xmath2 . ( in the following we are using units with electron charge @xmath3 and mass @xmath4 equal to unity ) . indeed , a hamiltonian of electron in a magnetic field is equivalent to an oscillator , it has a magneto - plasmon resonance at @xmath5 but in a linear oscillator there are no matrix elements at @xmath6 and hence a relatively weak microwave field is not expected to affect electron dynamics and resistivity properties of transport . of course , one can argue that impurities can generate harmonics being resonant at high @xmath7 but zrs is observed only in high mobility samples and thus the density of impurities is expected to be rather low . it is also important to note that zrs appears at high landau levels @xmath8 so that a semiclassical analysis of the phenomenon seems to be rather relevant . in this work we develop the theoretical approach proposed in @xcite . this approach argues that impurities produce only smooth potential variations inside a bulk of a sample so that zrs at high @xmath9 appear from the orbits moving along sharp sample boundaries . it is shown @xcite that collisions with boundaries naturally generate high harmonics and that a moderate microwave field gives stabilization of edge channel transport of electrons in a vicinity of @xmath10 producing at these @xmath9 a resistance going to zero with increasing microwave power . this theory is based on classical dynamics of electrons along a sharp edge . the treatment of relaxation processes is modeled in a phenomenological way by a dissipative term in the newton equations . additional noise term in the dynamical equations takes into account thermal fluctuations . the dissipation leads to synchronization of cyclotron phase with a phase of microwave field producing stabilization of edge transport along the edges in a vicinity of resonant @xmath11 values . thus , according to the edge stabilization theory @xcite the zrs phase is related to a universal synchronization phenomenon which is a well established concept in nonlinear sciences @xcite . while the description of edge transport stabilization @xcite captures a number of important features observed in zrs experiments it assumes that the contribution of bulk orbits in transport is negligibly small . this assumption is justified for smooth potential variations inside the bulk of a sample . however , a presence of isolated small scale scatterers inside the bulk combined with a smooth potential component can significantly affect the transport properties of electrons ( see e.g. @xcite ) . also the majority of theoretical explanations of zrs phenomenon considers only a contribution of scattering in a bulk @xcite . thus it is necessary to analyze how a scattering on a single impurity is affected by a combined action of magnetic and microwave fields . in this work we perform such an analysis modeling impurity by a rigid circular disk of finite radius . we show that the dynamics in a vicinity of disk has significant similarities with dynamics of orbits along a sharp edge leading to appearance of zrs type features in a resistivity dependence on @xmath9 . the paper is composed as follows : in section ii we discuss the dynamics in edge vicinity , in section iii we analyze scattering on a single disk , in section iv we study scattering on many disks when their density is low , here we determine the resistivity dependence on @xmath9 and other system parameters , physical scales of zrs effects are analyzed in section v , effects of two microwave driving fields and other theory predictions are considered in section vi , discussion of the results is given in section vii . we study various models which we list here for a reader convenience : wall model described by the newton equations ( [ eq1 ] ) , ( [ eq2 ] ) with microwave field polarization perpendicular to the wall ( model ( w1 ) equivalent to model ( 1 ) in @xcite ) ; the chirikov standard map description ( [ eq3 ] ) of the wall model dynamics called model ( w2 ) ( equivalent to model ( 2 ) in @xcite at parameter @xmath12 ) ; the single disk model with radial microwave field called model ( dr1 ) ; the chirikov standard map description ( [ eq3 ] ) of model ( dr1 ) called model ( dr2 ) ( here @xmath13 in ( [ eq3 ] ) , @xmath14 ) ; the model of a single disk in a linearly polarized microwave field and static electric field called model ( d1 ) ; the model of transport in a system with many disks called model ( d2 ) which extends the model ( d1 ) ; extension of model ( d2 ) with disk roughness and dissipation in space called model ( d3 ) ; the wall model ( w2 ) extended to two microwave fields is called model ( w3 ) . we remind first the approach developed in @xcite . here , the classical electron dynamics is considered in a proximity of the fermi surface and in a vicinity of sample edge modeled as a specular wall . the motion is described by newton equations @xmath15 where a dimensionless vector @xmath16 describes microwave driving field @xmath17 . here an electron velocity @xmath18 is measured in units of fermi velocity @xmath19 and @xmath20 describes a relaxation processes to the fermi surface . we also use the dimensionless amplitude of velocity oscillations induced by a microwave field @xmath21 . as in @xcite , in the following we use units with @xmath22 . the last two terms @xmath23 and @xmath24 in ( [ eq1 ] ) account for elastic collisions with the wall and small angle scattering . disorder scattering is modeled as random rotations of @xmath25 by small angles in the interval @xmath26 with poissonian distribution over time interval @xmath27 . the amplitude of noise is assumed to be relatively small so that the mean free path @xmath28 is much larger than the cyclotron radius @xmath29 . we note that the dissipative term is also known as a gaussian thermostat @xcite or as a landau - stuart dissipation @xcite . the dynamical evolution described by eq . ( [ eq1 ] ) is simulated numerically using the runge - kutta method . following @xcite we call this system model ( w1 ) ( equivalent to model ( 1 ) in @xcite ) . we note that for typical experimental zrs parameters we have : electron density @xmath30 , effective electron mass @xmath31 , microwave frequency @xmath32 , fermi energy @xmath33 , corresponding to @xmath34 , with fermi velocity @xmath35 . at such a frequency the cyclotron resonance @xmath36 takes place at @xmath37 with the cyclotron radius @xmath38 . at such a magnetic field we have the energy spacing between landau levels @xmath39 corresponding to a landau level @xmath40 . for a microwave field strength @xmath41 we have the parameter @xmath42 . with these physical values of system parameters we can always recover the physical quantities from our dimensionless units with @xmath43 . examples of orbits running along the edge of specular wall are given in @xcite ( see fig.1 there ) . a microwave field creates resonances between the microwave frequency @xmath1 and a frequency of nonlinear oscillations of orbits colliding with the wall . due to a specular nature of this collisions the electron motion has high harmonics of cyclotron frequency that leads to appearance of resonances around @xmath44 ( there is an additional shift of approximate value @xmath45 to @xmath11 values due to a finite width of nonlinear resonance ) . to characterize the dynamical motion it is useful to construct the poincar section following the standard methods of nonlinear systems @xcite . we consider the hamiltonian case at @xmath46 in absence of noise . also we choose a linear polarized microwave field being perpendicular to the wall which is going along @xmath47-axis ( same geometry as in @xcite ) . in this case the generalized momentum @xmath48 is an integral of motion since there are no potential forces acting on electron along the wall ( here we use the landau gauge with a vector potential @xmath49 ) . the momentum @xmath50 determines a distance @xmath51 between a cyclotron center and the wall , which also remains constant in time . the hamiltonian of the system has the form : @xmath52 where @xmath53 is the wall potential being zero or infinity for @xmath54 or @xmath55 . thus , we have here a so called case of one and half degrees of freedom ( due to periodic time dependence of hamiltonian on time ) and the poincar section has continuous invariant curves in the integrable regions of phase space @xcite . ) for @xmath56 ( left column ) and @xmath57 ( right column ) and different amplitudes of microwave field @xmath58 ( from bottom to top ) . here , the integral @xmath59 , trajectories start from wall with fixed @xmath60 . data for model ( w1 ) at @xmath46 , @xmath61 . ] the poincar sections for ( [ eq1 ] ) , ( [ eq2 ] ) at @xmath62 and various amplitudes of microwave field @xmath63 are shown in fig . it shows a velocity @xmath64 at moments of collision with the wall at @xmath65 as a function of microwave phase @xmath66 at these moments of time . all orbits initially start at the wall edge @xmath65 with the initial velocity @xmath67 . the value of @xmath68 is the integral of motion . however , the kinetic energy of electron @xmath69 varies with time . we see that at a small @xmath70 the main part of the phase space is covered by invariant curves corresponding to integrable dynamics . however , a presence of chaotic component with scattered points is also visible in a vicinity of separatrix of resonances , especially at large @xmath71 . the points at @xmath72 close to zero correspond to orbits only slightly touching the wall , while the orbits at @xmath73 have a large cyclotron radius and collide with the wall almost perpendicularly . there are also sliding orbits which have the center of cyclotron orbit inside the wall @xmath74 but we do not discuss they here . indeed , the orbits , which only slightly touch the wall ( @xmath75 ) , play the most important role for transport since the scattering angles in the bulk are small for high mobility samples and an exchange between bulk and edge goes via such type of dominant orbits @xcite . we note that the section of fig . [ fig1 ] at @xmath57 , @xmath76 is in a good agreement with those shown in fig.1b of @xcite . however , here we have single invariant curves while in @xcite the curves have a certain finite width . this happens due to the fact that in @xcite the poincar section was done with trajectories having different values of the integral @xmath68 that gave some broadening of invariant curves . for a fixed integral value we have no overlap between invariant curves as it is well seen in fig . [ fig1 ] here . the phase space in fig . [ fig1 ] has a characteristic resonance at a certain @xmath77 value which position depends on @xmath9 @xcite . an approximate description of the electron dynamics and phase space structure can be obtained on a basis of the chirikov standard map @xcite,@xcite . in this description developed in @xcite an electron velocity has an oscillating component @xmath78 ( assuming that @xmath79 ) and a collision with the wall gives a change of modulus of @xmath64 by @xmath80 ( like a collision with a moving wall ) . for small collision angles the time between collisions is @xmath81 . indeed , @xmath82 is the cyclotron period . however , the time between collisions is slightly smaller by an amount @xmath83 : at @xmath84 an electron moves in an effective triangular well created by the lorentz force and like for a stone thrown against a gravitational field this gives the above reduction of @xmath85 ( formally this expression for @xmath85 is valid for sliding orbits but for orbits slightly touching the wall we have the same @xmath85 but with minus that gives the correction @xmath86 ) . the same result can be obtained via semiclassical quantization of edge states developed in @xcite . it also can be found from a geometric overlap between the wall and cyclotron circle . this yields an approximate dynamics description in terms of the chirikov standard map @xcite : @xmath87 with the chaos parameter @xmath88 . usually we are in the integrable regime with @xmath89 due to small values of @xmath63 used in experiments . a developed chaos appears at @xmath90 @xcite . here bars mark the new values of variables going from one collision to a next one , @xmath64 is the velocity component perpendicular to the wall , @xmath91 is the microwave phase at the moment of collision . here we introduced a dimensionless parameter @xmath92 which is equal to @xmath12 for the case of the wall model @xmath93 considered here . however , we will show that for the dynamics around disk with a radial field in model ( dr1 ) we have the same map ( [ eq3 ] ) with @xmath94 . due to that it is convenient to write all formula with @xmath92 . we note that a similar map ( [ eq3 ] ) describes also a particle dynamics in a one - dimensional triangular well and a monochromatic field @xcite . the term @xmath95 in ( [ eq3 ] ) describes dissipation and noise . the later gives fluctuations of velocity @xmath64 at each iteration ( @xmath96 ; corresponding to random rotation of velocity vector in ( [ eq1 ] ) ) . damping from electron - phonon and electron - electron collisions contribute to @xmath97 . the poincar sections of this map are in a good agreement with those obtained from the hamiltonian dynamics as it is seen in fig . [ fig1 ] here and fig.1 in @xcite . following @xcite we call this system model ( w2 ) ( equivalent to model ( 2 ) in @xcite ) . a phase shift of @xmath98 by @xmath99 does not affect the dynamics and due to that the phase space structure changes periodically with integer values of @xmath9 . indeed , the position of the main resonance corresponds to a change of phase by an integer number of @xmath99 values @xmath100 that gives the position of resonance at @xmath101 where @xmath4 is the nearest integer of @xmath102 and @xmath103 is the fractional part of @xmath9 . due to this relation we have the different resonance position for @xmath56 and @xmath104 being in agreement with the data of fig . [ fig1 ] at small values of @xmath63 when nonlinear corrections are small ( we have here @xmath12 ) . thus at @xmath57 we have the resonance position at @xmath105 in agreement with fig . [ fig1 ] ( right bottom panel ) . for @xmath106 we have @xmath107 and at @xmath56 the resonance position moves to negative value @xmath108 . thus , at @xmath109 the resonance separatrix easily moves particles out from the edge at @xmath110 where they escape to the bulk due to noise . in contrast at @xmath111 particles move along separatrix closer to the edge being then captured inside the resonance which gives synchronization of cyclotron phase with the microwave phase . this mechanism stabilizes the transmission along the edge . of electrons as a function of their dimensionless cyclotron center position @xmath112 between two walls and the frequency ratio @xmath0 . the distance between specular walls is @xmath113 . the amplitude of microwave field is @xmath114 with polarization parallel ( left panel ) and perpendicular ( right panel ) to the walls ( see text for more details ) . here @xmath115 , @xmath116 , @xmath27 , the variation of @xmath0 is obtained by changing magnetic field ( @xmath117 ) keeping @xmath118 ; 100 electrons are simulated at each @xmath9 up to time @xmath119 . density is proportional to color changing from zero ( black ) to maximal density ( white ) . data for model ( w1 ) . ] in @xcite it is shown that the orbits started in edge vicinity are strongly affected by a microwave field that leads to zrs type oscillations of transmission along the edge and longitudinal resistivity @xmath120 . the zrs structure appears both in the frame of dynamics described by ( [ eq1 ] ) ( model ( w1 ) ) and map description ( [ eq3 ] ) ( model ( w2 ) ) . the physical mechanism is based on synchronization of a cyclotron phase with a phase of microwave driving that leads to stabilization of electron transport along the edge . an extensive amount of numerical data has been presented in @xcite and we think there is no need to add more . here , we simply want to illustrate that even those orbits which start in the bulk are affected by this synchronization effect . for that we take a band of two walls with a band width between them being @xmath121 . initially 100 trajectories are distributed randomly in a bulk part between walls when a cyclotron radius is not touching the walls @xmath122 . their dynamics is followed during the run time @xmath119 according to eq . ( [ eq1 ] ) and a density distribution @xmath123 averaged in a time interval @xmath124 is obtained for a range of @xmath125 ( 261 values of @xmath9 are taken homogeneously in this interval ) . the value of @xmath126 approximately corresponds to a distance propagation along the wall of @xmath127 at typical values @xmath128 , @xmath129 . this is comparable with a usual sample size used in experiments @xcite . similar values of @xmath130 were used in @xcite . as a function of @xmath112 for microwave polarization parallel ( left panels ) and perpendicular ( right panels ) to the walls . here we have no microwave at top panels , @xmath131 , @xmath132 at middle panels , @xmath131 , @xmath133 at bottom panels . in all panels we have noise amplitude @xmath116 as in fig . [ fig2 ] , dashed curves in bottom panels are obtained with @xmath134 . simulations are done with 500 trajectories , other parameters are the same as in fig . data for model ( w1 ) . ] the dependence of density @xmath135 on @xmath51 and @xmath9 are shown in fig . [ fig2 ] for two polarizations of microwave field . the data show that orbits from a bulk can be captured in edge vicinity for a long time giving an increase of density in a vicinity of edge . this capture is significant around resonance values @xmath136 . this is confirmed by a direct comparison of density profiles in fig . [ fig3 ] at @xmath137 and @xmath138 . in the later case we have a large density peak due to trajectories trapped in a resonance ( see fig . [ fig1 ] ) where they are synchronized with a microwave field . an increase of noise amplitude @xmath139 gives a significant reduction of the amplitude of these resonant peaks ( fig . [ fig3 ] bottom panels ) . the increase of density is more pronounced for polarization perpendicular to the wall in agreement with data shown in fig.2 of @xcite . we also performed numerical simulations using eq . ( [ eq1 ] ) with a smooth wall modeled by a potential @xmath140 . for large values @xmath141 ( e.g. @xmath142 ) we find the poincar sections to be rather similar to those shown in fig . [ fig1 ] that gives a similar structure of electron density as in figs . [ fig2],[fig3 ] . a finite wall rigidity can produce a certain shift of optimal capture conditions appearing as a result of additional correction to a cyclotron period due to a part of orbit inside the wall . the data presented in this section show that electrons from the bulk part of the sample can be captured for a long time in edge vicinity thereby increasing the electron density near the edge . this effect is very similar to the accumulation of electrons on the edges of the electron cloud under zrs conditions that was reported for surface electrons on helium in @xcite . however we have to emphasize that the confinement potential for surface electrons is very different from the hard wall potential assumed in our simulations , as a consequence our results can not be applied directly to this case . it is possible that the formation of ballistic channels on the edge of the sample combined with the redistribution of the electrons density can effectively short the bulk contribution and induce directly a vanishing @xmath120 . however , it is also important to understand how a scattering on impurities inside the bulk is affected by a microwave radiation . we study this question in next sections . it should be noted that resistivity properties of a regular lattice of disk antidots in 2deg had been studied experimentally @xcite and theoretically @xcite . but effects of microwave field were not considered till present . in our studies we model an impurity as a rigid disk of fixed radius @xmath143 keeping @xmath144 and changing @xmath117 . in a magnetic field a cyclotron radius moves in a free space only due to a static _ dc-_electric field @xmath145 . we fix the direction of @xmath145 along @xmath146axis and measure its strength by a dimensionless parameter @xmath147 . even in absence of a microwave field a motion in a vicinity of disk in crossed static electric and magnetic fields of moderate strength is not so simple . the studies presented in @xcite and @xcite show that dynamics in disk vicinity is described by a symplectic disk map which is rather similar to the map ( [ eq3 ] ) . it is characterized by a chaos parameter @xmath148 where @xmath149 is the drift velocity ; @xmath150 gives an amplitude of change of radial velocity at collision . orbits from a vicinity of disk can escape for @xmath151 @xcite . at @xmath152 ( from top to bottom ) ; right column panels : @xmath153 at @xmath76 . here the integral of orbital momentum is @xmath154 , trajectories start from disk with fixed tangent velocity component @xmath155 . data for model ( dr1 ) . ] we start our analysis from the construction of the poincar section in presence of microwave field at zero static field . to have a case with one and half degrees of freedom we start from a model case when a microwave field is directed only along radius from a disk center . the dynamics is described by eq . ( [ eq1 ] ) with a dimensionless microwave amplitude @xmath63 . the dynamical evolution is obtained numerically by the runge - kutta method . at first we consider a case without dissipation and noise . due to radial force direction the orbital momentum is an additional integral of motion ( as @xmath68 for the wall case ) and thus we have again 3/2 degrees of freedom . we call this disk model with radial microwave field as model ( dr1 ) . the poincar sections at the moments of collisions with disk are shown in fig . [ fig4 ] , [ fig5 ] for model ( dr1 ) . here , @xmath156 is the radial component of electron velocity and @xmath98 is a microwave phase both taken at the moment of collision with disk . we see that the phase space structure remains approximately the same when @xmath9 is increased by unity ( compare @xmath157 panels in fig . [ fig4 ] ) . this happens for orbits only slightly touching the disk ( small @xmath156 ) since the microwave phase change during a cyclotron period is shifted by an integer amount of @xmath99 ( in a first approximation at @xmath158 ) . the similarity between the wall and disk cases is directly seen from fig . [ fig5 ] as well as periodicity with @xmath159 . in fact in the case of disk with a radial field the dynamics can be also described by the chirikov standard map ( [ eq3 ] ) where @xmath64 should be understood as a radial velocity @xmath156 at the moment of collision . the second equation has the same form since the change of the phase between two collisions is given by the same equation but with the parameter @xmath94 . this expression for @xmath92 is obtained from the geometry of slightly intersecting circles of radius @xmath160 for disk and radius @xmath161 for cyclotron orbit ( the angle segment of cyclotron circle is @xmath162 ) . for @xmath163 this expression naturally reproduces the wall case while at @xmath158 we have the correction term proportional to @xmath164 going to zero that also well corresponds to the geometry of two disks . after such modification of @xmath92 we find that the resonance positions @xmath165 are proportional to @xmath92 . thus the model ( dr1 ) reduced to the map description ( [ eq3 ] ) at @xmath166 is called model ( dr2 ) . the expression for @xmath167 works rather well . indeed , for @xmath168 in fig . [ fig5 ] we obtain @xmath169 for model ( w2 ) and @xmath170 for model ( dr2 ) . these values are in a good agreement with numerical values @xmath171 for model ( w2 ) and @xmath172 for model ( dr1 ) . in the later case the agreement is less accurate due to a larger size of nonlinear resonance . the comparison of poincar sections given by the chirikov standard map ( [ eq3 ] ) and the dynamics from newton equations , shown in fig . [ fig5 ] , confirms the validity of map description . according to the well established results for the chirikov standard map @xcite we find for models ( w1 ) , ( w2 ) and ( dr1 ) , ( dr2 ) the width of separatrix @xmath173 and the corresponding resonance energy width @xmath174 : @xmath175 where @xmath176 is the resonance shift produced by a resonance half width @xmath177 . this relation shows that for the disk case this energy is increased by a factor @xmath92 compared to the wall case . in majority of our numerical simulations we have @xmath178 . thus a radial field models ( dr1 ) , ( dr2 ) represent a useful approximation to understand the properties of dynamics in a disk vicinity but a real situation corresponds to a linear microwave polarization and the poincar section analysis should be modified to understand the dynamics in this case . and @xmath168 ( top , middle ) and @xmath179 ( bottom ) . top panels are obtained from the chirikov standard map ( [ eq3 ] ) at @xmath12 ( left top panel ) , corresponding to the wall model ( w2 ) , and at @xmath180 ( right top panel ) , corresponding to the disk model ( dr2 ) . middle and bottom panels are obtained from solution of newton equations ( [ eq1 ] ) for wall ( left ) and disk ( right ) . other parameters are as in fig . [ fig1 ] and fig . [ fig4 ] , @xmath64 and @xmath156 are expressed in units of @xmath19 . there is no dissipation and no noise . ] due to that we start to analyze the scattering problem on a disk in presence of weak static field @xmath181 and microwave field @xmath63 using eq . ( [ eq1 ] ) . for the scattering problem we find more simple to have dissipation to work only at the time moments of electron collisions with disk : at such time moments the radial component of electron velocity is reduced by a factor @xmath182 , the reduction is done only if the kinetic energy of electron is larger than the fermi energy . such a dissipation can be induced by phonon excitations inside the antidot disk . we fix geometry directing _ dc-_field along @xmath146axis and microwave along @xmath183axis . the noise is modeled in the same way as above in eq . ( [ eq1 ] ) . we call this system disk model ( d1 ) . , @xmath57 . _ top right panel _ : temporary captured path at @xmath152 , @xmath57 . _ bottom left panel _ : path captured forever at @xmath152 , @xmath57 . _ bottom right panel _ : no capture at @xmath152 , @xmath106 . the trajectory part colliding with disk is shown by black curve , its part before and after collisions is shown in gray . the red ( light gray ) points and curves show the trajectory of cyclotron center . here the dissipation parameter is @xmath184 ; the static electric field is directed along @xmath146axis and @xmath185 ; microwave field is directed along @xmath183axis . there is no noise here . coordinates @xmath186 are expressed in units of @xmath160 . data for model ( d1 ) . ] examples of electron cyclotron trajectories scattering on disk are shown in fig . [ fig6 ] . in absence of microwave field a trajectory escapes from disk rather rapidly . a similar situation appears at @xmath106 and microwave field with @xmath152 . in contrast for @xmath57 and @xmath152 a trajectory can be captured for a long time or even forever depending on initial impact parameter . for some impact parameters a trajectory can be captured for a very long time @xmath187 , in certain cases in absence of noise we have @xmath188 . at such long capture times the collisions with disk become synchronized with the phase @xmath98 of microwave field at the moment of collisions . this is directly illustrated in fig . [ fig7 ] where we show the angle @xmath189 of a collision point on disk , counted from @xmath146axis , in dependence on @xmath98 . indeed , the dependence @xmath189 on @xmath98 forms a smooth curve corresponding to synchronization of two phases . at the same time the radial velocity at collisions @xmath156 moves along some smooth invariant curve @xmath190 in the phase space @xmath191 . however , to make a correct comparison with the radial field models ( dr1 ) , ( dr2 ) we should take into account that the cyclotron circle rotates around disk so that we should draw the poincar section in the rotational phase @xmath192 . in this representation we see the appearance of the resonance ( see right column of fig . [ fig7 ] ) that is similar to those seen in figs . [ fig4 ] , [ fig5 ] for the radial field models . and microwave phase @xmath98 . _ left column _ : dependence of angle @xmath189 of collision point on disk , counted from @xmath146axis , and radial velocity @xmath156 , taken at collision , on microwave phase @xmath98 . _ right column _ : same as in left column with @xmath192 . here @xmath193 , @xmath152 , @xmath194 , @xmath184 , @xmath195 , there is no noise ; points are shown for times @xmath196 , the capture time of this orbit is @xmath197 . data for model ( d1 ) . ] a more direct correspondence between radial field models ( dr1 ) , ( dr2 ) and the model ( d1 ) with a linearly polarized microwave field is well seen from the poincar sections shown in the rotation frame of phase @xmath192 in fig . [ fig8 ] . in this frame we see directly the resonance at @xmath198 being very similar to the wall case and the radial field model . however , the positions of resonance at @xmath199 are different from those in fig . [ fig5 ] . of course in the rotation frame the orbital momentum is only approximately conserved that gives a broadening of invariant curves in fig . [ fig8 ] . with @xmath192 for @xmath168 , @xmath200 ( left top ) ; @xmath193 , @xmath200 ( right top ) ; @xmath201 , @xmath76 ( left middle ) ; @xmath193 , @xmath76 ( right middle ) ; @xmath201 , @xmath152 ( left bottom ) ; @xmath193 , @xmath152 ( right bottom ) ; @xmath156 is expressed in units of @xmath19 . data for model ( d1 ) , no noise no dissipation . ] we explain this as follows . for the linear polarized field of model ( d1 ) the radial component of microwave field is proportional to @xmath202 where we kept only slow frequency component of radial field ( the neglected term with @xmath203 gives resonant values @xmath204 ) . the radial field @xmath205 gives kicks to the radial velocity component at collisions with disk similar to the case of model ( dr2 ) described by eq . ( [ eq3 ] ) : @xmath206 , @xmath207 . here we use the radial field component phase @xmath208 at a moment of collision with disk ( the tangent component does not change @xmath156 and can be neglected ) . the phase variation @xmath209 has the first term @xmath210 being the same as for the radial field model ( dr2 ) , and an additional term related to rotation around disk with @xmath211 which comes from geometry . indeed , the segment angles of intersections of circles @xmath160 and @xmath161 are : for disk radius @xmath160 it is @xmath212 and for cyclotron radius @xmath161 it is @xmath213 . thus , their ratio is @xmath214 in agreement with the geometrical scaling . this result can be obtained from the expression for @xmath215 by interchange of two disks that gives the above expression for @xmath216 ( at @xmath217 both shifts @xmath218 and @xmath219 are equal ) . thus again the dynamical description is reduced down to the chirikov standard map with slightly modified parameters giving us for the model ( d1 ) the chaos parameter @xmath220 being usually smaller than unity , resonance position @xmath167 , resonance width @xmath173 and the resonance energy width @xmath174 : @xmath221 where @xmath176 is a shift of resonance produced by a finite separatrix half width @xmath222 . for our numerical simulations we have @xmath178 with @xmath223 , @xmath224 and @xmath225 . at @xmath226 eq . ( [ eq5 ] ) gives the values @xmath227 at @xmath168 while the numerical data of fig . [ fig8 ] give @xmath228 , and we have at @xmath193 the theory value @xmath229 being in a good agreement with the numerical value @xmath230 of fig . [ fig8 ] . for @xmath201 we have the resonance position at @xmath231 corresponding to the bulk and thus the resonance is absent . the resonance width in fig . [ fig8 ] at @xmath193 , @xmath200 can be estimated as @xmath232 that is in a satisfactory agreement with the theoretical value @xmath233 from ( [ eq5 ] ) . we remind that in model ( d1 ) we have only approximate conservation of orbital momentum that gives a broadening of invariant curves and makes determination of the resonance width less accurate . in spite of this broadening we see that the resonance description by the chirikov standard map works rather well . in fig . [ fig9 ] we show the dependence of average capture time @xmath187 on @xmath9 in the model ( d1 ) . the averaging is done over @xmath234 trajectories scattered on disk at all such impact parameters that cyclotron orbit can touch the disk . here , we show the ratio @xmath235 where @xmath236 is an average capture time in absence of microwave . according to our numerical data we have approximate dependence @xmath237 corresponding to a period of nonlinear oscillations in a disk map description discussed in @xcite . on @xmath9 at @xmath152 shown at various static fields : @xmath238 ( top left ) , @xmath239 ( top right ) , @xmath240 ( bottom left ) , @xmath241 ( bottom right ) . here @xmath184 , there is no noise . data for model ( d1 ) . ] the data of fig . [ fig9 ] show a clear periodic dependence of capture time @xmath187 on @xmath9 corresponding to the periodicity variation of poincar section with @xmath9 ( see figs . [ fig4 ] , [ fig5 ] . [ fig8 ] ) . this structure is especially well visible at weak static fields . with an increase of @xmath181 this structure is suppressed . indeed , at large @xmath181 even without microwave field the trajectories can escape from disk as it discussed in @xcite and microwave field does not affect the scattering in this regime . the distributions of capture times are shown in fig . [ fig10 ] . we clearly see that at resonant values of @xmath9 a microwave field leads to appearance of long capture times . for example , we have the probability to be captured for @xmath242 being @xmath243 at @xmath193 while at @xmath106 we have @xmath244(left panel in fig . [ fig10 ] ) ; and we have @xmath245 at @xmath246 while at @xmath247 we have @xmath248 ( right panel in fig . [ fig10 ] ) . these data confirm much stronger capture at certain resonant values of @xmath9 . of capture times @xmath187 for parameters of fig . [ fig9 ] . _ left panel : _ @xmath106 ( dashed curve for minimum of capturing probability ) and @xmath193 ( full curve at maximum of capturing probability ) at @xmath194 . _ right panel : _ similar cases at @xmath247 ( dashed curve ) and @xmath246 ( full curve ) at @xmath249 . data are obtained with @xmath250 trajectories started at different impact parameters and running up to time @xmath251 . here @xmath187 is expressed in units of @xmath252 . data for model ( d1 ) . ] the data of figs . [ fig9 ] , [ fig10 ] show that the scattering process on disk is strongly modified by a microwave field . however , to determine the conductivity properties of a sample we need to know what is an average displacement @xmath253 along static field after a scattering on a single disk . indeed , in our model a dissipation is present only during collisions with disk while in a free space between disks the dynamics is integrable and hamiltonian . hence during such a free space motion there is no displacement along the static field ( the dissipative part of conductivity or resistivity appears only due to dissipation on disk ) . the dependence of @xmath253 on @xmath9 is shown in fig . [ fig11 ] . in absence of microwave field at @xmath254 we find @xmath255 that corresponds to a simple estimate @xmath256 . the numerical data show that @xmath253 is practically independent of @xmath181 and that @xmath257 in absence of dissipation at @xmath258 . in presence of microwave field we see that the displacement along the static field has strong periodic oscillations with @xmath9 . the striking feature of fig . [ fig11 ] is the appearance of windows of zero displacement @xmath259 at resonance values @xmath260 . we discuss how this scattering on a single disk modifies the resistance of a sample with large number of disks in next section . along static field after scattering on a single disk shown as a function of @xmath9 at various amplitudes of static and microwave fields : @xmath261 , @xmath262 ( gray dashed curve ) ; @xmath194 , @xmath262 ( black dashed curve ) ; @xmath261 , @xmath152 ( gray full curve ) ; @xmath194 , @xmath152 ( black full curve ) . the data are obtained by averaging over @xmath263 scattered trajectories with random impact parameters ; here @xmath184 , noise amplitude @xmath264 . data for model ( d1 ) . ] to determine a resistance of a sample with many disks we use the following scattering disk model . the scattering on a single disk in a static electric field @xmath181 is computed as it is described in the previous section with a random impact parameter inside the collision cross section @xmath265 . after that a trajectory evolves along @xmath266-axis according to the exact solution of hamiltonian eq . ( [ eq1 ] ) ( no dissipation and no noise ) up to a collision with next disk which is taken randomly on a distance between @xmath267 and @xmath268 where @xmath269 is a mean free path along @xmath183axis and @xmath270 is a two - dimensional density of disks ( of course @xmath271 ) . in a vicinity of disk the dynamical evolution is obtained by runge - kutta solution of dynamical equations as it was the case in previous section . we use low disk density with @xmath272 . the collision with disk is done with a random impact parameter in the @xmath146axis of disk vicinity : the impact parameter is taken randomly in the interval @xmath273 $ ] around disk center . noise acts only when a center of cyclotron radius of trajectory is on a distance @xmath274 from disk center so that a collision with disk is possible . after scattering on a disk a free propagation follows up to next collision with disk . along such a trajectory we compute the average displacement @xmath275 and @xmath276 after a time interval @xmath277 . in this way the number of collisions with disks is @xmath278 and a total displacement in @xmath146axis is @xmath279 where @xmath253 is an average displacement on one disk discussed in previous section . we compute the global displacements @xmath280 on a time interval @xmath281 averaging data over 200 trajectories . we call this system model ( d2 ) . and @xmath282 on magnetic field @xmath283 ( resistivity is expressed in arbitrary numerical units ) . blue and gray curves show respectively @xmath120 and @xmath282 in absence of microwave field . black curve with points show @xmath120 dependence of @xmath284 at microwave field @xmath152 ( here @xmath284 is expressed in units of @xmath285 ) . here @xmath194 , @xmath184 , noise amplitude @xmath264 , @xmath27 . resonant values @xmath11 are shown by arrows ; bars show statistical errors for @xmath120 . data are obtained by averaging over 200 trajectories propagating up to time @xmath286 . data for model ( d2 ) . ] then the current components are equal to @xmath287 , @xmath288 and conductivity components are @xmath289 , @xmath290 ( the current is computed per one electron ) . we work in the regime of weak _ dc-_field where @xmath291 scales linearly with @xmath145 . the current @xmath292 is determined by the drift velocity @xmath293 . since the mean free path is large compared to disk size @xmath294 we have an approximate relation @xmath295 , @xmath296 . as in 2deg experiments @xcite we have in our simulations @xmath297 ( see fig . [ fig12 ] ) . the resistivity is obtained by the usual inversion of conductivity tensor with @xmath298 , @xmath299 . the dependence of @xmath120 , @xmath282 , expressed in arbitrary numerical units , on magnetic field @xmath300 is shown in fig . [ fig12 ] . in absence of microwave field we find @xmath301 and @xmath302 similar to experiments @xcite . for small noise amplitude ( e.g. @xmath264 ) we have @xmath120 growing linearly with @xmath284 ( see fig . [ fig12 ] ) but at larger amplitudes ( e.g. @xmath303 ) its increase with @xmath284 becomes practically flat showing only @xmath304 increase in a give range of @xmath284 variation . ( left panel ) and @xmath282 ( right panel ) as function of @xmath0 at various noise amplitudes @xmath264 ( black curve ) , @xmath305 ( blue / dark curve ) , @xmath306 ( red / gray curve ) . here @xmath152 , @xmath194 , other parameters are as in fig . [ fig12 ] . curves are drown though numerical points obtained with a step @xmath307 . data for model ( d2 ) . ] in presence of microwave field the dependence of @xmath120 on @xmath284 is characterized by periodic oscillations with minimal @xmath120 values being close to zero at resonant values of @xmath308 well visible in fig . [ fig12 ] . the dependence of @xmath120 , @xmath282 rescaled to their values @xmath309 , @xmath310 in absence of microwave field are shown in fig . [ fig13 ] at various amplitudes of noise and fixed @xmath63 , and in fig . [ fig14 ] at various @xmath63 and fixed noise amplitude @xmath139 . we see that increase on noise leads to an increase of minimal values of @xmath120 at resonant values @xmath11 . in a similar way a decrease of microwave power leads to increase of minimal values of @xmath120 at @xmath11 . at the same time the hall resistance @xmath282 is only weakly affected by microwave radiation as it also happens in zrs experiments . at various microwave amplitudes @xmath200 ( red / gray curve ) , @xmath306 ( blue / dark curve ) , @xmath311 ( black curve ) ; amplitude of noise is fixed at @xmath264 . data for model ( d2 ) . ] these results are in a qualitative agreement with the zrs experiments . on the basis of our numerical studies we attribute the appearance of approximately zero resistance at @xmath11 values in our bulk model of disk scatterers to long capture times of orbits in disk vicinity at these @xmath11 values ( see fig . [ fig9 ] ) . during this time @xmath187 noise gives fluctuations of collisional phase @xmath189 and due to that a cyclotron circle escapes from disk practically at random displacement @xmath253 that after averaging gives average @xmath257 . since resistivity is determined by the average value of @xmath253 this leads to appearance of zrs . we note that this mechanism is different from the one of edge transport stabilization discussed here and in @xcite . however , both mechanisms are related to a long capture times near edge or near disk that happens due to synchronization of cyclotron phase with microwave field phase and capture inside the nonlinear resonance . of trajectories at the moment of collisions with disks for parameters of fig . [ fig14 ] at @xmath152 : @xmath168 ( top left ) , @xmath193 ( top right ) , @xmath201 ( bottom left ) , @xmath312 ( bottom right ) . here @xmath192 where @xmath91 is a microwave phase at the moments of collisions with disk and @xmath189 is the angle on disk at collision moment , counted from @xmath47-axis ( same as in figs . [ fig7 ] , [ fig8 ] ) . data are obtained from @xmath313 trajectories iterated up to time @xmath286 . density of points is shown by color with black at zero and white at maximum density . the average number of collisions per disk per trajectory is @xmath314 , @xmath315 , @xmath316 , @xmath317 respectively for @xmath168 , @xmath318 , @xmath319 , @xmath320 . data are obtained for model ( d2 ) . ] to illustrate the capture inside the resonance we present the distributions of trajectories from fig . [ fig14 ] shown in the phase space plane @xmath321 at the moments of collisions with disks in fig . [ fig15 ] . this is similar to the poincar sections of fig . [ fig8 ] however , now we consider the real case of diffusion and scattering on many disks in the model ( d2 ) with noise and dissipation . we see that for @xmath193 orbits are captured in a vicinity of the center of nonlinear resonance at @xmath322 well seen in fig . for @xmath168 we have a density maximum located at smaller values of @xmath167 and @xmath322 even if there is a certain shift of @xmath167 produced by a significant resonance width at @xmath152 . at @xmath201 we have a density maximum at @xmath323 corresponding to an unstable fixed point of separatrix . the total number of collision points @xmath324 in this case is by a factor @xmath325 smaller than in the case of stable fixed point at @xmath193 . a similar situation is seen in the case of wall model ( w1 ) ( see fig.1d , f in @xcite ) even if there the ratio between number of captured points was significantly larger . the results of fig . [ fig15 ] show that in the zrs phase the collisions with disk indeed create synchronization of cyclotron and microwave phases and capture of trajectories inside the nonlinear resonance . however , there are also some distinctions between bulk disk model ( d2 ) and experimental observations . the first one is that there are minima for @xmath326 but there are no peaks which are well visible around integer @xmath9 values in zrs experiments @xcite,@xcite and numerical simulations of transport along the edge @xcite . the second one is appearance of small negative values of @xmath120 at @xmath11 values . on @xmath9 in the disk model with dissipation at disk collisions at rate @xmath327 and dissipation in disk vicinity with rate @xmath328 ; a disk roughness gives additional angle rotations with amplitude @xmath329 ( see text ) ; the amplitude of noise in disk vicinity is @xmath330 . here we have @xmath152 , @xmath194 ; 51 numerical points in @xmath9 are connected by lines to adapt an eye . data are obtained by averaging over 100 trajectories propagating up to time @xmath286 . data for model ( d3 ) . ] we attribute the absence of peaks to a specific dissipation mechanism which takes place only at disk collisions . it is rather convenient to run long trajectories using exact solution for free propagation between disks . indeed , in this scheme there is no dissipation during this free space propagation and thus these parts of trajectories have no displacement along static field . we also tested a dissipation model with additional @xmath331 for @xmath332 and @xmath333 for @xmath334 . this dissipation works only in a disk vicinity when the distance between disk center and cyclotron center is smaller than @xmath335 . the dissipation @xmath336 on disk remains unchanged . we also added a certain roughness of disk surface modeled as an additional random angle rotation of velocity vector in the range @xmath337 , done at the moment of collision with disk . we call this system disk model ( d3 ) . the results for the resistivity ratio @xmath326 are shown in fig . [ fig16 ] . they show an appearance of clear peaks of @xmath326 in presence of such additional dissipation in vicinity of integer @xmath9 . there is also a small shift of minima from integer plus @xmath45 to integer plus @xmath338 . the second point of distinctions from zrs experiments is a small negative value of @xmath120 at resonant @xmath9 values . it is relatively small for disk model ( d2 ) ( see figs . [ fig13 ] , [ fig14 ] ) and it becomes more pronounced for disk model ( d3 ) ( see fig . [ fig16 ] ) . it is possible that a scattering on disk in presence of dissipation , noise , static and microwave field gives a negative displacement @xmath253 which generates such negative @xmath120 values . we expect that in the limit of static field going to zero this effect disappears . indeed , the negative values become smaller at smaller @xmath181 according to data of fig . [ fig11 ] but unfortunately the small @xmath181 limit is also very difficult to investigate numerically . we consider that at this stage of the theory the presence of negatives values for @xmath120 does not constitute a critical disagreement . the escape parameters for electrons that have been captured on an impurity for a time long enough to make many rotations around it , are likely to strongly depend on the model for the electron impurity interaction and further theoretical work on a more microscopic model is needed . in general a zero average displacement along the field direction seems natural for a smooth distribution of trapping times with a characteristic time scale much larger than the rotation time around the impurity ( this assumption does not seem to hold for our model , see for example the sharp features on fig . finally in section vii we propose a slightly different mechanism by which the combination of trapping on impurities investigated here and electron - electron interactions can lead to zrs . the zrs experiments @xcite show that the resistance @xmath120 in the zrs minima scales according to arrehenius law @xmath339 with a certain energy scale dependent on a strength of microwave field . in typical experimental conditions one finds very large @xmath340 at @xmath341 ( see e.g. fig.3 in @xcite ) . these data also indicate the dependence @xmath342 . this energy scale @xmath343 is very large being only by a factor @xmath344 smaller than the fermi energy @xmath345 . at the same time the amplitude of microwave field is rather weak corresponding to @xmath346 at field of @xmath347 or ten times larger at @xmath348 ( unfortunately it is not known what is an amplitude of microwave field acting on an electron ) . as in @xcite we argue that the arrehenius scale is determined by the energy resonance width ( [ eq4 ] ) with @xmath349 . indeed , the resonance forms an energy barrier for a particle trapped inside the resonance by dissipative effects being analogous to a wash - board potential . an escape from this potential well requires to overcome the energy @xmath350 leading to the arrehenius law for @xmath120 dependence on temperature . assuming the case of the wall with @xmath12 we obtain at @xmath351 the activation temperature @xmath352 being is a satisfactory agreement with the experimental observations . the theoretical relation ( [ eq4 ] ) also reproduces the experimental dependence @xmath353 at @xmath12 . in his relation @xmath354 being confirmed by the numerical simulations presented in @xcite . this dependence is in a satisfactory agreement with the power dependence found in experiments @xcite . in other samples one finds that the dependence @xmath355 works in a better way . we think that higher terms in a nonlinear resonance can be responsible for scaling @xmath355 being different from the relation ( [ eq4 ] ) . also a finite rigidity of the wall or disk scatterers can be responsible for appearance of higher power of @xmath63 . the energy scale @xmath350 on disks is enhanced by a factor @xmath94 for the case of radial field ( [ eq4 ] ) . however , we showed that for a linear polarization the scale @xmath350 is given by eq.([eq5 ] ) and thus there is no enhancement at large @xmath92 . indeed , we performed direct simulations at parameters of fig . [ fig14 ] with the reduced value of disk radius by a factor 2 . the numerical data give approximately the same traces @xmath326 vs. @xmath9 at @xmath356 without visible signs of deeper minima at small @xmath63 . this confirms the theoretical expressions ( [ eq5 ] ) . in any case , for small values @xmath158 one should analyze the quantum scattering problem which is significantly more complex compared to the classical case . we may assume that in a quantum case one should replace @xmath160 by a magnetic length @xmath357 . in such a case we are getting @xmath358 that gives @xmath359 at @xmath360 and microwave amplitude @xmath361 . however , in this case we obtain the scale @xmath362 being practically independent of @xmath9 which differs from experimental data . in any case in experiments the size of impurities is small compared to @xmath161 and a quantum treatment is required to reproduce the correct picture for @xmath120 dependence on parameters in the zrs phase . another point is related to the positions of zrs minima on @xmath9 axis . we remind that that for the wall model the resonance is located at @xmath363 ( [ eq4 ] ) and that the separatrix width is @xmath364 . the capture of trajectories from the bulk is most efficient when a half width of separatrix touch the border of bulk at @xmath365 with @xmath366 that gives the expression @xmath367 for the wall case . at @xmath368 , @xmath193 this gives @xmath369 being in a good agreement with the numerical data @xmath370 for @xmath120 dependence on @xmath9 ( see figs.2,3 in @xcite with a visible tendency of @xmath103 growth with @xmath9 ) . for the data presented here in fig . [ fig14 ] for the disk case at @xmath152 , @xmath178 , @xmath193 we obtain from ( [ eq5 ] ) @xmath371 that is slightly less than the numerical value @xmath372 for minima location . we attribute this difference to an approximate nature of expression for the resonance width at relatively strong microwave fields . we also note that in experiments an additional contribution to the value of @xmath103 can appear due to a finite rigidity of disk and wall potentials . the theoretical models presented here and in @xcite reproduce main experimental features of zrs experiments @xcite,@xcite . however , it would be useful to have some additional theoretical predictions which can be tested experimentally . a certain characteristic feature of both wall and disk models is appearance of nonlinear resonance . for example , according to the wall model ( w2 ) described by eq . ( [ eq3 ] ) the dynamics inside the resonance is very similar to dynamics of a pendulum . the frequency of phase oscillations inside the resonance is @xmath373 @xcite . here the frequency is expressed in number of map iterations and since the time between collisions is approximately @xmath82 we obtain the physical frequency @xmath374 of these resonant oscillations being @xmath375 . at @xmath376 this frequency is significantly smaller than the driving microwave frequency . the dynamics inside the resonance should be very sensitive to perturbations at frequency @xmath377 that gives : @xmath378 to check this theoretical expectation we study numerically the effect of additional microwave driving with dimensional amplitude @xmath379 ( @xmath380 ) and frequency @xmath381 . we use the wall model ( w2 ) based on the chirikov standard map described here and in @xcite . as for the additional driving frequency we have @xmath382 where @xmath383 is the field strength of microwave frequency @xmath381 , we assume that both main and additional fields are collinear and perpendicular to the wall . in presence of second frequency the map ( [ eq3 ] ) takes the form : @xmath384 the only modification appears in the first equation since now the change of velocity at collision depends on both fields ; the second equation remains the same as in ( [ eq3 ] ) . as in the model ( w2 ) the term @xmath385 describes the effects of dissipation with rate @xmath97 and noise with amplitude @xmath386 of random velocity angle rotations . we call this system model ( w3 ) . on frequency ratio @xmath387 in model ( w3 ) described by the map ( [ eq7 ] ) . here , the test driving at frequency @xmath381 has fixed amplitude @xmath388 ; the main microwave driving is located in the zrs phase at @xmath389 , and its amplitude takes values @xmath390 ( blue curve ) , @xmath391 ( green curve ) , @xmath311 ( magenta curve ) , @xmath392 ( red curve ) ( these curves follow from top to bottom at @xmath393 ) . the values of @xmath120 are computed at fixed @xmath388 and corresponding @xmath63 ; the values of @xmath394 are computed at @xmath395 and @xmath76 . the data are obtained at noise amplitude @xmath396 and dissipation @xmath397 ; averaging is done over 2000 trajectories for 5000 iterations of map ( [ eq7 ] ) . _ right panel : _ dependence of peak position @xmath387 on main microwave driving amplitude @xmath63 obtained from data of left panel at @xmath388 and additional data at @xmath398 ( blue points ) , the theory dependence ( [ eq6 ] ) at @xmath399 is shown by the straight red line . data for model ( w3 ) . , title="fig : " ] on frequency ratio @xmath387 in model ( w3 ) described by the map ( [ eq7 ] ) . here , the test driving at frequency @xmath381 has fixed amplitude @xmath388 ; the main microwave driving is located in the zrs phase at @xmath389 , and its amplitude takes values @xmath390 ( blue curve ) , @xmath391 ( green curve ) , @xmath311 ( magenta curve ) , @xmath392 ( red curve ) ( these curves follow from top to bottom at @xmath393 ) . the values of @xmath120 are computed at fixed @xmath388 and corresponding @xmath63 ; the values of @xmath394 are computed at @xmath395 and @xmath76 . the data are obtained at noise amplitude @xmath396 and dissipation @xmath397 ; averaging is done over 2000 trajectories for 5000 iterations of map ( [ eq7 ] ) . _ right panel : _ dependence of peak position @xmath387 on main microwave driving amplitude @xmath63 obtained from data of left panel at @xmath388 and additional data at @xmath398 ( blue points ) , the theory dependence ( [ eq6 ] ) at @xmath399 is shown by the straight red line . data for model ( w3 ) . , title="fig : " ] in the model ( w3 ) the resistance @xmath120 is computed numerically in the same way as in the model ( w2 ) described in @xcite : the displacement along the edge between collisions is @xmath400 ; it determines the total displacement @xmath253 along the edge during the total computation time @xmath401 ; then @xmath402 where @xmath403 is an effective diffusion rate along the edge . to see the effect of additional weak test driving @xmath379 at frequency @xmath381 we place the system in the zrs phase at @xmath389 and measure the variation of rescaled resistance @xmath404 . here @xmath120 is the resistance in presence of both microwave fields @xmath63 and @xmath379 while @xmath394 is the resistance at @xmath395 and a certain fixed @xmath63 . the dependence of @xmath404 on the frequency ratio is shown in fig . [ fig17 ] at left panel . the main feature of this data is appearance of peak at low frequency ratio @xmath405 . in the range @xmath406 the testing field @xmath379 is nonresonant and does not affect @xmath120 however at @xmath405 it becomes resonant to the pendulum oscillations in the wall vicinity and hence strongly modifies @xmath120 value . the dependence of this resonance ratio @xmath387 on amplitude of main driving field @xmath63 is shown in right panel of fig . [ fig17 ] . the numerical data are in a good agreement with the above theoretical expression ( [ eq6 ] ) . the theoretical dependence ( [ eq6 ] ) allows to check the synchronization theory of edge state stabilization . it also allows to measure the strength of main microwave driving force acting on an electron that still remains an experimental challenge . the experimental testing of relation ( [ eq6 ] ) requires to work with good zrs samples which have every low resistance in zrs minima since this makes the effect of testing field @xmath381 to be more visible . we note that the recent experiments in a low frequency regime @xmath407 @xcite demonstrate that @xmath120 is sensitive to low frequency driving . the expression ( [ eq6 ] ) is written for the case when @xmath120 is mainly determined by a transport along edges . if the dominant contribution is given by bulk disk scatterers then a certain numerical coefficient @xmath408 should be introduced in the right part of the expression . according to the data of fig . [ fig8 ] and eqs . ( [ eq5 ] ) we estimate @xmath409 ( the separatrix width is smaller for the disk case compared to the wall case at the same @xmath63 ) . another interesting experimental possibility of our theory verification is to take a hall bar of high mobility 2deg sample and put on it antidots with regular or disordered distribution ( it is important to have no direct collision - less path for a cyclotron radius in crossed _ dc-_electric and magnetic fields ) with a low density of antidot disks @xmath410 ( as in our numerical studies ) so that an average distance between antidots is larger than the cyclotron radius @xmath161 . the regular antidot lattices have been already realized experimentally @xcite . the effect of microwave field on electron transport in a regular lattice has been studied in the frame of ratchet transport in asymmetric lattices @xcite . even a case of symmetric circular antidots has been studied in @xcite but the lattice was regular and no special attention was paid to analysis of resistivity at zrs resonant regime with @xmath136 . we think that the experimental condition of @xcite can be relatively easy modified to observe the zrs effect on disk scatterers discussed here . above we presented theoretical and numerical results which in our opinion explain the appearance of microwave induced zrs in high mobility samples . the synchronization theory of zrs proposed in @xcite and extended here is based on a clear physical picture : high harmonics @xmath411 are generated by collisions with sharp edge boundary or isolated impurities which are modeled here by specular disks . the zrs phases appear in a vicinity of resonant values @xmath412 . at these @xmath11 values the cyclotron phase of electron motion becomes synchronized with the microwave phase due to dissipative processes present in the system . for trajectories at the edge vicinity this synchronization gives stabilization of propagation along edge channels that creates an exponential drop of resistivity contribution of these channels with decreasing amplitude of thermal noise and increasing amplitude of microwave field . the contribution to resistivity from trajectories in the bulk is analyzed in the frame of scattering on many well separated disk impurities . here again the synchronization of cyclotron phase with the microwave phase takes place approximately at the same resonant @xmath11 values . at these @xmath11 values the synchronization leads to long time capture of trajectories in disk vicinity . during this long time an initial cyclotron phase is washed out by noisy fluctuations and many rotations around disk and thus an electron escapes from a disk with an average zero displacement along the applied _ dc-_field even if dynamics in disk vicinity is dissipative . this provides the main mechanism of suppression of dissipative resistivity contribution from isolated impurities in the bulk . as a result the contribution of bulk to dissipative conductivity @xmath413 is suppressed , as it was assumed in @xcite , and the main contribution to current is given by electron propagation along edge states stabilized by a microwave field . as we showed above the resonance width or resonance energy scale @xmath350 are approximately the same for the disk and wall cases ( see eqs . ( [ eq4 ] , [ eq5 ] ) ) . we note that for the disk case the energy @xmath350 is not sensitive to the disk radius as soon as it is significantly smaller than the cyclotron radius . thus we expect that at @xmath11 values the conductivity @xmath413 in the bulk is suppressed by a microwave field and at these fields the current is flowing essentially along stabilized edge states . in the case of corbino geometry we have radial conductivity @xmath414 which is determined by the bulk scattering and now the minima of @xmath414 are located at @xmath11 values ( see e.g figs . [ fig12 ] , [ fig13],[fig14 ] where @xmath415 ) . the zrs experiments performed in the corbino geometry give minima of @xmath414 at these @xmath11 values ( see e.g. @xcite ) being in agreement with the synchronization theory . it is interesting to note that the nonlinear dynamics in vicinity of edge and disk impurity is well described by the chirikov standard map @xcite . the map description explains the location of resonances at integer values of @xmath9 with an additional shift @xmath370 produced by a finite separatrix width of nonlinear resonance . a finite rigidity of wall or disk potential can give a modification of this shift @xmath103 . our results show that the zrs phases at @xmath11 appear only at weak noise corresponding to high mobility samples . strong noise destroys synchronization and trajectories are no more captured at edge or disk vicinity . we also note that internal sample potentials with significant gradients act like a strong local _ dc-_field which destroys stability regions around disk impurities or near edge . due to that the zrs effect exists only in high mobility samples . the resistance at zrs minima drops significantly with the growth of microwave field strength since it increases the amplitude of nonlinear resonance which captures the synchronized trajectories . the synchronization theory of zrs is based on classical dynamics of noninteracting electrons . it is possible that electron - electron interaction effects can also suppress the contribution to resistivity from neutral short range range scatterers ( interface roughness , adatoms , ... ) . indeed , long capture times can increase the electron density around these short ranged impurities transforming them into long range charged scatters that the other electrons can circumvent by adiabatically following the long range component of the disorder potential thereby avoiding a scattering event . however , the theoretical description of this short - ranged impurity cloaking mechanism for the zrs effect remains a serious challenge . another important step remains the development of a quantum synchronization theory for zrs . even if in experiments the landau level is relatively high @xmath416 , there are only about ten landau levels inside a nonlinear resonance @xcite and quantum effects should play a significant role . the general theoretical studies show that the phenomenon of quantum synchronization persists at small effective values of planck constant @xmath417 but it becomes destroyed by quantum fluctuations at certain large values of @xmath417 @xcite . the importance of quantum zrs theory is also related to a short range nature of the impurities considered here , typically on a scale of a few nanometers or even less . we have modeled these impurities by disks with a radius that was only several times ( in fact @xmath9 times ) smaller than the cyclotron radius which is not so close to microscopic reality . we could argue that in the quantum case a nanometer sized impurity would act effectively as an impurity of a size of quantum magnetic length @xmath418 . this gives a ratio @xmath419 which is comparable with the one used in our simulations with @xmath420 but of course a quantum treatment of scattering on nanometer size impurity in crossed electric and magnetic and also microwave fields remains a theoretical challenge . we note that such type of scattering can be efficiently analyzed by tools of quantum chaotic scattering ( see e.g. @xcite ) and we expect that these tools will allow to make a progress in the quantum theory development of striking zrs phenomenon . we hope that the synchronization theory of microwave induced zrs phenomenon described here can be tested in further zrs experiments . this work was supported in part by anr france pnano project nanoterra ; ovz was partially supported by the ministry of education and science of russian federation . mani , j.h . smet , k. von klitzing , v. narayanamurti , w.b . johnson , and v. umansky , nature * 420 * , 646 ( 2002 ) . zudov , r.r . du , l.n . pfeiffer , and k.w . west , phys . lett . * 90 * , 046807 ( 2003 ) . dorozhkin , jetp lett . * 77 * , 577 ( 2003 ) j. h. smet , b. gorshunov , c. jiang , l. pfeiffer , k. west , v. umansky , m. dressel , r. meisels , f. kuchar , and k. von klitzing , phys . lett . * 95 * , 116804 ( 2005 ) a.a . bykov , a.k . bakarov , d.r . islamov , and a.i . toropov , jetp lett . * 84 * , 391 ( 2006 ) d. konstantinov , and k. kono , phys . lett . * 105 * , 226801 ( 2010 ) d.konstantinov , a.d.chepelianskii , and k.kono , jour . japan * 81 * , 093601 ( 2012 ) i.a . dmitriev , a.d . mirlin , d.g . polyakov , and m.a . zudov , rev . mod . phys . * 84 * , 1709 ( 2012 ) a.d . chepelianskii , and d.l . shepelyansky , phys . b * 80 * , 241308(r ) ( 2009 ) a. pikovsky , m. rosenblum , and j. kurths , _ synchronization : a universal concept in nonlinear sciences _ , cambridge univ . press , cambridge ( 2001 ) a.d . chepelianskii , _ non linear transport in drift - diffusion equations under magnetic field _ , + arxiv:1110.2033[cond - mat.mes - hall ] ( 2011 ) w. g. hoover , _ time reversibility , computer simulation , and chaos _ , world sci . , singapore ( 1999 ) b.v . chirikov , phys . rep . * 52 * , 263 ( 1979 ) a.j . lichtenberg , and m.a . lieberman , _ regular and chaotic dynamics _ , springer , berlin ( 1992 ) . b.chirikov , and d.shepelyansky , scholarpedia * 3(3 ) * , 3550 ( 2008 ) y. avishai , and g. montambaux , eur . j. b * 79 * , 215 ( 2011 ) f. benvenuto , g. casati , i. guarneri , and d. l. shepelyansky , z. phys . b cond . mat . * 84 * , 159 ( 1991 ) d. weiss , m. l. roukes , a. menschig , p. grambow , k. von klitzing , g. weimann , phys . * 66 * , 2790 ( 1991 ) g. m. gusev , v. t. dolgopolov , z. d. kvon , a. a. shashkin , v. m. kudryashov , l. v. litvin and yu . nastaushev , pisma zh . . fiz . * 54 * , 369 ( 1991 ) r. fleischmann , t. geisel , and r. ketzmerick , phys . lett . * 68 * , 1367 ( 1992 ) t. geisel , r. ketzmerick and o. schedletzky , phys . rev . lett . * 69 * , 1680 ( 1992 ) n. berglund , a. hansen , e. h. hauge , and j. piasecki , phys . lett . * 77 * , 2149 ( 1996 ) a.d . chepelianskii , and d.l . shepelyansky , phys . b * 63 * , 165310 ( 2001 ) a.d . chepelianskii , j. laidet , i. farrer , h.e . beere , d.a . ritchie , and h. bouchiat , phys . b * 86 * , 205108 ( 2012 ) s. sassine , yu . krupko , j .- c . portal , z.d . kvon , r. murali , k.p . martin , g. hill , and a.d . wieck , phys . b * 78 * , 045431 ( 2008 ) c.l . yang , m.a . zudov , t.a . knuuttila , r.r . du , l.n . pfeiffer , and k.w . west , phys . * 91 * , 096803 ( 2003 ) a.a . bykov , jetp lett . * 87 * , 551 ( 2008 ) o.v . zhirov , and d.l . shepelyansky , eur . j. d * 38 * , 375 ( 2006 ) j. wiersig , and m. hentschel , phys . lett . * 100 * , 033901 ( 2008 ) a. eberspcher , j. main , and g. wunner , phys . e * 82 * , 046201 ( 2010 )

in this paper we study an application of left - invariant total ordering of braid group @xmath1 to knot theory . an left - invariant total ordering , called dehornoy ordering is an interesting structure of braid groups found in 1990 s by dehornoy ( @xcite ) . today the ordering is interpreted in both algebraic and geometric way and these interpretation shows the dehornoy ordering is a quite natural structure of braid groups . braid groups are very important tool for studying links . many link invariants such as jones polynomial are first constructed via closed braid representative of links , so it is natural to study an application of this interesting structure of braid groups to knot theory . in @xcite,@xcite it is found that braid ordering can be seen as some kinds of restriction for admissibility of braid moves . braid moves are moves of closed braids which do not change their representing link types . most typical ( and important ) example of braid moves are _ exchange move _ or _ destabilization , stabilization _ , which is shown in figure [ braidmove ] . these works suggest that there would be some relationships between birman - menasco s braid foliation theory and ordering of braid groups because braid moves appears naturally in braid foliation theory ( @xcite ) . for example , exchange move is used to make braid foliation simple and it is one of the most fundamental operation in birman - menasco s braid foliation theory . indeed , we investigate relationship between braid foliation and left invariant total ordering of braid groups , and obtain following result : [ thm : main1 ] let @xmath2 be a braid and @xmath0 be an essential , genus @xmath3 closed surface in the complement of closed braid @xmath4 . 1 . if @xmath0 is tiled , then @xmath5_{d } < g+2 $ ] . 2 . if @xmath0 is mixed - foliated , @xmath5_{d } < 2g+1 $ ] . 3 . if @xmath5_{d } \geq 2g+1 $ ] , then @xmath0 is circular foliated . in above statements , @xmath6_{d}$ ] represents _ dehornoy floor _ of @xmath7 , which is a non - negative integer defined in section 2 . this integer is determined by the dehornoy ordering of braid groups , and efficiently computable . the notion of tiled , mixed foliated or circular foliated are derived from birman - menasco s braid foliation theory , which will be explained in section 3 . proof of theorem [ thm : main1 ] will be given in section 4 . theorem [ thm : main1 ] means braid ordering gives some information about the geometry of closed braid complements , and especially if dehornoy floor is relatively large with respect to its genus , then the surface can be positioned to special position , which is simple . we remark that dehornoy floor might vary under even conjugation , stabilization or destabilization , so it is far from knot invariant . indeed , dehornoy floor of braids obtained by performing stabilization always vanish . nevertheless theorem [ thm : main1 ] seems to suggests that braid ordering provide some important information about its closure which can not be obtained from known invariants of links . as an application of theorem [ thm : main1 ] we obtain a sufficient condition for nielsen - thurston classification determines the geometry of its closure s complement . nielsen - thurston theorem states braids are classified into three types by their dynamics ; _ periodic , reducible _ and _ pseudo - anosov_. on the other hand , a famous thurston s hyperbolization theorem shows knots are classified into following three types by the geometry of their complements ; _ torus knots , satellite knots , _ and _ hyperbolic knots_. although there are many similarities between these two classifications , in general there are no one - to - one correspondence of them . in figure [ fig : example ] , we give an example of periodic , reducible and pseudo - anosov braids whose closure represents @xmath8-torus knot . therefore the following question naturally arises : _ when the nielsen - thurston classification of braids completely determine the geometrical types of its closure ? _ theorem [ thm : main2 ] partially answers the question . -torus knots , width=226 ] [ thm : main2 ] let @xmath2 be a braid whose closure is a knot . if @xmath5_{d } \geq 3 $ ] , following holds : 1 . @xmath7 is periodic if and only if @xmath4 is a torus knot . 2 . @xmath7 is reducible if and only if @xmath4 is a satellite knot . 3 . @xmath7 is pseudo - anosov if and only if @xmath4 is a hyperbolic knot . especially , under this assumption , satellite knots are all braided - satellite . we recall that satellite knot @xmath9 is called braided satellite if @xmath9 admit closed braid representatives whose axis does not intersect some essential tori in the complement of @xmath9 . this theorem implies , under the condition @xmath5_{d } \geq 3 $ ] , the nielsen - thurston classification of braid groups and the geometric classification of their closures are in one - to - one correspondence . in @xcite , los studied the relationship between these two classifications in the case knots are represented by closed braid with minimum braid index . main result of @xcite shows under the condition of braid index , two classifications are related each other by performing exchange move . however , in general to determine braid index of given knots or links are very difficult question and to see whether or not given braids admit exchange move is also difficult , so the relationships between two classifications can not be seen directly . in this regard , theorem [ thm : main2 ] gives more general and direct condition for the question . we will give a proof of theorem [ thm : main2 ] in section 5 , and in this section we also give concrete method to create hyperbolic knots via nielsen - thurston theory . the author would like to thank his adviser toshitake kohno for many helpful suggestions and advices for preparing this manuscript . he also would like to thank eiko kin for useful conversations . in this section we review some of fundamental facts of dehornoy ordering and dehornoy floor . braid group @xmath1 is , by definition , a group which is presented by @xmath10 geometrically , elements of braid group @xmath2 are represented by braided @xmath11-strands which move from above to below , oriented downward . by connecting their endpoints of strands canonically , we obtain an oriented link . we denote this link by @xmath4 and call _ closed braid _ or _ closure _ of @xmath7 . we call the braid @xmath12 _ garside s fundamental braid _ and denote it by @xmath13 . the garside fundamental braid@xmath13 plays an important role in braid groups . it is known that @xmath14 generates the center of @xmath1 , which is an infinite cyclic group . for these basic facts about braid group , see @xcite . in this section we give a brief exposition of left - invariant total ordering of braid groups called dehornoy ordering , and define dehornoy floor . dehornoy floor is first introduced in @xcite , though they do not use term dehornoy floor . this is a measure of how far braid is from the identity elements with respect to dehornoy ordering and can be seen as some kinds of complexity of braids . we describe fundamental results about dehornoy floor , proved in @xcite,@xcite . a braid @xmath2 is called @xmath15-@xmath16 if @xmath7 can be represented by a word which contains at least one @xmath17 , and contains no @xmath18 for some @xmath19 . we say @xmath20 is true if and only if the braid @xmath21 is @xmath15-positive . it is known that the relation @xmath22 defines left - invariant total ordering of @xmath1 ; that is , @xmath22 is total ordering and if braids @xmath23 satisfy @xmath24 , then for all braid @xmath25 , @xmath26 holds . we call this ordering _ dehornoy ordering_. dehornoy ordering has following nice property , called _ property @xmath27 _ or _ subword property_. [ prop : property - s ] for any braids @xmath28 and @xmath19 , @xmath29 holds . as we mentioned at introduction , there are many other interpretation and definition of dehornoy ordering . see @xcite for those other definitions and meanings . we will describe one of the geometrical definition later . using dehornoy ordering , we define dehornoy floor of braids @xmath5_{d}$ ] as follows . dehornoy floor @xmath5_{d}$ ] of braid @xmath2 is a minimal non - negative integer @xmath30 which satisfies @xmath31 , where @xmath32 . now we remark one important facts . we must be careful that when we think about dehornoy floor , which @xmath1 a braid @xmath7 belongs to is very important . for example , for a braid @xmath33 , @xmath5_{d}=1 $ ] if we consider @xmath34 , and @xmath5_{d}=0 $ ] if we consider @xmath35 . fortunately , if we use braid groups to describe links , the number of strands are always implicit , so there might be no confusion about the number of braid strands . so in most cases , we omit to write which @xmath1 a braid @xmath7 belongs to . we also remark that dehornoy floor is not a conjugacy invariant of braids . for example , 3-braid @xmath36 has dehornoy floor @xmath37 but its conjugation @xmath38 has zero dehornoy floor . this is mainly because dehornoy ordering is not right - invariant ; that is , it does not hold that for every @xmath25 , @xmath39 whenever @xmath20 . thus , dehornoy floor is far from knot invariant . the most important property of dehornor floor is that it can be seen as some kinds of restriction for admissibility of braid moves . we summarize the basic properties of dehornoy floor , all of which are proved in @xcite , or @xcite . [ prop : dehornoyfloor ] let @xmath23 . then following holds . 1 . if braid @xmath2 is represented by braid word which contains @xmath40 occurrence of @xmath41 and @xmath42 occurrence of @xmath43 , then @xmath5_{d } < max\{s , k\}$ ] . @xmath44_{d}-[\beta']_{d}|\leq 1 $ ] if @xmath7 and @xmath45 are conjugate . 3 . @xmath46_{d } \leq 1+[\alpha]_{d}+[\beta]_{d}$ ] . 4 . if closure of braid @xmath2 admit destabilization or exchange move , @xmath5_{d } \leq 1 $ ] . 5 . for every @xmath11 , there exist positive integer @xmath47 such that for every @xmath2 , if @xmath5_{d}>r(n)$ ] then @xmath4 is the unique closed @xmath11-braid representative of link types represented by its closure . that is , if @xmath48 holds for some @xmath49 , @xmath7 and @xmath45 are conjugate . we remark that dehornoy floor is not a conjugacy invariant but proposition [ prop : dehornoyfloor ] ( 2 ) shows taking conjugate changes dehornoy floor at most by one . our definition of dehornoy floor is slightly different from one in @xcite . our definition is rather after @xcite . we adopt this definition because for application of knot theory , this formulation seems to be more natural and makes statement of results simple . there exist an infinite family of left - invariant total ordering of braid groups , containing dehornoy ordering , called _ thurston - type ordering _ ( see @xcite or @xcite ) . most of results in this paper can be stated by using thurston - type ordering instead of using dehornoy ordering , especially using `` thurston floor '' defined by the similar way . now we give a geometric description of dehornor ordering given in @xcite . let @xmath50 be a graph in @xmath51 consisting of horizontal line which connects puncture points . see figure [ fig : braidorder_meaning ] . we identify braid groups as a relative mapping class group @xmath52 , which is a group of isotopy classes of homeomorphisms of @xmath51 whose restriction to boundary @xmath53 is an identity . for braid @xmath54 , consider the image of graph @xmath50 . by performing isotopy , we can assume @xmath55 and @xmath56 have minimal intersections . if @xmath57 , these graphs must be different , so at some puncture point two graphs will diverge . we say @xmath58 if at divergence point , @xmath56 moves left side of @xmath55 . see figure [ fig : braidorder_meaning ] . in @xcite it is proved that this relation @xmath59 coincide with dehornoy ordering @xmath60 . since @xmath14 corresponds to rotation of the disc @xmath51 , dehornoy floor of @xmath7 is the number of how many times the initial segment of graph @xmath56 winding around @xmath53 under neglecting orientation . and examples , width=377 ] for later use , we prove slightly more precious estimation of dehornoy floor than proposition [ prop : dehornoyfloor ] ( 1 ) . [ prop : newestimation ] if a braid @xmath7 is conjugate to a braid @xmath61 which is represented by braid word which contains @xmath40 occurrence of @xmath41 and @xmath42 occurrence of @xmath43 , then @xmath5_{d } < max\{s , k\}$ ] . we show @xmath62 . the proof of @xmath63 is identical . by assumption , we can write @xmath64 where @xmath65 is a braid word which contains no @xmath41 and @xmath25 . we remark that we can assume @xmath66 is empty because we can amalgamate @xmath66 into conjugating element @xmath67 . let @xmath68 . from property s , we can find a braid @xmath69 by inserting @xmath70 to braid @xmath7 , for sufficiently large @xmath71 . we show @xmath72 , that is equivalent to an inequality @xmath73 . first we consider the case @xmath74 . we use the geometric interpretation of dehornoy ordering . let @xmath75 be a sub - disc of @xmath51 which contains @xmath76-th puncture of @xmath51 . since @xmath77 is p - times twists around @xmath78 and @xmath14 is a rotation along @xmath53 , we conclude that the image of @xmath79 moves more left than @xmath80 for all @xmath81 , thus @xmath82 . induction about @xmath42 gives desired result . this proposition gives more accurate estimation of dehornoy floor than using lemma [ prop : dehornoyfloor ] ( 1 ) . proposition [ prop : newestimation ] makes results in @xcite a bit stronger form because in @xcite they use estimation via lemma [ prop : dehornoyfloor ] ( 1 ) . especially , we conclude that if @xmath4 admit destabilization or exchange move , then @xmath83=0 $ ] . in this section , we summarize basic machinery of birman - menasco s braid foliation theory in case of an embedded essential surface . for details of these techniques , see @xcite or @xcite . fix an unknot @xmath84 , called _ axis _ and choose a meridian fibration @xmath85\}$ ] of the solid torus @xmath86 . a link @xmath87 in @xmath86 is called _ closed braid _ with axis @xmath88 if @xmath87 intersects every fiber @xmath89 transversely and each fiber is oriented so that all intersections of @xmath87 are positive . it is easy to see closed braid @xmath87 intersects every fiber @xmath89 in the same number of points , and we call this number _ braid index _ of @xmath87 . note that closed braids @xmath4 obtained by a braid @xmath90 as described in section 2 is indeed closed braid with @xmath91axis and braid index is @xmath11 . conversely , if we cut solid torus @xmath92 along the fiber @xmath93 , we obtain a braid @xmath7 . it is known that isotopy in @xmath92 does not change conjugacy class of @xmath7 and if the isotopy fixes @xmath94 , obtained braid @xmath7 does not change as a element of braid group though geometrical configuration of strands are varied . a closed embedded surface @xmath0 in @xmath86 is called _ essential _ if @xmath0 is incompressible and non - boundary parallel . let @xmath0 be a essential closed surface in @xmath95 . then the intersections of fiber @xmath96 with @xmath0 induce a singular foliation of @xmath0 . leaves of this foliation are connected components of intersection with fiber . braid foliation techniques are , in short , changing this foliation simpler as possible and obtain standard and simple position of the surface and braids . by the argument in @xcite , @xmath0 can be isotoped to `` general '' position with respect to the fibration satisfying : ( 1 ) : : axis @xmath88 pierces @xmath0 transversely in finitely many points . ( 2 ) : : for each point @xmath97 , there exists neighborhood @xmath98 of @xmath99 such that @xmath100 is radially foliated disc . ( 3 ) : : all but finitely many fibers @xmath89 intersect @xmath0 transversely , and each of the exceptional fiber is tangent to @xmath0 exactly one point . moreover , each point of tangency is saddle tangency and lies in the interior of @xmath101 . notice that the condition above is a bit stronger than usual general position arguments , since usual general position arguments only tell us that each tangency is local minimum , maximum , or saddle . in fact , this strong sense of general position is achieved from usual general position by deleting local minimum or maximum tangency . we say fiber @xmath89 _ regular _ if @xmath89 transverse @xmath0 and _ singular _ if @xmath89 tangent to @xmath0 . in the foliation of @xmath0 , every non - singular leaf of induced foliation is either simple closed curve or arc which has both endpoints on axis @xmath88 . we call former _ c - circle _ and the latter _ b - arc_. each b - arc in a fiber @xmath89 separates @xmath89 to two components , and we say b - arc is _ essential _ if both of these components are pierced at least once by @xmath87 . for c - circle in a fiber @xmath89 , c - circle bound a disc @xmath102 in @xmath89 and we say c - circle is _ essential _ if @xmath102 is pierced at least once by @xmath87 . as shown in @xcite , we can assume further more condition about foliation of @xmath0 . ( 4 ) : : every fiber is essential . ( 5 ) : : every c - circle is homotopically non - trivial . this condition is achieved by deleting all inessential leaves or homotopically trivial c - circle . from now on , we always assume that @xmath0 satisfies condition ( 1)@xmath103(5 ) . we call an intersection point of @xmath88 with @xmath0 _ vertex_. for each vertex @xmath81 , the valance of vertex @xmath81 is , by definition , the number of singular leaves which pass @xmath81 . we say singular point is @xmath104 if the singular point derived from two b - arcs . @xmath105 and @xmath106 are defined by the same way . each type of singularity has a neighborhood in @xmath0 as shown in figure [ fig : regions ] , and we say such a neighborhood _ regions_. a foliation of @xmath0 is classified into following three types according to its containing leaves ; we say @xmath0 is _ tiled _ if @xmath0 is foliated entirely b - arc and _ circular _ if @xmath0 is foliated entirely c - circle . the remaining case , if @xmath0 is foliated by both b - arcs and c - circles , we say @xmath0 is _ mixed_. in tiled surface , there are only bb - singularity and circular surface has only cc - singularity . mixed foliation might have all types of singularity and always contains bc - singularity . now we are ready to prove theorem [ thm : main1 ] . from now on , we always assume @xmath0 satisfies all conditions ( 1)-(5 ) described in section 3 . first we consider the case @xmath107 . in this case , @xmath0 is never circular foliated . the main result of @xcite shows in this case @xmath4 admit exchange move if @xmath0 is tiled and @xmath7 has word representatives which contains no @xmath108 for some @xmath42 , so @xmath5_{d}=0 $ ] . thus , in this case theorem holds . thus , we always assume @xmath109 . first we prove statement ( 1 ) , tiled case . assume @xmath0 is tiled . since @xmath0 is tiled , every singular point is bb - singularity , and @xmath0 can be decomposed into bb - tiles . a decomposition into bb - tiles determines cellular - decomposition of @xmath0 , and we call this _ tiling_. the notion of vertex and valance we defined in section 3 coincides to usual definition of vertex and valance in this cellular decomposition . first we estimate a valance of each vertex , using euler characteristic argument which is standard in braid foliation theory . [ lem : eulerform - tiled ] let @xmath7 be a braid and @xmath0 be a tiled essential genus @xmath3 closed surface in the complement of closed braid @xmath4 , and @xmath110 be a number of vertex whose valance is @xmath111 . then , @xmath112 holds . let @xmath113 be a number of 1-cell and 2-cell of tiling . since @xmath0 has an euler characteristic @xmath114 , @xmath115 . on the other hand , each 1-cell is the boundary of exactly two 2-cells and each 1-cell has two vertices as their boundary , so @xmath116 and @xmath117 . by combining these equation , we establish the claim . since each term of equation in lemma [ lem : eulerform - tiled ] is positive , lemma [ lem : eulerform - tiled ] implies @xmath110 are bounded except @xmath118 . next we see that dehornor floor can be estimated from above by a valance of vertex . [ lem : estimate - tiled ] let @xmath0 be a tiled essential genus @xmath3 closed surface in the complement of closed braid @xmath119 . if @xmath0 contains a valance @xmath99 vertex , then @xmath5_{d } < \frac{v}{2}+1 $ ] . let @xmath81 be a valance @xmath99 vertex and @xmath120 be a sequence of singular fibers such that each @xmath121 contains singular leaves which pass the vertex @xmath81 . let @xmath122 be a b - arc in the fiber @xmath89 which passes @xmath99 . by taking isotopy near each singular fiber @xmath121 , we can always assume that for sufficiently small @xmath123 , braiding of strands does not occur in the interval @xmath124 $ ] and each fiber @xmath125 is pierced by @xmath87 exactly the same points @xmath126 . notice that this modification can be done without changing @xmath94 so this modification does not change braid @xmath7 as a element of braid groups . after this modification , we can always consider sub - braiding in each interval @xmath127 $ ] , @xmath128 $ ] . since @xmath129 is chosen to sufficiently small , we can assume there exist no other singular points in each interval @xmath124 $ ] . now we see a sequence of @xmath122 as a move of b - arc in @xmath130 . then we can also assume by taking isotopy of surface and braids , in the interval @xmath128 $ ] , @xmath122 do not move when we see @xmath122 as a move of the arcs . first we investigate a sub - braiding @xmath131 in @xmath132 $ ] . by taking appropriate conjugacy of @xmath131 , we can assume @xmath133 consists of several number of vertical arcs as depicted in left of figure [ fig : model ] and bb - singularity in @xmath134 is modeled as shown in the right of figure [ fig : model ] . we denote this braid by @xmath135 then , by seeing braids as a move of puncture point of disc , we can write this modified braiding @xmath136 in @xmath132 $ ] as shown in figure [ fig : braiding ] , which contains only one @xmath137 . the original braiding @xmath131 in @xmath132 $ ] is conjugate to the braid @xmath136 so proposition [ prop : newestimation ] shows @xmath138_{d } = 0 $ ] . $ ] : bb - singularity case , width=302 ] iterating this argument , we conclude that sub - braid @xmath139 in the interval @xmath140 $ ] for @xmath141 has zero dehornoy floor . since original braid @xmath7 is conjugate to the braid @xmath142 , by proposition [ prop : dehornoyfloor ] , @xmath143_{d } < \frac{v}{2 } $ ] . combining above two lemmas , now it is easy to estimate dehornor floor via information of tiling . assume @xmath0 is tiled . let @xmath110 be the number of valance @xmath111 vertices . if some of @xmath110 is non - zero for @xmath144 , then lemma [ lem : estimate - tiled ] means @xmath5_{d } < 2 < g+1 $ ] , so conclusion holds . therefore we can assume @xmath145 for @xmath144 , and euler characteristic formula in lemma [ lem : eulerform - tiled ] become @xmath146 since @xmath3 is not zero , there exist at least two b - arcs in each regular fiber , so at least four vertices in the tiling of @xmath0 . therefore the minimal valance of vertices is smaller than @xmath147 , that is , @xmath148 . then lemma [ lem : estimate - tiled ] establishes @xmath5_{d } < g+2 $ ] . next we proceed to @xmath0 is mixed foliated case . the strategy is almost the same as tiled case , but we need to do some additional arguments to evaluate valance of vertices because the decomposition by regions do not determine cellular decomposition . first we see the same conclusion of lemma [ lem : estimate - tiled ] holds . [ lem : estimate - mixed ] let @xmath0 be a mixed foliated essential genus @xmath149 closed surface in the complement of closed braid @xmath4 . if @xmath7 contains a valance @xmath99 vertex , then @xmath5_{d } < \frac{v}{2}+1 $ ] . let @xmath81 be a valance @xmath99 vertex . we can use the same arguments appeared in proof of lemma [ lem : estimate - tiled ] . the main difference is a presence of c - circles and bc - singularity around @xmath81 . first we modify closed braid as in the proof of lemma [ lem : estimate - tiled ] so that we can consider sub - braiding in each interval @xmath124 $ ] , @xmath128 $ ] . assume that singular point in the fiber @xmath150 is bc - singularity . then by taking appropriate conjugation of sub - braiding @xmath132 $ ] , bc - singularity is now modeled as in the figure [ fig : model2 ] . then braiding in the interval @xmath132 $ ] can be written as shown in figure [ fig : braiding2 ] , which contains only one @xmath151 . thus , the same argument of the proof of lemma [ lem : estimate - tiled ] gives desired estimation . $ ] : bc - singularity case , width=302 ] then , we want to estimate the minimal valance of vertices . since there might be all types of singularity in the foliation of @xmath0 , in general decomposition into regions dose not define usual sense of cellular decomposition : bc - region is annulus , and cc - region is pair of pants . however , in special case of normal mixed foliation , defined below , one can constructs canonical cellular decomposition via decomposition into regions as shown in @xcite . first we describe this . we say mixed foliation is _ normal _ if there exists no cc - singularity . in normal mixed foliation case , we can construct canonical cellular decomposition as follows . first we see that in normal mixed foliation , bc - annuli always occur in pairs . this is because the c - circle boundary of bc - annuli must be attached to c - circle boundary of another bc - annuli , since there is no cc - regions and boundary of bb - tile are all b - arcs . now , take annulus @xmath152 obtained by attaching two bc - annuli along their c - circle boundaries . each component of @xmath153 has two vertices . by cutting @xmath152 along the arc which connects two vertices which lies different components of @xmath154 , as shown in figure [ fig : be - tile ] , we obtain two 2-cells , called _ be - tile_. these cutting arcs are called _ e - edge_. now decomposition into be - tile and bb - tile determines cellular decomposition of @xmath0 . note that there might exist be - tile which has no singular point or two singular points , but from construction of be - tile , be - tile with zero singularity and be - tile with two singularity must occur in pairs in each neighbor of vertex , so the notion of vertex and valance defined in section 3 coincide with the usual meaning of vertex and valance in this cellular decomposition . then we see the same euler characteristic formula holds : [ lem : eulerform - nmixed ] let @xmath0 be a normal mixed foliated surface with genus @xmath155 and @xmath156 be a number of vertex whose valance is @xmath99 in cellular decomposition described above . then , @xmath157 holds . a be - tile has exactly four 1-cells in its boundary and each 1-cell is boundary of two 2-cells , so the same argument in proof of lemma [ lem : eulerform - tiled ] gives desired result . so we can estimate minimal valance of vertices in case of normal mixed foliation by using lemma [ lem : eulerform - nmixed ] as in the tiled case . to estimate general mixed foliated case , we perform `` preimage surgery '' to delete cc - singularities . let @xmath0 be a mixed foliated surface and we denote @xmath158 by the preimage of @xmath0 ; that is , @xmath159 is a foliated surface forgetting information of embedding . let @xmath160 be an one of the c - circle boundary of cc - pants @xmath161 . @xmath160 is also a boundary of some other cc - pants or bc - annulus @xmath162 . we modify foliations of @xmath161 and @xmath162 as shown in figure [ fig : preimagesurgery ] by filling @xmath160 by discs and changes foliation in @xmath161 and @xmath162 , according to type of regions . since modified regions are trivially foliated , these regions are amalgamated to some other regions and they disappears . moreover , this operation can be done so that obtained surface is connected by choosing c - circle boundary @xmath160 appropriately and reduces genus of @xmath159 because all c - circle in the foliation is non - null homotopic in @xmath159 . thus , this modification deletes two regions @xmath161 and @xmath162 , hence two singular points . we call this operation _ preimage surgery _ because we can regard this modification as surgery of @xmath159 along c - circle @xmath160 . we remark that preimage surgery is not realizable as a surgery of embedded surface @xmath0 , because we can not choose take a disc bounded by @xmath160 without intersecting original surface @xmath0 in @xmath163 . since each preimage surgery reduces a genus of @xmath159 , after @xmath164 preimage surgery we obtain normal mixed foliated surface @xmath165 . as we remarked before , @xmath165 can not be realized as an embedded surface , but we can use the arguments of proof of lemma [ lem : eulerform - nmixed ] , because the proof uses only information from cellular decomposition of @xmath165 . therefore we can use lemma [ lem : eulerform - nmixed ] for @xmath165 to estimate valance of vertices of preimage - surgered surface @xmath165 . assume @xmath0 is mixed foliated and after performing preimage surgery @xmath42-times described above , we create abstract foliated surface @xmath165 with no cc - singularity , that is , normal mixed foliated . let @xmath30 be a minimal valance of vertices of @xmath165 . then , we trace preimage surgery procedure backwards . each surgery deletes at most one bc - singularity , so original surface @xmath0 has at most more than @xmath42 bc - singular points than that of @xmath165 . since singular leaf containing bc - singularity has two vertices as its endpoints and there exist at least two vertices in @xmath0 , we conclude minimal valance of vertices in @xmath0 is smaller than @xmath166 . thus , we only need to estimate @xmath30 to estimate minimal valance of vertices in original surface @xmath0 . let @xmath156 be the number of valance @xmath99 vertices of @xmath165 . if @xmath110 is non - zero for some @xmath167 , then form the argument above , we conclude the minimal valance of vertices of @xmath0 is smaller than @xmath168 , so @xmath5_{d } < \frac{3+g}{2 } \leq 2g+1 $ ] , so we can assume @xmath110 is zero for all @xmath167 . in this case , the euler characteristic formula for @xmath165 become @xmath169 mixed foliated means there exist at least one b - arc in the foliation , so there exist at least two vertices in @xmath165 . therefore the minimal value of valance of vertices in @xmath165 is less than @xmath170 . therefore , from our earlier argument , minimal valance of vertices in @xmath0 is smaller than @xmath171 . now , lemma [ lem : estimate - mixed ] establishes an inequality @xmath5_{d } < 2g-\frac{3k}{2 } < 2g+1 $ ] . from statement ( 1 ) and ( 2 ) , we conclude if @xmath5_{d } \geq 2g+1 $ ] , then @xmath0 is neither tiled nor mixed - foliated , hence circular - foliated . this completes proof of theorem [ thm : main1 ] in this section , we describe an application of theorem [ thm : main1 ] for genus one closed surface , namely , essential torus . a description of essential torus in closed braid complements is studied in @xcite,@xcite by using braid foliation machinery . since an essential torus deeply involves the geometry of link complements , we can extract some of geometrical information of closed braids from theorem [ thm : main1 ] . first of all , we recall the nielsen - thurston classification of braids . nielsen - thurston theorem states that an isotopy class of orientation preserving homeomorphism @xmath172 $ ] of compact oriented surface @xmath0 , that is , an element of mapping class group @xmath173 of @xmath0 , is classified into following three types : 1 . _ periodic _ : there exists a positive integer @xmath30 such that @xmath174 is isotopic to the identity . _ reducible _ : there exist an essential 1-submanifold @xmath175 of @xmath0 , which is invariant under @xmath176 . pseudo - anosov _ : @xmath176 is isotopic to a pseudo - anosov homeomorphism . we do not describe precise definition of pseudo - anosov homeomorphism , because we do not use it.(see @xcite for more details about nielsen - thurston theory . ) let @xmath51 be a @xmath11-punctured disc . since braid group @xmath1 is identified with the group of boundary - fixing isotopy class of homeomorphisms of @xmath51 which is identity on @xmath53 , there exist natural projection @xmath177 regarding an element of @xmath1 as an isotopy class of homeomorphism of @xmath51 whose restriction on @xmath53 are not necessarily identity . therefore we can define the nielsen - thurston classification of braids by taking pull - back of the nielsen - thurston classification of @xmath178 . concretely , the nielsen - thurston classification of @xmath1 can be described as following way : 1 . _ periodic _ : some powers of @xmath7 is equal to some powers of garside fundamental braid @xmath13 . pseudo - anosov _ : its representing homeomorphism is pseudo - anosov . reducible _ : its representing homeomorphism is reducible . we remark that the nielsen - thurston classification of mapping class groups has very beautiful connection to the geometry of its mapping torus . let @xmath0 be a closed surface , @xmath181 be a homeomorphism of @xmath0 and @xmath182 be the mapping torus of @xmath181 . the topology of mapping torus @xmath182 only depends on the isotopy class of @xmath181 . it is known that @xmath0 admit a complete hyperbolic structure if and only if @xmath181 is pseudo - anosov and @xmath182 is seifert - fibered if and only if @xmath181 is periodic . moreover , @xmath182 contains an essential torus ( hence decomposable with respect to geometric decomposition ) if and only if @xmath181 is reducible . thus , there exist one - to - one correspondence between the nielsen - thurston classification of mapping class group and the geometry of its mapping torus . hence , in this regard , theorem [ thm : main2 ] can be seen as some generalization of this well - known fact in the knot theory settings . first we investigate non - braided satellite knots . recall that a satellite knot @xmath9 is called braided - satellite if its complements has essential tori @xmath183 which is circular foliated under appropriate choice of closed braid representatives of @xmath9 . in general it is difficult to handle non - braided satellite knot in braid theory , but fortunately they do not appear under assumption of theorem [ thm : main2 ] . in @xcite , it is proved that if a closure of @xmath7 is a composite knot , then @xmath5_{d}<2 $ ] . proposition [ prop : nbdsatellite ] can be seen some kind of generalization of this result because composite knots are special types of non - braided satellite knots . torus knots are exchange - reducible . that is , for every closed braid representative @xmath4 of @xmath186-torus knot , there exists a sequence of closed braids @xmath187 such that @xmath188 is obtained by performing exchange move or destabilization or isotopy of @xmath189 , and @xmath190 . first assume @xmath2 and @xmath191 is a torus knot . if @xmath11 is not the braid index of knot @xmath9 , that is , if @xmath9 can be represented as a closure of other braids which have strictly smaller number of strands , then menasco s theorem says @xmath9 admit exchange move or destabilization , hence by the proposition [ prop : dehornoyfloor ] ( 5 ) , we conclude @xmath5_{d}\leq 1 $ ] , which contradicts our assumption @xmath5_{d } \geq 3 $ ] . therefore we can assume that @xmath11 is braid index of @xmath9 . in this case , also menasco s theorem says @xmath7 is conjugate to braid @xmath192 for some @xmath81 , hence @xmath7 is periodic . on the other hand , it is known that periodic braids are conjugate to braids @xmath193 or @xmath194 for some @xmath40 , both of which represent rotation of disc . since closure of @xmath194 never become a knot , it is obvious that closure of periodic braids represent torus knots . next assume @xmath184 is a satellite knot . let @xmath183 be a essential torus in the complement of @xmath9 . of course there might be some of choice for such torus , we can take arbitrary one . our assumption @xmath5_{d } \geq 3 $ ] and theorem [ thm : main1 ] shows @xmath183 is circular - foliated , so intersection of @xmath183 with some fiber @xmath89 gives reducing 1-submanifolds of @xmath7 as a homeomorphism of @xmath51 , hence @xmath7 is reducible . the converse is trivial . as a corollary , we get an infinite , almost disjoint family of hyperbolic knots for each pseudo - anosov element of mapping class group of punctured disc . although it is known that `` almost all '' prime knots are hyperbolic knot , in general it is not so easy to determine given not is hyperbolic or not . especially , most of known family of knots known to be hyperbolic are defined through knot diagrams such as alternativeness . we now construct many families of hyperbolic knot through braid - theoretical approach . [ cor : c1 ] suppose @xmath172 \in mcg(d_{n})$ ] be a pseudo - anosov element of mapping class group of @xmath11-punctured disc and @xmath195 be the natural projection . let @xmath196)=\ { \widehat{\beta } \;| \;\beta \in \pi^{-1 } ( [ f]),[\beta]_{d } \geq 3 \ } $ ] . then @xmath197)$ ] consists of infinite number of distinct hyperbolic knots . moreover , let @xmath198 \in mcg(d_{n})$ ] be another pseudo - anosov element which is not conjugate to @xmath172 $ ] in @xmath178 . then , the intersection of @xmath197)$ ] and @xmath199)$ ] are finite . since @xmath197)$ ] consists of pseudo - anosov braids with dehornoy floor larger than three , theorem [ thm : main2 ] shows their closures are all hyperbolic knots . proposition [ prop : dehornoyfloor ] @xmath200 says braids having sufficiently large dehornoy floor is a unique closed representative of its closure in braid index @xmath11 , therefore @xmath197)$ ] consists of infinite number of distinct hyperbolic knots and @xmath197)$ ] and @xmath199)$ ] have finite intersection unless @xmath198 $ ] is conjugate to @xmath172 $ ] . since every hyperbolic knots are represented as a closure of pseudo - anosov braids , one might expects our construction of hyperbolic knots produces _ all _ hyperbolic knots ; that is , for every hyperbolic knots , there exists pseudo - anosov mapping class @xmath172 $ ] such that @xmath201)$ ] . unfortunately , this too optimistic conjecture is not true . for example , we can prove figure - eight knot , the most simple hyperbolic knots , do not appear as an element of @xmath197)$ ] . more generally , we can show genus one hyperbolic knots can not be constructed from our method . this is a corollary of results of our subsequent paper @xcite . since @xmath81 is prime and @xmath4 is a knot , @xmath7 is never reducible braid . assume @xmath4 is satellite . then proposition [ prop : nbdsatellite ] , @xmath4 is braided - satellite , so theorem [ thm : main2 ] means @xmath7 is reducible , which contradicts above . thus , under the assumption of corollary , @xmath4 is either hyperbolic or torus knot and @xmath7 is either periodic or pseudo - anosov , so theorem [ thm : main2 ] establishes the result . although there is an algorithm to determine nielsen - thurston classification , it requires exponential times with respect to word length of braid . on the other hand , since to determine given braids are periodic or not is easy and it requires only quadratic times , corollary [ cor : c2 ] produces very many hyperbolic knots explicitly and rapidly . in particular , we conclude that for prime @xmath81 , closure of braids @xmath202 satisfying @xmath5_{d } \geq 2 $ ] are almost all hyperbolic knots . [ b ] j.birman , _ braids , links , and mapping class groups , _ annals of math . studies * 82 * , princeton univ . press ( 1975 ) . j.birman , e.finkelstein , _ studying surfaces via closed braids , _ journal of knot theory and its ramifications , * 7 * , no.3 ( 1998 ) , 267 - 334 . j.birman , w.menasco , _ studying links via closed braid iv : split links and composite links , _ invent . math . * 102*,(1990 ) , 115 - 139 . j.birman , w.menasco , _ special positions for essential tori in link complements , _ topology , * 33 * , no.3 ( 1994 ) , 525 - 556 . p.dehornoy , _ braid groups and left distributive operations , _ transactions american mathematical society , * 345 * ( 1994 ) , 115 - 150 . p.dehornoy , i.dynnkov , d.rolfsen and b.wiest , _ why are the braids orderable ? , _ panoramas et synthses * 14 * , soc . r.fenn , m.greene , d. rolfsen , c.rourke and b.wiest , _ ordering the braid groups , _ pacific journal of mathematics , * 191 * , no.1 ( 1999 ) , 49 - 73 . a. fathi , f. laudenbach , v. poenaru , _ travaux de thurston sur les surfaces , _ asterique , * 66 - 67 * , 1979 . t.ito , _ braid ordering and genus of knots , _ e - print , arxiv . math/0805.2042v1 j.los , _ knots , braid index and dynamical type , _ topology , * 33 * , no.2 ( 1994 ) , 257 - 270 . a.malyutin , n.netsvetaev , _ dehornoy s ordering on the braid group and braid moves , _ st.peterburg math . j. * 15 * , no.3 ( 2004 ) , 437 - 448 . a.malyutin , _ twist number of ( closed ) braids , _ st.peterburg math . j. * 16 * , no.5 ( 2005 ) , 791 - 813 . w.menasco , _ on iterated torus knots and transversal knots , _ geometry & topology , * 5 * , ( 2001 ) , 651 - 682 . w.menasco , _ an addendum to iterated torus knots , _ e - print , arxiv . math.gt/0610566 k.ng , _ essential tori in link complements , _ journal of knot theory and its ramifications , * 7 * , no.2 ( 1998 ) , 205 - 216 . h.short , b.wiest , _ ordering of mapping class groups after thurston , _ lenseignement mathmatique , * 46*,(2000 ) , 279 - 312 .

carrier dynamics in metals and semiconductors in the presence of external electromagnetic fields , the potentials of which usually vary on scales considerably large than the interatomic spacing , have been conveniently described by semiclassical transport theories . the semiclassical dynamics together with the boltzmann equation produce accurate descriptions of electrical and thermal conduction@xcite . in the larger picture , semiclassical approaches are indispensable in problems involving both position and momentum , since in quantum mechanics position and momentum can not be determined simultaneously . in recent years efforts have been made to extend the semiclassical theory to spin transport and generation@xcite . the attempts to resolve the challenges inherent in treating the transport of non - conserved quantities constitute a vibrant ongoing effort@xcite . a fundamental feature of semiclassical transport is its accounting for the finite extent of particles in real and reciprocal space . this feature is most naturally incorporated into the dynamics of wave packets@xcite where the notion of a wave - packet center in real space and @xmath0-space is retained . the carrier dynamics are described in terms of the displacement of these points under the action of external fields . the finite extent of the wave packet has important consequences for transport theory . for example , our recent research on spin transport has shown that , due to the fact that the spin and charge centers of a wave packet do not coincide , the expressions for the spin density , torque and current distributions are , in the language of wave packets , expressed as series of multipole terms@xcite . it has been demonstrated , in addition , that the wave packet formalism captures the physics connected with adiabatic motion and the berry phase , in particular the berry - curvature correction to the semiclassical equations of motion@xcite . the berry phase , which until berry s seminal article@xcite was not taken into account , appears naturally as part of the wave packet distribution function . the berry - curvature correction to the wave packet equations of motion is believed to play an important role in the anomalous hall effect@xcite and in spin transport@xcite , among other phenomena . semiclassical transport theory is not restricted to wave packet dynamics . wave packets , which may be constructed out of one band eigenstate@xcite or out of a superposition of eigenstates of several degenerate bands@xcite , represent pure states . in order to treat mixed states ( incoherent superpositions of eigenstates ) one must resort to a more general formalism . this is the principal motivation behind the current article . the most general description of a quantum mechanical system is based on the density operator . in this article we start from the density operator and formulate a theory of carrier dynamics in metals and semiconductors . we focus on non - interacting systems in which disorder is weak , strong spin - orbit interactions are present and a weak slowly - varying electric field is acting . these systems have come under intense scrutiny in recent years along with the take - off of spintronics@xcite . in such non - interacting systems the formalism may be simplified by defining a reduced one - particle density operator@xcite . since the bloch bands are clearly resolved the reduced density operator for this system may be expanded in a basis of bloch wave functions . the density matrix which emerges from this expansion may be defined as a wigner function . this function is used to study single particle dynamics and formulate definitions of macroscopic quantities . we focus on a number of fundamental aspects of adiabatic particle dynamics and demonstrate their relevance to transport phenomena . we pay particular attention to the band mixing induced by the electric field , which gives rise to a non - adiabatic correction to the wave functions . we also address several important questions concerning the relationship between the wigner function formalism and the wave - packet formalism@xcite . we discuss the way the carrier position is to be found and , where possible , compare the result with the expression for the real - space center of a wave packet . the requirement that the particle position be gauge invariant results in a gauge - dependent shift in the position at which the wigner function is evaluated . this shift was also found by littlejohn and flynn@xcite in the study of coupled - wave equations . this gauge - dependent shift is a consequence of the freedom of choosing the real - space center of a wave packet by changing the phase of the wave function@xcite . the gauge - dependent shift in the position of the wigner function must be taken into account when many - particle distributions , such as the particle number density , are expressed in the crystal - momentum representation . the carrier velocity may be derived directly from the particle position , recovering the berry - phase physics known from previous work@xcite . we also show that a spin velocity may be defined in such a way that spin transport may be described in an analogous fashion to charge transport . we discuss , in addition , important consequences of finite particle size . the wigner distribution is a quantum entity which takes into account the finite extent of the particles in real and reciprocal space . the distribution of , for example , spin for a single carrier , may not coincide with that of charge and the macroscopic spin distribution will be composed of a series of multipoles . the spin current and spin torque distributions are in turn composed of series of multipoles which we discuss and compare with those found in our previous work on spin transport@xcite . the outline of this article is as follows . we present the fundamentals of the single particle density matrix formalism in section ii . we determine the carrier position and velocity , emphasizing the gauge - dependent shift in the former and the berry - phase correction in the latter . we calculate the particle spin , torque and spin velocity and define a modified spin velocity which satisfies an equation analogous to the charge velocity . in section iii we demonstrate the modifications which must be made to extend the theory to many non - interacting particles . we define the charge and current densities and show the equation of continuity they satisfy . in the case of spin we define the spin , spin current and spin torque distributions and show the equation of continuity satisfied by them in the clean limit . in section iv we demonstrate the effect of local gauge transformations and the modifications which must be made to the dipoles in the charge- and spin - related distributions in order to make them gauge covariant . we conclude with a brief summary of our findings . we consider systems described by a hamiltonian having the following general form : @xmath1 the term @xmath2 is composed of the usual kinetic - energy contribution and the contribution due to the lattice - periodic potential . the term @xmath3 represents the spin - orbit interaction term involving the carrier spins and the lattice - periodic potential . we restrict our attention to the limit in which this spin - orbit interaction is sizably stronger than the disorder broadening and the thermal broadening . in this limit the system may also be described in terms of well - defined bands . the eigenstates of @xmath4 , which have the periodicity of the crystal , are given by : @xmath5 these wave functions are of the bloch form , that is @xmath6 , where the functions @xmath7 represent the lattice - periodic parts of the @xmath8 . the @xmath7 are spinors with the full periodicity of the lattice . since the hamiltonian contains strong spin - orbit interaction terms , which may depend on wave vector and position , it is not illuminating to decompose the eigenfunctions into an orbital and a spin part . we take the system under study to be described by a density operator @xmath9 . it is not our concern in this article to determine an expression for the density operator for a given system , since methods of finding the density operator have been studied extensively in the past@xcite , much work being done in the context of spin orientation of carriers in the non - degenerate limit . we assume the form of @xmath10 to be known and study the role it plays in the dynamics of charge and spin . the density operator may be expanded in the hilbert space spanned by a complete orthonormal set of bloch wave functions @xmath11 as @xmath12 in the approach we follow , all the time dependence of the density operator is contained in the wave functions , so that @xmath13 does not have time dependence . in thermal equilibrium it is diagonal and its elements , @xmath14 , are equal to the fermi - dirac function @xmath15 , where @xmath16 is the band energy . the expectation value of any operator @xmath17 is found from the formula @xmath18 = & \displaystyle { \rm tr } \ , \sum_{{\bf k } , { \bf k } ' } \rho({\bf k } , { \bf k } ' ) a({\bf k } ' , { \bf k } ) . \end{array}\ ] ] the notation @xmath19 stands for the matrix elements of the operator @xmath17 , namely @xmath20 . the operation denoted by tr is simply the matrix trace , which does not include the wave - vector summation . if the operator @xmath17 is replaced by the identity matrix we obtain @xmath21 . for a single particle , the normalization condition on the density operator is @xmath22 . in order to make transparent the analogy with the language of wave packets , center of mass and relative coordinates may be defined in @xmath0-space such that * q * = @xmath23 and @xmath24 . the density operator as a function of these coordinates can be re - expressed as @xmath25 where @xmath26 and @xmath27 is the periodic part of the bloch wave at @xmath28 . the wigner function corresponding to the one particle density matrix is found through the transformation @xmath29 for the sake of concreteness the integrals are represented as three dimensional . nevertheless , the theory applies to systems of any dimensionality . the wigner function plays the role of a distribution function in the variables * q * and * r*. technically , however , it is not a distribution since it may take negative values@xcite . the inverse transformation is @xmath30 finally , replacing the vector summations by integrations we are able to represent the density operator in the following form @xmath31 where @xmath32 . for the remainder of this section , the wigner function @xmath33 will frequently be abbreviated to @xmath34 . the variables * q * and * r * are simply labels for the carriers , not physical observables . in particular , the dummy variable * r * in the fourier expansion of the density matrix is simply the fourier dual of * q * and must not be confused with the position operator @xmath35 appearing in the density operator . it does not correspond to an actual position . if we consider a particle in one band , labelled @xmath36 , @xmath34 reduces to a scalar , which will be denoted by @xmath37 . the position of the particle can be found as the expectation value of the position operator , which yields @xmath38 here @xmath39 . the integrand is not gauge invariant but the integral can be shown to be by changing the variable of integration * r * to @xmath40 . the connection @xmath41 has no position dependence , therefore the jacobian of the transformation is unity . the expectation value of the position operator is @xmath42 where @xmath43 is defined in the same way as @xmath44 but with * r * replaced by @xmath45 . the gauge invariance of ( [ rnew ] ) will emerge below . we conclude that @xmath46 is to be interpreted as a label for the charge carrier , while the effective particle position is @xmath40 . neither the label @xmath46 nor the gauge field @xmath47 are by themselves gauge invariant , but together they form a gauge - invariant quantity which represents the true position of the carrier . this result was also found earlier , in somewhat different circumstances , in the work of littlejohn and flynn@xcite . in the one - band limit , a clear connection can also be made with the dynamics of wave packets . the gauge - dependent shift in * r * reflects the freedom of changing the phase of the wave functions @xmath11 , @xmath48 in ( [ rhodef ] ) . it is the same freedom one has in defining the center of mass of a wave packet , demonstrated by sundaram and niu@xcite , by changing the overall phase of the wave function@xcite . for multiple bands , the expression for the particle position is ( the einstein summation convention will be used henceforth ) @xmath49 one can rewrite this expression as @xmath50 and make the substitution @xmath51 as in the single band case . although @xmath47 is a matrix the jacobian of this transformation is unity . finally , the expectation value of the position operator can be expressed formally as @xmath52 unlike the single band case , the expression @xmath53 is a formal abbreviation for the taylor expansion about @xmath45 , that is @xmath54 @xmath55 higher order . expression ( [ rnew ] ) identifies the position of a particle . to determine if a particle is localized at its position one must calculate the variance of the position operator , the expectation value @xmath56 , and ensure that is does not diverge . it is shown in the appendix that the variance does not diverge and that this result applies to any number of bands . we consider a system acted on by a weak external electric field . the effect of this electric field is incorporated fully into the gauge invariant crystal momentum through the addition of the electromagnetic vector potential @xmath57 . because the bloch functions retain translational symmetry the electromagnetic vector potential does not enter the travelling - wave part of the wave functions , which have the form @xmath58 . the lattice - periodic functions @xmath59 depend implicitly on time only through the crystal wave vector @xmath60 . however , the presence of the external electric field results in a non - adiabatic mixing of the bands with the result that the perturbed lattice - periodic functions , @xmath61 , have the following form to first order in the electric field@xcite @xmath62 the phase @xmath63 includes the dynamical phase and the berry phase . the differential @xmath64 is equivalent to @xmath65 since @xmath66 . the result expressed by ( [ nonadb ] ) is general . moreover , although its derivation relies on the assumption that the bands are non - degenerate it can be shown that , when calculating _ intrinsic _ contributions to transport ( for the definition of intrinsic please see the appendix and our recent work@xcite ) , the result also holds for degenerate bands with the difference that the sum must exclude _ all _ bands which are degenerate in energy with band @xmath67 . the proof of this statement is given in the appendix . henceforth , @xmath68 and @xmath69 will be abbreviated to @xmath70 and @xmath71 respectively . as stated above , given that the @xmath72 are functions of @xmath73 only , one may replace @xmath64 in ( [ nonadb ] ) by @xmath65 , where @xmath74@xmath75 . equation ( [ nonadb ] ) can then be written as @xmath76 where the connection @xmath77 . the @xmath78 form a complete set . they are , however , _ not _ eigenstates of the _ time - dependent _ hamiltonian @xmath79 . in the evaluations of matrix elements in this paper the only property of the basis functions that is used is their bloch periodicity . therefore the results which are expressed in terms of the lattice periodic bloch functions hold as well for the @xmath80 as for the @xmath81 . in the absence of disorder , since all the time dependence is contained in the wave functions , the density matrix @xmath13 in ( [ rhodef ] ) can only depend on the wave vector * k * , not on the crystal wave vector @xmath82 . as a result the wigner function @xmath83 only depends on the wave vector * q * , not on @xmath73 . it is customary to consider only a subset of the hilbert space which contains the bands that are relevant to transport , which in semiconductors usually refers to the topmost filled valence bands and/or the lowest filled conduction bands . since the gauge of the @xmath78 is not fixed we impose the following gauge - fixing condition _ in the subspace _ under consideration @xmath84 & \displaystyle 0 , \,\,\ , { \rm for } \,\,\ , n \,\,\ , { \rm out \,\,\ , of \,\,\ , subspace } , \end{array}\ ] ] where @xmath85 . this condition fixes the phase(s ) @xmath86 in ( [ nonadb ] ) . we shall henceforth work only with the basis set @xmath87 . the particle velocity can be derived directly from ( [ rmb ] ) by evaluating the time derivative @xmath88 using equation ( [ gf ] ) , the differential @xmath89 becomes , after a little straightforward algebra , @xmath90 the abbreviation @xmath91 stands for the matrix elements of the time - dependent hamiltonian @xmath92 in the presence of an electric field , @xmath93 . in one band it is easily shown that ( [ drdt ] ) becomes @xmath94 the berry curvature for band @xmath36 , @xmath95 , is given by @xmath96 and @xmath97 represents the curl operator in reciprocal space . the quantity @xmath98 is defined as @xmath99 . equation ( [ vel ] ) is the analog of the semiclassical equation of motion found by sundaram and niu@xcite . writing @xmath100@xmath101 we obtain for electrons in a single band : @xmath102 for multiple bands , an elegant result is obtained by introducing the covariant derivatives@xcite @xmath103 and @xmath104 , where @xmath105 . making use of these derivatives , equation ( [ drdt ] ) can be written in the manifestly gauge covariant form @xmath106_{n'n } - i[\frac{d}{d\tilde{q}_\alpha } , \frac{d}{dt}]_{n'n}. \end{split}\ ] ] it is important to point out that , from the gauge - fixing condition ( [ gf ] ) which defines @xmath107 , it is evident that @xmath107 is gauge covariant , implying that the time derivative @xmath64 is itself gauge - covariant . this is a peculiarity of the gauge - fixing condition we have chosen . the fact that carriers have a finite extent in real and wave - vector space has profound implications for particle dynamics and transport . these implications were pointed out in a previous publication @xcite using the semiclassical language of wave - packets and will be elaborated in this section from the density matrix point of view . for the sake of clarity we will specialize in spin , although the discussion in this section applies to any quantity . when considering the transport of spin in a system composed of many carriers , one must associate a spin distribution with each individual carrier . a center of spin may be defined for each particle by @xmath108 . this definition may be troublesome if the expectation value of the spin operator were zero , but its only purpose is to illustrate a physical principle . evidently , if one replaced the spin operator with the charge or mass one would obtain the particle position as given by ( [ r ] ) , which we will refer to as @xmath109 . it is obvious from the definition of @xmath110 that in the general case there is no reason for the center of the spin distribution of an individual particle to be the same as the actual particle position . this center will be different for each component of the spin operator . furthermore , since spin is not conserved in the presence of e.g. spin - orbit interactions , @xmath110 may also be a function of time . this suggests that the spin distribution of one carrier will in general have a different shape than its charge distribution , and that the time development of the two distributions may be quite different . these facts are illustrated in figs . 1 and 2 . to evaluate the average carrier spin consider the expectation value of the operator @xmath111 , which stands for any one component of the spin operator . the result is @xmath112 where @xmath113 are the matrix elements of the spin operator . in the presence of spin - orbit interactions a spin torque is associated with the spin of every carrier , which accounts for the non - conservation of spin , as shown in fig . 2 . the average carrier torque is found by evaluating the expectation value of @xmath114 $ ] . the result is @xmath115 where @xmath116 . as a result of the distinction between @xmath110 and @xmath117 , in calculations of spin - related quantities which use the center of charge as the reference , multipole terms must be taken into account in addition to the average quantities calculated above . for example , in the distribution of spin a spin dipole will be present , as well as higher order multipole terms , which are assumed small . the spin dipole is found as the expectation value @xmath118 , evaluated in the appendix , and yields the gauge field @xmath119 in contrast to @xmath110 , @xmath120 is well defined . similarly , in the distribution of torque of an individual particle a torque dipole will be present . the torque gauge field @xmath121 is given , as shown in the appendix , by @xmath122 @xmath120 and @xmath123 are gauge dependent . it will prove useful to define gauge - covariant dipoles as @xmath124 and @xmath125 respectively . the remainder of the discussion of these gauge - covariant spin and torque dipoles is deferred to section iv . we define the spin velocity as the expectation value @xmath126 , where products of non - commuting operators are assumed to be symmetrized . the result is @xmath127 the spin velocity @xmath128 is given by @xmath129 = & \displaystyle { \frac{d \bar{\bf m}^s}{d t } } - \bar{\bf m}^\tau . \end{array}\ ] ] in order to write the velocity in a gauge covariant form , we replace in the above @xmath130 and @xmath131 . this yields immediately @xmath132 equation ( [ vsps ] ) is the gauge - covariant form of equation ( 10 ) in ref . [ 2 ] , the integrand of which represents the spin velocity . the first term in ( [ vsps])is a convective contribution and represents a moving electron transporting its average spin along with it . the second term is the time derivative of the spin dipole while the last term is the torque dipole which takes into account the non - conservation of spin . we may incorporate @xmath133 into a modified spin velocity@xcite , which we shall call @xmath134 , by @xmath135 . the modified spin velocity is given simply by : @xmath136 the importance of this definition of the velocity will be seen in the following section , when the macroscopic spin current is introduced . we will consider the macroscopic distributions of charge and spin starting from the formalism we have developed . in particular , our discussion of the corrections to the spin density , current and torque distributions is motivated by the observation that the existing literature has omitted various contributions to spin transport . several works have used semiclassical concepts but did not arrive at answers containing all the terms we have derived . the particle number density is defined by the formula : @xmath137.\ ] ] the trace operation tr involves integration over * r*. the procedure we follow has affinities with the coarse graining of electrodynamics in material media @xcite . the size of the carriers is taken to be smaller than the length scale of the wigner function . we regard the @xmath138-function as a sampling function@xcite with a width somewhere between the microscopic scale of the carriers and the macroscopic scale of the wigner function . the rationale for regarding the @xmath138-function as a sampling function comes from the realization that often the physics of the problem does not require absolute resolution of position , provided that the resolution is finer than the length scale of the external field . since the variance of @xmath35 is finite , the @xmath138-function may be expanded in a taylor series about the dummy variable * r * in the form @xmath139 $ ] . the expansion is truncated at the first order for simplicity but we will show in the next section that all the terms may be recovered in a concise and elegant fashion . the number density can be expressed as : @xmath140 in the above @xmath34 stands for @xmath141 , an abbreviation which will be frequently used in the remainder of the article . note that the gauge field @xmath142 in ( [ ndip ] ) plays the role of a dipole . the charge density is defined by the formula @xmath143.\ ] ] the charge @xmath144 is not to be confused with the wave vector * q*. the charge current density @xmath145 is defined by the equation @xmath146.\ ] ] the charge equation of continuity in the absence of external sources and disorder , @xmath147 is readily verified from the first - principles definitions of the charge and current densities . when the @xmath138-function is expanded the current density takes the form @xmath148 the velocity matrix elements @xmath149 as shown in the appendix , where @xmath150 . since in the equation of continuity it is the gradient of the current that appears , we have not included in ( [ jq ] ) corrections to the current which arise from the fact that the velocity distribution of a single carrier is different from its charge distribution . these corrections are in principle present but have been omitted for simplicity . in homogeneous systems the gradient term in ( [ jq ] ) drops out of the equation of continuity and , based on equations ( [ drdt ] ) and ( [ jq ] ) , the charge current can be expressed as the time rate of change of the gauge field @xmath142 @xmath151 the spin density is defined as @xmath152,\ ] ] while the torque density takes the form @xmath153\ ] ] and the spin current density is defined as @xmath154.\ ] ] in the absence of disorder the spin equation of continuity is @xmath155 which is verified from the first - principles definitions introduced above . the @xmath156-functions are expanded in the same way as for the particle number density , whereupon the spin density can be expressed as @xmath157 and the torque density is @xmath158 ignoring the gradient term , the spin current is @xmath159 in analogy with the modified spin velocity of the previous section , a modified spin current @xmath160 may be defined by : @xmath161 = & \displaystyle { \rm tr } \int \frac{d^3q}{(2\pi)^3 } \,\rho\,\left(\frac{1}{2}\{{\frac{d \bar{\cal r}}{d t } } , \bar s\ } + { \frac{d \bar{\bf p}^s}{d t}}\right ) . \end{array}\ ] ] the equation of continuity satisfied by this current in the clean limit is : @xmath162 in many models , such as the spherical four - band luttinger model@xcite , the rhs vanishes when all bands in the subspace are taken into account . in that case the spin current @xmath160 is conserved . much effort has been devoted to finding a conserved spin current . in our previous work @xcite we have argued , based on semiclassical ideas , that the closest one can come to a conserved spin current is by including the torque dipole in the definition of the current . in this article we have derived this current from a density matrix point of view and show that it supports our earlier conclusions . to date , the definition @xmath160 is the closest one has come to a conserved spin current . moreover , as shown also by zarea and ulloa @xcite , the behavior of the modified spin current may be rather different from that of @xmath163 . we have argued in section ii that in discussing transport of an observable one must take into account the fact that the distribution of that observable for each individual carrier in general involves a dipole correction , and in principle also higher - order corrections . if the quantity being transported is not conserved then the determination of its rate of change also involves a dipole correction . the existence of what appears to be a sound physical argument for the inclusion of these corrections suggests that objects such as the spin dipole and the torque dipole are not simply mathematical artifacts meant to facilitate calculations . it is appropriate to investigate whether they are in fact fundamental quantities with a true physical meaning . if that were the case , they ought to be expressible in a gauge - covariant way and to give rise to observable effects . in addition to these considerations , our aim of providing a physically transparent formalism requires a test of the gauge covariance of the macroscopic quantities and equations of motion we have derived in this article . the physics contained in them must be independent of the choice of basis functions . furthermore , the physical discussion can become cluttered with futile and meaningless objects if it is formulated in terms of gauge - dependent objects . in contrast , the gauge - covariant expressions we derive below are simple , elegant and transparent . a general local gauge transformation is represented by @xmath164 , with @xmath165 and @xmath67 and @xmath36 are indices of bands in the subspace under consideration . under this operation @xmath166 in which @xmath167 stands for @xmath168 . the berry curvature for one band @xmath95 is discussed extensively in the paper of sundaram and niu@xcite . one important additional detail which emerges from the transformation given by ( [ rtrans ] ) and the definition @xmath169 is the fact that , if the curvature is non - zero , one can not make a gauge transformation to eliminate @xmath170 . it can be easily shown that , whereas @xmath170 is gauge dependent , the berry curvature is gauge covariant . therefore if @xmath95 is non - zero in one gauge it is non - zero in all gauges and there is no gauge in which @xmath170 can be zero . as a result , in systems in which the curvature is non - zero the gauge - dependent position shift is necessarily present in the wigner function . it can be regarded as a ` penalty ' for working with the position operator in a basis of definite wave vector . the wigner function itself changes under the gauge transformation defined above . using ( [ wigout ] ) , we find that , to first order in @xmath171 , the wigner function changes as @xmath172 where @xmath173 ( the opposite of the gauge connection @xmath142 ) . all the macroscopic densities defined in the previous sections are covariant under a local gauge transformation . this will be demonstrated for the spin density . under the above gauge transformation , the gauge field @xmath120 transforms as @xmath174 the matrix elements of the transformed gauge field @xmath175 are given by @xmath176 and the transformed spin operator @xmath177 . to first order the extra terms acquired by the spin density under a gauge transformation are @xmath178 so that the spin density remains gauge covariant . the cancellation remains true for all orders in the expansion . similarly , the charge and current densities do not acquire additional terms under the local gauge transformation introduced above . the extra terms appearing as a result of the transformation cancel when the trace is taken . when the change in the wigner function under a gauge transformation is taken into account the overall expressions are gauge covariant . equation ( [ ndip ] ) for the particle number density can be formally written the following way @xmath179 re - expressing the integrand in this manner is tantamount to making , in the density matrix , the replacement @xmath180 . to ensure the gauge covariance of the number density the fourier dual * r * is replaced by the true position @xmath181 , as illustrated in fig . this is the same position as found in section ii . if the subspace contains one band expression ( [ nmod ] ) is an exact result , not simply a formal way of writing the number density . examining more closely the spin density as given in ( [ s ] ) it is evident that , whereas the spin density itself is gauge covariant , its individual constituents are not . if one were to consider the integrand of the dipole term in ( [ s ] ) without the wigner function , this quantity would not by itself be gauge covariant . we have already constructed a gauge - covariant spin dipole @xmath182 . in terms of the gauge - covariant spin dipole , the spin density can be re - expressed as @xmath183 it can be formally written the following way @xmath184.\ ] ] the gauge - covariant spin dipole @xmath185 , as pointed out above , is the result of a carrier s center of spin being different from its center of charge . in the same way the gauge - covariant torque dipole has been defined by @xmath186 . carrying out an identical manipulation to that for the spin density , use of the gauge - covariant torque dipole allows us to rewrite the torque density formally as @xmath187.\ ] ] a similar formal expression exists for the spin current . evidently , since these expressions are simply a rewriting of the spin , torque and current densities , the equation of continuity satisfied by them is unaltered . we remark that the gauge - covariant quantities introduced throughout this paper are well defined . in fact , even though the density matrix was expressed in a basis of extended bloch states , by means of various manipulations we have been able to obtain well - defined formulas for all the objects relevant in transport . finally , as described by dudarev _ et al . _ @xcite in the context of atomic physics , optical lattices can be constructed which mimic the spin - orbit interaction . in such a lattice , a wave packet can be constructed using cold atoms and the evolution of its centers of mass and spin can be followed . in the same spirit , kato _ et al . _ @xcite have shown that it is possible to follow the motion of a spin packet in solid state systems , which demonstrates the feasibility of an analogous experiment in ingaas . this would provide a way to measure a spin dipole directly . we have formulated a theory suitable for describing the dynamics of particles in both single and multiple bands . the one - band results known from the wave - packet formalism , including the terms connected to the berry phase , emerge from our theory . the formalism can be applied to any clean system regardless of the dimensionality of the hilbert space under consideration . as a result of the gauge degree of freedom of the basis functions , the position vector which is used as a label of the carrier must be modified by a gauge - dependent shift in order to obtain the true particle position . we have shown the way to define macroscopic density distributions for conserved and non - conserved operators . in the case of spin we have highlighted the correspondence between the multipole terms which appear in the density - matrix formalism and those found earlier in the wave - packet formalism and we have shown that experiments can be performed to identify the effect of a spin dipole . dc was supported by the nsf under grant number dmr-0404252 . qn was supported by the doe under grant number de - fg03 - 02er45958 . we will present in this appendix some of the proofs and evaluations which require lengthier calculations and would interrupt the flow if incorporated into the main text . for all derivations except the first , the @xmath78 and @xmath72 are interchangeable . the perturbed eigenfunctions are given by eq.([nonadb ] ) : @xmath188 for a set of degenerate bands the equilibrium part of the density matrix is proportional to the identity matrix @xmath189 and will have the form @xmath190 , where @xmath191 is usually the fermi - dirac distribution . transport theory often distinguishes between intrinsic effects , which are due to the equilibrium part of the density matrix , and extrinsic effects , which are due to the non - equilibrium correction to it . if we wish to evaluate the expectation value of an operator @xmath17 using the equilibrium part of the density matrix for a set of degenerate bands , which for simplicity here we will take as being two dimensional , the following quantity must be evaluated : @xmath192 the perturbed wave functions are given by : @xmath193 & \displaystyle { | \bar u_{2 } \rangle } = { | u_{2 } \rangle } - \frac{{\langle u_{1 } | i\hbar{\frac{d } { d t } } | u_{2 } \rangle}}{\varepsilon_2 - \varepsilon_1}{| u_{1 } \rangle } - \sum_{n\ne 1,2}\frac{{\langle u_{n } | i\hbar{\frac{d } { d t } } | u_{m } \rangle}}{\varepsilon_m - \varepsilon_n}{| u_{n } \rangle}. \end{array}\ ] ] the expectation values are : @xmath194 & \displaystyle { \langle \bar u_2 | \hat a | \bar u_2 \rangle } = a_{22 } - \frac{{\langle u_{1 } | i\hbar{\frac{d u_{2}}{d t } } \rangle}}{\varepsilon_2 - \varepsilon_1}a_{21 } - \frac{{\langle u_{2 } | i\hbar{\frac{d u_{1}}{d t } } \rangle}}{\varepsilon_2 - \varepsilon_1}a_{12 } + \sum_2^{out}. \end{array}\ ] ] in the above @xmath195 and @xmath196 stand for the sums involving the bands outside the degenerate manifold and @xmath197 . the denominators of the terms involving @xmath198 and @xmath199 have opposite signs . therefore , adding up the expectation values we obtain : @xmath200 therefore the terms with diverging denominators cancel out . in the general case , the expectation value of the position operator is given by @xmath201 we can expand the integrand in the above equation and obtain @xmath202 & \displaystyle + { \langle i{\frac{\partial \bar u_{n'-}}{\partial { \bf q } } } | e^{i{\bf q}\cdot\hat{\bf r } } | \bar u_{n+ } \rangle } \\ [ 2.2ex ] & \displaystyle + { \langle \bar u_{n'- } | e^{i{\bf q}\cdot\hat{\bf r } } | i{\frac{\partial \bar u_{n+}}{\partial { \bf q } } } \rangle } \\ [ 2.2ex ] & \displaystyle + { \bf r}{\langle \bar u_{n'- } | e^{i{\bf q}\cdot\hat{\bf r } } | \bar u_{n+ } \rangle}. \end{array}\ ] ] all four brackets represent integrals of products of lattice periodic functions and exponentials . they will all eventually be proportional to @xmath203 . because of this fact the first term above integrates to zero . the partial derivatives in the second and third terms are evaluated by expanding the lattice periodic bloch wave functions about * q * , treating * q * as a small parameter @xmath204 in the limit in which @xmath205 @xmath206 consequently @xmath207 and the expectation value of the position vector yields , finally @xmath208 the spin gauge field is found in an analogous fashion to the expectation value of the position operator @xmath209 expanding the above , @xmath210 & \displaystyle + { \langle i{\frac{\partial \bar u_{n'-}}{\partial { \bf q } } } | \hat s e^{i{\bf q}\cdot\hat{\bf r } } | \bar u_{n+ } \rangle } \\ [ 2.2ex ] & \displaystyle + { \langle \bar u_{n'- } | \hat s e^{i{\bf q}\cdot\hat{\bf r } } | i{\frac{\partial \bar u_{n+}}{\partial { \bf q } } } \rangle } . \end{array}\ ] ] using the same arguments as in the previous section , the first term integrates to zero , and after evaluating the brackets involving the exponential and the lattice - periodic functions , we obtain in the limit @xmath211 @xmath212 an almost identical derivation applies for the expectation value @xmath213 . the only difference is that , since @xmath214 may be a function of wave vector , an additional term involving @xmath215 is generated . however , that term drops out when the anti - commutator is taken , leaving us with the result @xmath122 since the expectation value of the carrier position has been shown above to contain no divergences , if the expectation value @xmath216 is not divergent then the variance @xmath217 is also finite . to see that the variance of the particle position does not contain any divergences , we need to evaluate the expectation value @xmath216 this requires us to evaluate the matrix element @xmath218 first note that an expression of the form @xmath219 can be written as @xmath220 & \displaystyle + 2{\frac{\partial } { \partial { \bf q}}}\cdot[{\frac{\partial a}{\partial { \bf q } } } bcd + ab { \frac{\partial } { \partial { \bf q}}}(cd ) ] \\ [ 2.2ex ] & \displaystyle - \frac{\partial^2 a}{\partial q^2}bcd - 2{\frac{\partial a}{\partial { \bf q } } } \cdot b{\frac{\partial } { \partial { \bf q}}}(cd ) \\ [ 2.2ex ] & \displaystyle - ab \frac{\partial^2 b}{\partial q^2 } ( cd ) . \end{array}\ ] ] in the case we are considering , all the products involve brackets which are proportional to @xmath203 . as a result , all the terms on the first line vanish under integration with respect to * q * and only the terms on the second line need to be evaluated . writing them out explicitly , @xmath221 & \displaystyle - { \langle \frac{\partial^2 \bar u_{n'-}}{\partial q^2 } | e^{i{\bf q}\cdot\hat{\bf r } } | \bar u_{n+ } \rangle } - { \langle \bar u_{n'- } | e^{i{\bf q}\cdot\hat{\bf r } } | \frac{\partial^2 \bar u_{n+}}{\partial q^2 } \rangle } \\[2.2ex ] & \displaystyle + 2i{\bf r}\cdot({\langle { \frac{\partial \bar u_{n'-}}{\partial { \bf q } } } | e^{i{\bf q}\cdot\hat{\bf r } } | \bar u_{n+ } \rangle } + { \langle \bar u_{n'- } | e^{i{\bf q}\cdot\hat{\bf r } } | { \frac{\partial \bar u_{n+}}{\partial { \bf q } } } \rangle})\\[2.2ex ] & \displaystyle - 2{\langle { \frac{\partial \bar u_{n'-}}{\partial { \bf q } } } | e^{i{\bf q}\cdot\hat{\bf r } } | { \frac{\partial \bar u_{n+}}{\partial { \bf q } } } \rangle}. \end{array}\ ] ] in order to evaluate the differentials , the wave functions are expanded as in ( [ uexpand ] ) except that now the expansion must be made to second order in * q*. the final result is @xmath222 & \displaystyle + \frac{1}{2}{\langle { \frac{\partial \bar u_{n'}}{\partial \tilde{\bf q } } } |}\cdot{| { \frac{\partial \bar u_{n}}{\partial \tilde{\bf q } } } \rangle } - \frac{1}{4}({\langle \frac{\partial^2\bar u_{n'}}{\partial \tilde q^2 } | \bar u_n \rangle } \\ [ 2.2ex ] & \displaystyle + { \langle \bar u_{n ' } | \frac{\partial^2 \bar u_{n}}{\partial \tilde q^2 } \rangle } ) ] . \end{array}\ ] ] this is clearly finite so the variance of the position operator is finite . the velocity matrix element @xmath223 is easily evaluated as @xmath224 & \displaystyle = \int d^3r ' \ , \bar u_{n'- } ( { \bf r } ' ) e^{-i{\bf q}_-\cdot{\bf r } ' } \hat { \bf v } e^{i{\bf q}_+\cdot { \bf r } ' } \bar u_{n+}({\bf r } ' ) \\ [ 2.2 ex ] & \displaystyle = \int d^3r ' \ , e^{i{\bf q}\cdot { \bf r } ' } \bar u_{n'- } ( { \bf r } ' ) ( e^{-i{\bf q}_+\cdot\hat{\bf r } ' } \hat { \bf v } e^{i{\bf q}_+\cdot\hat{\bf r } ' } ) \bar u_{n+}({\bf r } ' ) \\[2.2 ex ] & \displaystyle = \delta({\bf q})\int d^3r ' \ , \bar u_{n ' } ( { \bf r } ' ) \tilde { \bf v } \bar u_{n}({\bf r } ' ) = \delta({\bf q}){\langle \bar u_{n ' } | \tilde { \bf v } | \bar u_{n } \rangle}. \end{array}\ ] ] in the above we have abbreviated @xmath225 . we also need to evaluate the commutator @xmath226 . one half of the commutator is evaluated as @xmath227 & \displaystyle = { \langle \bar u_{n'- } | e^{-i{\bf q}_-\cdot ( \hat{\bf r } - { \bf r } ) } \hat { \bf v}[-i{\frac{\partial } { \partial { \bf q}_+ } } e^{i{\bf q}_+\cdot ( \hat{\bf r } - { \bf r } ) } ] | \bar u_{n+ } \rangle } \\[2.2 ex ] & \displaystyle = -i{\frac{\partial } { \partial { \bf q}_+ } } { \langle \bar u_{n'- } | e^{-i{\bf q}_-\cdot ( \hat{\bf r } - { \bf r } ) } \hat { \bf v } e^{i{\bf q}_+\cdot ( \hat{\bf r } - { \bf r } ) } | \bar u_{n+ } \rangle } \\[2.2 ex ] & \displaystyle + { \langle \bar u_{n'- } | e^{-i{\bf q}_-\cdot ( \hat{\bf r } - { \bf r } ) } { \frac{\partial \hat { \bf v}}{\partial { \bf q}_+ } } e^{i{\bf q}_+\cdot ( \hat{\bf r } - { \bf r } ) } | \bar u_{n+ } \rangle } \\ [ 2.2 ex ] & \displaystyle + { \langle \bar u_{n'- } | e^{-i{\bf q}_-\cdot ( \hat{\bf r } - { \bf r } ) } \hat { \bf v } e^{i{\bf q}_+\cdot ( \hat{\bf r } - { \bf r } ) } | i{\frac{\partial \bar u_{n+}}{\partial { \bf q}_+ } } \rangle } , \end{array}\ ] ] while the other is @xmath228 & \displaystyle = { \langle \bar u_{n'- } | [ i{\frac{\partial } { \partial { \bf q}_- } } e^{-i{\bf q}_-\cdot ( \hat{\bf r } - { \bf r } ) } ] \hat { \bf v } e^{i{\bf q}_+\cdot ( \hat{\bf r } - { \bf r } ) } | \bar u_{n+ } \rangle } \\[2.2 ex ] & \displaystyle = i{\frac{\partial } { \partial { \bf q}_- } } { \langle \bar u_{n'- } | e^{-i{\bf q}_-\cdot ( \hat{\bf r } - { \bf r } ) } \hat { \bf v } e^{i{\bf q}_+\cdot ( \hat{\bf r } - { \bf r } ) } | \bar u_{n+ } \rangle } \\[2.2 ex ] & \displaystyle - { \langle i{\frac{\partial \bar u_{n-}}{\partial { \bf q}_- } } | e^{-i{\bf q}_-\cdot ( \hat{\bf r } - { \bf r } ) } \hat { \bf v } e^{i{\bf q}_+\cdot ( \hat{\bf r } - { \bf r } ) } | \bar u_{n'+ } \rangle } \\[2.2 ex ] & \displaystyle - { \langle \bar u_{n'- } | e^{-i{\bf q}_-\cdot ( \hat{\bf r } - { \bf r } ) } { \frac{\partial \hat { \bf v}}{\partial { \bf q}_- } } e^{i{\bf q}_+\cdot ( \hat{\bf r } - { \bf r } ) } | \bar u_{n+ } \rangle}. \end{array}\ ] ] all the brackets in the above two equations are proportional to @xmath229 , causing most of the terms to cancel . we are left with : @xmath230 the operator @xmath232 must be invariant under gauge transformations . from the expansion of the density operator in bloch eigenstates , @xmath233 it is evident that , if the wave functions change according to @xmath234 , then @xmath235 must transform to @xmath236 . using the definition of @xmath13 in terms of the wigner function , @xmath237 and remembering that @xmath238 and @xmath239 , we obtain @xmath240 the matrices @xmath241 and @xmath242 may be expanded about their arguments as : @xmath243&\displaystyle o(\tilde{\bf q}_- ) = o(\tilde{\bf q } ) - \frac{{\bf q}}{2 } \cdot { \frac{\partial o ( \tilde{\bf q})}{\partial \tilde{\bf q } } } + o(q^2 ) . \end{array}\ ] ] then , to first order in @xmath244 , abbreviating @xmath245 to @xmath171 and @xmath246 to @xmath247 , @xmath248&\displaystyle = o^{-1 } \rho_q o + \frac{{\bf q}}{2 } \cdot ( { \frac{\partial o^{-1}}{\partial \tilde{\bf q}}}\rho_q o - o^{-1 } \rho_q { \frac{\partial o}{\partial \tilde{\bf q } } } ) \\[2.2ex]&\displaystyle = \tilde\rho_q + \frac{{\bf q}}{2 } \cdot ( { \frac{\partial o^{-1}}{\partial \tilde{\bf q}}}o \tilde\rho_q - \tilde \rho_q o^{-1}{\frac{\partial o}{\partial \tilde{\bf q } } } ) , \end{array}\ ] ] where @xmath249 . the last line is obtained by inserting @xmath250 or @xmath251 as appropriate . since @xmath252 , we have @xmath253 and @xmath254 finally , writing @xmath255 , we recover formula ( [ rhogt ] ) for the transformation of @xmath231 . n. w. ashcroft and n. d. mermin , _ solid state physics _ , 2@xmath256 edition ( saunders college publishing , new york , 1976 ) . d. culcer , j. sinova , n. a. sinitsyn , t. jungwirth , a. h. macdonald , and q. niu , phys . * 93 * , 046602 ( 2004 ) . d. culcer , y. g. yao , a. h. macdonald , and q. niu , phys . b ( 2005 ) . j. sinova , d. culcer , q. niu , n. a. sinitsyn , t. jungwirth , and a. h. macdonald , phys . lett . * 92 * , 126603 ( 2004 ) . s. murakami , n. nagaosa , and s .- c . zhang , science * 301 * , 1348 ( 2003 ) . s. murakami , n. nagaosa , and s .- c . zhang , phys . b * 69 * , 235206 ( 2004 ) . s. zhang , phys lett * 85 * , 393 ( 2000 ) . s. zhang and z. yang , phys . lett . * 94 * , 066602 ( 2005 ) . l. hu , j. gao , and s. q. shen , phys . b * 70 * , 235323 ( 2004 ) . j. inoue , g.w.bauer , l. w. molenkamp , phys . b * 67 * , 033104 ( 2003 ) . e. g. mishchenko , a. v. shytov , and b. i. halperin , phys . lett . * 93 * , 226602 ( 2004 ) . o. v. dimitrova , phys . b * 71 * , 245327 ( 2005 ) . g. y. guo , y. yao , and q. niu , phys . lett . * 94 * , 226601 ( 2005 ) . y. yao and z. fang , cond - mat/0502351 . j. callaway , _ quantum theory of the solid state _ ( academic press , new york , 1976 ) . g. sundaram and q. niu , phys . b * 59 * , 14915 ( 1999 ) . m. v. berry , proc . london ser . a , * 392 * , 45 ( 1984 ) . d. culcer , a. h. macdonald , and q. niu , phys . b * 68 * , 045327 ( 2003 ) . t. jungwirth , q. niu , and a. h. macdonald , phys . lett . * 88 * , 207208 ( 2002 ) . yugui yao , leonard kleinman , a. h. macdonald , jairo sinova , t. jungwirth , ding - sheng wang , enge wang , and qian niu , phys . lett . * 92 * , 037204 ( 2004 ) . d. culcer , y. g. yao , and q. niu , phys . b ( 2005 ) . r. shindou and k. imura , nucl . b * 720 * [ fs ] , 399 ( 2005 ) . s. a. wolf , d. d. awschalom , r. a. buhrman , j. m. daughton , s. von molnar , m. l. roukes , a. y. chtchelkanova , and d. m. treger , science * 294 * , 1488 ( 2001 ) . hammar and m. johnson , phys . lett . * 88 * , 066806 ( 2002 ) . y. k. kato , r. c. myers , a. c. gossard , and d. d. awschalom , science * 306 * , 1910 ( 2004 ) . j. wunderlich , b. kaestner , j. sinova , and t. jungwirth , phys . lett . * 94 * , 047204 ( 2005 ) . y. kato , r. c. myers , a. c. gossard , and d. d. awschalom , phys . * 93 * , 176601 ( 2004 ) . l. e. reichl , _ a modern course in statistical physics _ ( university of texas press , austin , 1980 ) . an interesting recent article , g. a. fiete and e. g. heller , phys . a * 68 * , 022112 ( 2003 ) , discusses this topic from a somewhat different point of view . r. g. littlejohn and w. g. flynn , phys . a * 44 * , 5239 ( 1991 ) . m. marder , _ condensed matter physics _ ( john wiley , new york , 2000 ) . dyakonov and v. i. perel , jetp lett . * 35*a , 459 ( 1971 ) . dyakonov and v. i. perel , sov . jetp * 33 * , 1053 ( 1971 ) . dyakonov and v. i. perel , sov . solid state * 13 * , 3023 ( 1972 ) . dyakonov and v. y. kachorovskii , sov . . semicond . * 20 * , 110 ( 1986 ) . e. l. ivchenko , y. b. lyanda - geller , and g. e. pikus , sov . jetp * 71 * , 550 ( 1990 ) . g. e. pikus and a. n. titkov in _ optical orientation _ ( elsevier science publications , amsterdam , 1984 ) . a. g. aronov , yu . b. lyanda - geller , and g. e. pikus , sov . jetp * 73 * , 537 ( 1991 ) . y. qi and s. zhang , phys . b * 65 * , 214407 ( 2002 ) , ibid . * 67 * , 052407 ( 2003 ) . u. fano , rev . * 29 * , 74 ( 1957 ) . a. messiah , _ quantum mechanics _ ( dover publications , new york , 2000 ) . ping zhang , junren shi , di xiao , and qian niu , phys . lett * 96 * , 076604 ( 2006 ) . jackson , _ classical electrodynamics _ , 3@xmath257 edition ( john wiley @xmath258 sons , 1999 ) , section 6.6 ( pp . 248 - 258 ) . j. m. luttinger , phys . rev . * 102 * , 1030 ( 1956 ) . m. zarea and s. e. ulloa , phys . b * 73 * , 165306 ( 2006 ) . a. m. dudarev , r. b. diener , i. carusotto , and q. niu phys . rev . lett . * 92 * , 153005 ( 2004 ) y. kato _ et al . _ , nature * 427 * , 50 ( 2004 ) .

in 1974 , t hooft described the behavior of the large @xmath1 limit of qcd in @xmath0 dimensions @xcite . he showed that the model confined quarks and was exactly soluable . his work was followed quickly by several papers @xcite discussing various aspects of the model . the work by einhorn @xcite discussed the t hooft model predictions for deep inelastic scattering ( dis ) . einhorn found that the dis cross section was proportional to the square of the undressed quark masses , and hence would approach zero as the quark masses approach zero . this pathlogical result made it difficult to apply the t hooft model to the phenomenology of dis ; clearly someting major was missing . of course , limiting the physics to @xmath0 dimensions also excludes transverse spin degrees of freedom , so @xmath3 was zero and the callan - gross relation @xcite did not hold . since then qcd in @xmath0 diminsions has been studied from a variety of points of view @xcite and applied recently to the study of duality in heavy meson decays @xcite . however , to the best of our knowledge , no one has followed up on einhorn s study @xcite of dis . a few years ago we found that a very simple toy model of scalar and spinor fields in @xmath0 dimension gave nice phenomenological results for dis @xcite . by extending the spinor degrees of freedom to 1 + 3 dimensions we were able to recover the callan - gross relation . we decided to extend our approach to the t hooft model , and learned of the pathology in its description of dis . it turns out that this pathology can be removed by extending the spinor degrees of freedom to 1 + 3 dimensions , just as we did before , and this paper grew out of that study . this paper is divided into five sections , with seven appendices that include many of the details . following this brief introduction , sec . ii reviews the t hooft model results for the strong sector . our discussion focuses on ( i ) a method of treatment of the confining interaction that removes all singularities from the theory and gives finite dressed quark masses , and ( ii ) the consequence of including transverse degrees of freedom for the spin of the quarks . we calculate the dressed quark mass , and the properties and spectrum of @xmath4 bound states , and also study the consequences of the completness of the bound states . we show how to construct an off - shell @xmath4 scattering matrix , and show that the on - shell scattering matrix must be zero , as required for confined particles . sections iii and iv study electromagnetic interactions and deep inelastic scatterng . here the presence of transverse spin degrees of freedom completly alters the einhorn discussion , giving us phenomonologically useful results . we conclude with a brief discussion . in a subsequent paper @xcite we plan to present a numerical study of the approach to scaling in the dis limit . we propose that the action have the following form @xmath5 where the qcd part of the action ( gluons and quark momenta or coordinates ) live in @xmath0 dimensions only , but the quark spin and the qed part ( leptons and the photon ) live in the full @xmath2 dimensions @xmath6 in his original paper , t hooft @xcite discussed the strong interactions only , and in this section we review the 1 + 1 dimensional model of qcd that he presented . the qed part of the action ( [ a2 ] ) , and reduced qcd ( rqcd ) , will be discussed in sec . [ rqcd ] . the qcd lagrangian density is @xmath7 + \sum_i \bar{q}_i\left(i d_{\mu } \gamma^{\mu}-m_{0i } \right)q_i\ , , \label{1eq1}\ ] ] where @xmath8 is the quark field with flavor @xmath9 and bare mass @xmath10 . the gluon field quantities are @xmath11 \\ \nonumber d_{\mu}&=&\partial_{\mu}+ig_0 a_{\mu } , \label{1eq2}\end{aligned}\ ] ] where @xmath12 is the gluon field tensor and @xmath13 are the gluon fields with the lorentz index @xmath14 and the color index @xmath15 . the gluon fields have components in the 0 and 3 direction only , so the sum over the index @xmath14 is restricted to 0 and 3 . the qcd coupling constant is @xmath16 and @xmath17 are the generators of the @xmath18 color group , normalized to @xmath19 = 2\,\delta^{ab}\nonumber\\ & & \sum_{a}\lambda^a\lambda^a= \frac{2(n_c^2 - 1)}{n_c}\;\openone \ , , \label{1eq2aa}\end{aligned}\ ] ] where matrix multiplication of the @xmath20 matrices is implied . as @xmath21 , finite results are obtained if @xmath22 as @xmath23 where the effective coupling @xmath24 is a constant . while the space - time coordinates , momenta , and lorentz vector sums in eq . ( [ 1eq1 ] ) are restricted to 1 + 1 dimension , we assume that the gamma matrices have the usual @xmath25 dirac structure . the quark fields are therefore a direct product of four dirac dimensions and @xmath1 color dimensions . note that the action ( [ a2 ] ) is lorentz invariant under the subgroup @xmath26 of lorentz transformations that leave the @xmath27 plane invariant . specifically , the group is generated by the hamiltoninian @xmath28 , and the boost @xmath29 , momentum @xmath30 , and angular momentum @xmath31 , operators that generate boosts and translations along the @xmath32 axis , and rotations about the @xmath32 axis . four of the six commutation relations between these generators are zero , and the other two close @xmath33=-i{\cal h } \qquad [ { \cal k}_3,{\cal h}]=-i{\cal p}_3 \ , , \label{lgroup}\end{aligned}\ ] ] guaranteeing that @xmath26 is a group . following t hooft @xcite we introduce light cone variables : @xmath34 for any arbitrary vector @xmath35 ( in the qcd sector , the only perpendicular components come from matrix elements of @xmath36 ) . note that the scalar product of any two vectors @xmath15 and @xmath35 is @xmath37 the derivatives are defined @xmath38 so that the divergence of a two - vector is @xmath39 in the same way , we can define the @xmath40 , @xmath41 , and @xmath42 components of the @xmath43 matrices . the anticommutation relations are all zero except for @xmath44 since the gluon fields are confined to 1 + 1 dimensions , there is only one nonvanishing component of the gluon field strength tensor @xmath45\ , .\ ] ] the qcd part of the theory is simplified if we choose the light cone gauge , where @xmath46 , so that the commutator contained in the field tensor @xmath47 disappears and there is only one nonzero component ( @xmath48 ) of the gluon field . the lagrangian density ( [ 1eq1 ] ) then reduces to @xmath49 + \sum_i \bar{q}_i\left(i \partial_+ \gamma_- + i \partial_- \gamma_+ -g_0 \gamma_{-}a_{+ } -m_{0i } \right)q_i\ , . \label{1eq4}\ ] ] the equation of motion for this field is then @xmath50 the solution of ( [ 1eq5 ] ) is @xmath51 where the green s function @xmath52 is @xmath53 the coefficients @xmath54 and @xmath55 can not be determined without knowing the boundary conditions , so they are free parameters . the gauge condition did not eliminate all superfluous degrees of freedom , just as the coulomb gauge , or the lorentz gauge , do not determine uniquely the photon propagator in qed ( gribov ambiguity ) . we can therefore set the coefficients @xmath54 and @xmath55 equal to zero ( a specific choice of gauge ) in order to simplify our calculations . einhorn discussed the gauge issues related to these two parameters and showed that the eigenvalues of the two body bound state equation are independent of the choice of @xmath54 and @xmath55 . however , the dressed quark mass does depend on the choice of @xmath55 , and this in turn implies that the location of the quark mass pole is gauge dependent . since any physically meaningful quantity is gauge invariant , we can conclude that the location of the mass pole is not physically meaningful , and this can happen only if the quarks are confined , so that free quark states do not exist . this is an indirect consequence of confinement . direct consequences of confinement will be discussed in the next subsection when the two - body bound state equation is discussed . equation ( [ 1eq6 ] ) shows that the gluonic field is no longer a dynamical variable , and that there are therefore no ghosts . the fourier transform of the green s function ( [ 1eq7 ] ) gives us the gluon `` propagator '' , or more precisely the momentum dependence of the effective quark - quark interaction . with @xmath54 and @xmath55 equal to zero , it can be written @xmath56 this is a singular operator that is well defined only when it is part of an integral over @xmath57 . the second term in eq . ( [ 1eq8 ] ) was introduced by gross and milana @xcite in a different context . its purpose is to preserve the condition @xmath58 and it also insures that the potential is finite for any finite value of @xmath59 . the only variables remaining are the quark fields . the feynman rule for the ( undressed ) quark propagator is @xmath60 and the quark-``gluon '' coupling is @xmath61 leading to the following result for the exchange of gluons with momentum @xmath57 between quarks with dirac indicies 1 and 2 @xmath62\nonumber\\ & & = -ig_0 ^ 2\ , \frac{2(n_c^2 - 1)}{n_c } \gamma^1_{-}{\cal o}_{12}\gamma^2_-\ , \tilde{\cal g}(k_- ) % & & { \displaystyle\lim_{n_c\to\infty } } \to -ig^2 \gamma^1_{-}{\cal o}_{12}\gamma^2_-\ , \tilde{\cal g}(k_- ) \ , , \label{1eq11}\end{aligned}\ ] ] where we used eq . ( [ 1eq2aa ] ) ( with matrix multiplication of the @xmath20 matrices implied ) , and took the limit as @xmath21 using the definition ( [ 1eq2a ] ) . the quark self - energy and the dressed quark mass are calculated in the next subsection . we determine the dressed single quark propagator , @xmath63 , using the ( one body ) dyson schwinger equation ( dse ) : @xmath64s_{0}(p)\ , , \label{dse}\end{aligned}\ ] ] shown graphically in fig . [ ds ] . the gluon interaction does not mix quark flavors , and the calculation is identical for each flavor of quark , so the flavor index is suppressed . the rainbow approximation ( undressed vertices and the absence of the quark loops from the gluon propagator ) is justified in the large @xmath1 limit @xcite . since for every internal loop there is a factor of @xmath65 , and a multiplicative factor of @xmath66 , the color dependence disappears . the vertex corrections and the quark - gluon vertices do not have a multiplicative factor of @xmath1 , and are therefore supressed in the large @xmath1 limit . in eq . ( [ dse ] ) , @xmath67 , and since @xmath68 does not depend on @xmath69 , it follows immediately that the self - energy integral does not depend on @xmath70 either , and must have the form @xmath71 . hence the dressed propagator is of the form @xmath72 \gamma_- -i\epsilon } \ , , \label{1eq11a}\ ] ] where , using @xmath73 , the self - energy contribution is then @xmath74 performing the @xmath69 integral gives @xmath75 and substituting this back into eq . ( [ 1eq12 ] ) gives @xmath76 using ( [ 1eq8 ] ) for @xmath77 gives @xmath78 substituting this back into eq . ( [ 1eq11 ] ) gives @xmath79 \gamma_- -i\epsilon } = \frac{m_0+p_-\gamma_+ + \left(p_+ + { \displaystyle\frac{g^2}{2\pi p_- } } \right ) \gamma_- } { m_0 ^ 2-{\displaystyle\frac{g^2}{\pi}}-2 p_+p_- -i \epsilon}\ , . \label{1eq16}\end{aligned}\ ] ] note that the quark mass pole has been shifted to a _ smaller _ value @xmath80 where @xmath81 is the dressed quark mass . it turns out that choosing @xmath82 will not effect the dressing of the mass , but choosing @xmath83 would give @xmath84 this gives a mass that is momentum dependent and not covariant . hence the dressed mass is gauge dependent and unphysical . with these cautionary remarks , we choose @xmath85 because the mass is covariant . one undesirable feature of this choice is that @xmath86 in the chiral limit . this could be corrected by choosing @xmath55 large and positive . having obtained the dressed propagator , we are able to proceed with the two body bound state calculation . consider a bound state of a @xmath4 pair . the quark has dressed mass @xmath87 and electric charge @xmath88 , and the antiquark ( which might be of a different flavor ) has dressed mass @xmath89 and charge @xmath90 . the momentum of the bound state is @xmath91 , the momentum of the quark is @xmath92 and the momentum of the outgoing antiquark is @xmath93 , as shown in fig . [ bs ] . in much of the following discussion , we will treat the outgoing antiquark as an incoming quark with momentum @xmath94 , and following this convention label the bound state vertex function @xmath95 . for a color neutral state we may carry out the color sums using ( [ 1eq11 ] ) , giving the bethe - salpeter equation for the bound state vertex function : @xmath96 where @xmath97 and @xmath98 are dressed quark propagators , and the kernel is the singular operator @xmath99 with the substitution @xmath100 @xcite , eq . ( [ bse ] ) becomes : @xmath101 \left[m_2 ^ 2-(k - r)^2 -i\epsilon\right]}\ , . \label{bse2}\ ] ] the formula shows that @xmath102 does not depend on @xmath70 , and hence the @xmath69 integration can be carried out immediately . integrations of the variable @xmath69 over two quark propagators occurr frequently , and are worked out in detail in appendix [ appen : a ] . it is convenient to introduce the momentum fractions @xmath103 if the fraction @xmath32 lies in the interval [ 0,1 ] , eq . ( [ bse2 ] ) reduces to @xmath104\right ) \ , , \label{phiin}\end{aligned}\ ] ] where @xmath105 is the principal value integral , and @xmath106 , @xmath107 , and @xmath108 are dimensionless parameters @xmath109 with @xmath110 the anticipated mass eigenvalues that will emerge from the solution of the equation , and @xmath111 the _ dressed _ quark masses . the two - body `` wave function '' @xmath112 , is _ defined _ by @xmath113 where @xmath114 is the anticipated eigenfunction solution for the vertex corresponding to the @xmath115th bound state , and @xmath116 is a normalization constant . it is chosen so that @xmath117 and will be calculated in subsection [ elasticff ] . note that eq . ( [ phiin ] ) _ defines _ the two - quark hamiltonian , @xmath118 , on the interval @xmath119 . the _ bare _ quark masses enter into the hamiltonian , while the _ dressed _ quark masses enter into the relation ( [ 2eq3d ] ) between the vertex function and the wave function . if the momentum fraction @xmath32 lies _ outside _ of the interval @xmath119 , the vertex function defined by eq . ( [ bse2 ] ) is not zero , but can be obtained by quadrature from the vertex function defined _ inside _ of the interval [ 0,1 ] . using ( [ 2eqac ] ) and the definition ( [ 2eq3d ] ) gives @xmath120\right ) \ , .\qquad \label{phiout}\end{aligned}\ ] ] these definitions and equations will be used in the following sections when the hadronic currents are computed . an alternative form of eq . ( [ phiin ] ) follows if we integrate the last term by parts : @xmath121 where @xmath122 . if @xmath123 , which is true in all but the chiral limit , this form of the equation follows directly from t hooft s principle value prescription @xcite , so that our two - body equation is identical with t hooft s , even though we have finite dressed quark masses . to avoid confusion later , we call attention to the fact that the wave function defined by eqs . ( [ 2eq3d ] ) and ( [ phiout ] ) is not identical to the bethe - salpeter wave function . the latter can be defined @xmath124 where the factor @xmath125 in the numerator comes from the @xmath126 factors in the quark propagators @xmath127 . this wave function carries the @xmath70 dependence of the quark propagators . if we integrate ( [ bswave ] ) over @xmath70 ( using the fact that @xmath128 does not depend on @xmath70 ) we obtain @xmath129 $ \cr 0 & if $ z\notin [ 0,1]$ } \ , . \label{equalpp}\ ] ] this is the `` equal @xmath130 '' wave function , and equals the wave function we are using _ only in the region _ @xmath131 $ ] . as we have seen , the wave function @xmath132 that solves the t hooft equation is non zero _ for all _ @xmath32 . we conclude this discussion of the two body wave equation by returning to eq . ( [ bse ] ) . in place of the replacement @xmath133 , it could equally well have been assumed that @xmath134 where @xmath135 is either @xmath136 or @xmath137 ( recall that we are assuming that the dirac matrices live in 1 + 3 dimensional space ) . note that the bethe - salpeter equations with the substitutions ( [ 2eqeee ] ) reduce to the same equation ( [ bse2 ] ) for @xmath138 , and hence _ each two - body state is triplely degererate , with spin structure given by @xmath126 , @xmath139 , or @xmath140_. in each of these three cases the wave functions and bound state mass are identical . the additional states with the @xmath141 or @xmath142 structure are outside the scope of the original t hooft model , but must be considered in reduced qcd . they are states with quark spins in the transverse direction , and will play a major role in the discussion of dis below . the chiral limit is exactly soluable and of some interest . if @xmath143 , then eq . ( [ phiin ] ) has the normalized ground state solution @xmath144 outside of the interval @xmath145 $ ] , eq . ( [ phiout ] ) gives the same result , so that @xmath146 for all @xmath32 in @xmath147 . the t hooft model illustrates how confinement can be realized in two ( apparently different ) ways . if the singularities of the confining interaction are left unregularized , as they were in t hooft s original paper @xcite , the quark masses are infinite . even if the singularity is regularized as in this paper , whenever @xmath148 the dressed quark mass will be unphysical , and there is no real quark mass pole . this insures that the quarks are confined . in the second case ( @xmath149 and @xmath150 ) the dressed quark mass is real and positive and the there is a quark mass pole . what keeps the quarks confined in this case ? the answer is that the quark vertex function @xmath151 has a _ zero _ at precisely the value of @xmath32 where all of the quarks in the bound state could be on mass shell . this requirement is illustrated diagramatically in fig . [ confinement ] , and seems miraculous . however , as discussed in ref . @xcite , it is only a consequence of fact that the _ wave function _ is finite at the pole ( because the wave equation permits no singularities ) , and therefore the vertex function @xmath152 \frac{\phi(z , r)}{{\cal n } r_-}\end{aligned}\ ] ] must be zero . precisely the same mechanism works in the schrdinger description of two nonrelativistic particles bound by a linear ( or any other ) confining potential . in summary , quark mass poles are always unphysical because the dressed mass is gauge dependent , and bound state poles are physical because they are gauge invariant . in this respect the undressed quark mass is the only physically significant parameter . the two ways of viewing confinement are therefore not exclusive , but equivalent , since they are related through gauge transformations . the bound state equation was solved numerically using modified cubic splines . the splines and the technique are fully described in appendix [ appen : splines ] . each spline is a smooth function ( continuous up to its third derivative ) and has support over a small region of the @xmath32 axis . the splines overlap , so that by independently adjusting the height of each spline it is possible to smoothly approximate any function defined on the interval @xmath145 $ ] . .numerical values of @xmath153 for @xmath154 and various @xmath155 ( number of splines ) and @xmath156 ( number of guass points for each spline segment ) . [ cols=">,>,^,^,^,^,^,^,^,^,^,^,^,^ " , ] [ tab:1 ] we have modified the standard spline technique to insure that the numerical solutions satisfy the boundary conditions exactly . these boundary conditions are determined by examing eq . ( [ phiin ] ) in the vicinity of @xmath157 and @xmath158 @xcite , and require the wave function go like @xmath159 note that the @xmath160 in the chiral limit ( when @xmath161 ) , consistent with the exact solution ( [ chiralsol ] ) . to fit these boundary conditions the splines closest to @xmath162 and @xmath163 have these fractional powers built into their functional form , as described in appendix [ appen : splines ] . the wave function is expanded into @xmath155 splines , each with four segments ( except for those at the boundary , which have only three ) . the splines overlap , and taking into account that there must be at least three splines ( one regular spline in the center and one modifed spline at each end ) , the total number of segments is @xmath164 , dividing the interval into segments of length @xmath165 . the equation is turned into a matrix equation by integrating over each segment using @xmath156 gaussian points . tables [ tab:2 ] and [ tab:1 ] show the numerical stability . if the quark mass is larger than unity , better than 1% accuracy is achieved with only two gauss points per interval , provided @xmath166 . as the level number @xmath115 increases above @xmath167 toward @xmath155 , the reliability of the calculation decreases . to study highly excited states , it is sufficient to use @xmath168 , and push @xmath155 as high as possible . the situation is somewhat different if the bare quark masses are less than unity ( table [ tab:1 ] ) . in this case the wave function is very steep at the boundaries and 4 or 6 gaussian points per interval are needed to get 1% accuracy for the low lying states . in a subsequent paper @xcite , we will limit our numerical discussion of duality to cases with @xmath169 where the solutions are very stable , and there are no poles in the region outside of [ 0,1 ] [ see the brief remarks following eq.([rlimit ] ) ] . figure [ mass ] shows the linearly rising regge trajectory characteristic of the t hooft model . figures [ 1st]-[10th ] show the ground , first excited and 10th level for three cases with equal quark masses of different values . note the `` tails '' of the wave functions outside of the region [ 0,1 ] , given by eq . ( [ phiout ] ) . these tails are large for small quark masses , but shrink quickly as the quark mass increases , and are also small for highly excited states . figure [ 20th ] shows the 20th level for spline numbers 20 , 40 , and 60 . note that the state is not fully described even with 40 splines , supporting the observation that an accurate description of states of order @xmath115 requires a spline number of approximately @xmath170 . the hamiltonian defined in eq . ( [ phiin ] ) is hermitian on the interval @xmath145 $ ] and it is straightforward to show that the different solutions are orthogonal on this interval . with the normalization given in eq . ( [ norm ] ) @xmath171 the completeness of the eigenfunctions implies that @xmath172\ , . \label{complete}\ ] ] the greens function is therefore @xmath173 where @xmath174 is any external momentum . the completness relation shows that @xmath128 has the property of a greens function @xmath175 \ , .\end{aligned}\ ] ] the same argument also gives @xmath176 \ , .\end{aligned}\ ] ] in preparation for the study of electromagnetic interactions , it is useful to introduce the @xmath4 scattering matrix , defined by the following infinite series @xmath177_1 \otimes\left[\gamma_{-}s_2(k - q)\gamma_-\right]_2 \nonumber\\ & & \qquad+i^2\int\frac{d^2k'}{(2\pi)^2}\int\frac{d^2k}{(2\pi)^2 } v(p',k')\,v(k',k)\,v(k , p)\,\nonumber\\ & & \qquad\qquad\times \left[\gamma_-s_1(k')\gamma_{-}s_1(k)\gamma_-\right]_1 \otimes\left[\gamma_{-}s_2(k - q)\gamma_-s_2(k'-q)\gamma_-\right]_2 \nonumber\\ & & \qquad + \cdots \ , .\quad \label{eq248}\end{aligned}\ ] ] the momentum are defined as for the two - body bound state , except that here @xmath174 is the unconstrained momentum of the @xmath4 pair and @xmath178 is not an eigenvalue . the kernel @xmath179 is the singular operator ( [ kernel2 ] ) . the singularities of this operator will never be exposed because we will limit application of this scattering equation to cases where we integrate over either the initial or final momentum . note that the structure of the equation anticipates the reduction of the dressed propagators @xmath180_i=\frac{k_-\ , [ \gamma_-\gamma_+\gamma_-]_i } { m_i^2 - 2k_-k_+}=\frac{2k_-}{m_i^2 - 2k_-k_+}\,\gamma_{i-}\ , , \end{aligned}\ ] ] so that the factor of @xmath181 is common , and the series ( [ eq248 ] ) reduces to @xmath182 where @xmath183 was defined in eq . ( [ d2 ] ) . from this result we can conclude that @xmath184 does not depend on @xmath185 and @xmath70 . in applications below , we will encounter this scattering series in the following form @xmath186 where @xmath187 is an operator that , by assumption , does not depend on @xmath70 . in appendix [ appen : b ] we show how to use the greens function to write this series as @xmath188 where we replaced @xmath189 and @xmath190 . this equation has a nice physical interpretation . it displays the `` scattering amplitude '' as the sum over the propagation of the bound states , which is all confinement will allow ( without meson decay mechanisms , which are ignored in this paper , no cuts are possible ) . the only singularities in @xmath178 that can occur are poles at the bound state masses @xmath153 . with these tools in place we are able to study the electromagnetic interaction of hadrons . we return to the discussion of the lagrangian density introduced in eq . ( [ a2 ] ) and look at the electromagnetic sector . in rqcd the electromagnetic fields ( and the electron currents ) are extended to 1 + 3 dimensions . the lagrangian density is therefore @xmath191 where @xmath192 is the electromagnetic field ( with the subscript 0 distinguishing it from the gluon field ) , the vector sums are now over 4 dimensions with @xmath193 , @xmath194 is the electromagnetic charge of quarks with flavor @xmath9 , @xmath195 is the electron current that produces the electromagnetic field , and @xmath196 is the electromagnetic field tensor . because the electron current exists in four dimensions , all four components of the electromagnetic field will be non - zero , in general , and all four components of the quark current will be excited by electromagnetic scattering . we now apply this discussion to electron scattering , where @xmath197 , so that @xmath198 . the initial particle ( quark or hadron ) has momentum @xmath199 and the final particle momentum @xmath200 , with momentum fractions @xmath201 if the final particle is on shell , then @xmath202 and @xmath203 , and @xmath204 as @xmath205 . we will use the notation @xmath206 to denote the _ bare _ quark current _ operator _ , which is @xmath207 while the dressing of the strong quark - gluon vertex is supressed by the large @xmath1 limit @xcite , the quark electromagnetic vertex can be dressed by gluon exchanges . the longitudinal part of the dressed electromagnetic vertex ( @xmath208 ) can be computed directly from the ward - takahashi identity @xmath209 where @xmath210 . using the dressed propagator ( [ 1eq16 ] ) , we find that @xmath211 consequently the @xmath212 component is unmodified . it is instructive to obtain the full quark current directly from the dressing of the quark - photon vertex . the vertex is dressed by successive gluon exchanges , as shown diagramatically in fig . [ quarkcurrent ] . when coupling to a photon , there is no flavor change , and hence , for each flavor @xmath9 , this series is @xmath213 this series is evaluated in appendix [ appen : c ] using the methods developed in appendix [ appen : b ] . the final results are conveniently expressed in term of a quark form factor , defined to be @xmath214 where @xmath215 and @xmath216 in anticipatation of applications to electron scattering . note that the form factor is expressed as a sum of bound state poles at @xmath217 , arising from the propagation of the bound states that couple to the photon . vector dominance is a rigorous consequence of this model . using this form factor , the currents reduce to @xmath218 note that the terms independent of the quark form factor @xmath219 are identical to the expected result ( [ current1 ] ) , and that the contribution from the form factor is _ purely transverse _ , in that @xmath220\big|_{f_i\;{\rm term } } & = & \gamma_- \,f_i(z , q^2)\,\left\ { q_+-q_-\frac{q_+}{q_-}\right\}=0\ , .\qquad \label{ff5}\end{aligned}\ ] ] this result mimics the method for insuring current conservation developed in ref . @xcite , and gives some evidence that that method is dynamically sound . these results are used in the next subsection to calculate the hadronic transition current . in the presence of confinement , and without any meson decay mechanisms , the quarks in a @xmath4 bound state must remain bound , even after they absorb an energetic photon . in this section we calculate the transition current for the process @xmath221 , where @xmath222 is the mass of a ( possibly excited ) final bound state and @xmath223 is the mass of the initial ( ground ) state . the transition current consists of the sum of two contributions : one in which the photon is absorbed by the quark with dressed mass @xmath87 , and a second in which the photon is absorbed by the antiquark with mass @xmath89 . the momentum used to label each of these processes are shown in fig . [ hadronicamplitude ] ; the initial bound state has momentum @xmath224 and the final state has momentum @xmath225 . the transition current coming from the first term is @xmath226\nonumber\\ & & \qquad\quad = i \int\frac{d^2k}{(2 \pi)^2}\frac { e_1n_c\,g_f\left(-p_f - k ,- k\right ) g_i\left(p+k , k \right)\,n^\mu } { [ m_1 ^ 2-(p_f+k)^2-i\epsilon ] [ m_1 ^ 2-(p+k)^2-i\epsilon ] [ m_2 ^ 2-k^2-i\epsilon ] } \ , , \qquad \label{3eq1-}\end{aligned}\ ] ] where @xmath227 is the quark current operator for quark 1 , as worked out in subsection [ quarkem ] , @xmath228 is the spin dependent numerator discussed below , and we have allowed for the possibility that the final state may have _ either _ the @xmath126 _ or _ the @xmath139 structure discussed above ( we will not consider the @xmath140 states ; they are needed only if we consider the @xmath137 component of the transverse current and give results identical to the results for the @xmath136 components ) . note the factor of @xmath1 coming from the sum over all possible quark colors . only the term in the numerator of the @xmath98 propagator equal to @xmath229 can give a nonzero result , but the terms that contribute from the @xmath97 propagators depend on the matrix form of the final state , and of @xmath227 . because the trace of a product of gamma matrices with only one factor of @xmath141 or @xmath142 is zero , the transverse currents ( with @xmath141 or @xmath142 ) will be zero unless the final state also includes @xmath141 ( or @xmath142 for the @xmath137 component of the current ) . hence we have the rules : ( i ) @xmath136-type transverse currents couple @xmath126 states to @xmath139 states only , and ( ii ) longitudinal currents couple @xmath126 states to each other . because of the exact degeneracy of these states , the results in both of these cases use the same momentum - space wave functions . the integral ( [ 3eq1- ] ) is evaluated in appendix [ appen : d ] . the result for the @xmath41 component of the current is @xmath230 \ , , \qquad\qquad\qquad\qquad \label{3eq1b}\end{aligned}\ ] ] where @xmath231 [ for definitions of all the momentum fractions see eqs . ( [ frac2 ] ) and ( [ frac3 ] ) ] , the state @xmath232 has the same structure as the initial state ( assumed to be @xmath126 ) , @xmath233 is the quark form factor defined in eq . ( [ ff1 ] ) and @xmath234 is given in eq . ( [ rlimit ] ) . the plus and transverse components of the current are @xmath235 } { m_{01}\,\eta\,(1-\eta)\,q^2 } \nonumber\\ \left < f_-|{\cal j}_+(p_f , p)|i_-\right > \big|_{e_1\,{\rm term}}&=&\frac{16 e_1n_c\,p_{-}}{{(2 \pi)\cal n}^2 } \left\ { \int_{y}^1 + \int_{0}^{y } { \cal r}\right\ } d\xi \;\phi_f(\xi',p_f)\phi_i(\xi , p)\nonumber\\ & & \qquad\qquad\times \frac{q_+\,\left[m_1 ^ 2- \eta(1-\eta)\,q^2 \,f_1(\eta , q^2)\right ] } { q_-\,\eta\,(1-\eta)\,q^2 } \ , , \qquad\qquad \label{currentd}\end{aligned}\ ] ] where the state @xmath236 has the structure @xmath139 . these results are used in the next subsection , and in the discussion of dis . using the definitions of the momenta shown in fig.[hadronicamplitude ] , it is easy to see that the second term in the transition current can be obtained from the first by letting @xmath237 , as discussed in appendix [ appen : d ] . the exact expression for the @xmath136 component of the transition current , which is the sum of the @xmath88 and @xmath90 contributions , can therefore be written @xmath238 where the reduced current @xmath239 is @xmath240 with @xmath241 and @xmath242 and @xmath243 \ , . \label{qffeff}\end{aligned}\ ] ] finally , the function @xmath244 , was defined in eq . ( [ rlimit ] ) ; the substitutions that convert the @xmath88 term into the @xmath90 term therefore give @xmath245 . if the scattering is elastic , @xmath246 and the momentum fraction @xmath247 must be computed from the exact expressions for @xmath174 and @xmath224 in the c.m . frame @xcite . the result , for elastic scattering from the state @xmath115 , is @xmath248 as @xmath205 , @xmath249 and @xmath250 [ defined in eq . ( [ fraccur ] ) ] approaches @xmath251 , @xmath252 , and the _ exact _ result for the elastic curent becomes @xmath253 ^ 2=2\,e_1\,p_-\ , { \cal f}_n(0)\ , , \label{3eq1c}\end{aligned}\ ] ] where charge normalization requires that the form factor @xmath254 be unity at @xmath255 . hence @xmath256 gives the normalization condition ( [ norm ] ) . furthermore , elastic scattering requires that , as @xmath205 , @xmath257 and hence @xmath258 which follows from the wave function identity ( [ idet3 ] ) . note that _ the charge is properly normalized and conserved for _ both _ components because we used the correct dressed quark current _ ( [ current1 ] ) . finally , we point out that the elastic transverse current is zero for all @xmath259 , as required by the summetries of the states @xmath260 we now turn to a discussion of deep inelastic scattering . using the standard conventions ( also defined in ref . @xcite ) , the dis cross section is @xmath261\nonumber\\ & = & \sigma_m\left[w_2 + 2w_1\tan^2\left(\frac{\theta}{2}\right ) \right]\ , , \end{aligned}\ ] ] where the structure functions @xmath3 and @xmath262 are part of the hadronic tensor , defined by @xmath263 where @xmath264 is the target , or ground state mass . the structure function @xmath3 is immediately extracted from the @xmath265 ( or @xmath266 ) component of the tensor @xmath267 here we anticipate that @xmath268 scales in the deep inelastic limit to the function @xmath269 of the variable @xmath270 . note that @xmath3 must be identically zero if the @xmath136 ( and @xmath137 ) components of the current are zero . the structure function @xmath262 can be extracted from the @xmath40 and @xmath41 components of the current using @xmath271 in the deep inelastic limit this gives @xmath272 the contribution from @xmath273 will be shown to go like @xmath274 in the following sections . hence , _ if @xmath233 is non - zero _ , we obtain the callan - gross relation . however , if the transverse components of the current are omitted , @xmath275 and @xmath276 does not scale . in this case it is the quanity @xmath277 that scales , and we recover the results of einhorn @xcite . we now turn to a calculation of the structure functions . we calculate them first ( i ) in the _ partonic picture _ , where all final state interations are ignored and the quarks in the final state are assumed to be free , and then ( ii ) in the _ hadronic _ picture , where confinement ensures that the only possible final states are the bound @xmath4 states we have already discussed . we emphasize that the partonic picture can never actually occur because the confining interaction , which also acts in the final state , can never be ignored . consider deep inelastic scattering ( dis ) in the partonic picture , which assumes a final state composed of free quarks with no interaction . this picture is illustrated in fig . [ partonicamplitude ] . the full result is the sum of a contributions in which the photon is absorbed by the quark with charge @xmath88 , and one in which it is absorbed by the antiquark with charge @xmath90 . in the center of mass of the ejectiles , either particle can go forwards or backwards and the unpolarized cross section is the sum over both helicities of the particles . begin by calculating the transverse component ( chosen be be @xmath136 ) of the current . the outgoing quark has a mass of @xmath87 and momentum @xmath278 in the center of mass frame of the ejectiles , while the antiquark has a mass of @xmath89 and momentum of @xmath279 ( in this section @xmath279 will be the physical momentum of the outgoing antiquark , opposite in sign to that used in the earlier sections of this paper ) . the current is the sum of two diagrams , only one of which is shown in fig . [ partonicamplitude ] . since the @xmath4 vertex contains the factor of @xmath126 , only the factor @xmath280 in the numerator of the dressed propagator ( [ 1eq16 ] ) survives , but we will retain only the factor of @xmath281 which dominates the result at large @xmath259 . the following result is obtained for the @xmath136 component of the current for the disintegration of the ground state ( the subscript @xmath282 usually attached to the ground state will be omitted in this section ) : @xmath283 \frac{(p_1-q)_-\,g(p_1-q ,- p_2)}{m_1 ^ 2-(p_1-q)^2 } \nonumber\\ & & + e_2\,[{\bar u}(p_1)\,\gamma_-\gamma_+ \gamma_x\ , v(p_2 ) ] \frac{(p_2-q)_-\ , g(p_1,-p_2+q ) } { m_2 ^ 2-(p_2-q)^2 } \ , . \label{5eq1}\end{aligned}\ ] ] introducing the momentum of the bound state , @xmath284 , and defining @xmath285 [ as in eq . ( [ defz ] ) ] , the first denominator can be written @xmath286(p - p_2)_- \nonumber\\ & = & m_1 ^ 2-z\,\left(m^2-\frac{m_2 ^ 2}{1-z}\right ) = z\left[\frac{m_1 ^ 2}{z}+\frac{m_2 ^ 2}{1-z}-m^2\right ] \ , . \label{5eq1a}\end{aligned}\ ] ] here we have made use of the constraints @xmath287 and @xmath288 . a similar result follows for the second denominator if we define a new momentum fraction @xmath289 @xmath290 with these definitions we obtain @xmath291 \;\frac { p_- g(z , p)}{\delta(z , m^2 ) } \nonumber\\ & & + e_2 [ { \bar u}(p_1)\ , \gamma_-\gamma_+ \gamma_x\ , v(p_2 ) ] \ ; \frac { p_- g(1-z',p ) } { % { \displaystyle\frac{m_2 ^ 2}{z'}+\frac{m_1 ^ 2}{1-z'}- m^2 } \delta(1-z ' , m^2 ) } \nonumber\\ \nonumber\\ & = & -\frac{e_1}{{\cal n } } \left[{\bar u}(p_1)\,\gamma_x\,\gamma_+\gamma_-\ , v(p_2 ) \right ] \phi(z , p ) \nonumber\\ & & + \frac{e_2}{\cal n}\left[{\bar u}(p_1)\ , \gamma_-\gamma_+ \gamma_x\ , v(p_2 ) \right]\phi(1-z',p ) \ , , \label{5eq1b}\end{aligned}\ ] ] where the answer has been expressed in terms of the wave function ( [ 2eq3d ] ) . the current and the dis cross section are evaluated in the c.m . frame , using the kinematics defined in appendix [ appen : e ] . the unpolarized cross section is obtained by squaring the current and summing over spins using @xmath292\nonumber\\ & = & 32 \,p_{1+}p_{2- } \ , , \end{aligned}\ ] ] and neglecting the interference term ( which vanishes in the large @xmath259 limit ) . observing that @xmath293 in the dis limit , and summing over all colors of the outgoing quarks ( which do not interfere ) gives @xmath294 where the momentum are evaluated in appendix [ appen : e ] . the hadronic tensor defined in eq . ( [ wmn ] ) is related to the square of the currents by @xmath295 the @xmath296 component of the tensor is @xcite is therefore @xmath297 this gives the familiar parton model result for the structure function @xmath3 @xmath298 the structure function @xmath262 , given in eq . ( [ w2 ] ) , is then @xmath299 where we anticipate the result @xmath300 as @xmath301 . this is the callan - gross relation . now look at the plus and minus components of the current . if the dirac space were restricted to two dimensions , only these components of the current would exist . the minus component can be obtained immediately from ( [ 5eq1b ] ) by replacing @xmath141 with @xmath126 , giving @xmath302\,\frac{2}{\cal n}\,\left\{e_1 \phi(z , p)-e_2 \phi(1-z',p)\right\ } \ , , \label{5eq1d}\end{aligned}\ ] ] the plus component could include the two contributions from the quark current , eq . ( [ current1 ] ) , but in the partonic picture the outgoing quark is not dressed , so the second part of the current proportional to @xmath303 should be ignored . however , in order to study the contribution of this term we will retain it for now . the leading contribution from both terms in the @xmath304 current is @xmath305 + \frac{g^2}{\pi}\frac{\phi(z , p ) } { p_{1-}^{(1)}(p_{1-}^{(1)}-q_-)}\,\left[{\bar u}(p_1)\,\gamma_-\ , v(p_2)\right ] \right\}\nonumber\\ & & + \frac{e_2}{\cal n}\left\{m_{02 } \frac{\phi(1-z',p)}{p_{2-}^{(2)}-q_-}\,\left[{\bar u}(p_1)\ , \gamma_-\,\gamma_+\ , v(p_2 ) \right ] + \frac{g^2}{\pi } \frac{\phi(z , p ) } { p_{2-}^{(2)}(p_{2-}^{(2)}-q_-)}\ , \left[{\bar u}(p_1)\,\gamma_-\ , v(p_2)\right ] \right\ } .\qquad\quad \label{5eq1e}\end{aligned}\ ] ] the @xmath273 component of the hadron tensor is then @xmath306 using ( [ 5eq1d ] ) and ( [ 5eq1e ] ) , and summing over colors , the @xmath307 contribution to the current sum becomes , at large @xmath259 , @xmath308\left(m_{01}^2-\frac{g^2}{\pi}\right ) \nonumber\\ & = & 8\pi\,e_1 ^ 2\,m_{1}^2 \left(\frac{1-x}{x}\right ) \;\phi^2(x , p)\ , . \label{5eq1g}\end{aligned}\ ] ] hence @xmath309 and this term is subleading at high @xmath259 . note that the effect of neglecting the term proportional to @xmath303 in the @xmath40 component of the quark current ( as we are instructed to do in the partonic picture ) is to replace @xmath310 by the bare quark mass @xmath311 . if @xmath312 , then @xmath313 scales as @xmath314 up to a factor , this is the result originally obtained by einhorn @xcite . in that paper , the dirac space was restricted to 2 dimensions , @xmath3 was necessarily zero , and @xmath277 scaled as ( [ einhorn ] ) . this distribution amplitude does not have the correct physical interpretation . in particular , it predicts that the dis cross section depends on the square of bare quark mass , a clearly unphysical result . only by considering the full four dimensional dirac space are we able to obtain scaling for @xmath276 , and the familiar physical result ( [ 5eq2c ] ) for @xmath315 . another consequence of the fact that @xmath273 is subleading is that we can avoid coming to grips with the fact that the current components @xmath316 and @xmath304 , as defined above , do not conserve current . since these components do not contribute to the final result , we need not discuss how they can be redefined in order to conserve current . the _ exact _ hadronic tensor ( [ wmn ] ) for electroproduction of a _ single _ final state @xmath317 with mass @xmath318 ( denoted by @xmath115 ) from the ground state @xmath9 with mass @xmath319 ( denoted by 0 ) is @xmath320 this is a delta function . in any physical measurement , the detectors will accept a finite range of values of the final electron energy , @xmath321 , requiring that this theoretical cross section be averaged over @xmath321 . this averaging process will be discussed in subsection [ sec : duality ] below , where the deta function will be written as a delta function in the bjorken variable @xmath136 , which spikes at values of @xmath322 corresponding to the excitation of the final state @xmath115 . in the c.m . system , this leads to the following _ exact _ transformation @xmath323 with @xmath322 the value of @xmath136 at the bound state mass @xmath318 , given in eq . ( [ xf ] ) below . in the dis limit , the transverse currents dominate the scattering . the exact result for the transverse tensor , @xmath296 , that describes the scattering from the ground state with @xmath126 structure , to a _ single _ transverse final state @xmath115 with @xmath139 structure , is @xmath324 where the reduced transverse current was defined in eq.([tranff ] ) . ( here the @xmath325 subscript is replaced by @xmath115 , and denotes a transition between the ground and @xmath115th state . ) next we observe that the value of @xmath136 at which the final state is excited is @xmath326 and that the spacing between neighboring levels is @xmath327 with @xmath328 . hence the _ exact _ result for the total inelastic tensor @xmath296 is @xmath329 where the sum is over all possible final states @xmath115 . the dis limit is defined by the requirement that @xmath301 with @xmath136 held constant . in this limit both @xmath259 and @xmath153 become large , with @xmath330 study of the dis limit therefore requires estimating the transition form factors for large final state number @xmath115 , but for all @xmath136 ( since @xmath322 depends on @xmath259 ) . in the dis limit , @xmath331 [ easily seen from the expansions ( [ 5eq5 ] ) ] , so @xmath137 will be replaced by @xmath136 in the following discussion . when @xmath115 is large the `` tails '' of the wave functions are very small , and the exact transition current ( [ tranff ] ) can be approximated by neglecting the contribution from the region @xmath332 $ ] ( where @xmath333 ) , giving @xmath334,p ) } { \xi'[\xi'(1-x)+x]}\ , f_{1}^{\rm eff}(\eta , q^2 ) \ , , \qquad\label{translim}\end{aligned}\ ] ] where we continue to suppress the label @xmath282 on the ground ( initial ) state wave function , and @xmath335 in the region of integration . when @xmath336 and @xmath301 , the relation ( [ ff3 ] ) may be used to approximate the quark form factor by @xmath337 giving @xmath338 equation ( [ translim ] ) is now further reduced by expanding @xmath339 around @xmath340 @xmath341 where the higher order terms are proportional to powers of @xmath342 @xmath343 this replacement gives the following series for @xmath344 @xmath345 where @xmath346 using the identity ( [ idet4 ] ) and the other estimates given in appendix [ appen : a2 ] it follows that @xmath347 hence the large @xmath115 approximation to ( [ translimseries ] ) is @xmath348 the plus component of the transition current is smaller than the transverse current by a factor of @xmath349 [ compare the two results in eq . ( [ currentd ] ) ] , and hence is obtained immediately from the result ( [ translim2 ] ) . using eq . ( [ 5eq5 ] ) for @xmath350 , and eq . ( [ xf2 ] ) to replace @xmath351 , gives @xmath352 the minus component can also be obtained from eq . ( [ translim2 ] ) . reviewing the derivation , we see that the minus component , in the dis limit , is obtained from the transverse component by replacing @xmath353 hence , multiplying ( [ translim2 ] ) by @xmath354 @xmath355 where the @xmath356 integral was evaluated using identity ( [ b10 ] ) . the transition currents can be shown to be gauge invariant by useing the ward - takahashi identity on the quark - photon vertex , and reducing the result using the bethe - salpeter equation . these currents will now be used to determine the structure functions @xmath3 and @xmath262 in the hadron picture . the cross section in the dis limit can now be obtained from the exact formula ( [ tensorxx2 ] ) . as long as @xmath136 is not too close to unity each term in the sum must correspond to some state with large @xmath115 ( the states with small @xmath115 `` pile up '' near @xmath246 where their total contribution is very small ) , and the approximate result for the transverse current , ( [ translim2 ] ) , may be used . furthermore , for large @xmath115 , @xmath357 may be approximated by using ( [ b11 ] ) , and @xmath358 . hence the final hadronic result [ denoted @xmath359 to distinguish it from the partonic result @xmath317 of eq . ( [ 5eq2c ] ) ] in the dis limit is @xmath360 note that @xmath361 is a series of separate spikes , zero for all @xmath136 except at particular values @xmath322 , where it is infinite . the partonic function , @xmath269 , is a smooth function of @xmath136 . to compare @xmath361 to @xmath269 , we must average over @xmath136 . this will lead to the concept of duality and will be discussed in the next subsection below . now look at the @xmath273 component of the tensor . replacing the transverse components of the current in eq . ( [ tensorxx2 ] ) with plus and minus components , and taking the dis limit , gives @xmath362 substituting eqs . ( [ translimplus ] ) and ( [ translimminus ] ) , and droping the term proportional to @xmath363 ( which averages to zero ) gives @xmath364 this is a nonleading term and can be neglected in the dis limit . hence the hadronic picture also gives the callan - gross relation , and it is sufficient to compare the functions @xmath233 and @xmath365 only . if @xmath3 were zero , eq . ( [ wpm2 ] ) would give the following scaling relation for @xmath313 @xmath366 this is to be compared with ( [ einhorn ] ) . hence , the duality of @xmath3 and the duality of the nonleading terms in @xmath313 depends on comparison of the same functions , @xmath315 and @xmath367 . we discuss this comparison now . in any physical measurement , the detectors which define the final state will accept a range of final electron energies @xmath368 . for fixed @xmath259 , this can be converted into the acceptance of a range of values of @xmath136 centered at @xmath369 with width @xmath370 so that @xmath136 varies over the interval bounded by @xmath371 . then the _ experimentally measured _ hadronic structure function in the dis limit can be computed from ( [ f1 ] ) @xmath372 note that this is the sum of smooth terms proportional to @xmath307 and @xmath373 , and a rapidly oscillating interference term proportional to @xmath374 . in the dis limit , the separation @xmath375 between states approaches zero , and therefore a large number of states are necessarily included in any interval @xmath370 . the interference term , which changes sign as @xmath363 , therefore averages to zero . this cancellation of the interference term was pointed out originally by einhorn @xcite , and empahsized recently by close and isgur in the context of the nonrelativistic quark model @xcite . the smooth terms can be approximated by their value at the center of the interval . using the fact that the number of states in the interval times the @xmath136 spacing between them must necessarily equal the width @xmath370 , so that @xmath376 we can reduce ( [ f1hadronb ] ) to @xmath377 where @xmath315 was defined in eq . ( [ 5eq2c ] ) . this result should be compared with the partonic result , averaged in the same way . since the partonic result is smooth ( and the interference term already neglected ) the partonic average is trival @xmath378 hence we see that @xmath379 proving duality . note that a similar identity holds for the @xmath115th moments of @xmath317 . in the high @xmath259 limit when the final states are numerous and close enough together so that @xmath380 , the @xmath81th moment of the hadronic structure function is @xmath381 \nonumber\\ & \to & \int_0 ^ 1 dx\ ; x^{m}\;\left[e^2_1\,\phi^2(x , p ) + e^2_2\,\phi^2(1-x\ , , p ) \right]\nonumber\\ & = & \int_0 ^ 1dx\;x^m f(x ) \ , . \label{f1a}\end{aligned}\ ] ] quark - hadron duality means that properly averaged hadronic observables in a certain kinematic regime ( high @xmath259 ) can be described by perturbative qcd , as schematically represented in fig . [ hadronic ] and demonstrated in eq.([f1hadron ] ) . recent data from jefferson laboratory exhibit duality , and this has focused attention on this old subject . recent studies show that the weak decays within the t hooft model exhibit duality @xcite , and in the framework of dis , close and isgur @xcite studied its emergence in the context of realistic nonrelativistic quark models . a recent paper by isgur , jeschonnek , melnitchouk and van orden @xcite studied duality using a relativistic klein - gordon equation with a confining interaction . this work is complementary to these other studies . while we have proved that duality emerges in the dis limit , and have corrected the pathologies of earlier work , we have not yet studied how duality emerges as a function of @xmath259 . a future paper will present numerical studies of the onset of duality for rqcd @xcite . we have shown that a very satisfactory description of deep inelastic scattering can be obtained from large @xmath1 qcd in 1 + 1 dimensions , _ provided _ the quark spin degrees of freedom are extended to the full 1 + 3 dimensions ( along with the leptons and the electromagnetic field ) . we refer to this new model as reduced qcd , and suggest that rqcd may be used to provide insight into the behavior of many other physical processes . in the applications to dis discussed in this paper , we are able to confirm the callan - gross relation , and that duality emerges in the deep inelastic limit . our theoretical discussion teaches several lessons . electromagnetic gauge invariance and duality emerge only if the quark current is fully and consistently dressed . the way in which a quark form factor emerges from this dynamical model mimics the phenomenlogical approach of gross and riska @xcite , showing that their prescription has a dynamical justification . our treatment of confinement removes all singularities , giving a finite mass for the dressed quark . this mass is gauge dependent , showing that it is not a physical quantity , which can be taken as a demonstration of confinement . but the fact that we have finte dressed quark masses in the presence of confinement provides justification for the constituent quark model . the simultaneous presence of confinement and finite quark mass is possible only if the `` on - shell '' quark scattering amplitude is identicallyt zero , and we have shown that this is indeed true , as modeled previously in ref . @xcite . the treatment of reduced qcd presented here lays a foundation for much further study . this work has been supported in part by the doe through grant no . de - fg05 - 88er40435 . the southeastrn universities research association ( sura ) operates the thomas jefferson national accelerator facility under doe contract de - aco5 - 84er40150 . one of us ( z. b. ) has been supported by a predoctoral fellowship at the college of william & mary and by the funda~ ao para a ci^ encia e a tecnologia under the post doctoral grant sfrh / bpd/5661/2001 the project grant cfif - plurianual . integrals over the variable @xmath69 like the one encountered in eq . ( [ bse2 ] ) appear several places in this paper . here we evaluate it in full generality . consider the integral @xmath382 \left[m_2 ^ 2-(k - r)^2 -i\epsilon\right]}\ , . \label{int1}\end{aligned}\ ] ] substituting for @xmath179 gives @xmath383 where the denominator @xmath183 is @xmath384 \left[\frac{m_2 ^ 2}{(k - r)_-}-2(k - r)_+ -i\,\frac{\epsilon}{(k - r)_-}\right]\nonumber\\ & = & \left[\frac{m_1 ^ 2}{y\,r_-}-2k_+ -i\,\frac{\epsilon}{y\,r_-}\right ] \left[-\frac{m_2 ^ 2}{(1-y)\,r_-}-2(k - r)_+ + i\,\frac{\epsilon}{(1-y)\,r_-}\right]\ , , \qquad \label{d2}\end{aligned}\ ] ] and we have introduced the momentum fractions ( [ defz ] ) . this shows that the denominator has only two poles in @xmath69 , and that they will both be in the same half of the complex plane ( giving zero for the integral ) unless @xmath385 doing the integral over @xmath69 , and expressing the answer in terms of the momentum fractions @xmath137 and @xmath32 gives @xmath386 where the new denominator @xmath387 is @xmath388 \ , .\label{d1}\ ] ] [ note that this result holds for _ both _ signs of @xmath389 . this is because the sign of the countor integral , which depends on the sign of @xmath389 , is cancelled by the sign of the integral over @xmath390 , which must be changed from @xmath391 when @xmath392 . ] it is important to note that the integral over @xmath137 is not restricted in the term in ( [ 2eqaa ] ) that is proportional to @xmath393 because for that term the restriction ( [ r1 ] ) applies to @xmath32 and not @xmath137 . this term not only regulates the singularity in the first term ( which arises only if @xmath394 ) , but also contibutes an additional contribution . furthermore , if @xmath32 lies outside of the interval @xmath394 , the first term is non singular ( except at the end points of the interval ) and is nonzero ! hence , in the region @xmath394 the integral ( [ 2eqaa ] ) becomes @xmath395\right)\ , , \qquad\quad \label{2eqab}\end{aligned}\ ] ] and outside of this region it is @xmath396\right)\ , , \qquad \label{2eqac}\end{aligned}\ ] ] where we have defined @xmath397$\cr { \rm v}_0(z , y)\equiv { \displaystyle\frac{g^2}{\pi } \frac{1}{(z - y)^2 } } & if $ z\notin[0,1]$ } \ , .\qquad \label{eq251a}\end{aligned}\ ] ] the multiplication by @xmath398 in ( [ eq251a ] ) is needed to scale the momenta from @xmath399 , etc . we emphasize that the identities ( [ 2eqab ] ) and ( [ 2eqac ] ) hold for both @xmath400 and @xmath392 . the first case is needed for the bound state equation and the second for electron scattering . in this appendix we derive a number of identities that hold when @xmath123 ( i.e. in all cases but the chiral limit ) . equation ( [ 2eqee ] ) leads immediately to the following identity @xmath401 integrating by parts gives @xmath402 an additional identity is derived by multiplying ( [ 2eqee ] ) by @xmath403 and integrating . first observe that @xmath404 hence , integrating by parts and using the normalization condition ( [ norm ] ) for the wave function , we obtain @xmath405 similarly we can prove that @xmath406 additional identities are needed for the study of duality . first , integrate eq . ( [ phiin ] ) over @xmath32 to obtain @xmath407 next , the commutator of the operator @xmath408 with the hamiltonian defined in ( [ phiin ] ) is @xmath409\,\phi_n(x , r)=\int_0 ^ 1dy \left [ \frac{m^2_{01}}{xy } -\frac{m^2_{02}}{(1-x)(1-y)}\right]\phi_n(y , r ) \ , , \label{com1}\end{aligned}\ ] ] as discussed in refs . @xcite . multiplying this by @xmath410 and integtating over @xmath136 gives zero on the l.h.s . , and hence the relation @xmath411 taking the square root of both sides of this equation gives @xmath412 where the sign is positive if the phase of @xmath413 is the same as @xmath414 and negative if it is opposite . since the states are non - degenerate , the phase of the @xmath115th eigenstate is @xmath415 . combining this with eq . ( [ b6 ] ) gives @xmath416 c_n\ , , \label{b10}\end{aligned}\ ] ] where the constant @xmath417 is yet to be determined . in this paper we have chosen the sign of @xmath112 so that it is always positive as @xmath418 , so that @xmath419 for all @xmath115 . our demonstration of duality depends upon the relations @xmath420 which can be derived approximately using arguments given by t hooft @xcite , and callan , coote , and gross @xcite , given here for completeness . first , for large @xmath115 , the boundary conditions require that the normalized wave function go like @xmath421 one test of this approximation is shown in fig . [ fig : idet1 ] . if @xmath115 is very large , so that @xmath422 , the behavior of the wave function at the endpoints can be ignored , and near @xmath423 eq . ( [ 2eqee ] ) can be approximated @xmath424 substituting the ansatz ( [ ansatz ] ) into this equation gives @xmath425 hence @xmath426 as @xmath427 and identity ( [ b11 ] ) is proved . to prove ( [ identb ] ) we consider the matrix element of ( [ com1 ] ) @xmath428\,\phi_n(x , r)&= & 2c_{n+1}c_n \nonumber\\ & = & \left(m_{n+1}^2-m_n^2\right)\dashint_0 ^ 1 dxdy \frac{\phi_{n+1}(x , r)\phi_n(y , r)}{y - x}\ , .\end{aligned}\ ] ] where we used the fact that the phases of @xmath429 and @xmath430 are opposite . the integral is evaluated in the large @xmath115 approximation @xmath431\pi x)\sin(n\pi y)}{y - x - i\epsilon } \nonumber\\ & = & 2 % \frac{2\pi}{n_c } \int_0 ^ 1 dx \sin([n+1]\pi x)\cos(n\pi x)\nonumber\\ & = & 2 % \frac{2}{n_c } \left(1 + \frac{1}{2n+1}\right ) \to 2 % \frac{2}{n_c } \ , .\end{aligned}\ ] ] hence , @xmath432 and ( [ identb ] ) is proved . a numerical demonstration of the identities ( [ idet4 ] ) , ( [ identb ] ) , and ( [ b11 ] ) is given in figs.[fig : idet1 ] . the identities are satisfied to better than a few percent for state numbers between 40 and 80 , provided the number of splines is at least 160 ( twice the number of the maximum state of interest ) . two additional relations that follow from the completeness relation ( [ complete ] ) and the bound state equation are needed . if @xmath433 $ ] , then : @xmath434 the proof of the second relation ( [ id2 ] ) follows from application of the bound state equation @xmath435\int_0 ^ 1dz'\ , \phi_n(z',r_n)\nonumber\\ & & \qquad=\sum_n\left\{\int_0^{z-\epsilon } + \int_{z-\epsilon}^{z+\epsilon}+\int_{z+\epsilon}^1\right\ } dy\ , \left[\frac{\phi_n(y , r_n)-\phi_n(z , r_n ) } { ( y - z)^2 } \right]\int_0 ^ 1dz'\ , \phi_n(z',r_n)\nonumber\\ & & \qquad\to\sum_n\frac{1}{2}\int_{z-\epsilon}^{z+\epsilon}dy\ ; \phi_n''(z , r_n)\int_0 ^ 1dz'\,\phi_n(z',r_n ) = \nonumber\\ & & \qquad= \epsilon\,\sum_n \phi_n''(z , r_n)\int_0 ^ 1dz'\,\phi_n(z',r_n)\to0 \ , , \end{aligned}\ ] ] where the identity ( [ id1 ] ) is used to eliminate the nonsingular integrals over the intervals @xmath436 $ ] and @xmath437 $ ] , and the singular part of the integral is proportional to @xmath438 which vanishes as @xmath439 . if eq . ( [ phiout ] ) is used , modifications of the completeness relation ( [ complete ] ) and the identity ( [ id1 ] ) , which hold in the region @xmath440 $ ] , can be derived . the results are @xmath441 \label{complete2}\\ & & % n_c \sum_n\left(\frac{m_1 ^ 2}{z}+\frac{m^2_2}{1-z}- m_n^2\right)\phi_n(z , r_n)\int_0 ^ 1dz'\ , \phi_n(z',r_n)= -\frac{g^2/\pi}{z(1-z ) } \label{id3 } \ , .\end{aligned}\ ] ] note that the completeness relation and all of these identities hold for _ any _ two quark hamiltonian . in particular , the hamiltonian may describe bound states with flavor ( where , in general , @xmath442 ) as well as flavorless bound states where @xmath443 . in some applications we will study the coupling of a photon to a @xmath4 pair . such a coupling necessarily involves _ flavorless _ states only , and we will use the identities for the flavorless states . here we show how to sum this series ( [ eq249 ] ) and obtain a useful form for @xmath184 . since @xmath179 is independent of the plus components of momentum , the integrations over the plus components of momentum can be carried out . using the identity ( [ 2eqab ] ) [ with @xmath198 in this case ] , the first term on the rhs of ( [ eq249 ] ) becomes @xmath444\ , , \quad\quad\quad\label{first}\end{aligned}\ ] ] or @xmath445\ , .\quad\quad\quad\label{firsta}\end{aligned}\ ] ] here the replacement ( [ eq251a ] ) has been used to convert @xmath446 ( or v@xmath447 ) . similarly , the second term becomes @xmath448 \ , , \quad\quad\quad \label{second}\end{aligned}\ ] ] or @xmath449 \ , , \quad\quad\quad \label{seconda}\end{aligned}\ ] ] and the third term follows the same pattern . replacing @xmath189 and @xmath190 , the series ( [ eq249 ] ) , in the interval @xmath450 $ ] , is clearly summed by the following equation @xmath451 \ , .\qquad\qquad \label{eq253}\end{aligned}\ ] ] introducing the shorthand notation @xmath452 eq . ( [ eq253 ] ) can be rearranged into the following form @xmath453 which requires that @xmath454 satisfy the equation @xmath455 this equation is easily solved with the use of the greens function : @xmath456 putting it all together gives eq . ( [ eq267 ] ) . if @xmath457 $ ] , the series reduces to the same formal result @xmath458 where the completeness relation ( [ complete ] ) and eq . ( [ phiout ] ) were used in the last step . this shows that ( [ eq267 ] ) can be applied for _ any _ @xmath459 $ ] . the series ( [ current1a ] ) for the dressed quark current is evaluated in this appendix . the `` first '' integral over @xmath69 in each term in the series is evaluated as in eq . ( [ first ] ) or ( [ firsta ] ) @xmath460$\cr { \displaystyle\int_0 ^ 1dy\,{\rm v}_0(z',y)\,\frac{n_i^\mu(k_-,q_-)}{\delta_i(y , q^2 ) } } & if $ z'\notin[0,1]\ , , \qquad\quad$ } \label{current1b}\end{aligned}\ ] ] where @xmath461\,\gamma^\mu\,\left[m_{0i } + ( k_--q_-)\gamma_+\right]\,\gamma_-}{4\,k_-(k_--q_-)}\ , , \label{current1bb}\end{aligned}\ ] ] and @xmath462 and @xmath463 are ( [ d2 ] ) and ( [ d1 ] ) with equal masses @xmath464 . note that the only dependence of the integrand on @xmath69 was in the denominator @xmath462 , so the @xmath69 integral could be evaluated using the methods of appendix [ appen : b ] . the second integral is similarily evaluated . if @xmath450 $ ] , the following result is obtained for the series eq . ( [ current1a ] ) @xmath465 where @xmath466 , and the series is summed using eq . ( [ eq253 ] ) for the scattering matrix m. using the solution ( [ eq267 ] ) for the scattering matrix the result , for all values of @xmath215 , is @xmath467 where @xmath468 , and @xmath463 is the @xmath469 of eq . ( [ d1 ] ) with equal masses @xmath464 . the same expression also holds for @xmath457 $ ] , but the derivation differs . as before , the full result can be deduced from the form of the first two terms of the series ( [ current1a ] ) . guided by ( [ current1b ] ) , and using the steps already displayed in eq . ( [ eq267a ] ) , we obtain @xmath470 we see , as we did with eq . ( [ eq267 ] ) , that ( [ current2a ] ) holds for _ all _ @xmath289 . equation ( [ current2a ] ) expresses the dressed current as a sum over contributions from the bound states that couple to the photon , showing that vector dominance is a rigorous consequence of this model . the structure of the numerator n@xmath471 depends on the bare current : @xmath472 note the interesting structure of the correction term to the bare current @xmath141 , which will be of central importance in our discussion below . substituting these results into ( [ current2a ] ) gives the following @xmath473 we organize these expressions by introducing the quark form factor , defined in eq . ( [ ff1 ] ) . this immediately gives eq.([ff3aa ] ) for the @xmath212 current . the other currents in ( [ current5 ] ) can also be written in terms of @xmath219 . to this end note that the bound state equation for equal mass quarks implies that @xmath474 hence the quark mass can be removed from the second term in @xmath475 , giving @xmath476 where @xmath477 defined by this equation should not be confused with the two body greens function or the bound state vertex function . if @xmath450 $ ] , the second term can be reduced using the identities ( [ id1 ] ) and ( [ id2 ] ) @xmath478 \phi_n(z',r_n)\ ; \int_0 ^ 1dy\,\phi_n(y , r_n ) = -\frac{g^2}{\pi}\frac{1}{z'(1-z ' ) } \ , .\qquad\quad \label{ff3}\end{aligned}\ ] ] the same result is obtained directly from identity ( [ id3 ] ) if @xmath457 $ ] . hence , for all @xmath289 , @xmath479 finally , the @xmath480 component of the current can be similarly reduced . the second term in @xmath480 is proportional to @xmath481 the part of the second term in @xmath482 is the same constant derived above , and combining the results gives eq . ( [ ff6 ] ) . kinematics for dis in the c.m . frame are given in this appendix . when @xmath259 is very large , the components of @xmath224 and @xmath174 in this frame are ( see ref . @xcite ) @xmath483 \nonumber\\ q_-=&&-q\sqrt{\frac{x}{2(1-x ) } } \left[1+\frac{m_0 ^ 2 x ( 1 - 2x)}{2q^2(1-x)}+ ... \right ] \nonumber\\ p_+=&&\frac{m_0 ^ 2}{q}\sqrt{\frac{x(1-x)}{2 } } \left[1-\frac{m_0 ^ 2x(1 - 2x^2)}{2q^2(1-x)}+ ... \right ] \nonumber\\ p_-=&&\frac{q}{\sqrt{2x(1-x ) } } \left[1+\frac{m_0 ^ 2x(1 - 2x^2)}{2q^2(1-x)}+ ... \right]\ , . \label{5eq5}\end{aligned}\ ] ] the energy and momentum of particles 1 and 2 in the final state are @xmath484 \nonumber\\ p_{10}&=&\sqrt{m_1 ^ 2+p_z^2}= \frac{q}{2}\sqrt{\frac{1-x}{x } } \left[1+\frac{x(m_0 ^ 2 + 2m_1 ^ 2 - 2m_2 ^ 2)}{2q^2(1-x)}+\cdots\right ] \nonumber\\ p_{20}&=&\sqrt{m_2 ^ 2+p_z^2}= \frac{q}{2}\sqrt{\frac{1-x}{x } } \left[1+\frac{x(m_0 ^ 2 - 2m_1 ^ 2 + 2m_2 ^ 2)}{2q^2(1-x)}+\cdots\right ] \ , .\end{aligned}\ ] ] there are two possibilities corresponding to the two terms in the current ( [ 5eq1b ] ) . the first term in the current will be large only if the momentum of particle 1 is in the direction of @xmath485 ( i.e. @xmath486 ) . [ choosing it in the opposite direction requires a very large momentum flow through the wave function , and is suppressed . ] in this case @xmath487 \nonumber\\ p^{(1)}_{1-}&=&\sqrt{\frac{x}{2(1-x)}}\ ; \frac{m_1 ^ 2}{q } \left[1-\frac{x(m_0 ^ 2 - 2m_2 ^ 2)}{2q^2(1-x)}+\cdots\right ] \nonumber\\ p^{(1)}_{2+}&=&\sqrt{\frac{x}{2(1-x)}}\ ; \frac{m_2 ^ 2}{q } \left[1-\frac{x(m_0 ^ 2 - 2m_1 ^ 2)}{2q^2(1-x)}+\cdots\right ] \nonumber\\ p^{(1)}_{2-}&=&q\sqrt{\frac{1-x}{2x } } \left[1+\frac{x(m_0 ^ 2 - 2m_1 ^ 2)}{2q^2(1-x)}+\cdots\right ] \ , , \label{5eq5a}\end{aligned}\ ] ] where the superscript @xmath488 is a reminder that these relations hold only for the _ first _ term in the current . the momentum relations for the _ second _ term in the current ( [ 5eq1b ] ) follow by interchanging 1 and 2 on both sides of these equations . we see that the first term in the current gives @xmath489 and @xmath490 , while the second term gives @xmath491 and @xmath492 . the two terms describe kinematically distinct regions of phase space , and their interference can safely be neglected when the total cross section is computed . using the definitions ( [ defza ] ) and ( [ defzp ] ) , @xmath32 and @xmath289 can be easily related to the bjorken variable @xmath136 . we have @xmath493 this appendix includes details of the evaluation of the transition currents shown in fig . [ hadronicamplitude ] . the @xmath88 term will be evaluated first . the @xmath90 term can then be obtained by a simple substitution , as discussed at the end of this appendix . the @xmath88 term , given in eq . ( [ 3eq1- ] ) , is reduced by introducing the momentum fractions @xmath494 so that @xmath495 in this notation the @xmath136-type transverse quark current , evaluated in eq . ( [ ff6 ] ) , is @xmath496 first evaluate the numerators , @xmath228 , of the traces in ( [ 3eq1- ] ) . the numerator @xmath497 is @xmath498\nonumber\\ & = & 16\,k_- \,\left[m_{01}\,q_- + 2(p_f+k)_-(p+k)_- h_1(\eta , q)\right]\nonumber\\ & = & 16\,k_-q_- \frac{m_1 ^ 2}{m_{01}}-32(p_f+k)_-(p+k)_-\frac{q_+}{m_{01}}\ , f_1(\eta , q^2)\nonumber\\ & = & \frac{16k_-q_-}{m_{01}}\left(m_1 ^ 2- \eta(1-\eta)\,q^2 \,f_1(\eta , q^2)\right)\ , , \label{nx}\end{aligned}\ ] ] where the trace has been evaluated in a dirac space of four dimensions . the numerator of the minus component of the current has only one term @xmath499\,\left[1+f_1(\eta , q^2)\right]\nonumber\\ & = & 32\,k_- ( p_f+k)_-(p+k)_-\,\left[1+f_1(\eta , q^2)\right ] \nonumber\\ & = & 16k_-\,\frac{q_-}{q_+}\,\eta(1-\eta)\,q^2\ , \left[1+f_1(\eta , q^2)\right]\ , . \label{minus}\end{aligned}\ ] ] the trace for the plus component is @xmath500\nonumber\\ & = & 16k_-m_1 ^ 2 - 32k_-(p_f+k)_-(p+k)_-\,\frac{q_+}{q_-}\ , f_1(\eta , q^2 ) \nonumber\\ & = & 16k_-\left(m_1 ^ 2-\eta(1-\eta)\,q^2\ , f_1(\eta , q^2)\right)\ , .\label{plusc}\end{aligned}\ ] ] inserting the numerator ( [ minus ] ) into the general result ( [ 3eq1- ] ) gives the following result for the minus component of the current @xmath501[m_1 ^ 2-(p+k)^2-i\epsilon ] [ m_2 ^ 2-k^2-i\epsilon]}\,\left[1+f_1(\eta , q^2)\right ] \nonumber\\ & & \!\!\!\!=32i e_1\int \frac{d^2k}{(2 \pi)^2 } \;\frac{1}{d_3 } g_f\left(-p_f - k ,- k\right)\,g_i\left(p+k , k\right ) \,\left[1+f_1(\eta , q^2)\right ] \ , , \label{3eq1}\end{aligned}\ ] ] where the subscript - on @xmath232 reminds us that the final state must have a pure @xmath126 structure ( if the ground state does ) , @xmath502 is the product of the denominators of the three quark propagators , with three poles in @xmath69 @xmath503 \left[\frac{m_1 ^ 2}{\xi\,p_-}-\frac{m_i^2}{p_-}-2k_+-i \epsilon_b\right ] % \nonumber\\&&\times \left[\frac{m_2 ^ 2}{(\xi-1)\,p_-}-2k_+-i\epsilon_c\right]\ , , \qquad \label{3poles}\end{aligned}\ ] ] and the @xmath438 s change sign according to @xmath504\nonumber\\ \epsilon_b&=&\epsilon/(p+k)_-=\epsilon/(\xi\,p_- ) \nonumber\\ \epsilon_c&=&\epsilon / k_-=\epsilon/[(\xi-1)\,p_- ] \ , . \label{epsilon}\end{aligned}\ ] ] since the vertex functions @xmath128 do not depend on @xmath69 , we can evaluate the @xmath69 integral . it will be nonzero only when the three poles of ( [ 3poles ] ) do not all lie in the same half plane . since @xmath198 for electron scattering , @xmath505 , which implies that @xmath506 hence @xmath507 and all three poles will be in the same half of the complex plane ( giving zero for the integral ) unless @xmath508 there are two terms , depending on the sign of @xmath509 . closing the @xmath69 contour in the upper half plane , and using @xmath510 and @xmath511 gives @xmath512 \nonumber\\ & & -16 e_1\,p_-^2p_{f- } \int_{0}^{y}\frac{d\xi}{(2\pi ) } \;\frac{g_f\left(\xi',p_f\right)\,g_i\left(\xi , p\right ) \left[1+f_1(\eta , q^2)\right ] } { \delta(\xi',p_f^2)\,\left(\delta(\xi , p^2)-\displaystyle { \frac{\delta(\xi',p_f^2)}{(1-y ) } } \right ) } \ , , \qquad\quad \label{3eq1a}\end{aligned}\ ] ] where the first term is the contribution from the pole due to the zero in the third term in ( [ 3poles ] ) above , and the second from the pole due to the zero in the first term . using the definition ( [ 2eq3d ] ) of the wave function and combining the two contributions in the region @xmath513 $ ] we get eq . ( [ 3eq1b ] ) with @xmath514 } { ( 1-\xi')\left\{\xi\,\xi'\left(m_f^2-(1-y)m_i^2 \right)-y\ , m_1 ^ 2\right\ } } \quad\to 1\quad{\rm if}\quad m_f^2\to \infty\ , . \label{rlimit}\end{aligned}\ ] ] this factor is needed for positive values of @xmath515 in the interval @xmath513 $ ] and _ negative _ values of @xmath356 in the interval @xmath516 $ ] . in this region the denominator has a zero only if @xmath517 . in order to avoid the discussion of such cases we limit numerical applications to cases with @xmath518 . results for the other components of the transition current , eq.([currentd ] ) , are obtained by a similar arguement using ( [ nx ] ) and ( [ plusc ] ) in place of ( [ minus ] ) . now consider the modifications required in order to evaluate the @xmath90 term . using the momenta defined in fig.[hadronicamplitude ] , and the same definitions of momentum fractions ( [ frac2 ] ) and ( [ frac3 ] ) , the @xmath90 term is obtained from the @xmath88 term simply by substituting @xmath519 and @xmath520 . however , the momentum fractions in the wave functions are , by convention , the fraction of the momentum carried by the quark @xmath87 , and the momentum fraction in the quark form factor is that of the the _ outgoing _ quark , and hence these functions must be written in terms of @xmath521 these observations lead immediately to the final result ( [ exacttran ] ) . the two body equations were solved using a modification of the standard cubic splines employed previously in many problems . the standard splines are defined on 4 segments of length @xmath522 , bounded by the 5 points @xmath523 : @xmath524 & if $ b < x < c$\cr 1 + 3\left[\frac{(d - x)}{h}+\frac{(d - x)^2}{h^2}-\frac{(d - x)^3}{h^3}\right ] & if $ c < x < d$\cr \frac{(e - x)^3}{h^3 } & if $ d < x < e$\ , .}\end{aligned}\ ] ] if the interval [ 0,1 ] is spanned by @xmath525 standard splines , there must be @xmath164 segments , of length @xmath526 . the first spline is numbered @xmath527 beginning at @xmath528 and the last is numbered @xmath529 ending at @xmath246 . the modified splines used in this paper consist of the standard splines plus one additional , non - standard spline inserted at the beginning and end of the interval [ 0,1 ] . these non - standard splines are defined on only three segments , and will be numbered 1 and @xmath155 . they are @xmath530 & if $ h < x<2h$\cr \frac{(3h - x)^3}{4h^3 } & if $ 2h < x<3h$ } \nonumber\\ s_{n_s}(x)&=&\cases{\frac{(x-1 + 3h)^3}{4h^3 } & if $ 1 - 3h < x<1 - 2h$\cr \frac{1}{4 } + \frac{3}{4}\left[\frac{(1-h - x)}{h}+\frac{(1-h - x)^2}{h^2 } -\frac{(1-h - x)^3}{h^3}\right ] & if $ 1 - 2h < x<1-h$\cr d_1\frac{(1-x)^{\beta_2}}{h^{\beta_2}}+d_2\frac{(1-x)^2}{h^2}+ d_3\frac{(1-x)^3}{h^3 } & if $ 1-h < x<1$\ , . } \end{aligned}\ ] ] where the @xmath531 are the fractional exponents determined by the boundary conditions [ given in eq . ( [ betas ] ) ] , and the coefficients @xmath532 depend on the exponents @xmath533 @xmath534 where @xmath535 for @xmath97 or @xmath536 for @xmath537 . these coefficients were fixed by the requirement that the spline and its first two derivatives be continuous . figure [ fig : spline ] shows two examples of the modified splines for the minimal number @xmath538 . in this case the central spline spans the full interval [ 0,1 ] and has 4 segments . 99 g. t hooft , nucl . * b75 * , 461 ( 1974 ) . c. g. callan , n. coote and d. j. gross , phys . d * 13 * , 1649 ( 1976 ) . m. einhorn , phys . d * 14 * , 3451 ( 1976 ) . m. i bars , m. b. green , phys . rev * d * 15 , ( 1978 ) . c. g. callan and d. j. gross , phys . * 22 * , 156 ( 1969 ) . m. burkardt , phys . d * 62 * , 094003 ( 2000 ) [ arxiv : hep - ph/0005209 ] . m. burkardt and n. uraltsev , phys . d * 63 * , 014004 ( 2001 ) [ arxiv : hep - ph/0005278 ] . v. schon and m. thies , contribution to the festschrift in honor of boris ioffe , m. shifman , ( ed ) . in _ at the frontier of particle physics , vol . 3 _ p. 1945 ( 2000 ) [ arxiv : hep - th/0008175 ] . m. burkardt , f. lenz and m. thies , phys . rev . d * 65 * , 125002 ( 2002 ) [ arxiv : hep - th/0201235 ] . y. s. kalashnikova and a. v. nefediev , phys . * 45 * , 347 ( 2002 ) [ usp . nauk * 172 * , 378 ( 2002 ) ] [ arxiv : hep - ph/0111225 ] . b. grinstein , r. f. lebed , phys . rev . d * 57*:1366 ( 1998 ) [ hep - ph/9708396 ] . r. f. lebed , talk given at 3rd international conference in quark confinement and hadron spectrum ( confinement iii ) , newport news , va , 7 - 12 jun 1998 ; published in _ newport news 1998 , quark confinement and the hadron spectrum iii _ , p.161 ( 1998 ) [ hep - ph/9808452 ] . r. f. lebed , n. g. uraltsev , phys . d * 62*:094011 ( 2000 ) [ hep - ph/0006346 ] . z. batiz and f. gross , phys . c * 58 * , 2963 ( 1998 ) . + note that the square bracket in the second of eqs . ( 5.8 ) should be @xmath539 $ ] instead of @xmath540 $ ] . + z. batiz and f. gross , in preparation . f. gross and j. milana , phys . d * 43*(1991 ) 2401 . + f. gross and j. milana , phys . d * 45 * ( 1992 ) 969 . + f. gross and j. milana , phys . d * 50 * ( 1994 ) 3332 . c. savkli and f. gross , phys . c * 63 * , 035208 ( 2001 ) . f. gross and d. o. riska , phys . c * 36 * , 1928 ( 1987 ) . f. e. close , n. isgur , phys . lett . b * 509 * , 81 ( 2001 ) [ hep - ph/0102067 ] . n. isgur , s. jeschonnek , w. melnitchouk and j. w. van orden , phys . d * 64*:054005 ( 2001 ) .

the scattering experiments on deep inelastic electron - proton ( dis ) at hera have provided measurements of the inclusive structure function @xmath0 as well as of the @xmath1 and @xmath2 in very small bjorken variable @xmath3 values ( @xmath4 ) . in these collisions the proton target is analysed through a hard probe with virtuality @xmath5 . the momentum fraction is @xmath6 , where @xmath7 and @xmath8 are the four - momenta of the incoming proton and of the virtual photon probe . in the kinematical region of small @xmath3 , the gluon is the leading parton driving the behavior of the deep inelastic observables . the standard qcd evolution @xcite gives a powerlike growth for the gluon distribution and related quantities , and this feature leads , in principle , to the unitarity violation at asymptotic energies , requiring a control of the gluon distribution in high energies . in the partonic language , in the infinite momentum frame , the small momentum fraction region corresponds to the high parton density domain , which is connected with the black disk limit of the proton target and with the parton recombination phenomenon . these issues can be addressed through a non - linear dynamics beyond the usual dglap formalism ( for a review , see @xcite ) . the complete knowledge about the non - linear dynamical regime plays an important role in the theoretical description of the reactions in the forthcoming experiments rhic and lhc . in the breit frame , the qcd factorization theorem allows to calculate the scattering processes through the convolution of the partonic subprocess with the parton distribution functions ( pdf s ) . there , the degrees of freedom of the theory are the quasi - free partons ( quarks and gluons ) . on the other hand , more recently the theoretical description of the small @xmath3 physics has been widely analyzed in the target rest frame , which is a powerful tool concerning an unified picture for both the inclusive and the diffractive scatterings , including vector meson production @xcite . now , the degrees of freedom are the color dipoles , which are the most simple configurations considering the virtual photon fock states expansion . its main appeal is a quite simplified picture for the different mentioned processes , based on general properties of quantum mechanics . the description of dis in the color dipole picture is quite intuitive , allowing a simple representation instead of the involved one from the breit ( infinite momentum ) frame , and such framework was proposed by gribov many years ago @xcite . considering small values of the bjorken variable @xmath3 , the virtual photon fluctuates into a @xmath9 pair ( dipole ) with fixed transverse separation @xmath10 at large distances upstream of the target and interacts in a short time with the proton . more complicated configurations should be considered for larger transverse size systems , for instance the photon fock state @xmath11 . an immediate consequence of the lifetime of the pair ( @xmath12 ) to be bigger than the interaction s one is the factorization between the photon wavefunction and the cross section dipole - proton in the @xmath13 total cross section . the wavefunctions are perturbatively calculable , namely through qed for the @xmath9 configuration @xcite and from qcd for @xmath14 @xcite . the effective dipole cross section should be modeled and encloses both perturbative and non - perturbative content . however , since the interaction strength relies only on the configuration of the interacting system the dipole cross section turns out to be universal and may be employed in a wide variety of small @xmath3 processes @xcite . currently , there are several models for the dipole cross section based either in pure phenomenological parametrizations or in a more theoretical ground ( for a review , see ref . the main feature in those models is the description of the energy dependence of the interaction , namely taking into account the interplay between hard and soft domains . the dipole cross section should be consistent with the sharp growth on energy at small transverse separation @xmath10 ( large gluon density ) and a softer behavior for larger @xmath10 ( regge - like phenomenology ) . an additional ingredient is the expectation for saturation effects in high energies as a consequence of the unitarity requirements . indeed , the growth of the gluon distribution should be tamed at very small @xmath3 and it has been found that the corrections are important already in the present hera kinematics @xcite . moreover , such effects are associated with higher twist contributions concerning the standard linear evolution equation , i.e. the dglap formalism @xcite . here , we take into account a formalism providing the unitarity corrections to the deep inelastic scattering at small @xmath3 , namely the glauber - mueller approach in qcd . it was introduced by a. mueller @xcite , who developed the glauber formalism to study saturation effects in the quark and gluon distributions in the nucleus considering the heavy onium scattering . afterwards , the authors of ref . @xcite extended that approach that has as a limit the glr results @xcite . it is obtained an evolution equation taking into account the unitarity corrections ( perturbative shadowing ) , generating a non - linear dynamics which is related with higher twist contributions . its main characteristic is to provide a theoretical framework for the saturation effects , relying on the multiscattering of the pqcd pomeron . the latter is represented through the usual gluonic ladder in the double logarithmic approximation . summarizing the glauber - mueller approach , the gluonic content of the nucleus or nucleons is obtained in the following way : in the rest frame , a virtual probe ( gluon ) decays into a gluon pair interacting with the nucleon inside the nucleus . the multiple scatterings of the pair give rise to the unitarization of the corresponding cross section . the calculations are performed in the double logarithmic approximation ( dla ) , corresponding to the condition @xmath15 , where @xmath16 and @xmath17 are the qcd coupling constant and the gluon anomalous dimension , respectively . in this approximation , the transverse separation of the pair remains fixed allowing an eikonal description for the interaction gluon pair - nucleon through incoherent multiscatterings @xcite . the cross section for the interaction can be expressed in terms of the nucleon gluon distribution @xmath18 and of the transverse separation @xmath10 of the gluon pair . the procedure for an initial state quark - antiquark pair is similar , up to proper color coefficients . such a formulation has produced comprehensive phenomenological applications : the inclusive structure function @xmath19 @xcite , the longitudinal @xmath1 and charmed @xmath2 ones @xcite , the logarithmic slope @xmath20 @xcite and other related quantities have been calculated in the breit system . the respective nuclear case has been investigated in refs . @xcite and the equivalence with other high density qcd approaches has been reported in refs . moreover , the asymptotic limit of the inclusive structure function and its logarithmic slope are estimated in @xcite . regarding the rest frame , glauber - mueller has been also used to estimate the saturation effects for dis and diffractive dissociation @xcite as well as it is considered as initial condition for a high energy evolution equation ( for a review , see @xcite ) . therefore , the glauber - mueller approach gives a good framework for the unitarity effects ( saturation ) in the nucleon and nuclear sectors , providing the dynamics of the observables in a quantitative level . in this work we make use of the parton saturation formalism to study the description of the observables driven by the gluonic content of the proton in the color dipole picture . the inclusive structure function @xmath19 is calculated properly , disregarding approximations commonly considered in previous calculations @xcite . the structure functions @xmath1 and @xmath2 are presented for the first time using the glauber - mueller approach and the rest frame in comparison with the experimental data . the saturation effects are included in the effective dipole cross section corresponding to the small size dipole contributions . the large dipole sizes are taken into account through an anstz for the non - perturbative region . here , we choose to freeze the gluon distribution at large distances @xcite . the dipole framework provides a clear identification in the transverse distance @xmath10 range where the perturbative and non - perturbative sectors ( soft domain ) contribute to the calculated quantities . large dipoles correspond to soft domain and the small ones are connected with the hard sector . as we will see , the photon wavefunctions play the role of a weight function selecting small dipole sizes , with a non - negligible contribution coming from the large dipole sizes . however , when considering the production of heavy quark pairs ( charm and bottom ) , their masses largely diminish the soft contribution . these issues are addressed throughout this paper . this work is organized as follows . in the next section we shortly review the deep inelastic scattering in the rest frame , introducing the main formulae for the further analysis . in sec . ( 3 ) one addresses the effective dipole cross section considering saturation effects ( unitarity corrections ) encoded in the glauber - mueller approach , pointing out its main properties and discussing the large transverse separation contribution . ( 4 ) is devoted to calculate the theoretical estimates for the hera observables , focusing on those ones dominated by the gluon content . in the last section we draw our conclusions and comments . in the reference frame where the target ( proton ) has infinite momentum the usual description of the dynamics evolution is the following : the emitted partons from the target remain quasi - free for enough time in such a way that the virtual probe ( photon ) detects them as real particles ( asymptotic freedom ) . on the other hand , in the small @xmath3 region it is suitable to describe the evolution in a system where the target is at rest : in this situation the evolution is related to the partonic fluctuations of the probe and their interactions with the nucleon . the rest frame physical picture is advantageous since the lifetimes of the photon fluctuation and of the interaction process are well defined @xcite . the simplest case is the quark - antiquark state ( color dipole ) , which is the leading configuration for small transverse size systems . its life time can be estimated by the uncertainty principle through the energy fluctuation associated with the emerging pair . the well known coherence length is expressed as @xmath21 , where @xmath3 is the bjorken variable and @xmath22 the proton mass . for instance , in deep inelastic at hera kinematics reaching at @xmath23 , the coherence length is about @xmath24 fm , which is a distance larger than the radius of any atomic nuclei . an immediate consequence of such a picture is the factorization between virtual photon wavefunctions and the interaction cross section on the corresponding amplitude in the impact parameter space representation . a striking consequence of the formulation above is that the photoabsortion cross section can be derived from the expectation value of the interaction cross section for the multiparticle fock states of the virtual photon weighted by the light - cone wave functions of these states @xcite . the scattering matrix has an exact diagonalization in the @xmath25 fock states representation , where the partial - wave amplitudes subjected to the @xmath26-channel unitarization are identified properly . then , the photoabsortion cross section can be cast into the quantum mechanical factorized form , @xmath27 the formulation above is valid even beyond perturbation theory , since it is determined from the space - time structure of the process . the @xmath28 are the photon wavefunctions ( for transverse , @xmath29 , and longitudinal , @xmath30 , polarizations ) describing the pair configuration , where @xmath31 and @xmath32 are the fraction of the photon s light - cone momentum carried by the quark and antiquark of the pair , respectively . the transverse separation of the pair is @xmath33 . the precise normalizations of the wavefunctions can be determined through the fock expansion @xmath34 , where @xmath35 is the @xmath36-wavefunction renormalization constant , @xmath37 denotes the bare state and @xmath38 is the coefficient determining the probability of the @xmath9 pair fluctuation in the photon @xcite . considering completely normalized states , then @xmath39 and the normalization for transverse photons is obtained from @xmath40 . the remaining normalizations , i.e. , longitudinal component and cross section , are consequently fixed @xcite . the explicit expressions are well known , @xmath41\:\varepsilon^2\,k_1 ^ 2\,(\varepsilon\ , r ) \,+\ , m_q^2\ : k_0 ^ 2\,(\varepsilon\ , r)\,\big\ } \label{eq9}\end{aligned}\ ] ] and @xmath42 clarifying the notation , the auxiliary variable @xmath43 , with @xmath44 the light quark mass , and @xmath45 and @xmath46 are the mc donald functions of rank zero and one , respectively . the quantity @xmath47 is interpreted as the cross section of the scattering of the effective dipole with fixed transverse separation @xmath48 @xcite . it is directly dependent on the unintegrated gluon distribution , @xmath49 the unintegrated gluon distribution , @xmath50 , vanishes at the gluon transverse momentum @xmath51 ( and similarly for the factor @xmath52 ) due to gauge invariance , and therefore the dipole cross section has to be infrared finite . the most important feature of the dipole cross section is its universal character , namely it depends only on the transverse separation @xmath48 of the color dipole . the dependence on the external probe , i.e. the photon virtuality , is included in the wavefunctions . the main technical difficulty in eq . ( [ eq11 ] ) is to model the unintegrated gluon distribution function in a suitable way , mainly in the small transverse momentum ( @xmath53 ) region ( infrared sector ) . in particular , to obtain these distributions one should solve numerically an evolution equation , which makes the procedure cumbersome for practical use ( instead , for a prompt parametrization see @xcite ) . in general , this is avoided introducing an anstz for the effective dipole cross section and analyzing the process in the impact parameter space . the main feature of the current models in the literature is to interpolate the physical regions of small transverse separations ( qcd - parton improved model picture ) and the large ones ( regge - soft picture ) . below , we quote two of them , which have connections with the work performed here . the phenomenological saturation model of golec - biernat and wsthoff ( hereafter gbw ) gives a good description of dis data @xcite . the corresponding cross section interpolates between color transparency , i.e. @xmath54 at small @xmath10 , and constant cross section @xmath55 at large @xmath10 ( confinement ) . such a procedure ensures @xmath56-saturation , while parton saturation at low @xmath3 is obtained with an eikonal - inspired shape for the dipole cross section , @xmath57 where the parameters @xmath58 mb , @xmath59 and @xmath60 are fitted to the hera dis inclusive data with @xmath61 , whereas @xmath62 gev@xmath63 sets the dimension . the @xmath3-dependent saturation radius @xmath64 scales the pair separation @xmath10 in the cross section and is associated with the mean separation between partons in the nucleon . this approach was used to describe diffractive dissociation in a parameter - free way , considering also the required @xmath14 configuration . some criticism to the gbw model , mainly concerning a better knowledge of the parton ( gluon ) distribution at low @xmath10 and its saturation at small @xmath3 are postponed to the next section . the most of them can also to be found in refs . @xcite . our work in this paper is closer to the mcdermott et al . one @xcite ( hereafter mcdfgs ) , where perturbative qcd relates the dipole cross section to the leading logarithmic gluon distribution function in the proton , at lla(@xmath56 ) accuracy , @xmath65 where the identification @xmath66 is allowed at leading - log level . a study of the dipole cross section is carried for all transverse separations @xmath10 in @xcite . the small @xmath10 region is described by the usual gluon pdfs , evoluting with the scale @xmath67 , while for large separations ( @xmath68 fm ) the dipole cross section is driven by the pion - proton contribution with the typical soft energy behavior from the hadronic sector . moreover , the taming of the parton density is implemented by hand starting at the named critical transverse separation @xmath69 , stated when the dipole cross section reaches one half of its maximum value labeled by the pion - proton cross section . in connection with the present work , the glauber - mueller approach gives the eq . ( [ eq14 ] ) at the born level . instead of using an ad hoc control of the gluon distribution , the glauber - mueller provides corrections required by unitarity in an eikonal expansion . for the large @xmath10 region , we choose to follow a similar procedure from the gbw model , namely saturating the dipole cross section ( @xmath10-independent constant value ) . having defined the notation and reviewed the main properties settled by the rest frame representation of deep inelastic process , in the next section we address the unitarity corrections formalism contained in the glauber - mueller approach , that we will consider in the calculation of @xmath19 , @xmath1 and @xmath2 . the glauber formalism concerns mainly interactions with a target nucleus , allowing to calculate the amount of unitarity corrections to the nuclear cross section . however this approach can be extended to take into account the evolution of the partonic densities ( saturation ) through the multiple scatterings . below , we review shortly the main properties of the glauber formalism in qcd , either in the nuclear case as its application to the nucleon case . we indicate the original papers @xcite for a complete presentation . it should be stressed that saturation effects mean unitarity corrections to the observables . indeed , the asymptotic calculations have produced an unified @xmath70 pattern for the cross section and gluon function instead of a effectively saturated one @xcite . since the small @xmath3 limit is driven by the gluonic interactions , we consider a virtual probe @xmath71 with invariant mass @xmath56 which decays into a gluon pair @xmath72 having a transverse separation @xmath48 and transverse momentum @xmath73 . then , the pair interacts with the target through a gluonic ladder at fixed transverse separation . following the discussion from the previous section , the photoabsortion cross section for the probe particle in terms of @xmath3 and virtuality @xmath56 in the nuclear case is @xmath74 where the variables have the same identification as in the @xmath9 pair discussion . the quantity @xmath75 is the light - cone wavefunction for the gluon pair . the glauber s multiscattering theory employs the phase shift method to describe processes at high energy for an incident particle undergoing into successive scatterings . when @xmath3 is small , the coherence length @xmath76 is bigger than the mean radius @xmath77 for all nuclei and the interaction of the initial parton stands through the entire nuclear path , providing coherent interactions with all target partons along the distance @xmath76 . these scatterings are coherent in an interaction of a hadronic fluctuaction , i.e. quark or gluon pairs , with the nucleus . there are interference effects among them , generating a reduction in the nuclear cross section , @xmath78 . otherwise , when @xmath79 , full incoherent scatterings occur leading to the expectation that the nuclear cross section equals to @xmath80 . the well known glauber formula for the total cross section of a hadronic state with the nucleus is @xmath81 where @xmath82 is a profile function containing the dependence on the impact parameter @xmath83 , which is the conjugate variable to the momentum transfer @xmath84 . it is related to the nucleon distribution inside the nucleus and encodes the information about the angular distribution of the scattering . we discuss its particular shape later on . ( [ glauber ] ) is quite general and allows to describe the hadronic fluctuations of the photon or any virtual probe , as saw in the last section . considering the scattering amplitude dependent on the usual mandelstan variables @xmath26 and @xmath84 , now written in the impact parameter representation @xmath83 , @xmath85 the corresponding total and elastic cross sections ( from optical theorem ) are rewritten in the impact parameter representation ( @xmath86 ) as @xmath87 a very important property when treating the scattering in the impact parameter space is the simple definition for the unitarity constraint @xcite . at fixed @xmath83 , the constraint can be expressed in the following way , @xmath88 with @xmath89 denoting the sum of contributions from all the inelastic channels . the constraint above has a simple solution . if the real part of the scattering amplitude vanishes at the high energy limit , corresponding to small @xmath3 values , the solution is @xmath90\,,\\ \sigma_{tot } & = & 2\,\int d^2{\bf b}\ , \left[\ , 1-\rm{e}^{-\frac{1}{2}\,\omega\,(s,\,{\bf b})}\ , \ , \right]\ , , \label{sigeik } \end{aligned}\ ] ] where the opacity @xmath91 is an arbitrary real function and it should be determined by a detailed model for the interaction . the opacity function has a simple physical interpretation , namely @xmath92 corresponds to the probability that no inelastic scatterings with the target occur . to provide the connection with the glauber formalism , the opacity function can be written in the factorized form @xmath93 , considering @xmath94 normalized as @xmath95 ( for a detailed discussion , see i.e. @xcite ) . from eqs . ( [ sigeik ] ) and ( [ glauber2 ] ) , we identify the opacity @xmath96 . moreover , it has been found that the same formalism for multiple scatterings can be applied to the nucleon case . to proceed , we should determine the @xmath72 cross section . the gluon pair cross section is equivalent to the quark pair one , up to a color factor ( @xmath97 ) . the ( @xmath9 pair ) dipole - proton cross section is well known @xcite , calculated starting from eq . ( [ eq11 ] ) , and in double logarithmic approximation ( dla ) has the following form @xmath98 with the @xmath10-dependent scale @xmath99 . considering eq . ( [ dipdla ] ) one can connect directly the dipole picture with the usual parton distributions ( gluon ) , since they are solutions of the dglap equations . in our case , we follow the calculations in ref . @xcite and consider the effective scale @xmath100 . such a value differs from @xcite , where it is @xmath101 , which is obtained by an averaging procedure on the transverse size integral of @xmath1 . however , in further studies in vector mesons it was found that @xmath102 ranges from 4 - 15 , and @xmath19 and @xmath1 are not sensitive to those variations . thus , these values are consistent in leading logarithmic @xmath56 approximation . from the above expression , one obtains a dipole cross section satisfying the unitarity constraint and a framework to study the unitarity effects ( saturation ) in the gluon dglap distribution function . hence , hereafter we use the glauber - mueller dipole cross section given by @xmath103 here , some comments are in order . the glauber - mueller approach is valid in the small @xmath3 region , and the gluon emission is described in the double logarithmic approximation of perturbative qcd . the interaction of the quark pair with the nucleon ( proton ) occurs through ladder diagrams exchange , satisfying the dglap equations in the dla limit . the high energy limit allows to treat successive scatterings as independent collisions , meaning the process described by the eikonal picture of a relativistic particle crossing the nucleus . moreover , as a consequence of no correlation among nucleons inside the nucleus ( in the nuclear case ) , there are no correlations among partons from different partonic cascades , stressing that only the fastest parton interacts with the target . the corrections coming from the slowest partons in the cascade ( emitted by the pair ) lead to the agl non - linear evolution equation @xcite , and they have been considered recently to describe diffractive dis in ref . @xcite . regarding criticisms to the glauber - mueller approach , we indicate the recent paper @xcite for a complete discussion about the eikonal - like models , concerning their advantages and limitations as well as pointing out the improvements to be taken into account to introduce the proper corrections . now , we proceed to calculate numerical estimates of the dipole cross section using the glauber - mueller approach through eqs . ( [ dipdla]-[dipolgm ] ) . then we are calculating saturation effects in the color dipole picture . firstly , we need to discuss the profile function @xmath104 . this function contains information about the angular distribution in the scattering , namely the @xmath84 dependence ( quark pair - ladder and proton - ladder couplings ) . both of them can be approximated by an exponential parametrization , leading to a simple gaussian shape in the impact parameter space , @xmath105 , where @xmath106 is the atomic number and @xmath77 is the target radius . we will keep this notation although we are only concerned with the nucleon case . the @xmath107 value should be determined from data , ranging between @xmath108 gev@xmath109 for the proton case ( see discussions in @xcite ) . for nuclear reactions , a more suitable shape for the profile should be taken into account , since the gaussian approximation is no longer appropriate to describe the nuclear profile for large @xmath106 . here , we have used the value ( @xmath110 gev@xmath109 ) obtained from a good description of both inclusive structure function and its derivative @xcite . such a value corresponds to significative unitarity corrections to the standard dglap input even in the current hera kinematics . now , we discuss in a detailed way the main characteristics emerging from the dipole cross section eq . ( [ dipolgm ] ) . in order to do this , in fig . ( [ dipsigmas ] ) one shows the glauber - mueller dipole cross section as a function of dipole transverse size @xmath10 at fixed momentum fraction @xmath3 . for a better illustration on the expected partonic saturation effects , we run the bjorken @xmath3 down to a quite small value @xmath111 ( thera region ) . hereafter , we are using the grv gluon distribution at leading order @xcite in the input eq . ( [ dipdla ] ) , whose choice we justify below . the solid lines correspond to the dipole cross section calculation , eq . ( [ dipolgm ] ) , whereas the dashed lines are the gbw model @xcite presented for comparison . the general shape in terms of the dipole size comes from the balancing between the color transparency @xmath112 behavior and the gluon distribution . here some comments about the large transverse separation are in order : although perturbative qcd provides reliable results at small distances ( small dipole sizes ) , the nonperturbative sector is not still completely understood . the usual pdf s are evoluted from a perturbative initial scale @xmath113 gev@xmath63 , and one has little information about the behavior at @xmath114 . in general one makes use of regge phenomenology to estimate those contributions ( see , for instance @xcite ) . thus , extrapolating to lower virtuality regions ( large dipole sizes ) one needs a parametrization regarding the nonperturbative sector . this is the main justification of the use the grv94 parametrization @xcite in our calculations . bearing in mind that @xmath115 , its evolution initial scale is @xmath116 gev@xmath63 allowing to scan dipole sizes up to @xmath117 gev@xmath118 (= 0.62 fm ) . for the most recent parametrizations , where @xmath119 gev@xmath63 ( @xmath120 fm ) the amount of nonperturbative contribution in the calculations should increase . an additional advantage is that grv94 does not include non - linear effects to the dglap evolution since the parametrization was obtained from rather large @xmath3 values . this feature ensures that the parametrization does not include unitarity corrections ( perturbative shadowing effects ) in the initial scale . now , we should introduce an ansatz for the large transverse separation region . a more phenomenological way is to match the pqcd dipole cross section with the typical hadronic one @xmath121 at @xmath122 as performed in @xcite . however , due to the significant growth of the pqcd dipole cross section at high energies , we choose an alternative ansatz : the gluon distribution is frozen at scale @xmath122 , namely @xmath123 . then , for the large distance contribution @xmath124 the gluon distribution reads as @xmath125 leading to the correct behavior @xmath126 as @xmath127 . in a more sophisticated case , one can substitute the freezed scale @xmath128 by the saturation scale @xmath129 to take into account a realistic value of the gluon anomalous dimension in all kinematic region ( see correlated issues in @xcite ) . recently , the phenomenological model of ref . @xcite has produced a good description of hera data in both inclusive and diffractive processes . it is constructed interplaying the color transparency behavior @xmath130 at small dipole sizes and a flat ( saturated ) behavior at large dipole sizes @xmath131 ( confinement ) . the expression has the eikonal - like form , @xmath132 \,,\end{aligned}\ ] ] where @xmath133 gev@xmath63 and the three fitted parameters are @xmath134 mb , @xmath135 and @xmath60 and the notation for the saturation radius @xmath136 . the gbw total cross section lies below the typical hadronic cross section : for instance in the pion - proton case , convoluting the pion wavefunction squared with the gbw dipole cross section we would have a constant cross section at high energies @xmath137 . despite describing data in good agreement , gbw has some details that deserve some discussions : the approach does not present a dynamical hypothesis for the saturation phenomena and does not match dglap evolution . in gbw , saturation is characterized by the @xmath3-dependent saturation radius @xmath138 instead of the scale coming from glauber - mueller , @xmath139 , which can be easily extended for the nuclear case @xcite . in fig . ( [ dipsigmas ] ) one shows the glauber - mueller dipole cross section and the gbw model @xcite . we choose to compare them due to the fact that gbw is actually a particular case of the glauber - mueller approach , considering an oversimplified profile function . we should show below that this fact allows to construct an extended saturation model with dglap evolution @xcite ( bgk model ) . in the lower plots , where @xmath140 and @xmath141 , the gm cross section underestimates the gbw one . however , as @xmath3 decreases the gluon distribution in the proton rises producing a bigger dipole cross section . this feature is clear in the upper plots , for smaller @xmath142 and @xmath111 , where gm overestimates gbw by a significant factor mostly at intermediate @xmath10 values . finally , we show the connection between the glauber - mueller approach with the saturation model with dglap evolution @xcite , observing that it is timely since the latter is quite efficient in describing data and its qualitative success corroborates quantitative qcd studies . considering the particular case of central collisions , namely scattering at impact parameter @xmath143 ( @xmath144 ) , the glauber - mueller approach produces , @xmath145 the integration over @xmath146 can be promptly carryied out , and introducing the notation for the proper normalization for the dipole cross section , @xmath147 the eq . ( [ dipolgmb0 ] ) recovers the simple expression for the saturation model dglap evoluted @xcite , @xmath148\ , . \end{aligned}\ ] ] for a phenomenological analysis , the parameter @xmath55 and the scale @xmath149 are determined from data in @xcite . in our case , it assumes the well defined value @xmath150 mb , using @xmath151 gev@xmath109 . for a larger radius , for instance @xmath152 gev@xmath109 , one obtains a value @xmath153 mb , closer to the gbw one . here , the virtuality scale is @xmath100 ) , whereas bgk choose the parametric form @xmath154 and a two - parameter initial condition for the gluon distribution function . this section is devoted to the study and estimate of the gluon driven observables measured at hera kinematical domain in the rest frame . the first one is the inclusive structure function @xmath156 , the main quantity testing the small @xmath3 physics . the unitarity corrections are well established for this observable considering glauber - mueller approach @xcite as well as its high energy asymptotics @xcite , namely the black disk limit . we review these issues considering the dipole picture ( rest frame ) , using a more complete analysis similar to @xcite , but mostly , discuss in detail the role played by the nonperturbative physics needed to describe the structure function , and where in the transverse separation @xmath10 range it starts to be important . the longitudinal structure function @xmath157 is also addressed , verifying the frame invariance in comparison with previous laboratory frame calculations @xcite . the longitudinal wavefunction strongly suppress large @xmath10 contributions , thus selecting smaller nonperturbative contribution in comparison with the @xmath19 case . moreover , @xmath1 is one of the main observables scanning possible higher twist corrections in the standard operator product expansion ( ope ) @xcite . therefore , a reasonable description of this quantity suggests that the glauber - mueller formalism ( or similar eikonal - like approaches ) take into account the most important contributions to the complete higher - twist corrections at current kinematical regimes . the structure function @xmath158 gives the charm quark content on the proton and is directly driven by the gluon distribution . therefore it is a powerful observable to scan saturation effects in the small @xmath3 region . however the current experimental status requires more dedicated measurements and a better statistics . we verify a consistent description in the rest frame corroborating the similar analysis in the dipole models @xcite and in those ones considering unitarity corrections in the laboratory frame @xcite . now we perform estimates for the inclusive structure function in the rest frame considering the glauber - mueller dipole cross section . the expression for @xmath19 , with the explicit integration limits on photon momentum fraction @xmath31 and transverse separation @xmath10 is , @xmath159 the notation has been introduced in the previous sections . in the fig . ( [ figf2 ] ) one shows the estimates for the structure function for representative virtualities @xmath56 from the latest h1 collaboration measurements @xcite . the longitudinal and transverse contributions are shown separately , the longitudinal one being subdominant as is well known . an effective light quark mass ( @xmath160 quarks ) was taken , with the value @xmath161 mev , and the target radius is considered @xmath110 gev@xmath109 , in agreement with ref . it should be stressed that this value leads to larger saturation corrections rather than using radius ranging over @xmath162 gev@xmath109 . the soft contribution comes from the freezing of the gluon distribution at large transverse separation as discussed at the previous section . the gluon distribution considered is grv94 at leading order @xcite , @xmath163 , whose choice has been justified in the previous section . from the plots we verify a good agreement in the normalization , however the slope seems quite steep . this fact is due to the modeling for the soft contribution and it suggests that a more suitable nonperturbative input should be taken . indeed , in ref . @xcite such a question is addressed , claiming that the correct input is the pion - proton cross section parametrized through the donnachie - landshoff pomeron . it is found that the large transverse separations give a larger contribution at low @xmath56 , whereas it vanishes concerning higher virtualities . to clarify this issue , we can calculate the integrand @xmath164 used in the @xmath10-integration for the cross section @xmath165 , that should have significant nonperturbative content . we plot in the fig . ( [ intpartial ] ) this quantity at low virtuality @xmath166 gev@xmath63 and at higher one @xmath167 gev@xmath63 for momentum fraction ranging on @xmath168 , verifying that the main contribution comes from an asymmetric peak at @xmath169 fm for @xmath166 gev@xmath63 , while it is shifted to @xmath170 at @xmath167 gev@xmath63 . in our calculation , the perturbative contribution holds up to @xmath171 fm , therefore the region @xmath172 gives a non marginal contribution to the cross section at low virtualities . indeed , one has found that it reaches about 10 % at @xmath166 gev@xmath63 and that when the virtuality increases the contribution gradually vanishes . in fact , using the most recent pdf s this situation is more critical since @xmath122 is smaller ( @xmath173 gev@xmath63 ) . this suggests that the photon piece @xmath174 multiplying the dipole cross section enhances the @xmath10-integrand to smaller @xmath10 at high @xmath56 , corroborating a similar conclusion already found in ref . @xcite . to clarify the role played by the soft nonperturbative contribution to the inclusive structure function and to verify the frame invariance of the approach , in the fig . ( [ f2comp ] ) we plot separately the perturbative contribution and parametrize the soft contribution introducing the nonperturbative structure function @xmath175 @xcite , which is added to the perturbative one . the soft piece normalization is @xmath176 . such procedure is done in order to compare explicitly the results found in ref . accordingly , we have used just shadowing corrections for the quark sector , taking into account only the transverse photon wavefunction and zero quark mass . the integration on the transverse separation is taken for @xmath177 , with @xmath116 gev@xmath63 for leading order grv94 gluon distribution . this leads to a residual contribution absorbed in the the soft piece coming from transverse separations @xmath178 . we considered the target radius being @xmath151 gev@xmath63 ( supported by @xcite ) , which produces a correction more important than the value @xmath179 gev@xmath63 . it is again verified that the soft contribution is important at small virtualities and decreasing as it gets larger . in the plots , the dot - dashed lines represent only the perturbative calculations using the particular approximations indicated above , and the solid lines represent the results when we add the soft term . concluding , we have a theoretical estimate , i.e. no fitting procedure , of the inclusive structure function @xmath0 through the glauber - mueller approach for the dipole cross section , detecting a non negligible importance of a suitable input for the large dipole size region . as we saw in the sec . ( [ sec2 ] ) , the inclusive structure function can be expressed in terms of the cross sections @xmath180 and @xmath181 for the absortion of virtual photons transversally and longitudinally polarized , @xmath182 . at small @xmath3 , the longitudinal structure function is @xmath183 from qed , the longitudinal photons have zero helicity ( @xmath184 ) and therefore they have a virtual character . in the naive parton model , the helicity conservation for the electromagnetic vertex implies to the callan - gross relation @xmath185 and consequently a vanishing value for the longitudinal structure function @xmath186 , considering the scattering photon - quarks ( spin 1/2 ) . from qcd theory , this quantity has a non - zero value due to the gluon radiation , as is encoded in the altarelli - martinelli expression ( see discussion in @xcite ) @xmath187\ , , \label{martinelli}\end{aligned}\ ] ] where @xmath188 is the inelasticity variable . therefore , the structure function @xmath1 is an auxiliary observable to detect saturation effects ( unitarity corrections ) in the gluon distribution . experimentally , the determination of the @xmath1 is quite limited , providing few data points . most recently , the h1 collaboration has determined the longitudinal structure function through the reduced double differential cross section @xcite , @xmath189 where @xmath190 . for large inelasticity , the reduced cross section becomes @xmath191 and the contribution of @xmath1 is enhanced for large @xmath192 . the longitudinal structure function should be obtained only in the region of large inelasticity , covered in a large range at hera . in ref . @xcite , two methods were used to perform the extraction : ( i ) for large @xmath193 gev@xmath63 , @xmath1 is obtained through the extrapolation method , using a nlo dglap qcd fit ( in the restrict kinematic range @xmath194 and @xmath195 gev@xmath63 ) to extrapolate @xmath19 into the high @xmath192 region . ( ii ) at low @xmath196 gev@xmath63 , the behavior of @xmath19 as a function of @xmath197 is obtained using the derivative method , based on the cross section derivative @xmath198 . the data points obtained are consistent with the previous measurements , however they are more precise and lying into a broader kinematical range . therefore , in the following we use only the new data points to analyse . in fig . ( [ flrframe ] ) we present the estimates for the @xmath1 structure function , in representative virtualities as a function of @xmath3 . for the calculations , it was considered light quarks ( @xmath199 ) with effective mass @xmath200 mev and the target radius @xmath110 gev@xmath109 . the large @xmath10 region is considered by the freezening of the gluon distribution at that region . our expression for the observable is then , @xmath201 the behavior is in agreement with the experimental result , either in shape as in normalization . a better description can be obtained by fine tunning the target size or the considered gluon distribution function , however it should be stressed that the present prediction is parameter - free and determined using the dipole picture taking into account unitarity ( saturation ) effects in the effective dipole cross section . we verify that the rest frame calculation , taking into account the dipole degrees of freedom and unitarity effects produces similar conclusions to those ones using the breit system . for instance , in a previous work @xcite , the unitarity corrections to the longitudinal structure function were estimated in the laboratory frame considering the eq . ( [ martinelli ] ) , with unitarized expressions for @xmath19 and @xmath202 , obtaining that the expected corrections reach to @xmath203 % as @xmath204 , namely on the kinematical sector of the upcoming thera project . the higher twist corrections to the longitudinal structure function have been pointed out . for instance , bartels et al . @xcite have calculated numerically the twist - four correction obtaining that they are large for @xmath205 and @xmath1 , although with opposite signs . this fact leads to remaining small effects to the inclusive structure function by almost complete cancellation between those contributions . the higher twist content is analyzed considering the model @xcite as initial condition . concerning @xmath1 , it was found that the twist - four correction is large and has negative signal , concluding that a leading twist analysis of @xmath1 is unreliable for high @xmath56 and not too small @xmath3 . the results are in agreement with the simple parametrization for higher twist ( ht ) studied by the mrst group in ref . @xcite , where @xmath206 . the second term would parametrize the higher twist content . in our case , the unitarity corrections provide an important amount of higher twist content , namely it takes into account some of the several graphs determining the twist expansion ( for recent discussions in these issues , see @xcite ) . in perturbative qcd , the heavy quark production in electron - proton interaction occurs basicaly through photon - gluon fusion , in which the emitted photon interacts with a gluon from the proton generating a quark - antiquark pair . therefore , the heavy quark production allows to determine the gluon distribution and the amount of unitarity ( saturation ) effects for the observable . in particular , charmed mesons have been measured at deep - inelastic at hera and the corresponding structure function @xmath158 is defined from the differential cross section for the @xmath207 pair production , @xmath208\,f_2^{c\bar{c}}(x , q^2)\,,\end{aligned}\ ] ] with @xmath192 being the inelasticity variable . in the laboratory frame , the dominant mechanism is the boson - gluon fusion @xmath209 . hence , the charm structure function is directly driven by the gluon distribution and provides constraints for the gluonic function . in leading order ( lo ) , it is written as @xcite , @xmath210 where @xmath211 . the mass factorization scale lies in the range @xmath212 . such a scale introduces an uncertainty of about 10 % , and an additional source of uncertainty is the charm mass , in general ranging on @xmath213 gev . the standard qcd coefficient function is labeled by @xmath214 . experimentally , the latest measurements of the charm structure function are obtained by measuring mesons @xmath215 production @xcite . from the theoretical input , it was used nlo coefficient functions , considering charm mass @xmath216 gev and factorization - normalization scale @xmath217 . the function @xmath158 shows an increase with decreasing @xmath3 at constant values of @xmath56 , whereas the rise becomes more intense at higher virtualities . the data are consistent with the nlo dglap calculations . concerning the ratio @xmath218 , the charm contribution to @xmath19 grows steeply as @xmath3 diminishes . it contributes less than 10% at low @xmath56 and reaches to about 30 % for @xmath219 gev@xmath63 @xcite . once more the color dipole picture will provide a quite simple description for the charm structure function in a factorized way . now , the glauber - mueller dipole cross section is weighted by the photon wavefunction constituted by a @xmath207 pair with mass @xmath220 . our expression for the charmed contribution in deep inelastic is thus written as @xmath221 where @xmath222 is the probability to find in the photon the @xmath207 color dipole with the charmed quark carrying fraction @xmath31 of the photon s light - cone momentum with @xmath223 polarizations . for the correspondent wavefunctions , the quark mass in eqs . ( [ eq9],[eq10 ] ) should be substituted by the charm quark mass @xmath220 . here , we should take care of the connection between the regge parameter @xmath224 and the bjorken variable @xmath225 . for calculations with the light quarks these variables are equivalent , however for heavier quarks the correct relation is @xcite , @xmath226 in fig . ( [ f2cfig ] ) we show the estimates for the charm structure function as a function of @xmath225 [ eq . ( [ xbj ] ) ] for representative virtualities . in our calculations , it was used the charm mass @xmath227 gev , the target size @xmath110 gev@xmath109 and the frozen gluon distibution at large @xmath10 . we have verified a small soft contribution as in the @xmath1 case , decreasing as the virtuality rises . there is a slight sensitivity to the value of the charm mass , increasing the overall normalization as @xmath220 diminishes . such a feature suggests that the charm mass is a hard scale suppressing the non - perturbative contribution to the corresponding cross section , which is in agreement with the recent bfkl color dipole calculations of nikolaev - zoller @xcite and donnachie - dosch @xcite . regarding the breit system description , in ref . @xcite it was found strong corrections to the charm structure function , which are larger than those of the @xmath19 ones making use of expression ( [ f2ccms ] ) . considering the ratio @xmath228 , the corrections predicted by the glauber - mueller approach would reach to 62 % at values of @xmath229 ( thera region ) . then , an important result is a large deviation of the standard dglap expectations at small @xmath3 for the ratio @xmath218 due to the saturation phenomena ( unitarization ) . with our calculation one verifies that it is obtained a good description of data in both reference systems , suggesting a consistent estimation of the unitarity effects for this quantity . we study the dipole picture for the description of deep inelastic scattering , focusing on observables driven directly by the gluon distribution . starting from the dipole cross section provided by the glauber - mueller approach in qcd , we perform estimates for the inclusive structure function @xmath19 , the longitudinal function @xmath1 and the charm structure function on the proton @xmath2 . for each of the observables discussed , we obtain theoretical estimates , in the rest frame , without further fitting procedure , in good agreement with the updated data from hera . the resulting calculations corroborate a quite consistent picture for the unitarity corrections from the glauber - mueller approach in both breit and rest reference systems . in the laboratory frame the unitarity effects are connected with the gluon distribution function , whereas in the color dipole framework the basic block is the dipole cross section which is corrected considering saturation effects . the small transverse separation @xmath10 region is dominated by the leading log dglap formalism , with the additional ingredient of unitarization phenomenon as the momentum fraction acquires quite small values . such corrections are associated with the taming of the gluon distribution in the very small @xmath3 region , in general named saturation regime . however , it should be stressed that the glauber - mueller approach and similar eikonal - like models provide a logarithmic @xmath70 asymptotic behavior for the inclusive structure function and gluon distribution , instead of a constant value for asymptotic energies . the large transverse separation is described by non - perturbative aspects of qcd . since this domain is not well determined at the moment , some modelling of the soft region is needed . in this work we choose the anstz in which the gluon distribution is frozen for virtualities above a cut radius @xmath230 , which corresponds to the region @xmath231 . a convenient choice for the gluon pdf in order to cover the widest possible kinematical window diminishes the uncertainty coming from the soft sector . the most appropriated input is the grv94 parametrization , where @xmath232 fm is found , whereas it can take values @xmath233 fm for the more recent pdf s . throughout the paper we used the target size @xmath151 gev@xmath109 , which corresponds to strong unitarity corrections . when performing a comparison with the phenomenological model gbw , we have found that the glauber - mueller approach underestimate the dipole cross section from gbw at not small @xmath234 . instead , for very low @xmath235 the glauber - mueller overestimates gbw due to the strong increasing of the gluon function in this region . concerning the saturation model with dglap evolution ( bgk ) , it is a particular case of the gm approach when considering central scattering @xmath143 . despite that the bgk model matches dglap evolution , glauber - mueller describes more properly the realistic impact parameter dependence of the process . moreover , in gm the extension to the nuclear case is built in . when considering the structure function @xmath19 , we have found that it is dominated by small transverse distances contributions . however , a non - negligible content from the soft sector is present . moreover , the photon wavefunctions enhances the dipole cross section into smaller dipole sizes , since the weight function selects smaller @xmath10 as the virtuality @xmath56 diminishes . our estimates here are parameter - free , however a fine tunning of the parameters can improve the data description . furthermore , we notice that in calculations from @xcite , only the aligned jet dipole configuration @xmath31 , @xmath236 ( and only transverse contribution ) is considered , whereas we take into account all configurations , including the symmetric ones . thus , all dipole sizes , even those from the non - perturbative region are included in our results . concerning @xmath1 , the estimates are consistent with the previous calculations in the breit system and are in good agreement with data . a remarkable feature is that the glauber - mueller approach in the color dipole framework gives important higher twist contributions to the leading twist calculation in a simple way . as is well known , @xmath1 is the main quantity to study the expected higher twist effects in low virtualities . the function @xmath2 is directly dependent of the gluon distribution and important unitarity corrections had been found when considering the breit frame . here , we verify consistent results in the rest frame in comparison with the previous ones in the fast proton system . we verified that the charm mass suppressed soft contributions in comparison with the @xmath19 case , and the results present a slight dependence with the specific value of @xmath220 . in conclusion , the glauber - mueller approach provides a well stablished formalism to take into account the unitarity effects . it allows to estimate the higher twist contributions to relavant observables in a simplified way , while matching dglap evolution equation at born level and including the impact parameter dependence properly . mvtm acknowledges martin mcdermott ( liverpool university - uk ) , igor ivanov ( ikp - forschungszentrum juelich , germany ) and victor gonalves ( ifm - ufpel , brazil ) for useful enlightenements . this work was partially financed by cnpq , brazil .

event shape variables are perhaps the most popular observables for testing qcd and for improving our understanding of its dynamics . event shape studies began in earnest towards the late seventies as a simple quantitative method to understand the nature of gluon bremsstrahlung @xcite . for instance , it was on the basis of comparisons to event shape data that one could first deduce that gluons were vector particles , since theoretical predictions employing scalar gluons did not agree with experiment @xcite . quite generally , event shapes parametrise the geometrical properties of the energy - momentum flow of an event and their values are therefore directly related to the appearance of an event in the detector . in other words the value of a given event shape encodes in a continuous fashion , for example , the transition from pencil - like two - jet events with hadron flow prominently distributed along some axis , to planar three - jet events or events with a spherical distribution of hadron momenta . thus they provide more detailed information on the final state geometry than say a jet finding algorithm which would always classify an event as having a certain finite number of jets , even if the actual energy flow is uniformly distributed in the detector and there is no prominent jet structure present in the first place . event shapes are well suited to testing qcd mainly because , by construction , they are collinear and infrared safe observables . this means that one can safely compute them in perturbation theory and use the predictions as a means of extracting the strong coupling @xmath1 . they are also well suited for determinations of other parameters of the theory such as constraining the quark and gluon colour factors as well as the qcd beta function . additionally they have also been used in other studies for characterising the final state , such as investigations of jet and heavy quark multiplicities as functions of event shape variables @xcite . aside from testing the basic properties of qcd , event shape distributions are a powerful probe of our more detailed knowledge of qcd dynamics . all the commonly studied event shape variables have the property that in the region where the event shape value is small , one is sensitive primarily to gluon emission that is soft compared to the hard scale of the event and/or collinear to one of the hard partons . such small - transverse - momentum emissions have relatively large emission probabilities ( compared to their high transverse momentum counterparts ) due to logarithmic soft - collinear dynamical enhancements as well as the larger value of their coupling to the hard partons . predictions for event shape distributions therefore typically contain large logarithms in the region where the event - shape is small , which are a reflection of the importance of multiple soft and/or collinear emission . a successful prediction for an event shape in this region requires all - order resummed perturbative predictions or the use of a monte - carlo event generator , which contains correctly the appropriate dynamics governing multiple particle production . comparisons of these predictions to experimental data are therefore a stringent test of the understanding qcd dynamics that has been reached so far . one other feature of event shapes , that at first appeared as an obstacle to their use in extracting the fundamental parameters of qcd , is the presence of significant non - perturbative effects in the form of power corrections that vary as an inverse power of the hard scale , @xmath2 . for most event shapes , phenomenologically it was found that @xmath3 ( as discussed in @xcite ) , and the resulting non - perturbative effects can be of comparable size to next - to - leading order perturbative predictions , as we shall discuss presently . the problem however , can be handled to a large extent by hadronisation models embedded in monte - carlo event generators @xcite , which model the conversion of partons into hadrons at the cost of introducing several parameters that need to be tuned to the data . once this is done , application of such hadronisation models leads to very successful comparisons of perturbative predictions with experimental data with , for example , values of @xmath1 consistent to those obtained from other methods and with relatively small errors . however since the mid - nineties attempts have also been made to obtain a better insight into the physics of hadronisation and power corrections in particular . theoretical models such as those based on renormalons have been developed to probe the non - perturbative domain that gives rise to power corrections ( for a review , see @xcite ) . since the power corrections are relatively large effects for event shape variables , scaling typically as @xmath4 ( no other class of observable shows such large effects ) , event shapes have become the most widely used means of investigating the validity of these ideas . in fact an entire phenomenology of event shapes has developed which is based on accounting for non - perturbative effects via such theoretically inspired models , and including them in fits to event shape data alongside the extraction of other standard qcd parameters . in this way event shapes also serve as a tool for understanding more quantitatively the role of confinement effects . the layout of this article is as follows . in the following section we list the definitions of several commonly studied event shapes in @xmath5 annihilation and deep inelastic scattering ( dis ) and discuss briefly some of their properties . then in section [ sec : perturbative ] we review the state of the art for perturbative predictions of event - shape mean values and distributions . in particular , we examine the need for resummed predictions for distributions and clarify the nomenclature and notation used in that context , as well as the problem of matching these predictions to fixed order computations . we then turn , in sections [ sec : exp ] and [ sec : distributions ] , to comparisons with experimental data and the issue of non - perturbative corrections , required in order to be able to apply parton level calculations to hadronic final state data . we discuss the various approaches used to estimate non - perturbative corrections , ranging from phenomenological hadronisation models to analytical approaches based on renormalons , and shape - functions . we also discuss methods where the standard perturbative results are modified by use of renormalisation group improvements or the use of ` dressed gluons ' . we study fits to event shape data , from different sources , in both @xmath5 annihilation and dis and compare the various approaches that are adopted to make theoretical predictions . in addition we display some of the results obtained by fitting data for various event shape variables for parameters such as the strong coupling , the qcd colour factors and the qcd beta function . lastly , in section [ sec : outlook ] , we present an outlook on possible future developments in the field . we list here the most widely studied event shapes , concentrating on those that have received significant experimental and theoretical attention . _ the _ canonical event shape is the thrust @xcite , @xmath6 : @xmath7 where the numerator is maximised over directions of the unit vector @xmath8 and the sum is over all final - state hadron momenta @xmath9 ( whose three - vectors are @xmath10 and energies @xmath11 ) . the resulting @xmath8 is known as the thrust axis . in the limit of two narrow back - to - back jets @xmath12 , while its minimum value of @xmath13 corresponds to events with a uniform distribution of momentum flow in all directions . the infrared and collinear safety of the thrust ( and other event shapes ) is an essential consequence of its linearity in momenta . often it is @xmath14 ( also called @xmath15 ) that is referred to insofar as it is this that vanishes in the @xmath16-jet limit . a number of other commonly studied event shapes are constructed employing the thrust axis . amongst these are the invariant squared jet - mass @xcite and the jet - broadening variables @xcite , defined respectively as @xmath17 where the plane perpendicular to the thrust axis is used to separate the event into left and right hemisphere , @xmath18 and @xmath19 . given these definitions one can study the heavy - jet mass , @xmath20 and wide - jet broadening , @xmath21 ; analogously one defines the light - jet mass ( @xmath22 ) and narrow - jet broadening @xmath23 ; finally one defines also the sum of jet masses @xmath24 and the total jet broadening @xmath25 and their differences @xmath26 , and @xmath27 . like @xmath14 , all these variables ( and those that follow ) vanish in the two - jet limit ; @xmath22 and @xmath23 are special in that they also vanish in the limit of three narrow jets ( three final - state partons ; or in general for events with any number of jets in the heavy hemisphere , but only one in the light hemisphere ) . another set of observables @xcite making use of the thrust axis starts with the thrust major @xmath28 , @xmath29 where the maximisation is performed over all directions of the unit vector @xmath30 , such that @xmath31 . the thrust minor , @xmath32 , is given by @xmath33 and is sometimes also known @xcite as the ( normalised ) out - of - plane momentum @xmath34 . like @xmath22 and @xmath23 , it vanishes in the three - jet limit ( and , in general , for planar events ) . finally from @xmath28 and @xmath32 one constructs the oblateness , @xmath35 . an alternative axis is used for the spherocity @xmath36 @xcite , @xmath37 while there exist reliable ( though algorithmically slow ) methods of determining the thrust and thrust minor axes , the general properties of the spherocity axis are less well understood and consequently the spherocity has received less theoretical attention . an observable similar to the thrust minor ( in that it measures out - of - plane momentum ) , but defined in a manner analogous to the spherocity is the acoplanarity @xcite , which minimises a ( squared ) projection perpendicular to a plane . it is also possible to define event shapes without reference to an explicit axis . the best known examples are the @xmath38 and @xmath39 parameters @xcite which are obtained from the momentum tensor @xcite . because of the quadratic dependence on particle momenta , these observable are collinear unsafe and so no longer widely studied . ] @xmath40 where @xmath41 is the @xmath42 component of the three vector @xmath43 . in terms of the eigenvalues @xmath44 , @xmath45 and @xmath46 of @xmath47 , the @xmath38 and @xmath39 parameters are given by @xmath48 the @xmath38-parameter is related also to one of a series of fox - wolfram observables @xcite , @xmath49 , and is sometimes equivalently written as @xmath50 the @xmath39-parameter , like the thrust minor , vanishes for all final - states with up to 3 particles , and in general , for planar events . the @xmath38-parameter can actually be considered ( in the limit of all particles being massless ) as the integral of a more differential observable , the energy - energy correlation ( eec ) @xcite , @xmath51 where the average in eq . ( [ eq : sigmachi ] ) is carried out over all events . another set of variables that characterise the shape of the final state are the @xmath52-jet resolution parameters that are generated by jet finding algorithms . examples of these are the jade @xcite and durham @xcite jet clustering algorithms . one introduces distance measures @xmath53 @xmath54 for each pair of particles @xmath55 and @xmath56 . the pair with the smallest @xmath53 is clustered ( by adding the four - momenta the @xmath57 recombination scheme ) and replaced with a single pseudo - particle ; the @xmath53 are recalculated and the combination procedure repeated until all remaining @xmath53 are larger than some value @xmath58 . the event - shapes based on these jet algorithms are @xmath59 , defined as the maximum value of @xmath58 for which the event is clustered to 3 jets ; analogously one can define @xmath60 , and so forth . other clustering jet algorithms exist . most differ essentially in the definition of the distance measure @xmath53 and the recombination procedure . for example , there are e0 , p and p0 variants @xcite of the jade algorithm , which differ in the details of the treatment of the difference between energy and the modulus of the 3-momentum . the geneva algorithm @xcite is like the jade algorithm except that in the definition of the @xmath53 it is @xmath61 that appears in the denominator instead of the total energy . an algorithm that has been developed and adopted recently is the cambridge algorithm @xcite , which uses the same distance measure as the durham algorithm , but with a different clustering sequence . we note that a number of variants of the above observables have recently been introduced @xcite , which differ in their treatment of massive particles . these include the @xmath62-scheme where all occurrences of @xmath11 are replaced by @xmath63 and the @xmath57-scheme where each 3-momentum @xmath43 is rescaled by @xmath64 . the former leads to a difference between the total energy in the initial and final states , while the latter leads to a final - state with potentially non - zero overall @xmath65-momentum . while such schemes do have these small drawbacks , for certain observables , notably the jet masses , which are quite sensitive to the masses of the hadrons ( seldom identified in experimental event - shape studies ) , they tend to be both theoretically and experimentally cleaner . ref . @xcite also introduced a ` decay ' scheme ( an alternative is given in @xcite ) where all hadrons are artificially decayed to massless particles . since a decay is by definition a stochastic process , this does not give a unique result on an event - by - event basis , but should rather be understood as providing a correction factor which is to be averaged over a large ensemble of events . it is to be kept in mind that these different schemes all lead to identical perturbative predictions ( with massless quarks ) and differ only at the non - perturbative level . a final question relating to event - shape definitions concerns the hadron level at which measurements are made . since shorter lived hadrons decay during the time of flight , one has to specify whether measurements were made at a stage before or after a given species of hadron decays . it is important therefore when experimental results are quoted , that they should specify which particles have been taken to be stable and which have not . as well as the @xmath66 variables discussed above , it is also possible to define , by analogy , event shapes in dis . the frame in which dis event shapes can be made to most closely resemble those of @xmath66 is the breit frame @xcite . this is the frame in which @xmath67 , where @xmath68 is the incoming proton momentum and @xmath69 the virtual photon momentum . one defines two hemispheres , separated by the plane normal to the photon direction : the remnant hemisphere ( @xmath70 , containing the proton remnant ) , and the current hemisphere ( @xmath71 ) . at the level of the quark - parton model , @xmath71 is like one hemisphere of @xmath0 and it is therefore natural to define event shapes using only the momenta in this hemisphere . in contrast , any observable involving momenta in the remnant hemisphere must take care to limit its sensitivity to the proton remnant , whose fragmentation can not be reliably handled within perturbation theory . a possible alternative to studying just the current hemisphere , is to take all particles except those in a small cone around the proton direction @xcite . a feature that arises in dis is that there are two natural choices of axis . for example , for the thrust @xmath72 one can either choose the unit vector @xmath73 to be the photon ( @xmath74 ) axis , @xmath75 , or one can choose it to be the true thrust axis , that which maximises the sum , giving @xmath76 . similarly one defines two variants of the jet broadening , @xmath77 for the jet - mass and @xmath38-parameter the choice of axis does not enter into the definitions and we have @xmath78 all the above observables can also be defined with an alternative normalisation , @xmath79 replacing @xmath80 , in which case they are named @xmath81 and so on . we note that @xmath81 was originally proposed in @xcite . for a reader used to @xmath0 event shapes the two normalisations might at first sight seem equivalent however when considering a single hemisphere , as in dis , the equivalence is lost , and indeed there are even events in which the current hemisphere is empty . this is a problem for observables normalised to the sum of energies in @xmath71 , to the extent that to ensure infrared safety it is necessary to exclude all events in which the energy present in the current hemisphere is smaller than some not too small fraction of @xmath82 ( see for example @xcite ) . additionally there are studies of variables that vanish in the 2 + 1-parton limit for dis , in particular the @xmath34 defined in analogy with the thrust minor of @xmath0 , with the thrust axis replaced by the photon axis @xcite , or an azimuthal correlation observable @xcite . rather than being examined just in @xmath71 , these observables use particles also in the remnant hemisphere ( for @xmath34 , all except those in a small cone around the proton ) . as in @xmath0 , jet rates are studied also in dis . unlike the event - shapes described above , the jet - shapes make use of the momenta in both hemispheres of the breit frame . their definitions are quite similar to those of @xmath0 except that a clustering to the proton remnant ( beam jet ) is also included @xcite . though only @xmath0 and dis are within the scope of this review , we take the opportunity here to note that related observables are being considered also for other processes . notably for drell - yan production an out - of - plane momentum measurement has been proposed in @xcite and various thrust and thrust - minor type observables have been considered in hadron - hadron dijet production in @xcite . the observables discussed above are all infrared and collinear safe they do not change their value when an extra soft gluon is added or if a parton is split into two collinear partons . as emerges from the original discussion of sterman and weinberg @xcite , this is a necessary condition for the cancellation of real and virtual divergences associated with such emissions , and therefore for making finite perturbative predictions . for an event shape that vanishes in the @xmath52-jet limit ( which we shall generically refer as an @xmath52-jet observable ) , the leading perturbative contribution is of order @xmath83 . for example the thrust distribution in @xmath0 is given by @xmath84 + { { \cal o}\left({\alpha_s}^2\right ) } \nonumber\end{aligned}\ ] ] ( see e.g. @xcite ) . when calculating perturbative predictions for mean values of event shapes , as well as higher moments of their distributions , one has integrals of the form @xmath85 we note that in general , fixed - order event - shape distributions diverge in the limit as @xmath86 goes to @xmath87 , cf . ( [ eq : thrustdist ] ) for @xmath88 . in eq . ( [ eq : meanv ] ) , the weighting with a power of @xmath86 is sufficient to render the singularity integrable , and the integral is dominated by large @xmath86 ( and so large transverse momenta of order @xmath82 , the hard scale ) . this dominance of a single scale ensures that the coefficients in the perturbative expansion for @xmath89 are well - behaved ( for example they are free of any enhancements associated with logarithms of ratios of disparate scales ) . beyond leading order , perturbative calculations involve complex cancellations between soft and collinear real and virtual contributions and are nearly always left to general - purpose `` fixed - order monte carlo '' programs . the current state of the art is next - to - leading order ( nlo ) , and available programs include : for @xmath0 to 3 jets , event @xcite , eerad @xcite and event2 @xcite ; for @xmath0 to 4 jets , menlo parc @xcite , mercutio @xcite and eerad2 @xcite ; for dis to @xmath90 jets , mepjet @xcite , disent @xcite and disaster++ @xcite ; and for all the above processes and additionally dis to @xmath91 jets and various hadron - hadron and photo - production processes , nlojet++ @xcite ( for photo - production , there is additionally jetvip @xcite ) . in recent years much progress has been made towards nnlo calculations , though complete results remain to be obtained ( for a review of the current situation , see @xcite ) . for the event shape distributions themselves , however , fixed order perturbative estimates are only of use away from the @xmath92 region , due to the singular behaviour of the fixed order coefficients in the @xmath93 limit . in this limit , at higher orders each power of @xmath94 is accompanied by a coefficient which grows as @xmath95 . these problems arise because when @xmath86 is small one places a restriction on real emissions without any corresponding restriction on virtual contributions the resulting incompleteness of cancellations between logarithmically divergent real and virtual contributions is the origin of the order - by - order logarithmic enhancement of the perturbative contributions . to obtain a meaningful answer it is therefore necessary to perform an all - orders _ resummation _ of logarithmically enhanced terms . when discussing resummations it is convenient to refer to the integrated distribution , @xmath96 quite generally @xmath97 has a perturbative expansion of the form @xmath98 one convention is to refer to all terms @xmath99 as leading logarithms ( ll ) , terms @xmath100 as next - to - leading logarithms ( nll ) , etc . , and within this hierarchy a resummation may account for all ll terms , or all ll and nll terms and so forth . such a resummation gives a convergent answer up to values of @xmath101 , beyond which terms that are formally subleading can become as important as the leading terms ( for example if @xmath102 , then the nnll term @xmath103 is of the same order as the ll term @xmath104 ) . at its limit of validity , @xmath105 , a n@xmath106ll resummation neglects terms of relative accuracy @xmath107 . an important point though is that for nearly all observables that have been resummed there exists a property of exponentiation : @xmath108 in some cases @xmath109 is written in terms of a sum of such exponentiated contributions ; in certain other cases the exponentiation holds only for a suitable integral ( e.g. fourier ) transform of the observable . the fundamental point of exponentiation is that the inner sum , over @xmath110 , now runs only up to @xmath111 , instead of @xmath112 as was the case in eq . ( [ eq : dumbsum ] ) . with the exponentiated form for the resummation , the nomenclature `` ( next - to)-leading - logarithmic '' acquires a different meaning n@xmath106ll now refers to all terms in the exponent @xmath113 . to distinguish the two classification schemes we refer to them as n@xmath106ll@xmath114 and n@xmath106ll@xmath115 . the crucial difference between n@xmath106ll@xmath114 and n@xmath106ll@xmath115 resummations lies in the range of validity and their accuracy . a n@xmath106ll@xmath115 resummation remains convergent considerably further in @xmath116 , up to @xmath102 ( corresponding usually to the peak of the distribution @xmath117 ) ; at this limit , neglected terms are of relative order @xmath118 . consequently a n@xmath106ll@xmath115 resummation includes not only all n@xmath106ll@xmath114 terms , but also considerably more . from now when we use the term n@xmath106ll it should be taken to mean n@xmath106ll@xmath119 . two main ingredients are involved in the resummations as shown in eq . ( [ eq : expresum ] ) . firstly one needs to find a way of writing the observable as a factorised expression of terms for individual emissions . this is often achieved with the aid of one or more integral ( mellin , fourier ) transforms . secondly one approximates the multi - emission matrix element as a product of individual independent emission factors . it is to be kept in mind that not all observables exponentiate . the jade jet resolution is the best known such example @xcite ( and it has not yet been resummed ) . the general criteria for exponentiation are discussed in @xcite . resummed results to nll accuracy exist for a number of @xmath16-jet observables in @xmath120 jets : the thrust @xcite , the heavy - jet mass @xcite and the single - jet and light - jet masses @xcite , jet broadenings @xcite , @xmath38-parameter @xcite , durham and cambridge jet resolutions @xcite , the thrust major and oblateness @xcite , eec @xcite and event - shape / energy - flow correlations @xcite ; for @xmath65-jet observables in @xmath121 jet events we have the thrust minor @xcite and the @xmath39-parameter @xcite . for dis to @xmath122 jets the current - hemisphere observables are resummed in refs . @xcite and the jet rates in @xcite ( though only to nll@xmath114 accuracy ) while the @xmath90 jet observables , @xmath34 and azimuthal correlations , have been resummed in @xcite . methods for automated resummation of arbitrary observables are currently in development for a range of processes@xcite , and techniques have also being developed for dealing with arbitrary processes @xcite . we also note that for the @xmath0 thrust and heavy - jet mass there have been investigations of certain classes of corrections beyond nll accuracy @xcite . the above resummations all apply to events with only light quarks . investigations to nll@xmath123 accuracy have also been performed for @xmath0 jet rates with heavy quarks @xcite . the above resummations are all for @xmath52-jet observables in the @xmath52-jet limit . for some @xmath65-jet observables it is useful to have resummations in the @xmath16-jet limit ( because it is here that most of the data lie ) , though currently little attention has been devoted to such resummations , with just nll@xmath123 results for jet - resolution parameters @xcite and nll@xmath115 results for the light - jet mass and narrow - jet broadening @xcite . furthermore in addition to the resummation of the @xmath38-parameter for @xmath124 there exists a ll resummation @xcite ( that could straightforward be extended to nll ) of a ` shoulder ' structure at @xmath125 ( related to a step function at lo ) , which corresponds to symmetric three jet events . some of the observables ( for example many single - jet observables ) have the property that they are sensitive to emissions only in some of the phase space . these are referred to as non - global observables and nll resummed predictions for them require that one account for coherent ensembles of energy - ordered large - angle gluons . this has so far been done only in the large-@xmath126 limit @xcite ( for reviews see @xcite ) , though progress is being made in extending this to finite @xmath126 @xcite . at @xmath127 . the resummed distribution has been matched ( @xmath128 scheme , see below ) so as to include the full lo and nlo contributions . the inset differs from the main plot only in the scales.,scaledwidth=60.0% ] to illustrate the impact of resummations we show in figure [ fig : thrlonlo ] the lo , nlo and resummed results for the thrust distribution . not only is the divergence in the lo distribution clearly visible , but one also observes a marked difference in the qualitative behaviour of the lo and nlo results , both of which show large differences relative to the nll resummed prediction at small @xmath14 . for the practical use of resummed results an important step is that of matching with the fixed - order prediction . at the simplest level one may think of matching as simply adding the fixed - order and resummed predictions while subtracting out the doubly counted ( logarithmic ) terms . such a procedure is however too naive in that the fixed - order contribution generally contains terms that are subleading with respect to the terms included in the resummation , but which still diverge . for example when matching nll and nlo calculations , one finds that the nlo result has a term @xmath129 , which in the distribution diverges as @xmath130 . this is unphysical and the matching procedure must be sufficiently sophisticated so as to avoid this problem , and in fact it should nearly always ensure that the matched distribution goes to zero for @xmath131 . several procedures exist , such as @xmath128 and @xmath109 matching @xcite and multiplicative matching @xcite and they differ from one - another only in terms that are nnlo and nnll , i.e. formally beyond the state - of - the - art accuracy . we note the matching generally ensures that the resummation , in addition to being correct at nll accuracy ( our shorthand for nll@xmath115 ) , is also correct at nnll@xmath123 accuracy . matching also involves certain other subtleties , for example the ` modification of the logarithm ' , whereby @xmath132 is replaced with @xmath133 , where @xmath134 is the largest kinematically allowed value for the observable . this ensures that the logarithms go to zero at @xmath134 , rather than at some arbitrary point ( usually @xmath135 ) , which is necessary in order for the resulting matched distribution to vanish at @xmath134 . [ sec : means ] [ sec : meanpc ] much of the interest in event shapes stems from the wealth of data that is available , covering a large range of centre of mass energies ( @xmath0 ) or photon virtualities ( dis ) . the @xmath0 data comes from the pre - lep experiments @xcite , the four lep experiments @xcite @xcite @xcite and sld @xcite . in addition , because many observables have been proposed only in the last ten or so years and owing to the particular interest ( see below ) in data at moderate centre of mass energies , the jade data have been re - analysed @xcite . have also been obtained from lep 1 @xcite by considering events with an isolated hard final - state photon and treating them as if they were pure qcd events whose centre of mass energy is that of the recoiling hadronic system . such a procedure is untrustworthy because it assumes that one can factorise the gluon production from the photon production . this is only the case when there is strong ordering in transverse momenta between the photon and gluon , and is not therefore applicable when both the photon and the gluon are hard . tests with monte carlo event generators which may indicate that any ` non - factorisation ' is small @xcite are not reliable , because the event generator used almost certainly does not contain the full 1-photon , 1-gluon matrix element . it is also to be noted that the isolation cuts on the photon will bias the distribution of the event shapes , in some cases @xcite similarly to an event - shape energy - flow correlation @xcite . therefore we would argue that before relying on these data , one should at the very least compare the factorisation approximation with exact lo calculations , which can be straightforwardly obtained using packages such as grace @xcite , comphep @xcite or amegic++ @xcite . ] results in dis come from both h1 @xcite ( mean values and distributions ) and zeus @xcite ( means ) , @xcite ( means and distributions in rapidity gap events ) . thrust , compared to lo , nlo and nlo+@xmath136 ; right : data for the durham jet resolution @xmath137 compared to the nlo prediction ( figure taken from @xcite).,title="fig:",scaledwidth=48.0% ] thrust , compared to lo , nlo and nlo+@xmath136 ; right : data for the durham jet resolution @xmath137 compared to the nlo prediction ( figure taken from @xcite).,title="fig:",scaledwidth=48.0% ] let us start our discussion of the data by examining mean values ( distributions are left until section [ sec : distributions ] ) . as we have already mentioned , one of the most appealing properties of event shapes , in terms of testing qcd , is the fact that they are calculable in perturbation theory and so provide a direct method for the extraction of @xmath1 as well as testing other parameters of the theory such as the colour factors @xmath138 and @xmath139 by using a wealth of available experimental data . however one obstruction to the clean extraction of these fundamental parameters is the presence in many cases , of significant non - perturbative effects that typically fall as inverse powers @xmath140 of @xmath82 , the hard scale in the reaction ( the centre - of mass energy in @xmath5 annihilation and the momentum transfer in dis ) . the importance of such _ power corrections _ varies from observable to observable . for example , it can be seen from fig . [ fig : meandata ] ( left ) that the data for the mean value of the thrust variable need a significant component @xmath141 in addition to the lo and nlo fixed order perturbative estimates , in order to be described . on the other hand the comparison ( right ) for the mean value of the durham jet resolution parameter @xmath137 with the nlo prediction alone is satisfactory , without the need for any substantial power correction term . the problem of non - perturbative corrections is so fundamental in event - shape studies that it is worthwhile giving a brief overview of the main approaches that are used . no statement can be made with standard perturbative methods about the size of these power corrections and hence the earliest methods adopted to quantify them were phenomenological hadronisation models embedded in monte carlo event generators @xcite . one of the main issues involved in using such hadronisation models is the existence of several adjustable parameters . this essentially means that a satisfactory description of data can be obtained by tuning the parameters , which does not allow much insight into the physical origin of such power behaved terms which are themselves of intrinsic theoretical interest ( for a more detailed critique of hadronisation models see @xcite ) . however since the mid - nineties there have been developments that have had a significant impact on the theoretical understanding of power corrections . perhaps the most popular method of estimating power corrections is based on the renormalon model . in this approach one examines high - order terms @xmath142 of the perturbative series that are enhanced by coefficients @xmath143 , where @xmath144 is the first coefficient of the @xmath145 function . the factorial divergence leads to an ambiguity in the sum of the series of order @xmath2 , where the value of @xmath62 depends on the speed with which the high - order terms diverge . this approach has seen far more applications than can possibly be described here and for a full discussion and further references on renormalons the reader is referred to @xcite . the most extensive phenomenological applications of renormalon based ideas have been in the estimation of power corrections for event - shape variables . the reason is that event shapes have much larger ( and so experimentally more visible ) power corrections ( typically @xmath146 ) than most other observables ( typically @xmath147 or smaller ) . for event shape variables in @xmath5 annihilation in the two - jet limit , renormalon - inspired ( or related ) studies of power corrections have been carried out in refs . for the @xmath122 jet limit of dis , results on power corrections based on the renormalon approach can be found in @xcite . studies in the @xmath65-jet limit have been given in @xcite . the approach that has been most commonly used in comparing theoretical predictions with experimental data is that initiated by dokshitzer and webber @xcite . here the full result for the mean value , @xmath148 , is given by @xmath149 where @xmath150 is the perturbative prediction for the mean value . the hadronisation correction is included through the term @xmath151 , where @xmath62 is related to the speed of divergence of the renormalon series ( as discussed above ) , @xmath152 is an observable - dependent ( calculable ) coefficient and @xmath153 is a non - perturbative factor , scaling as @xmath154 which is hypothesised to be common across a whole class of observables with the same value of @xmath62 ( strictly speaking the full story is a little more complicated , see e.g. @xcite ) . most event shapes have @xmath3 , implying a leading power correction scaling as @xmath146 , essentially a consequence of the observables linearity on soft momenta . for these observables with @xmath3 the form for @xmath155 that has become standard is @xcite , @xmath156 where @xmath157 , @xmath158 and @xmath159 . in eq . ( [ eq : cpfin ] ) an arbitrary _ infrared matching scale _ @xmath160 has been introduced , intended to separate the perturbative and non - perturbative regions . it is usually taken to be @xmath16 gev ( and for systematic error estimates it is varied between @xmath161 and @xmath65 gev ) . the only truly non - perturbative ingredient in eq . ( [ eq : cpfin ] ) is @xmath162 , which can be interpreted as the average value of an infrared finite strong coupling for scales below @xmath160 ( such a concept was first applied phenomenologically in @xcite ) . though one could imagine estimating it from lattice studies of the coupling ( such as @xcite ) , one should keep in mind that the coupling in the infrared is not a uniquely defined quantity . in practice the phenomenological test of the renormalon approach will be that @xmath163 has a consistent value across all observables ( in the @xmath3 class ) . the terms with negative sign in the parenthesis in eq . ( [ eq : cpfin ] ) are a consequence of merging standard nlo perturbative results , which include the small spurious contributions from the infrared region ( scales up to @xmath160 ) , with the non - perturbative power correction that accounts correctly for scales from zero to @xmath160 . carrying out the subtraction of the perturbative terms that arise from the infrared region , below @xmath160 , amounts to inclusion of the negative sign terms in parenthesis above . note that this subtraction procedure must be carried out to the level of accuracy of the corresponding perturbative estimate . in other words if a perturbative estimate becomes available at @xmath164 then an additional subtraction term proportional to @xmath165 is required . with the subtraction procedure carried out to @xmath166 , as above , one expects a residual @xmath160 dependence @xmath167 . the scale @xmath168 is the renormalisation scale , which should be taken of the same order as the hard scale @xmath82 . the ` milan factor ' , @xmath169 in eq . ( [ eq : cpfin ] ) , is @xcite @xmath170 where the numerical result is given for @xmath171 , since only light flavours will be active at the relevant ( low ) scales . it accounts for the fact that the usual , ` naive ' , calculation for @xmath152 is carried out on the basis of a decaying virtual gluon ( i.e. cutting a bubble in the chain of vacuum polarisation insertions that lead to the running of the coupling ) , but without fully accounting for the non - inclusiveness of that decay when dealing with the observable . it was pointed out in @xcite that this is inconsistent and full ( two - loop ) calculations revealed @xcite that if the ` naive ' @xmath152 coefficient is calculated in an appropriate scheme , then the factor @xmath169 comes out to be universal . as discussed in @xcite this is essentially a consequence of the fact that regardless of whether one accounts for the virtual gluon decay , the distribution of ` non - perturbative ' transverse momentum is independent of rapidity ( the ` tube - model ' of @xcite , based essentially on boost invariance ) . while the factor @xmath169 corrects the @xmath152 at two - loop level , the question of even higher order corrections is still open , although one can argue that such effects will be suppressed by a factor @xmath172 relative to the leading power correction . this argument relies on the hope that the strong coupling remains moderate even in the infrared region . .coefficients of @xmath4 power corrections for @xmath5 event shapes ; @xmath173 . [ cols="^,^,^,^,^,^,^",options="header " , ] tables 1 and 2 display the values of the @xmath152 coefficients obtained in the 2-jet @xmath5 and and @xmath122-jet dis cases respectively . note the different behaviour , proportional to @xmath174 , that arises in the case of the jet broadenings in both processes , a consequence of the fact that the broadening is more sensitive to recoil induced by perturbative radiation @xcite . the improved theoretical understanding that led to this prediction was strongly stimulated by experimental analysis @xcite of features of earlier predictions @xcite that were inconsistent with the data . since the origin of the @xmath174 enhancement is a perturbative , non - perturbative interplay best explained with reference to the power correction for the distribution , we delay its discussion to section [ sec : anldist ] . another consequence of this interplay of the broadening power correction with perturbative radiation is the fact that the dis broadening @xmath175 has a bjorken @xmath176 dependent term corresponding to dglap evolution of the parton densities , as indicated by the term proportional to @xmath177 @xcite , where @xmath178 denotes a parton density function . the coefficients for the other dis variables are @xmath176 independent since in those cases the power correction arises from soft emission alone . we also mention the result for the power correction to another interesting variable , the energy - energy correlation ( eec ) , in the back - to - back region . like the broadening this variable exhibits the impact of perturbative - non perturbative interplay . the results are fractional power corrections that vary as @xmath179 ( for the quark - gluon correlation ) and @xmath180 for the hard quark - antiquark correlation @xcite , instead of @xmath4 and @xmath181 contributions respectively , that would be obtained by considering np emission without the presence of harder perturbative emissions . observables for which the power corrections have yet to be understood include the jet resolution parameters . renormalon based predictions were given in @xcite suggesting a @xmath146 correction for @xmath182 and @xmath147 for @xmath183 . however it seems that in both cases there could be significant perturbative non - perturbative interplay which will complicate the picture . nevertheless , as we have seen in fig . [ fig : meandata ] ( right ) any correction to @xmath184 is certainly small compared to that for other observables . a general point to be kept in mind is that the universality of @xmath163 can be broken by contributions of order @xmath185 , where @xmath186 is the mass scale for hadrons @xcite . whether this happens or not depends on the hadron mass scheme in which the observable is defined ( cf . section [ sec : definitions ] ) . most observables are implicitly in the @xmath62-scheme , which generally involves a small negative breaking of universality ( which is almost observable - independent , so that an illusion of universality will persist ) . the jet masses are an exception and in their usual definitions they have a significant ( positive ) universality breaking component . the @xmath57-scheme is free of universality breaking terms , as is the decay scheme ( since hadron masses are zero ) . we also note that the power correction associated with hadron mass effects is enhanced by an anomalous dimension , @xmath187 with @xmath188 . the normal ` renormalon ' power correction would also be expected to have an anomalous dimension , but it has yet to be calculated . as well as studies of mean values ( first moments ) of observables there has also been work on higher - moments , @xmath189 with @xmath190 . simple renormalon - inspired arguments suggest that the @xmath191 moments of event shapes should have as their leading power correction at most a @xmath192 contribution ( essentially since @xmath193 vanishes at @xmath194 for small @xmath195 ) . however as was pointed out by webber @xcite , the fact ( to be discussed in section [ sec : anldist ] ) that the @xmath146 power correction essentially corresponds to a shift of the distribution ( for @xmath6 , @xmath38 and @xmath196 , but not for the broadenings ) , means that to a first approximation we will have @xmath197 given that @xmath198 is of order @xmath94 , all higher moments of event shapes will receive power corrections of order @xmath199 , which is parametrically larger than @xmath192 . strictly speaking it is not possible to guarantee the coefficient of @xmath199 given in eq . ( [ eq : momentpc ] ) , since the shift approximation on which it is based holds only in the @xmath16-jet region , whereas the dominant contribution to the relevant integral comes from the @xmath65-jet region . for the @xmath38-parameter an alternative coefficient for the @xmath199 power correction has been proposed in @xcite , however since it also is not based on a full calculation of the power correction in the three - jet region ( which does not yet exist ) , it is subject to precisely the same reservations as eq . ( [ eq : momentpc ] ) . we note also an interesting result by gardi @xcite regarding the exact renormalon analysis of @xmath200 . while it is clear that the physical answer @xmath199 will not appear in a leading renormalon analysis ( the extra factor of @xmath94 means that it is associated with a subleading renormalon ) the calculation @xcite has the surprising result that the leading renormalon contribution is not @xmath147 as naively expected but rather @xmath201 . considerable progress has been made in recent years in the calculation of power corrections for three - jet observables ( those that vanish in the two and the three - jet limits ) . in the three - jet limit , there exist explicit results for the thrust minor @xcite and the @xmath39-parameter @xcite in @xmath202 , the out of plane momentum in drell - yan plus jet @xcite and in @xmath90-jet dis @xcite , as well as azimuthal correlations in dis @xcite . except for the @xmath39-parameter , all these calculations involve perturbative , non - perturbative interplay in a manner similar to that of the broadenings and back - to - back eec . the explicit forms for the results are rather complicated and so we refer the reader to the original publications . simpler ( though numerical ) results have been obtained @xcite in the case of the @xmath39-parameter ( confirmed also by @xcite ) integrated over all 3-jet configurations , where one finds @xmath203 the power of @xmath94 comes from the matrix - element weighting of the @xmath65-jet configurations and the fact that the @xmath39-parameter vanishes in the @xmath16-jet limit . a final @xmath65-jet result of interest @xcite is that for the @xmath38-parameter just above the sudakov shoulder , at @xmath125 . this is the only case where a proper calculation exists for the power correction to a two - jet observable in the three jet limit : @xmath204 this is slightly less than half the correction that appears in the two - jet limit ( @xmath205 ) . a point that emerges clearly from eqs . ( [ eq : dpower ] ) and ( [ eq : c34power ] ) , but that is relevant for all @xmath65-jet studies is that the power correction acquires a dependence on @xmath139 , i.e. there is sensitivity to hadronisation from a gluon . in situations where one selects only three - jet events there is additionally non trivial dependence on the geometry , due to the coherence between the three jets . comparisons to data for such observables would therefore allow a powerful test of the renormalon - inspired picture . in particular , other models , such as the flux tube model @xcite ( based essentially on boost invariance along the @xmath206 axis ) , which in the @xmath16-jet limit give the same predictions as renormalon - inspired approaches , can not as naturally be extended to the @xmath65-jet case . one of the most widespread ways of testing the dokshitzer - webber approach to hadronisation is to carry out simultaneous fits to the mean value data for @xmath1 and @xmath163 . figure [ fig : meandata ] , where the term labelled @xmath4 is actually of the form @xmath207 , shows the good agreement that is obtained with the thrust data . the quality of agreement is similar for other observables ( see for example fig . 11 of @xcite ) . the true test of the approach however lies in a verification of the universality hypothesis , namely that @xmath163 is the same for all observables and processes . , one may wonder why one also fits for @xmath1 . the two principal reasons are ( a ) that one is in any case interested in the value of @xmath94 and ( b ) that the data and perturbative prediction differ also by higher - order terms in @xmath94 and fixing @xmath94 would mean trying to fit these higher - order terms with a power correction , which would be misleading . ] while there are strong general reasons for expecting universality within @xmath0 ( boost invariance and the flux - tube model ) , universality across difference processes is less trivial for example one could imagine dis hadronisation being modified by interactions between outgoing low - momentum gluons ( or hadrons ) and the ` cloud ' of partons that make up the proton remnant ( present even for a fast - moving proton , since the longitudinal size of the cloud is always of order @xmath208 ) . figure [ fig : meanfits ] shows @xmath161-@xmath209 contours for fits of @xmath1 and @xmath163 in @xmath0 @xcite ( left , all experiments combined ) and dis ( right , h1 @xcite and zeus @xcite merged into one plot ) . aside from the different scales used , some care is required in reading the figures because of different treatments of the errors in the different plots . for example h1 include experimental systematic errors whereas zeus do not . in the @xmath0 fit , the systematics are included but treated as uncorrelated , which may underestimate the final error . in neither figure are theoretical uncertainties included : over the available range of @xmath0 energies , renormalisation scale dependence gives an uncertainty of about @xmath210 on @xmath94 , while a ` canonical ' variation of the milan factor @xmath211 by @xmath212 ( to allow for higher - order terms ) leads to an uncertainty of @xmath213 on @xmath163 @xcite . in dis the corresponding uncertainties are larger for @xmath94 ( @xmath214 , essentially because the fits are dominated by lower @xmath82 values , and they vary substantially for @xmath163 ( larger for photon - axis observables , smaller for the others ) . also to be kept in mind is that the dis figure includes the jet mass , @xmath215 in the default scheme , and in @xmath0 the origins of the arrows indicate the default - scheme results for @xmath216 and @xmath196 . since the default scheme for jet masses breaks universality ( cf . section [ sec : dwdetails ] ) these results should not be compared directly to those for other observables . , @xmath163 fit to @xmath0 data on mean values of event shapes , with the jet masses converted to the @xmath62-scheme ( the arrow indicates the change in going from default to @xmath62-scheme ) @xcite . right : fits to dis mean event - shape data by h1 @xcite and zeus @xcite ( curves taken from figures in @xcite ) . contours indicate 1-@xmath209 uncertainties ( statistical and experimental systematic , except for the zeus results which are just statistical).,title="fig:",scaledwidth=48.0% ] , @xmath163 fit to @xmath0 data on mean values of event shapes , with the jet masses converted to the @xmath62-scheme ( the arrow indicates the change in going from default to @xmath62-scheme ) @xcite . right : fits to dis mean event - shape data by h1 @xcite and zeus @xcite ( curves taken from figures in @xcite ) . contours indicate 1-@xmath209 uncertainties ( statistical and experimental systematic , except for the zeus results which are just statistical).,title="fig:",scaledwidth=48.0% ] taking into account the uncertainties , which are not necessarily correlated from one observable to another , fig . [ fig : meanfits ] indicates remarkable success for the renormalon - inspired picture . the results are in general consistent with a universal value for @xmath163 in the range @xmath217 to @xmath218 . furthermore the @xmath0 results for @xmath94 are in good agreement with the world average . on the other hand the dis results for @xmath94 seem to be somewhat larger than the world average . the discrepancy is ( just ) within the theoretical systematic errors , however it remains a little disturbing and one wonders whether it could be indicative of large higher - order corrections , or some other problem ( known issues include the fact that the low-@xmath82 data can bias the fits , for example because of heavy - quark effects and @xmath219 corrections ; and also at high-@xmath82 there is @xmath220 interference which is not usually accounted for in the fixed - order calculations ) . there is potentially also some worry about the @xmath221 result which has an anomalously low @xmath163 and large @xmath94 ( and additionally shows unexpectedly substantial @xmath176 dependence @xcite ) . another point relates to the default scheme jet masses though the universality - breaking effects should be purely non - perturbative , they have an effect also on the fit results for @xmath1 . this is a consequence of the anomalous dimension that accompanies @xmath146 hadron mass effects . similar variations of @xmath94 are seen when varying the set of particles that are considered stable @xcite . though as discussed above some issues remain , they should not however be seen as detracting significantly from the overall success of the approach and the general consistency between @xmath0 and dis . we also note that the first moment of the coupling as extracted from studies of heavy - quark fragmentation @xcite is quite similar to the value found for event shapes . there have also been experimental measurements of higher moments of event shapes , notably in @xcite . the parameters @xmath1 and @xmath163 are fixed from fits to mean values and then inserted into eq . ( [ eq : momentpc ] ) to get a prediction for the second moments . for @xmath222 this gives very good agreement with the data ( though the fact that @xmath196 is in the default scheme perhaps complicates the situation ) , while @xmath223 and @xmath224 show a need for a substantial extra correction . in @xcite this extra contribution is shown to be compatible with a @xmath219 term , though it would be interesting to see if it is also compatible with a @xmath225 term with a modified coefficient . currently ( to our knowledge ) no fits have been performed for three - jet event shapes , though we understand that such fits are in progress @xcite for the mean value of the @xmath39-parameter . we look forward eagerly to the results . in the meantime it is possible to verify standard parameters for @xmath94 and @xmath163 against a single published point at @xmath127 @xcite and one finds reasonable agreement within the uncertainties . two other points of view that have also been used in analysing data on @xmath5 event shapes are the approach of _ dressed gluons _ employed by gardi and grunberg @xcite and the use of _ renormalisation group improved _ perturbation theory @xcite as carried out by the delphi collaboration @xcite . in the gardi - grunberg approach one uses the basic concept of renormalons ( on which the dw model is also based ) , but one choses to treat the renormalon integral differently to the dw model . this renormalon integral is ill - defined due to the landau singularity in the running coupling and instead of assuming an infrared finite coupling below some matching scale @xmath160 as was the case in the dw model , gardi and grunberg define the renormalon integral by its principal value ( a discussion of the theoretical merits of different approaches has been give in @xcite ) . in doing this they explicitly include higher order renormalon contributions in their perturbative result , rather than including them via power behaved corrections below scale @xmath160 that result from assuming an infrared finite coupling ( though they do also compare to something similar , which they call a ` cutoff ' approach ) . however since one is dealing with an ambiguous integral ( a prescription other than a principal value one would give a result differing by an amount proportional to a power correction ) one must still allow for a power correction term . gardi and grunberg studied the mean thrust in @xmath5 annihilation and used the following form for fitting to the data : @xmath226+\frac{\lambda}{q}\ ] ] where the subscript pv denotes the principal value of the renormalon integral @xmath227 for the thrust ( which includes the full lo contribution and parts of the higher - order contributions ) and @xmath228 is a piece that accounts for the difference between the true nlo coefficient for @xmath14 and the contribution included in @xmath227 . the best fit values obtained for @xmath229 and @xmath230 are respectively @xmath231 and @xmath232 . we note that @xmath233 in this approach corresponds to a smaller power correction than is required in the dw model . this is probably a consequence of the inclusion of pieces of higher perturbative orders via a principal value prescription , although it leads to a somewhat small value ( compared to the world average ) of the coupling at scale @xmath234 . next we turn to the renormalisation group improved ( rgi ) perturbative estimates @xcite that have also been used to compare with event shape data for the mean value of different event shapes in @xmath5 annihilation . the basic idea behind this approach is to consider the dependence of the observable on the scale @xmath82 , which can be expressed using renormalisation group invariance as @xmath235 where in the above formula @xmath236 , @xmath237 , with @xmath238 being the mean value of a given event shape and @xmath239 being the coefficient of @xmath240 in its perturbative expansion . from comparisons to @xmath5 event - shape data using the rgi approach . the band shows the mean value of @xmath94 and uncertainty , as obtained from the observables below the dashed line . figure taken from ref . @xcite.,scaledwidth=40.0% ] thus the first term in the perturbative expansion of @xmath109 is simply @xmath241 . the @xmath242 are renormalisation scheme independent quantities . in particular the quantity @xmath243 is simply the ratio of the first two coefficients of the qcd @xmath145 function while @xmath244 additionally depends on the first three ( up to nnlo ) perturbative coefficients in the expansion of @xmath245 . current studies using this method are therefore restricted to an nlo analysis involving @xmath246 alone . with this simplification , the solution of eq . ( [ eq : rgi ] ) simply corresponds to the introduction of an observable - specific renormalisation scheme and associated scale parameter @xmath247 which is such that it sets the nlo perturbative term to be zero ( and neglects higher perturbative orders ) . its relation to the standard @xmath248 is easily obtained : @xmath249 where @xmath250 and @xmath251 , with @xmath252 being the nlo coefficient in the @xmath253 expansion for the observable . the terms in the bracket account for different definitions of the coupling in terms of @xmath254 as used in the @xmath109 and @xmath253 schemes . in the specific case of fits to @xmath5 event shape mean values , the above approach has met with a considerable success @xcite . in particular the introduction of @xmath247 seems to remove any need for a significant power correction . the delphi collaboration @xcite have performed a combined fit of the parameter @xmath247 and a parameter @xmath255 that quantifies the power correction @xcite . the value of @xmath255 was found to be consistent with zero in most cases and rather small in all cases which indicates that there is no real need for a power behaved correction once the perturbative expansion is fixed through the rgi technique . the agreement between @xmath1 values extracted from different observables is impressive ( see figure [ fig : rgi ] ) with a spread that is only about half as large as that obtained using the standard perturbative result supplemented by a power correction term , using the dw initiated model . while it is clear that the size of the power correction piece inevitably depends on how one choses to define the perturbative expansion ( in fact all the methods discussed thus far , including the dw model and the gardi - grunberg approach , account for this effect ) , it is nevertheless very interesting that for several different observables , with significantly different perturbative coefficients , one observes after the introduction of rgi , hardly any need for a power correction term . certainly it is not clear on the basis of any physical arguments , why the genuine non - perturbative power correction should be vanishingly small and that a perturbative result defined in a certain way should lead to a complete description of the data . the clarification of this issue is still awaited and the above findings are worth further attention and study . a final point to be kept in mind about the rgi approach in the above form , is that it is valid only for inclusive observables that depend on just a single relevant scale parameter @xmath82 . this makes its applicability somewhat limited and for instance it is not currently clear how to extend the procedure so as to allow a meaningful study of event shape distributions in @xmath5 annihilation or dis event shape mean values and distributions ( which involve additional scales ) . so far our discussion of comparisons between theory and experiment has been limited to a study of mean values of the event shapes . as we have seen , there is some ambiguity in the interpretation of these comparisons , with a range of different approaches being able to fit the same data . this is in many respects an unavoidable limitation of studies of mean values , since the principal characteristic of the different models that is being tested is their @xmath82-dependence , which can also be influenced by a range of ( neglected ) higher - order contributions . in contrast , the full distributions of event shapes contain considerably more information and therefore have the potential to be more discriminatory . in section [ sec : perturbative ] we discussed the perturbative calculation of distributions . as for mean values though , the comparison to data is complicated by the need for non - perturbative corrections . these corrections however involve many more degrees of freedom than for mean values , and there are a variety of ways of including them . accordingly we separate our discussion into two parts : in section [ sec : evgen ] we shall consider studies in which hadronisation corrections are taken from monte carlo event generators , and where the main object of study is the perturbative distribution , with for example fits of the strong coupling . in section [ sec : anldist ] we shall then consider studies which involve analytical models for the non - perturbative corrections , and where it is as much the non - perturbative models , as the perturbative calculations that are under study . compared to a variety of event generators @xcite . the shaded band in the lower plot indicates the experimental statistical and systematic errors added in quadrature . figure taken from @xcite.,title="fig:",scaledwidth=48.0% ] compared to a variety of event generators @xcite . the shaded band in the lower plot indicates the experimental statistical and systematic errors added in quadrature . figure taken from @xcite.,title="fig:",scaledwidth=48.0% ] before considering the purely perturbative studies mentioned above , let us recall that one major use of event shape distributions has been in the testing and tuning @xcite of monte carlo event generators such as herwig @xcite , jetset / pythia @xcite and ariadne @xcite . figure [ fig : tuning ] shows comparisons for two event shapes , the thrust and thrust major , and the agreement is remarkable testimony to the ability of the event generators to reproduce the data . the use of event generators to probe the details of qcd is unfortunately rather difficult , essentially because they contain a number of parameters affecting both the non - perturbative modelling and in some cases the treatment of the perturbative shower . furthermore though considerable progress is being made in matching to fixed order calculations ( see for example @xcite ) , event generators are in general able to guarantee neither the nll accuracy nor the nlo accuracy of full matched nll - nlo resummed calculations . event - shapes ( global @xcite two - jet event shapes ) herwig @xcite is expected to be correct to nll accuracy . ] nevertheless , the good agreement of the event generators with the data suggests that the bulk of the dynamics is correct and in particular that a good model for the hadronisation corrections can be had by comparing parton and hadron ` levels ' of the generator . there has been a very widespread use of event generators in this way to complement the nll+nlo perturbative calculations . such a method has both advantages and drawbacks and it is worth devoting some space to them here . on one hand , the parton level of an event generator is not a theoretically well - defined concept it is regularised with some effective parton mass or transverse momentum cutoff , which already embodies some amount of non - perturbative correction . in contrast the nlo+nll partonic prediction integrates down to zero momenta without any form ( or need ) of regularisation . this too implies some amount of non - perturbative contribution , but of a rather different nature from that included via a cutoff . the resulting difference between the event - generator and the purely perturbative nlo+nll parton levels means that the ` hadronisation ' that must be applied to correct them to hadron level is different in the two cases . there are nevertheless reasons why event - generators are still used for determining the hadronisation corrections . the simplest is perhaps that they give a very good description of the data ( cf . [ fig : tuning ] ) , which suggests that they make a reasonable job of approximating the underlying dynamics . furthermore the good description is obtained with a single set of parameters for all event shapes , whereas as we shall see , other approaches with a single ( common ) parameter are currently able to give equally good descriptions only for a limited number of event shapes at a time . additionally , the objection that the monte carlo parton level is ill - defined can , partially , be addressed by including a systematic error on the hadronisation : it is possible for example to change the internal cutoff on the monte carlo parton shower and at the same time retune the hadronisation parameters in such a way that the monte carlo description of the hadron - level data remains reasonable . in this way one allows for the fact that the connection between monte carlo and nll+nlo ` parton - levels ' is not understood . this is the procedure that has been used for example in @xcite , and the results are illustrated in figure [ fig : sldhadr ] as a hadronisation correction factor @xmath256 with an uncertainty corresponding to the shaded band . , with uncertainties , as discussed in the text ) obtained from the ratio of jetset parton to hadron - level distributions . it is compared to nlo and nll+nlo predictions . also shown is the detector to hadron - level correction factor @xmath257 . figure taken from @xcite . ] two further points are to be made regarding monte carlo hadronisation corrections . firstly , as is starting now to be well known , it is strongly recommended not to correct the data to parton - level , but rather to correct the theoretical perturbative prediction to hadron level . this is because the correction to parton level entails the addition of extra assumptions , which a few years later may be very difficult to deconvolve from the parton - level ` data ' . it is straightforward on the other hand to apply a new hadronisation model to a perturbative calculation . secondly it is best not to treat the hadronisation correction as a simple multiplicative factor . to understand why , one should look again at figure [ fig : sldhadr ] ( which , dating from several years ago , corrects data to parton level ) . the multiplicative factor @xmath256 ( parton@xmath258hadron ) varies very rapidly close to the peak of the distribution and goes from above @xmath161 to below @xmath161 . this is because the distribution itself varies very rapidly in this region and the effect of hadronisation is ( to a first approximation , see below ) to shift the peak to larger values of the observable . but a multiplicative factor , rather than shifting the peak , suppresses it in one place and recreates it ( from non - peak - like structure ) in another . instead of applying a multiplicative correction , the best way to include a monte carlo hadronisation correction is to determine a transfer matrix @xmath259 which describes the fraction of events in bin @xmath56 at parton level that end up in in bin @xmath55 at hadron level . then for a binned perturbative distribution @xmath260 , the binned hadronised distribution @xmath261 is obtained by matrix multiplication , @xmath262 . this has been used for example in @xcite . having considered the basis and methods for including hadronisation corrections from event generators , let us now examine some of the perturbative studies made possible by this approach . the majority of them use nll+nlo perturbative calculations to fit for the strong coupling ( as discussed below some , e.g. @xcite , just use nlo results , which exist for a wider range of observables ) and such fits can be said to be one of the principal results from event - shape studies . in general there is rather good agreement between the data and the resummed predictions ( cf . [ fig : sldhadr ] ) . obtained by fitting nll+nlo event - shape distributions to l3 data at a range of centre of mass energies ( figure taken from @xcite ) . only experimental uncertainties are shown.,scaledwidth=60.0% ] an example of such a study is given in figure [ fig : l3asevln ] , which shows results for fits of @xmath1 to l3 data at a range of energies ( including points below @xmath263 from radiative events which , as discussed earlier , are to be taken with some caution ) . one important result from such studies is an overall value for @xmath1 , which in this case is @xmath264 @xcite , where the first error is experimental and the second is theoretical . similar results have been given by other collaborations , e.g. @xmath265 from aleph @xcite . there is also good evidence for the evolution of the coupling , though leaving out the potentially doubtful radiative points leaves the situation somewhat less clearcut , due to the limited statistics at high energies ( another resummed study which gives good evidence for the running of the coupling uses jade and opal data @xcite , however it is limited to jet rates ) . these and similar results from the other lep collaborations and jade highlight two important points . firstly at higher energies there is a need to combine results from all the lep experiments in order to reduce the statistical errors and so improve the evidence for the running of the coupling in the region above @xmath127 . work in this direction is currently in progress @xcite . so far only preliminary results are available @xcite , and as well as improving the picture of the high - energy running of the coupling , they suggest a slightly smaller value of @xmath1 than that quoted above , more in accord with world averages @xcite . secondly , in the overall result for @xmath266 , by far the most important contribution to the error is that from theoretical uncertainties . however theoretical uncertainties are notoriously difficult to estimate , since they relate to unknown higher orders . systematic investigations of the various sources of uncertainty have been carried out in @xcite and in particular @xcite proposes a standard for the set of sources of uncertainty that ought to be considered , together with an approach for combining the different sources into a single overall uncertainty on @xmath94 . one of the main principles behind the method is that while one may have numerous estimates of sources of theoretical uncertainty , it is not advisable to combine them in quadrature , as one might be tempted to do , because this is likely to lead to a double - counting of uncertainties . rather , one should examine the different sources of uncertainty across the whole range of the distribution and at each point , take the maximum of all sources to build up an uncertainty envelope or band , represented by the shaded area in figure [ fig : lepqcdbands ] ( shown relative to a reference prediction for a ` standard ' theory ) . the overall uncertainty on the coupling ( rather than the distribution ) is given by the range of variation of @xmath94 such that the prediction remains within the band . it is to be noted that this kind of approach is of relevance not just to event - shape studies but also quite generally to any resummed matched calculation , which is inherently subject to many sources of arbitrariness . allowed such that the prediction stays within the uncertainty band.,scaledwidth=63.0% ] one of the drawbacks of nll+nlo studies is that the nll resummed predictions exist for only a fraction of observables . recently there has been an extensive study by the delphi collaboration @xcite of 18 event - shape distributions compared to nlo calculations . since in most cases nlo calculations with a renormalisation scale @xmath267 tend to describe the data rather poorly , they have examined various options for estimating higher orders so as to improve the agreement . their main approach involves a simultaneous fit of the coupling and the renormalisation scale , i.e. the renormalisation scale dependence is taken as a way of parameterising possible higher orders . in such a procedure the choice of fit - range for the observable is rather critical , since the true higher - orders may involve have quite a different structure from that parameterised by the @xmath168 dependence . the delphi analysis restricts the fit range to be that in which a good fit can be obtained . with these conditions they find relatively consistent values of @xmath94 across their whole range of observables ( finding a scatter of a few percent ) , with a relatively modest theoretical uncertainty , as estimated from the further variation of @xmath168 by a factor of @xmath13 to @xmath16 around the optimal scale . their overall result from all observables treated in this way is @xmath268 ( including both experimental and theoretical errors ) . they also compare this approach to a range of other ways of ` estimating ' higher orders . they find for example that some theoretically based methods for setting the scale ( principle of minimal sensitivity @xcite , effective charge @xcite ) lead to scales that are quite correlated to the experimentally fitted ones , while another method ( blm @xcite ) is rather uncorrelated . they compare also to resummation approaches , though there is no equivalent renormalisation scale choice that can be made . instead the comparison involves rather the quality of fit to the distribution and the final spread of final @xmath1 values as a measure of neglected higher - order uncertainties . not surprisingly , they find that while the nll results work well in the two - jet region , in the three - jet region the combined nll+nlo predictions fare not that much better than the pure ( @xmath267 ) nlo results . they go on to argue that nll+nlo is somewhat disfavoured compared to nlo with an ` optimal ' scale . this statement is however to be treated with some caution , since the fit range was chosen specifically so as to obtain a good fit for nlo with the ` optimal ' scale one could equally have chosen a fit range tuned to the nll+nlo prediction and one wonders whether the nlo with optimal scale choice would then have fared so well . regardless of this issue of fit range , it is interesting to note that over the set of observables which can be treated by both methods , the total spread in @xmath1 values is quite similar , being @xmath269 in the case of nlo with optimal scale choice and @xmath270 for nll+nlo . this is similar to the spread seen for fits to mean values with analytical hadronisation models ( fig . [ fig : meanfits ] ) , though the rgi fits of @xcite have a somewhat smaller error . the jet rates ( especially the durham and cambridge algorithms and certain members of the jade family ) are among the few observables for which the pure nlo calculation gives a reasonable description of the distribution ( cf . table 3 of @xcite ) . one particularly interesting set of nlo studies makes use of the 3-jet rate as a function of @xmath271 ( this is just the integral of the distribution of @xmath137 ) in events with primary @xmath144-quarks as compared to light - quark events @xcite ( some of the analyses use other observables , such as the 4-jet rate or the thrust ) . using nlo calculations which account for massive quarks @xcite makes it possible , in such studies , to extract a value for the @xmath144 mass at a renormalisation scale of @xmath263 , giving first evidence of the ( expected ) running of the @xmath144-quark mass , since all the analyses find @xmath272 in the range @xmath273 to @xmath274 gev ( with rather variable estimates of the theoretical error ) . we have already seen , in section [ sec : meanpc ] , that analytical models for hadronisation corrections , combined with normal perturbative predictions , can give a rather good description of mean values . there the inclusion of a non - perturbative ( hadronisation ) correction was a rather simple affair , eq . ( [ eq : meancorrections ] ) , since it was an additive procedure . in contrast the general relation between the full distribution @xmath275 for an observable @xmath276 and the perturbatively calculated distribution @xmath277 is more complicated , @xmath278 where @xmath279 encodes all the non - perturbative information . ( [ eq : shapegenconvoluion ] ) can be seen as a generalised convolution , where the shape of the convolution function depends on the value @xmath86 of the variable as well as the coupling and hard scale . such a general form contains far more information however than can currently be predicted , or even conveniently parameterised . accordingly analytical hadronisation approaches generally make a number of simplifying assumptions , intended to be valid for some restricted range of the observable . the most radical simplification of eq . ( [ eq : shapegenconvoluion ] ) that can be made is to replace @xmath280 with a @xmath281-function , @xmath282 , leading to @xmath283 this was proposed and investigated in @xcite , and the combination @xmath284 is the same that appears in the power correction for the mean value @xcite , discussed in section [ sec : meanpc ] . ( [ eq : npshift ] ) holds for observables with an @xmath1 independent power correction in tables [ tab : cvs ] and [ tab : cvsdis ] , in the region @xmath285 , where the lower limit ensures that one can neglect the width of the convolution function @xmath286 , while the upper limit is the restriction that one be in the born limit ( in which the @xmath152 were originally calculated ) . for certain 3-jet observables a similar picture holds , but with a shift that depends on the kinematics of the 3-jet configuration @xcite . for the broadenings , the situation is more complicated because the power correction is enhanced by the rapidity over which the quark and thrust axes can be considered to coincide , @xmath287 , @xmath288 being the angle between thrust and quark axes this angle is strongly correlated with the value of the broadening ( determined by perturbative radiation ) and accordingly the extent of the shift becomes @xmath252-dependent @xcite . the simplest case is the @xmath0 wide - jet broadening , for which one has @xmath289 where the shift is written for the integrated distribution @xmath290 in order to simplify the expressions . the @xmath291 enhancement for the power correction to the mean broadening ( tables [ tab : cvs ] and [ tab : cvsdis ] ) comes about simply because the average value of @xmath292 , after integration over @xmath86 with the resummed distribution , is of order @xmath293 . the full form for @xmath294 and analogous results for the total @xmath0 and dis broadenings have been given in @xcite , with results existing also for the thrust minor @xcite , and the dis and drell - yan out - of plane momenta @xcite . yet subtler instances of perturbative , non - perturbative interplay arise for observables like the eec @xcite and dis azimuthal correlation @xcite , with the appearance of _ fractional _ powers of @xmath82 in the power correction , as mentioned before . -dependent non - perturbative shift in the jet broadening distribution for @xmath5 annihilation . the dashed curve is the perturbative prediction ( nll+nlo ) , while the solid curve includes a non - perturbative ( @xmath252-dependent ) shift . figure taken from @xcite . right : nll+nlo distribution with a @xmath4 shift for @xmath295 in dis at a range of @xmath82 values ( round dots indicate points used in fits ) @xcite . the @xmath82 values range from 15 gev ( bottom ) to 81 gev ( top).,title="fig:",scaledwidth=44.0%,scaledwidth=48.0% ] -dependent non - perturbative shift in the jet broadening distribution for @xmath5 annihilation . the dashed curve is the perturbative prediction ( nll+nlo ) , while the solid curve includes a non - perturbative ( @xmath252-dependent ) shift . figure taken from @xcite . right : nll+nlo distribution with a @xmath4 shift for @xmath295 in dis at a range of @xmath82 values ( round dots indicate points used in fits ) @xcite . the @xmath82 values range from 15 gev ( bottom ) to 81 gev ( top).,title="fig:",scaledwidth=48.0% ] one might a priori have thought that the formal domain of validity of the shift approximation , @xmath296 , would be somewhat limited . figure [ fig : broadpc ] ( left ) illustrates what happens in the case of @xmath297 quite remarkably the shift describes the data well over a large range of @xmath297 , with slight problems only in the extreme two - jet region ( peak of the distribution ) and in the four - jet region ( @xmath298 ) . similar features are seen in the right - hand plot of fig . [ fig : broadpc ] for @xmath299 in dis at a range of @xmath82 values : a large part of the distribution is described for all @xmath82 values , and problems appear only in the @xmath91 jet region ( @xmath300 ) , and in the @xmath122 jet region for low @xmath82 . this success is reproduced for quite a range of observables in @xmath0 and dis . -@xmath209 contours for simultaneous fits of @xmath94 and @xmath163 to @xmath0 event - shape distributions . the error contours account for statistic , systematical and theoretical uncertainties . the shaded band indicates the result for @xmath1 when using monte carlo hadronisation models . figure taken from @xcite . right : @xmath161-@xmath209 contours from fits to h1 dis event shape distributions @xcite with statistical and systematic errors added in quadrature . figure taken from @xcite.,title="fig:",scaledwidth=48.0% ] -@xmath209 contours for simultaneous fits of @xmath94 and @xmath163 to @xmath0 event - shape distributions . the error contours account for statistic , systematical and theoretical uncertainties . the shaded band indicates the result for @xmath1 when using monte carlo hadronisation models . figure taken from @xcite . right : @xmath161-@xmath209 contours from fits to h1 dis event shape distributions @xcite with statistical and systematic errors added in quadrature . figure taken from @xcite.,title="fig:",scaledwidth=48.0% ] as was the case for mean values , a more systematic test involves carrying out a simultaneous fit of @xmath94 and @xmath163 for each observable and then checking for consistency between observables and with the corresponding results for mean values . results from @xmath0 and dis are shown in figure [ fig : distcont ] , taken from @xcite and @xcite . other fits that have been carried out in recent years include @xcite .- parameter in @xmath65-jet events . ] just as for the mean values , one notes that ( with exception of @xmath301 and @xmath302 in @xmath0 , to be discussed shortly ) there is good consistency between observables , both for @xmath0 and dis . this statement holds holds also for the eec , fitted in @xcite , not shown in fig . [ fig : distcont ] . there is also good agreement with the results for the mean values , fig . [ fig : meanfits ] , except marginally for dis @xmath94 results , which for distributions are in better accord with the world average . the good agreement between distributions and mean values is important in the light of alternative approaches to fitting mean values , such as the rgi which , we recall , seems to suggest that the power correction for mean values can just as well be interpreted as higher order contributions . were this really the case then one would expect to see no relation between @xmath163 results for mean values and distributions . despite the generally good agreement , some problems do persist . examining different groups results for the @xmath1-@xmath163 fits one finds that for some observables there is a large spread in the results . for example for the @xmath0 @xmath297 , aleph @xcite find @xmath303 while the jade result @xcite shown in fig . [ fig : distcont ] corresponds to @xmath304 ( and this differs from a slightly earlier jade result @xmath305 @xcite ) . while the results all agree to within errors , those errors are largely theoretical and would be expected to be a common systematic on all results . in other words we would expect the results to be much closer together than @xmath161-@xmath209 . that this is not the case suggests that the fits might be unstable with respect to small changes in details of the fit , for example the fit range ( as has been observed when including data in the range @xmath306 in the fit corresponding to the right hand plot of figure [ fig : distcont ] @xcite ) . these differences can lead to contradictory conclusions regarding the success of the description of different observables and it would be of value for the different groups to work together to understand the origin of the differences , perhaps in a context such as the lep qcd working group . there do remain two cases however where there seems to be clear incompatibility between the data and theoretical prediction : the heavy - jet mass , @xmath196 ( also referred to as @xmath302 ) and the wide - jet broadening , @xmath301 . the heavy - jet mass is subject to @xmath146 hadron mass effects , as discussed already in the context of mean values , however even after accounting for this by considering @xmath62 or @xmath57-scheme data , as has been done by delphi @xcite , the results appear to be rather inconsistent with other observables : @xmath1 is too small and @xmath163 too large , as for @xmath301 , with a fixed @xmath307 lead to a consistent value for @xmath308 , though the overall fit quality is poor . ] though less extreme . the common point between these two observables is that they both select the heavier / wider hemisphere and so are less inclusive than other observables . studies with event generators @xcite suggest that whereas for more inclusive observables the 2-jet limit of the shift works well even into the 3-jet region , for heavy - wide observables the 3-jet limit behaves very differently from the 2-jet limit and these differences becomes relevant even for relatively low values of the observable ( some partially related discussion of the problem has been given also in @xcite ) . we will come back to this and other potential explanations below when we discuss shape functions . beyond the ` usual ' observables discussed above , there are some recent results also for the @xmath38-parameter shoulder at @xmath125 , where @xcite the shoulder position is seen to be consistent with a shift with the usual 2-jet coefficient @xmath205 this is in contradiction with the ( smaller ) result eq . ( [ eq : c34power ] ) for the shift to the shoulder ; however the analysis @xcite does nt take into account resummation effects at the shoulder which may also contribute to an effective shift . a first analysis of corrections to the @xmath309 distribution has been given in @xcite which shows evidence for @xmath310 contributions in the @xmath309 distribution at the two lowest jade centre of mass energies ( 14 and 22 gev ) . one last interesting point to discuss regarding the shift approximation for hadronisation is an application to the measurement of the qcd colour factors . the fact that the hadronisation is encoded via just a single parameter , with all the dependence on colour factors available explicitly for the perturbative and non - perturbative contributions , means that it is feasible to carry out a simultaneous fit not only for @xmath311 and @xmath163 but also for one or more of the qcd colour factors @xmath139 , @xmath138 and @xmath312 @xcite . this would have been much more difficult using monte carlo hadronisation because all the hadronisation parameters would have had to have been retuned for each change of the colour factors . the results for this fit are summarised and compared to other approaches in figure [ fig : colfact ] @xcite , which shows @xmath16-@xmath209 contours for @xmath138 and @xmath139 . while other approaches constrain the ratio @xmath313 , the event - shape approach gives by far the best independent measurement of @xmath139 and @xmath138 . as a result the combination of the results gives rather tight limits on the colour factors , which are in good agreement with the qcd expectation . ( shift ) power corrections and a variety of other methods . figure taken from @xcite . the ellipses show correlated measurements from studies of the angular distributions of 4-jet events @xcite and event shapes @xcite , while the lines are obtained from studies of the gluon fragmentation function @xcite and from multiplicities in gluon jets @xcite.,scaledwidth=51.0% ] we have already mentioned that the approximation of a simple shift is valid only in the region @xmath296 . to better understand the origin of the lower limit , let us expand eq . ( [ eq : npshift ] ) , @xmath314 where @xmath315 is the @xmath316 derivative of @xmath317 with respect to @xmath86 . since @xmath315 goes as @xmath318 one sees that when the shift @xmath319 all terms in the series are of the same order . while the @xmath320 term can be related to the power correction to the mean value of @xmath276 , the shift merely provides an ansatz for the higher terms . the breakdown for small values of @xmath321 is clearly visible for the ( dis ) thrust in fig . [ fig : broadpc ] ( right ) . a physically transparent way of dealing with the @xmath322 terms has been developed by korchemsky , sterman and collaborators @xcite . one approximates @xmath323 in eq . ( [ eq : shapegenconvoluion ] ) by a _ shape function _ @xcite @xmath324 ( rather than the @xmath281-function which corresponds to a pure shift ) , @xmath325 in eq . ( [ eq : expandedshift ] ) , @xmath326 is then replaced by @xmath327 where @xmath328 is the @xmath316 moment of @xmath329 . there is considerable freedom in one s choice of the form for the shape function . accordingly , success in fitting a given event - shape distribution for any single value of @xmath82 can not , alone , be considered strong evidence in favour of the shape - function picture . the true test of the approach comes by determining the shape function at one value of @xmath82 and then establishing whether it applies for all values of @xmath82 . this is illustrated for the @xmath0 thrust in fig . [ fig : thrustshape ] ( left ) , where the following shape - function has been used @xcite , @xmath330 with two free parameters , @xmath331 and @xmath254 ( it has been argued that the behaviour of this function for small @xmath176 and large @xmath176 is similar to that expected from a simple physical model for the dynamics at the origin of the shape function @xcite ) . one notes the remarkable agreement , far into the two - jet region , for the whole range of @xmath82 values . ) . shown for @xmath332 and @xmath333 gev . figure taken from @xcite . right : the thrust distribution at the perturbative ( dge ) level , with a non - perturbative shift , and with a shape function . figure taken from @xcite.,title="fig:",scaledwidth=48.0%,scaledwidth=48.0% ] ) . shown for @xmath332 and @xmath333 gev . figure taken from @xcite . right : the thrust distribution at the perturbative ( dge ) level , with a non - perturbative shift , and with a shape function . figure taken from @xcite.,title="fig:",scaledwidth=48.0%,scaledwidth=48.0% ] this is a clear success for the shape - function approach as compared to the simple shift discussed above . however the drawback of shape - functions is that a priori one loses all predictivity ( save for the first moment ) when going from one observable to another . this predictivity was one of the main strengths of the shift approach . considerable work has therefore been devoted to understanding both the general properties of shape functions and the relations between shape functions of different observables . one line of investigation has been to relate the shape functions for the thrust and heavy - jet mass ( @xmath196 ) this can be done with the help of the observation @xcite that in the two - jet limit the thrust is just the sum of the light and heavy - jet ( squared normalised ) masses . in @xcite korchemsky and tafat introduced a shape function @xmath334 describing the distribution of non - perturbative radiation in each of the two hemispheres . the shape function for the thrust is then obtained by @xmath335 instead for the heavy - jet mass one calculates the full ( np ) double differential distribution for the two jet masses and from that derives the heavy - jet mass distribution ( this is not quite equivalent to a simple shape - function for @xmath196 ) . this approach gave a good fit to the thrust and heavy - jet mass data , on the condition that there was significant correlation between the non - perturbative contribution to the two hemispheres .- parameter ( with an overall rescaling to ensure the correct first moment ) , however the data were less well described than for the thrust . this is perhaps not surprising : owing to different sensitivities to large - angle radiation , the thrust and @xmath38-parameter shape functions are expected to have a different structure of higher moments @xcite . ] it is difficult to interpret this result however , because as has already been mentioned , the default measurement scheme for the jet masses suffers from extra ( universality - breaking ) @xmath146 corrections , and it is possible that in @xcite these are being mimicked by inter - hemisphere correlations . the thrust and heavy - jet mass have been investigated in great detail also by gardi and rathsman @xcite . in analogy with the approach for the mean thrust @xcite , they attempt not only to account for nll+nlo and power correction contributions , but also for all ( perturbative ) enhanced effects associated with the running of the coupling , within a formalism that they refer to as dressed gluon exponentiation ( dge ) . among the interesting theoretical results ( see also @xcite ) is that the renormalon analysis suggests that the second moment of the shape function should be precisely that expected from a pure shift ( the same is true for the @xmath38-parameter @xcite ) . this observation seems to be consistent with fits to the data , and would explain why the shift approximation works so well , at least for the relevant observables . the right - hand plot of fig . [ fig : thrustshape ] shows the result of their analysis for the thrust , with both a shift and a shape - function . while the shape - function is clearly needed in the immediate vicinity of the peak and below , it is striking to see , above the peak , how well it is approximated by the shift . one of the other interesting results in @xcite relates to the analysis of the heavy - jet mass . in contrast to @xcite , a common hadron - mass scheme ( decay - scheme ) is used for the thrust and the heavy - jet mass , and there are no inter - hemisphere correlations in their shape function . using the same parameters for the thrust and @xmath196 leads to reasonable fits , suggesting that the true inter - hemisphere correlations may actually be small . ( a fit for @xmath196 alone does however prefer a lower value of @xmath94 than for the thrust , as is seen also with the usual shifted nll+nlo approach @xcite ) . a puzzle that emerges in the work of @xcite is that all the fits lead to quite small values of the coupling , @xmath336 with a theoretical error of about @xmath210 . a similarly low result was obtained in the analogous analysis of the mean thrust @xcite . in contrast a study of @xmath144-fragmentation using dressed gluon exponentiation finds more standard values , @xmath337 @xcite . let us close our discussion of shape functions with a recent result by berger & sterman @xcite . given that there are no simple relations between the shape functions for most common event shapes , they instead consider a new class of event shapes @xcite whose definition involves a parameter @xmath331 which allows the nature of the event shape to be continuously varied from a broadening ( @xmath338 ) to a thrust - type observable ( @xmath339 ) and beyond . their observation is that for all variants of the observable with @xmath340 , the logarithm of any moment of the shape functions scales as @xmath341 . in practice this scaling is found to hold in monte carlo studies of this class of observables it would therefore be interesting to have data for these observables with which to make a comparison . conceptually , event shapes are rather simple observables , yet they are sensitive to a range of characteristics of qcd radiation . this combination of theoretical simplicity and experimental sensitivity is perhaps one of the main reasons why event shapes have found so many applications . as we have seen , they provide vital inputs in studies of the ingredients of the qcd lagrangian , such as the coupling and the colour factors ; they play an important role in the tuning and testing of event generators ; and they are at the heart of recent investigations into analytical approaches to understanding the dynamics of hadronisation . there remain several directions in which progress may be expected ( or at the very least hoped for ) in coming years . all the experimental comparisons discussed here have been for event - shapes that vanish in the @xmath16-jet limit , yet there have been significant theoretical developments in recent years for observables that vanish also in the @xmath65-jet limit , examples being the @xmath39-parameter and thrust minor @xcite . studies with jet rates @xcite suggest that going to such ` three - jet ' observables could reduce theoretical errors , though as discussed in @xcite this needs to be investigated more systematically . additionally three - jet observables will allow much more stringent tests of analytical hadronisation approaches since for the first time they introduce sensitivity to predictions about hadronisation from a gluon @xcite , not just from a quark . one of the frontiers of event shape studies is their extension to new processes . while @xmath0 is the traditional domain for event - shape studies , we have seen that studies in dis are in many respects competitive with @xmath0 results . additionally they provide important input on the question of universality of analytical hadronisation models , ruling out for example significant modifications of the hadronisation picture stemming from interactions with the proton remnant . though so far the question has received only limited study , one interesting application of dis studies would be in the use of event shapes for obtaining information on parton distribution functions , as suggested in @xcite . currently an option that is attracting growing experimental @xcite and theoretical @xcite interest is the development of event shapes in hadron - hadron collisions . though much work remains to be done on this subject , it seems that the sensitivity to parton distributions will be somewhat stronger than in most dis observables . furthermore one expects a rich hadronisation structure associated both with the four - jet structure of hadronic dijet events and with the properties of the underlying event . it is to be hoped that the availability of automated resummation methods @xcite will make the theoretical aspects of the study of new observables and processes more straightforward than has been the case in the past . another frontier on which progress is expected in near future is with respect to the accuracy of theoretical predictions . nnlo perturbative calculations are making rapid progress ( for a recent review see @xcite ) with the main outstanding problem being that of a full subtraction procedure for combining zero , one and two - loop contributions and its implementation in a monte carlo integrator . progress on nnll resummations is also being made , albeit so far only in inclusive cases @xcite . though technically more involved , an extension to event shape resummations is certainly conceivable in the near future . aside from the expected gains in accuracy that are the main motivation for these calculations , it will also be interesting to compare the exact higher - order calculations with predictions from approaches such as @xcite which aim to identify the physically dominant higher - order contributions and in some cases @xcite claim to significantly reduce the theoretical uncertainties on @xmath94 . less predictable is what development can be expected on the subject of hadronisation corrections . as we have seen , much has been learnt from event - shape studies . yet a variety of open questions remain . at a practical level it would for example be of interest to see an extension of the shape - function approach to a wider range of variables , including the quite subtle case of the broadenings and also to dis event shapes . the wealth of dis data at low @xmath82 , where shape functions are most important , makes the latter especially interesting . there remain also important issues that are poorly understood even at the conceptual level . for example predictions are currently limited to the domain in which the event - shape is close to vanishing ( e.g. @xmath14 in the @xmath16-jet limit ) . it is tempting to suggest that problems that persist for a couple of ` recalcitrant ' observables ( specifically , @xmath196 , @xmath301 ) may actually be due to large differences between the power correction in the three - jet region ( a significant part of the fit - range ) and that in the well - understood two - jet limit ( see e.g. fig . 3 of @xcite ) . but techniques that would allow the calculation of the power correction to , say , the thrust or the broadening in the @xmath65-jet limit do not yet exist . another ( partially related ) issue is that of anomalous dimensions these have so far only been calculated for @xmath146 effects that are associated with hadron masses @xcite . yet , as has been pointed out also in @xcite , they are bound to be present for all classes of power suppressed contributions . one clear physical origin for them is that soft - gluons ( or hadrons ) with transverse momenta of order @xmath254 are radiated not just from the @xmath206 system but also from all soft perturbative gluons with transverse momenta between @xmath254 and @xmath82 . accounting for radiation from the latter will clearly lead to enhancements of power corrections by terms @xmath342 . and a final point , not to be forgotten , is that most of the analytical hadronisation approaches are essentially perturbative methods ` in disguise ' . an understanding of how they really relate to _ _ hadron__isation is very much lacking . let us close this review by remarking that one of the characteristics of event shape studies is that progress is made not merely through a better understanding of existing observables and data , but also through critical experimental studies of new observables with ( supposedly ! ) better theoretical properties . the dialogue between experimenters and theorists is crucial in this respect . we look forward to its being as fruitful in coming years as it has been up to now . it goes without saying that we are grateful to our collaborators and numerous other colleagues , including members of the lep qcd working group , for the many stimulating discussions that we have had about event shapes . for discussions on issues that arose specifically while writing this review , we thank in particular g. heinrich , f. krauss , c. j. maxwell , g. rodrigo , m. h. seymour , h. stenzel and g. zanderighi . furthermore we are indebted to a. banfi , yu . l. dokshitzer and g. zanderighi for their careful reading of and helpful comments on the manuscript . one of us ( gps ) is grateful also to the cern th division for hospitality during the course of this work . 999 j. r. ellis , m. k. gaillard and g. g. ross , nucl . b * 111 * ( 1976 ) 253 [ erratum - ibid . b * 130 * ( 1977 ) 516 ] . e. farhi , phys . * 39 * ( 1977 ) 1587 . h. georgi and m. machacek , phys . * 39 * ( 1977 ) 1237 . g. c. fox and s. wolfram , phys . * 41 * ( 1978 ) 1581 ; ibid . b * 149 * ( 1979 ) 413 [ erratum - ibid . b * 157 * ( 1979 ) 543 ] . c. l. basham , l. s. brown , s. d. ellis and s. t. love , phys . * 41 * ( 1978 ) 1585 ; c. l. basham , l. s. brown , s. d. ellis and s. t. love , phys . d * 19 * ( 1979 ) 2018 . r. k. ellis , w. j. stirling and b. r. webber , _ qcd and collider physics _ , cambridge university press , 1996 . g. marchesini , b. r. webber , g. abbiendi , i. g. knowles , m. h. seymour and l. stanco , comput . commun . * 67 * ( 1992 ) 465 . g. corcella _ et al . _ , jhep * 0101 * ( 2001 ) 010 [ hep - ph/0011363 ] . t. sjstrand , comput . commun . * 82 * ( 1994 ) 74 ; t. sjstrand , p. eden , c. friberg , l. lnnblad , g. miu , s. mrenna and e. norrbin , hep - ph/0010017 . l. lnnblad , comput . * 71 * ( 1992 ) 15 . m. beneke , phys . * 317 * ( 1999 ) 1 [ hep - ph/9807443 ] . s. brandt , c. peyrou , r. sosnowski and a. wroblewski , phys . * 12 * ( 1964 ) 57 . l. clavelli , phys . b * 85 * ( 1979 ) 111 ; t. chandramohan and l. clavelli , nucl . b * 184 * ( 1981 ) 365 ; l. clavelli and d. wyler , phys . b * 103 * ( 1981 ) 383 . s. catani , g. turnock and b. r. webber , phys . b * 295 * ( 1992 ) 269 . d. p. barber _ et al . _ [ mark - j collaboration collaboration ] , phys . * 63 * ( 1980 ) 337 . a. banfi , g. marchesini , y. l. dokshitzer and g. zanderighi , jhep * 0007 * ( 2000 ) 002 [ hep - ph/0004027 ] . a. banfi , y. l. dokshitzer , g. marchesini and g. zanderighi , phys . b * 508 * ( 2001 ) 269 [ hep - ph/0010267 ] . a. de rujula , j. r. ellis , e. g. floratos and m. k. gaillard , nucl . b * 138 * ( 1978 ) 387 . r. k. ellis , d. a. ross and a. e. terrano , nucl . b * 178 * ( 1981 ) 421 . g. parisi , phys . b * 74 * ( 1978 ) 65 ; j. f. donoghue , f. e. low and s. y. pi , phys . rev . d * 20 * ( 1979 ) 2759 . j. d. bjorken and s. j. brodsky , phys . d * 1 * ( 1970 ) 1416 ; g. hanson _ et al . _ , phys . * 35 * ( 1975 ) 1609 . c. berger _ [ pluto collaboration ] , phys . b * 82 * ( 1979 ) 449 . w. bartel _ et al . _ [ jade collaboration ] , z. phys . c * 33 * ( 1986 ) 23 . s. catani , yu . l. dokshitzer , m. olsson , g. turnock and b. r. webber , phys . b * 269 * ( 1991 ) 432 . s. bethke , z. kunszt , d. e. soper and w. j. stirling , nucl . b * 370 * ( 1992 ) 310 [ erratum - ibid . b * 523 * ( 1998 ) 681 ] . y. l. dokshitzer , g. d. leder , s. moretti and b. r. webber , jhep * 9708 * ( 1997 ) 001 [ hep - ph/9707323 ] . g. p. salam and d. wicke , jhep * 0105 * ( 2001 ) 061 [ hep - ph/0102343 ] . e. gardi and j. rathsman , nucl . b * 638 * ( 2002 ) 243 [ hep - ph/0201019 ] . r. p. feynman , _ photon hadron interactions _ , w. a. benjamin , new york , 1972 ; r. d. peccei and r. ruckl , nucl . b * 162 * ( 1980 ) 125 ; l. v. gribov , y. l. dokshitzer , s. i. troian and v. a. khoze , sov . jetp * 67 * ( 1988 ) 1303 [ zh . fiz . * 94 * ( 1988 ) 12 ] . k. h. streng , t. f. walsh and p. m. zerwas , z. phys . c * 2 * ( 1979 ) 237 . a. banfi , g. marchesini , g. smye and g. zanderighi , jhep * 0111 * ( 2001 ) 066 [ hep - ph/0111157 ] . m. dasgupta and g. p. salam , jhep * 0208 * ( 2002 ) 032 [ hep - ph/0208073 ] . a. banfi , g. marchesini and g. smye , jhep * 0204 * ( 2002 ) 024 [ hep - ph/0203150 ] . s. catani , y. l. dokshitzer and b. r. webber , phys . b * 285 * ( 1992 ) 291 . a. banfi , g. marchesini , g. smye and g. zanderighi , jhep * 0108 * ( 2001 ) 047 [ hep - ph/0106278 ] . i. a. bertram [ d0 collaboration ] , acta phys . b * 33 * ( 2002 ) 3141 . a. banfi , g. p. salam and g. zanderighi , hep - ph/0304148 ; _ ibid . _ in preparation . a. banfi , hep - ph/0309161 . g. sterman and s. weinberg , phys . * 39 * ( 1977 ) 1436 . w. t. giele and e. w. glover , phys . d * 46 * ( 1992 ) 1980 . s. catani and m. h. seymour , phys . b * 378 * ( 1996 ) 287 [ hep - ph/9602277 ] ; s. catani and m. h. seymour , nucl . b * 485 * ( 1997 ) 291 [ erratum - imbid . b * 510 * ( 1997 ) 503 ] [ hep - ph/9605323 ] . l. j. dixon and a. signer , phys . d * 56 * ( 1997 ) 4031 [ hep - ph/9706285 ] ; a. signer , comput . commun . * 106 * ( 1997 ) 125 . s. weinzierl and d. a. kosower , phys . d * 60 * ( 1999 ) 054028 [ hep - ph/9901277 ] . j. m. campbell , m. a. cullen and e. w. glover , eur . j. c * 9 * ( 1999 ) 245 [ hep - ph/9809429 ] . z. nagy and z. trocsanyi , nucl . b * 486 * ( 1997 ) 189 [ hep - ph/9610498 ] ; z. nagy and z. trocsanyi , phys . * 79 * ( 1997 ) 3604 [ hep - ph/9707309 ] ; z. nagy and z. trocsanyi , phys . d * 59 * ( 1999 ) 014020 [ erratum - ibid . d * 62 * ( 2000 ) 099902 ] [ hep - ph/9806317 ] ; z. nagy and z. trocsanyi , phys . lett . * 87 * ( 2001 ) 082001 [ hep - ph/0104315 ] ; z. nagy , phys . * 88 * ( 2002 ) 122003 [ hep - ph/0110315 ] ; z. nagy , hep - ph/0307268 . b. potter , comput . phys . commun . * 133 * ( 2000 ) 105 [ hep - ph/9911221 ] ; b. potter , comput . commun . * 119 * ( 1999 ) 45 [ hep - ph/9806437 ] . t. gehrmann , hep - ph/0310178 . m. dasgupta and g. p. salam , phys . b * 512 * ( 2001 ) 323 [ hep - ph/0104277 ] . n. brown and w. j. stirling , phys . b * 252 * ( 1990 ) 657 . g. leder , nucl . b * 497 * ( 1997 ) 334 [ hep - ph/9610552 ] . s. catani , g. turnock , b. r. webber and l. trentadue , phys . b * 263 * ( 1991 ) 491 . s. catani , l. trentadue , g. turnock and b. r. webber , nucl . b * 407 * ( 1993 ) 3 . s. catani , g. turnock and b. r. webber , phys . b * 272 * ( 1991 ) 368 . s. j. burby and e. w. glover , jhep * 0104 * ( 2001 ) 029 [ hep - ph/0101226 ] . y. l. dokshitzer , a. lucenti , g. marchesini and g. p. salam , jhep * 9801 * ( 1998 ) 011 [ hep - ph/9801324 ] . s. catani and b. r. webber , phys . b * 427 * ( 1998 ) 377 [ hep - ph/9801350 ] . g. dissertori and m. schmelling , phys . b * 361 * ( 1995 ) 167 . a. banfi , g. p. salam and g. zanderighi , jhep * 0201 * ( 2002 ) 018 [ hep - ph/0112156 ] . y. l. dokshitzer and d. diakonov , phys . b * 84 * ( 1979 ) 234 . g. parisi and r. petronzio , nucl . b * 154 * ( 1979 ) 427 . j. kodaira and l. trentadue , phys . b * 123 * ( 1983 ) 335 . j. c. collins and d. e. soper , phys . * 48 * ( 1982 ) 655 . p. e. rakow and b. r. webber , nucl . b * 187 * ( 1981 ) 254 . y. l. dokshitzer , g. marchesini and b. r. webber , jhep * 9907 * ( 1999 ) 012 [ hep - ph/9905339 ] . c. f. berger , t. kucs and g. sterman , phys . d * 68 * ( 2003 ) 014012 [ hep - ph/0303051 ] . a. banfi , y. l. dokshitzer , g. marchesini and g. zanderighi , jhep * 0105 * ( 2001 ) 040 [ hep - ph/0104162 ] . v. antonelli , m. dasgupta and g. p. salam , jhep * 0002 * ( 2000 ) 001 [ hep - ph/9912488 ] . m. dasgupta and g. p. salam , eur . j. c * 24 * ( 2002 ) 213 [ hep - ph/0110213 ] . r. bonciani , s. catani , m. l. mangano and p. nason , phys . b * 575 * ( 2003 ) 268 [ hep - ph/0307035 ] . e. gardi and j. rathsman , nucl . b * 609 * ( 2001 ) 123 [ hep - ph/0103217 ] . e. gardi and l. magnea , `` the c parameter distribution in e+ e- annihilation , '' jhep * 0308 * ( 2003 ) 030 [ hep - ph/0306094 ] . f. krauss and g. rodrigo , phys . b * 576 * ( 2003 ) 135 [ hep - ph/0303038 ] . s. catani and b. r. webber , jhep * 9710 * ( 1997 ) 005 [ hep - ph/9710333 ] . m. dasgupta and g. p. salam , jhep * 0203 * ( 2002 ) 017 [ hep - ph/0203009 ] ; c. f. berger , t. kucs and g. sterman , phys . d * 65 * ( 2002 ) 094031 [ hep - ph/0110004 ] ; a. banfi , g. marchesini and g. smye , jhep * 0208 * ( 2002 ) 006 [ hep - ph/0206076 ] . r. b. appleby and m. h. seymour , jhep * 0212 * ( 2002 ) 063 [ hep - ph/0211426 ] ; r. b. appleby and m. h. seymour , jhep * 0309 * ( 2003 ) 056 [ hep - ph/0308086 ] . y. l. dokshitzer and g. marchesini , jhep * 0303 * ( 2003 ) 040 [ hep - ph/0303101 ] . m. dasgupta , hep - ph/0304086 ; r. b. appleby and g. p. salam , hep - ph/0305232 . h. weigert , hep - ph/0312050 . i. h. park _ et al . _ ( amy collaboration ) , phys . lett . * 62 * ( 1989 ) 1713 ; + y. k. li _ et al . _ ( amy collaboration ) , phys . * d 41 * ( 1990 ) 2675 ; + h. j. behrend _ et al . _ ( cello collaboration ) , z. phys . * c44 * ( 1989 ) 63 ; + d. bender _ et al . _ ( hrs collaboration ) , phys . * d 31 * ( 1985 ) 1 ; + d. p. barber _ et al . _ ( mark j collaboration ) , phys . * 43 * ( 1979 ) 830 ; + a. petersen _ et al . _ ( mark ii collaboration ) , phys . d 37 * ( 1988 ) 1 ; + s. bethke _ et al . _ ( mark ii collaboration ) , z. phys . * c43 * ( 1989 ) 325 ; + c. berger _ ( pluto collaboration ) , z. phys . * c12 * ( 1982 ) 297 ; + w. braunschweig _ et al . _ ( tasso collaboration ) , phys . * b214 * ( 1988 ) 286 ; + w. braunschweig _ et al . _ ( tasso collaboration ) , z. phys . * c45 * ( 1989 ) 11 ; + w. braunschweig _ et al . _ ( tasso collaboration ) , z. phys . * c47 * ( 1990 ) 187 ; + i. adachi _ et al . _ ( topaz collaboration ) , phys . * b227 * ( 1989 ) 495 ; + k. nagai _ et al . _ ( topaz collaboration ) , phys . * b278 * ( 1992 ) 506 ; + y. ohnishi _ et al . _ ( topaz collaboration ) , phys . * b313 * ( 1993 ) 475 . d. decamp _ et al . _ [ aleph collaboration ] , phys . b * 255 * , 623 ( 1991 ) . d. decamp _ et al . _ [ aleph collaboration ] , phys . b * 284 * , 163 ( 1992 ) . d. buskulic _ et al . _ [ aleph collaboration ] , z. phys . c * 73 * ( 1997 ) 409 . r. barate _ [ aleph collaboration ] , phys . * 294 * ( 1998 ) 1 . aleph collaboration , `` qcd measurements in @xmath0 annihilations at centre - of - mass energies between 189 and 202 gev '' , aleph 2000 - 044 conf 2000 - 027 . p. abreu _ et al . _ [ delphi collaboration ] , z. phys . c * 73 * , 229 ( 1997 ) . p. abreu _ et al . _ [ delphi collaboration ] , phys . b * 456 * , 322 ( 1999 ) . p. abreu _ et al . _ [ delphi collaboration ] , eur . j. c * 14 * , 557 ( 2000 ) [ hep - ex/0002026 ] . j. abdallah _ et al . _ [ delphi collaboration ] , eur . j. c * 29 * ( 2003 ) 285 [ hep - ex/0307048 ] . o. adriani _ et al . _ [ l3 collaboration ] , phys . b * 284 * , 471 ( 1992 ) . b. adeva _ et al . _ [ l3 collaboration ] , z. phys . c * 55 * ( 1992 ) 39 . o. adriani _ et al . _ [ l3 collaboration ] , phys . rept . * 236 * , 1 ( 1993 ) . m. acciarri _ et al . _ [ l3 collaboration ] , phys . b * 411 * , 339 ( 1997 ) . m. acciarri _ et al . _ [ l3 collaboration ] , phys . b * 371 * , 137 ( 1996 ) . m. acciarri _ et al . _ [ l3 collaboration ] , phys . b * 404 * , 390 ( 1997 ) . m. acciarri _ et al . _ [ l3 collaboration ] , phys . b * 444 * , 569 ( 1998 ) . m. acciarri _ et al . _ [ l3 collaboration ] , phys . b * 489 * , 65 ( 2000 ) [ hep - ex/0005045 ] . p. achard _ et al . _ [ l3 collaboration ] , phys . b * 536 * ( 2002 ) 217 [ hep - ex/0206052 ] . p. d. acton _ et al . _ [ opal collaboration ] , z. phys . c * 55 * ( 1992 ) 1 . d. acton _ et al . _ [ opal collaboration ] , z. phys . c * 59 * , 1 ( 1993 ) . g. alexander _ et al . _ [ opal collaboration ] , z. phys . c * 72 * , 191 ( 1996 ) . k. ackerstaff _ et al . _ [ opal collaboration ] , z. phys . c * 75 * , 193 ( 1997 ) . g. abbiendi _ et al . _ [ opal collaboration ] , eur . j. c * 16 * ( 2000 ) 185 [ hep - ex/0002012 ] . p. a. movilla fernandez , o. biebel , s. bethke , s. kluth and p. pfeifenschneider [ jade collaboration ] , eur . j. c * 1 * , 461 ( 1998 ) [ hep - ex/9708034 ] . o. biebel , p. a. movilla fernandez and s. bethke [ jade collaboration ] , phys . b * 459 * , 326 ( 1999 ) [ hep - ex/9903009 ] . p. pfeifenschneider _ et al . _ [ jade collaboration ] , eur . j. c * 17 * , 19 ( 2000 ) [ hep - ex/0001055 ] . p. a. movilla fernandez , s. bethke , o. biebel and s. kluth , eur . j. c * 22 * ( 2001 ) 1 [ hep - ex/0105059 ] . s. kluth , m. blumenstengel , p. a. movilla fernandez , s. bethke , o. biebel , c. pahl and j. schieck [ jade collaboration ] , hep - ex/0305023 . j. schieck [ for the opal collaboration ] , hep - ex/0312036 . f. yuasa _ et al . _ , prog . * 138 * ( 2000 ) 18 [ hep - ph/0007053 ] . a. pukhov _ et al . _ , hep - ph/9908288 . f. krauss , r. kuhn and g. soff , jhep * 0202 * ( 2002 ) 044 [ hep - ph/0109036 ] . c. adloff _ et al . _ [ h1 collaboration ] , phys . b * 406 * ( 1997 ) 256 [ hep - ex/9706002 ] . c. adloff _ et al . _ [ h1 collaboration ] , eur . j. c * 14 * ( 2000 ) 255 [ addendum - ibid . c * 18 * ( 2000 ) 417 ] [ hep - ex/9912052 ] . h. u. martyn [ h1 collaboration ] , hep - ex/0010046 . s. chekanov _ et al . _ [ zeus collaboration ] , eur . j. c * 27 * ( 2003 ) 531 [ hep - ex/0211040 ] . j. breitweg _ et al . _ [ zeus collaboration ] , phys . b * 421 * ( 1998 ) 368 [ hep - ex/9710027 ] . b. r. webber , nucl . * 71 * ( 1999 ) 66 [ hep - ph/9712236 ] . a. v. manohar and m. b. wise , phys . b * 344 * ( 1995 ) 407 [ hep - ph/9406392 ] . g. p. korchemsky and g. sterman , nucl . b * 437 * ( 1995 ) 415 [ hep - ph/9411211 ] . g. p. korchemsky and g. sterman , hep - ph/9505391 . g. p. korchemsky and g. sterman , nucl . b * 555 * ( 1999 ) 335 [ hep - ph/9902341 ] . g. p. korchemsky and s. tafat , jhep * 0010 * ( 2000 ) 010 [ hep - ph/0007005 ] . b. r. webber , phys . b * 339 * ( 1994 ) 148 [ hep - ph/9408222 ] . y. l. dokshitzer and b. r. webber , phys . b * 352 * ( 1995 ) 451 [ hep - ph/9504219 ] . y. l. dokshitzer , g. marchesini and b. r. webber , nucl . b * 469 * ( 1996 ) 93 [ hep - ph/9512336 ] . r. akhoury and v. i. zakharov , phys . b * 357 * ( 1995 ) 646 [ hep - ph/9504248 ] . y. l. dokshitzer , a. lucenti , g. marchesini and g. p. salam , nucl . b * 511 * ( 1998 ) 396 [ erratum - ibid . b * 593 * ( 2001 ) 729 ] [ hep - ph/9707532 ] . y. l. dokshitzer , g. marchesini and g. p. salam , eur . j. directc * 1 * ( 1999 ) 3 [ hep - ph/9812487 ] . m. dasgupta , l. magnea and g. smye , jhep * 9911 * ( 1999 ) 025 [ hep - ph/9911316 ] . g. e. smye , jhep * 0105 * ( 2001 ) 005 [ hep - ph/0101323 ] . e. gardi and g. grunberg , jhep * 9911 * ( 1999 ) 016 [ hep - ph/9908458 ] . e. gardi , jhep * 0004 * ( 2000 ) 030 [ hep - ph/0003179 ] . c. w. bauer , a. v. manohar and m. b. wise , phys . * 91 * ( 2003 ) 122001 [ hep - ph/0212255 ] . z. trocsanyi , jhep * 0001 * ( 2000 ) 014 [ hep - ph/9911353 ] . m. dasgupta and b. r. webber , eur . j. c * 1 * ( 1998 ) 539 [ hep - ph/9704297 ] . m. dasgupta and b. r. webber , jhep * 9810 * ( 1998 ) 001 [ hep - ph/9809247 ] . g. p. salam and z. trcsnyi , _ limited automated predictor of power suppressed effects ( lappse ) _ , unpublished work . b. r. webber , hep - ph/9510283 . y. l. dokshitzer , `` why it seems too early to report on status of qcd , '' _ given at international school of subnuclear physics : 31th course : from supersymmetry to the origin of space - time , erice , italy , 4 - 12 jul 1993 _ j. m. campbell , e. w. n. glover and c. j. maxwell , phys . * 81 * ( 1998 ) 1568 [ hep - ph/9803254 ] . d. t. barclay , c. j. maxwell and m. t. reader , phys . d * 49 * ( 1994 ) 3480 . m. beneke , phys . b * 307 * ( 1993 ) 154 . p. abreu _ et al . _ ( delphi collaboration ) , z. phys . * c73 * ( 1996 ) 11 . d. buskulic _ et al . _ ( aleph collaboration ) , z. phys . * c55 * ( 1992 ) 209 . i. g. knowles _ , `` qcd event generators , '' hep - ph/9601212 . s. frixione and b. r. webber , jhep * 0206 * ( 2002 ) 029 [ hep - ph/0204244 ] . s. frixione , p. nason and b. r. webber , jhep * 0308 * ( 2003 ) 007 [ hep - ph/0305252 ] . s. gieseke , a. ribon , m. h. seymour , p. stephens and b. webber , hep - ph/0311208 . lep qcd working group , work in progress . h. stenzel , hep - ex/0309061 . s. bethke , nucl . * 121 * ( 2003 ) 74 [ hep - ex/0211012 ] . k. hagiwara _ et al . _ [ particle data group collaboration ] , phys . rev . d * 66 * ( 2002 ) 010001 . r. w. l. jones , m. ford , g. p. salam , h. stenzel and d. wicke , jhep * 0312 * ( 2003 ) 007 [ hep - ph/0312016 ] . p. m. stevenson , phys . d * 23 * ( 1981 ) 2916 . g. grunberg , phys . d * 29 * ( 1984 ) 2315 . s. j. brodsky , g. p. lepage and p. b. mackenzie , phys . d * 28 * ( 1983 ) 228 . p. abreu _ et al . _ [ delphi collaboration ] , phys . b * 418 * ( 1998 ) 430 . g. abbiendi _ et al . _ [ opal collaboration ] , eur . j. c * 21 * ( 2001 ) 411 [ hep - ex/0105046 ] . a. brandenburg , p. n. burrows , d. muller , n. oishi and p. uwer , phys . b * 468 * ( 1999 ) 168 [ hep - ph/9905495 ] . r. barate _ [ aleph collaboration ] , eur . j. c * 18 * ( 2000 ) 1 [ hep - ex/0008013 ] . p. bambade , m. j. coster , j. fuster and p. tortosa , delphi conference note 2003 - 022-conf-642 ( unpublished ) . w. bernreuther , a. brandenburg and p. uwer , phys . * 79 * ( 1997 ) 189 [ hep - ph/9703305 ] . g. rodrigo , a. santamaria and m. s. bilenky , phys . * 79 * ( 1997 ) 193 [ hep - ph/9703358 ] . p. nason and c. oleari , nucl . b * 521 * ( 1998 ) 237 [ hep - ph/9709360 ] . p. a. movilla fernandez , hep - ex/0209022 . t. kluge , hep - ex/0310063 . g. p. salam and g. zanderighi , nucl . * 86 * ( 2000 ) 430 [ hep - ph/9909324 ] . a. heister _ et al . _ [ aleph collaboration ] , eur . j. c * 27 * ( 2003 ) 1 . g. abbiendi _ et al . _ [ opal collaboration ] , eur . j. c * 20 * ( 2001 ) 601 [ hep - ex/0101044 ] . p. abreu _ et al . _ [ delphi collaboration ] , eur . j. c * 13 * ( 2000 ) 573 . g. abbiendi _ et al . _ [ opal collaboration ] , eur . j. c * 23 * ( 2002 ) 597 [ hep - ex/0111013 ] . g. p. korchemsky , hep - ph/9806537 . g. sterman , nucl . b * 281 * ( 1987 ) 310 ; + s. catani and l. trentadue , nucl . b * 327 * ( 1989 ) 323 ; + s. catani and l. trentadue , nucl . b * 353 * ( 1991 ) 183 ; + g. p. korchemsky , mod . lett . a * 4 * ( 1989 ) 1257 ; + g. p. korchemsky and g. marchesini , phys . b * 313 * ( 1993 ) 433 ; + i. i. y. bigi , m. a. shifman , n. g. uraltsev and a. i. vainshtein , int . j. mod . phys . a * 9 * ( 1994 ) 2467 [ hep - ph/9312359 ] ; + r. d. dikeman , m. a. shifman and n. g. uraltsev , int . j. mod . a * 11 * ( 1996 ) 571 [ hep - ph/9505397 ] ; + m. neubert , phys . d * 49 * ( 1994 ) 3392 [ hep - ph/9311325 ] ; + m. neubert , phys . d * 49 * ( 1994 ) 4623 [ hep - ph/9312311 ] ; + g. p. korchemsky and g. sterman , phys . b * 340 * ( 1994 ) 96 [ hep - ph/9407344 ] ; + a. g. grozin and g. p. korchemsky , phys . d * 53 * ( 1996 ) 1378 [ hep - ph/9411323 ] . m. cacciari and e. gardi , nucl . b * 664 * ( 2003 ) 299 [ hep - ph/0301047 ] . g. bozzi , s. catani , d. de florian and m. grazzini , phys . b * 564 * ( 2003 ) 65 [ hep - ph/0302104 ] . l. magnea , hep - ph/0211013 .

supersymmetric theories possess remarkable properties on the quantum level . these properties are provided by non - renormalization theorems according to which supersymmetric theories have a much better ultraviolet behavior than the non - supersymmetric ones . the most famous example is , certainly , the @xmath5 , @xmath6 supersymmetric yang mills ( sym ) theory , which is finite conformal invariant quantum field theory model @xcite . perturbative quantum corrections in @xmath5 , @xmath0 sym theories are finite starting from the two - loop approximation @xcite . @xmath5 , @xmath7 supersymmetric theories have less number of independent renormalization constants in comparison with the non - supersymmetric ones and the superpotential has no quantum corrections @xcite . moreover , it is possible to find an expression for a @xmath2-function of @xmath1 supersymmetric theories , which is exact in all orders @xcite . ( for theories containing chiral matter superfields this expression relates the @xmath2-function with the anomalous dimension of the matter superfields . ) this expression for the @xmath2-function is called the exact novikov , shifman , vainshtein , and zakharov ( nsvz ) @xmath2-function . the most elegant approach to the non - renormalization theorems is obtained in the framework of the superfield formulation of supersymmetric theories . in this case , the proof of the non - renormalization theorems is based on three points : ( i ) superspace structure of superpropagators , containing the delta - functions of anticommuting variables that allow to shrink the loops into dots in @xmath8 space , ( ii ) the superfield background field method for supersymmetric gauge theories , and ( iii ) an assumption about existence of a regularization manifestly preserving the supersymmetry ( see e.g. @xcite ) . first two points are realized in the explicit form . as to the last one , it is not very clear from the beginning , how to construct a regularization which preserves the supersymmetry and which will be convenient for practical computations of supergraphs ( see e.g. @xcite ) . therefore , a part of the proofs of the non - renormalization theorems based on the above assumption needs an additional justification . it is known that the usually used dimensional regularization @xcite , explicitly breaks the supersymmetry ( see e.g. @xcite ) , because numbers of bosonic and fermionic degrees of freedom differently depend on the space - time dimension . most calculations of quantum corrections in supersymmetric theories are done using the regularization by the dimensional reduction @xcite , which is a special modification of the dimensional regularization , and the @xmath9-scheme . using @xmath9-scheme the finiteness of the @xmath6 sym theory was verified by explicit calculations in one- @xcite , two- @xcite , three- @xcite , and four - loop @xcite approximations . vanishing of two- and three - loop contributions to the @xmath2-function of the @xmath10 sym theory was explicitly demonstrated in @xcite . the @xmath2-function for a general @xmath7 sym theory was calculated in one- @xcite , two- @xcite , three- @xcite , and four - loop @xcite approximations . ( the result agrees with the exact nsvz @xmath2-function only in the one- and two - loop approximations , where a @xmath2-function is scheme - independent . in the higher orders the nsvz @xmath2-function can be obtained only after a specially constructed finite renormalization @xcite . ) however , it is known that the dimensional reduction is not consistent from the mathematical point of view @xcite . due to this inconsistency any @xmath11 supersymmetry can be broken by quantum corrections in higher loops @xcite . this means that the non - renormalization theorems are not completely justified in framework of dimensional reduction and the problem of their justification is in general open . other methods can be also used @xcite for calculations of quantum corrections . in principle , it is possible even to use non - invariant regularizations , if a subtraction scheme is tuned in such a way that the slavnov taylor identities are valid for the renormalized effective action @xcite . however , for practical purposes it is much better to use an invariant regularization . moreover , the existence of a regularization which preserves supersymmetries of a theory is a key step for proving the non - renormalization theorems . in order to construct an invariant regularization it is convenient to formulate a theory in terms of @xmath1 superfields , because in this case @xmath1 supersymmetry is a manifest symmetry . a mathematically consistent invariant regularization , which does not break @xmath1 supersymmetry , is the higher covariant derivative regularization @xcite . in the supersymmetric case it can be formulated in terms of @xmath7 superfields @xcite and , therefore , does not break @xmath1 supersymmetry . in addition to the supersymmetric regularization , manifestly supersymmetric quantization of a theory also requires supersymmetric gauge fixing procedure . in the case of @xmath1 supersymmetric theories structure of divergences can be studied either using the component fields in the wess zumino gauge , or using the superfield formulation . in the latter case the gauge can be fixed without breaking @xmath1 supersymmetry @xcite and the quantum corrections are calculated in a manifestly @xmath1 supersymmetric way that makes this procedure very convenient . application of the higher covariant derivative regularization ( complemented by a supersymmetric gauge condition ) to calculation of quantum corrections for @xmath1 supersymmetric theories allows to explain naturally the origin of the exact nsvz @xmath2-function . loop integrals for the @xmath2-function appear to be integrals of total derivatives @xcite and even integrals of double total derivatives @xcite . a qualitative explanation of this fact can be given by analyzing feynman rules @xcite using a method proposed in @xcite . because the integrands in integrals which determine a @xmath2-function are total derivatives , at least one of the loop integrals can be calculated analytically . this gives the nsvz relation between the @xmath2-function and the anomalous dimension which are defined in terms of the bare coupling constant @xcite . for the @xmath1 supersymmetric electrodynamics this was proved in all orders @xcite . as a consequence , in the abelian case the nsvz @xmath2-function was obtained exactly in all orders of the perturbation theory for the renormalization group functions defined in terms of the bare coupling constant . if the renormalization group functions are defined in terms of the renormalized coupling constant , the nsvz @xmath2-function is obtained in a special subtraction scheme , which can be naturally constructed if the theory is regularized by higher covariant derivatives @xcite . this ( nsvz ) scheme is obtained by imposing the boundary conditions ( [ nsvz_scheme ] ) on the renormalization constants . however , so far there is no proof that in non - abelian theories the exact nsvz @xmath2-function is obtained with the higher covariant derivative regularization in all orders . nevertheless , arguments based on anomalies @xcite and explicit calculations in the lowest loops @xcite allow to suggest this . therefore , using the invariant regularization for @xmath1 supersymmetric theories it is possible to make general conclusions concerning the structure of divergences . existence of an invariant regularization is also needed for proving that the @xmath2-function of @xmath0 sym theories vanishes beyond the one - loop approximation @xcite . the higher covariant derivative regularization is formulated in terms of @xmath1 superfields . certainly , @xmath0 sym theories can be written in terms of @xmath1 superfields ( see e.g. @xcite ) . however , in this case only @xmath1 supersymmetry is manifest , the second supersymmetry being hidden and on - shell . versions of the higher covariant derivative regularization so far used for explicit calculations preserve only @xmath1 supersymmetry . ( a version of the higher derivative regularization for @xmath0 supersymmetric theories was constructed in @xcite , but the higher derivative term , which is invariant under both supersymmetries , was not presented . ) therefore , the @xmath1 higher derivative regularization being applied to @xmath11 extended supersymmetric theories can only state that @xmath1 supersymmetry is preserved . it does not guarantee that total @xmath0 supersymmetry is not broken by quantum corrections . therefore , the effective action is invariant only under the manifest supersymmetry , and it is not clear , whether it is invariant under the second ( hidden ) supersymmetry . in this paper , using the formulation of @xmath0 sym theories in terms of @xmath1 superfields , we construct a manifestly @xmath1 supersymmetric higher covariant derivative regularization , which is also invariant under the hidden supersymmetry . this regularization guarantees that all loop quantum corrections are automatically manifestly @xmath1 supersymmetric , and invariance under the hidden supersymmetry can be broken only by the gauge fixing procedure . ( this situation is similar to the using of the wess - zumino gauge in @xmath1 supersymmetric theories : supersymmetry is also broken only by the gauge fixing procedure . ) we find that if the effective action is invariant under the background gauge transformations and background @xmath0 supersymmetry , then all anomalous dimensions of the chiral superfields vanish . staring from the exact nsvz @xmath2-function we prove that beyond one loop the divergences are completely determined by the anomalous dimension @xmath12 of the chiral superfield @xmath3 ( which forms the @xmath0 vector supermultiplet together with the gauge superfield @xmath4 ) . as a result , vanishing the @xmath2-function beyond the one - loop approximation depends on whether the this anomalous dimension vanishes or not . this paper is organized as follows : in sect . [ section_regularization ] we construct the higher derivative regularization for @xmath0 supersymmetric theories which is manifestly @xmath1 supersymmetric and also possesses the additional hidden supersymmetry . a higher derivative term and the pauli villars determinants proposed in this section are invariant under both supersymmetries . ( a derivation of the @xmath0 supersymmetric higher derivative term by the noether method is described in appendix [ appendix_noether ] . ) however , the second supersymmetry is broken by the gauge fixing procedure . the renormalization of the considered model is discussed in sect . [ section_renormalization ] . in sect . [ section_invariance ] we prove that if the effective action is invariant under the background @xmath0 supersymmetry , then the anomalous dimension @xmath13 vanishes . in sect . [ section_finiteness_and_nsvz ] starting from the exact nsvz @xmath2-function we derive a relation between @xmath2-function and the function @xmath13 . according to this relation , the nsvz @xmath2-function gets no corrections beyond one - loop if @xmath13 vanishes . one - loop calculation of quantum corrections with the constructed regularization is presented in sect . [ section_one - loop ] . the results are briefly discussed in the conclusion . [ section_regularization ] in this paper we consider the @xmath0 sym theory with matter . it is convenient to describe this theory in terms of @xmath1 superfields @xcite . using this notation the action can be written as @xmath14 where @xmath15 is a bare coupling constant , and the real superfield @xmath4 contains the gauge field @xmath16 as a component . a superfield strength of the gauge superfield @xmath4 is defined by @xmath17 ( in our notation indices of right spinors are denoted by the latin letters , and indices of left spinors are denoted by latin letters with dots . vector indices are denoted by greek letters . ) the chiral superfield @xmath3 belongs to the adjoint representation of the gauge group . together with the superfield @xmath4 it forms the multiplet of the @xmath0 sym theory . the chiral superfields @xmath18 and @xmath19 form an @xmath10 hypermultiplet . the superfield @xmath18 lies in a representation @xmath20 , which can be , in general , reducible . the superfield @xmath19 lies in the conjugated representation @xmath21 . for simplicity the action is written for a theory with a single coupling constant ( i.e. the gauge group is simple ) and a single mass @xmath22 ( this corresponds to the irreducible representation @xmath20 ) . the results described below can be easily generalized to more complicated cases . the theory ( [ n=2_action ] ) is invariant under the supersymmetric gauge transformations @xmath23 where the parameter @xmath24 is an arbitrary chiral scalar superfield which takes values in the lie algebra of the gauge group . ( in the second string @xmath24 should be certainly presented as @xmath25 matrix . ) also the theory ( [ n=2_action ] ) is invariant under two supersymmetries . the first supersymmetry is a manifest symmetry , because the action is written in terms of the @xmath1 superfields . the transformations of this supersymmetry can be also written in terms of the @xmath1 superfields @xcite . for this purpose it is convenient to present the exponent of the gauge superfield as @xmath26 and define the right and left spinor gauge covariant derivatives @xmath27 respectively . then the transformations of the manifest supersymmetry in terms of the @xmath1 superfields can be written as @xmath28;\vphantom{\frac{1}{2}}\nonumber\\ & & \delta\phi = i \bar d^2\big(e^{-2v } d^a \xi\ , d_a(e^{2v}\phi ) + \frac{1}{2 } d^2\xi \phi\big);\nonumber\\ & & \delta\widetilde \phi = i \bar d^2\big(e^{2v^t } d^a \xi\ , d_a(e^{-2v^t}\widetilde \phi ) + \frac{1}{2 } d^2\xi \widetilde \phi\big),\end{aligned}\ ] ] where @xmath29 is a real scalar superfield which does not depend on the space - time coordinates . this superfield is a parameter of the transformations of the manifest supersymmetry . the action ( [ n=2_action ] ) is also invariant under the transformations of the second on - shell supersymmetry . in terms of the @xmath1 superfields these transformations have the form @xmath30 where @xmath31 is a chiral superfield independent of the space - time coordinates . in order to regularize the theory ( [ n=2_action ] ) we add to its action a term @xmath32 with higher covariant derivatives . certainly , this term is not uniquely defined , because a number of derivatives can be arbitrary . in this paper we construct the simplest variant of this term , which is proportional to @xmath33 , where @xmath34 is a dimensionful parameter with the dimension of a mass . in appendix [ appendix_noether ] using the noether method we construct an expression for the action @xmath32 invariant under @xmath0 supersymmetry . it can be written in the following form : @xmath35 \vphantom{\frac{1}{2}}\nonumber\\ & & + 4(e^{-\omega^+ } \bar w^{\dot a } e^{\omega^+ } ) \big[(e^\omega \phi e^{-\omega}),\bar\nabla_{\dot a}(e^{-\omega^+ } \phi^+ e^{\omega^+})\big ] -8 \big [ ( e^\omega \phi e^{-\omega } ) , ( e^{-\omega^+ } \phi^+ e^{\omega^+})\big]^2 \bigg)\bigg\}.\end{aligned}\ ] ] it is easy to see that after adding this term the divergences remain only in one - loop supergraphs , which is a typical feature of the higher derivative regularization @xcite . therefore , it is necessary to regularize the remaining one - loop divergences . usually for this purpose the pauli villars determinants should be inserted into the generating functional @xcite . however , in the considered case it is necessary to do this very carefully , because this procedure should not break @xmath0 supersymmetry . it is necessary to introduce two different sets of the pauli villars fields : the first set cancels one - loop divergences originated by the @xmath10 gauge supermultiplet and the second one cancels one - loop divergences originated by the hypermultiplet . taking into account absence of quadratic divergences for the @xmath0 sym theory , in order to cancel one - loop divergences of the gauge supermultiplet ( and ghosts ) , it is sufficient to use a single pauli villars determinant @xmath36 where the action for the pauli villars fields @xmath37 and @xmath38 ( in the adjoint representation of the gauge group ) is given by @xmath39 + m_0 \widetilde\varphi \varphi\big ) + \mbox{c.c . } \big).\end{aligned}\ ] ] we choose the mass @xmath40 of these pauli villars fields proportional to the dimensionful parameter @xmath34 in the higher derivative term : @xmath41 where the finite constant @xmath42 does not depend on the bare coupling constant . introducing the pauli villars fields @xmath37 and @xmath38 is motivated by the analogy with the @xmath6 sym theory . really , in the @xmath6 sym theory 3 chiral superfields in the adjoint representation of the gauge group compensate divergences from the gauge supermultiplet and ghosts . this allows to guess that two pauli villars fields @xmath37 and @xmath38 compensate at least a one - loop divergence originated by the gauge supermultiplet and one chiral superfield @xmath3 in the adjoint representation ( including one - loop divergences of the ghost loop ) . this statement is verified by the explicit calculation made in sect . [ section_one - loop ] . moreover , the action ( [ pv_action ] ) is evidently invariant under both supersymmetries , because it coincides with the action of the massive @xmath0 hypermultiplet in the adjoint representation of the gauge group . also we insert in the generating functional the pauli villars determinants which cancel one - loop divergences originated by the hypermultiplet : @xmath43 where the coefficients @xmath44 satisfy the conditions @xmath45 again , it is convenient to present the pauli villars determinants in the form @xmath46 where @xmath47 lies in the same representation @xmath20 as the fields @xmath18 , @xmath48 and the masses are proportional to the parameter @xmath34 : @xmath49 @xmath50 being independent of @xmath15 . both pauli villars actions are invariant under the transformations of @xmath0 supersymmetry , because they coincide with the actions for the massive @xmath0 hypermultiplets . therefore , the regularization procedure is also invariant under both supersymmetries . the next step is gauge fixing . we will do this in the framework of the @xmath1 superfield background field method @xcite , which allows to get a manifestly @xmath1 supersymmetric effective action preserving the classical gauge invariance . the gauge superfield @xmath4 is split into the background field and the quantum field by making the substitution @xmath51 then the background superfield @xmath52 and the quantum superfield @xmath4 are defined by @xmath53 the background - quantum splitting for the chiral superfield @xmath3 is trivial : @xmath54 where in the right hand side @xmath55 is the background superfield and @xmath3 is the quantum superfield . the gauge fixing term used in this paper does not include the superfields @xmath3 and @xmath55 . as a consequence , the effective action depends only on the sum @xmath56 . this can be easily verified by making the linear substitution @xmath57 in the generating functional . due to the same reason we do not use the background field method for the other chiral matter superfields . it is convenient to fix a gauge without breaking the background gauge invariance @xmath58 where @xmath59 is an arbitrary real scalar superfield and @xmath24 is an arbitrary chiral superfield . for this purpose we use the background covariant derivatives which are defined by @xmath60 note that the theory is also invariant under the quantum gauge transformations @xmath61 where @xmath62 and @xmath63 are arbitrary background-(anti)chiral superfields : @xmath64 the generating functional can be formally written as @xmath65 where the action @xmath66 and the pauli villars determinants are invariant under @xmath0 supersymmetry by construction . then according to standard procedure @xcite we insert into the generating functional @xmath67\cdot \int d\mbox{\boldmath$a$ } d\mbox{\boldmath$a$}^+ \delta(\mbox{\boldmath$\bar \nabla$}^2 v^{(\mbox{\scriptsize \boldmath$a$ } ) } - \mbox{\boldmath$f$ } ) \delta(\mbox{\boldmath$\nabla$}^2 v^{(\mbox{\scriptsize \boldmath$a$ } ) } - \mbox{\boldmath$f$}^+ ) , \ ] ] where @xmath68 and @xmath69 are background-(anti)chiral superfields which satisfy the conditions @xmath70 the quantum gauge superfield transformed under the infinitesimal quantum gauge transformations is denoted by @xmath71 the generating functional obtained after this insertion of @xmath72 is defined by @xmath73 $ ] . it is evidently equal to the original generating functional . then we perform the integration @xmath74 \to \int d\mbox{\boldmath$f$ } d\mbox{\boldmath$f$}^+ d\mbox{\boldmath$c$ } d\mbox{\boldmath$c$}^+\ , z[j,\mbox{\boldmath$f$}]\ , \exp\big(-\frac{i}{16 e_0 ^ 2 } \mbox{tr } \int d^4x\,d^4\theta\ , \mbox{\boldmath$f$}^+ \big(1-\frac{\mbox{\boldmath$\bar \nabla$}^2\mbox{\boldmath$\nabla$}^2}{16\lambda^2}\big ) \mbox{\boldmath$f$ } \big)\nonumber\\ & & \times \exp\big(-\frac{i}{16 e_0 ^ 2}\mbox{tr}\int d^4x\,d^4\theta\,\mbox{\boldmath$c$}^+ \big(1-\frac{\mbox{\boldmath$\bar \nabla$}^2\mbox{\boldmath$\nabla$}^2}{16\lambda^2}\big ) \mbox{\boldmath$c$ } \big).\qquad\end{aligned}\ ] ] the anticommuting background ( anti)chiral nielsen kallosh ghosts @xmath75 and @xmath76 ( in the adjoint representation of the gauge group ) can be expressed in terms of the ( anti)chiral superfields @xmath77 and @xmath78 as @xmath79 the integration over @xmath77 and @xmath78 cancels the determinant appearing after the integration over the fields @xmath80 and @xmath81 , which are defined using equations similar to ( [ background_chiral_fields ] ) . it is convenient to present the corresponding contribution in the form @xmath82 where @xmath83 is the action for the nielsen kallosh ghosts . ( it is assumed that the fields @xmath75 and @xmath76 are expressed in terms of @xmath77 and @xmath78 using eq . ( [ background_chiral_fields ] ) . ) substituting the explicit expression for the functional @xmath73 $ ] and taking the integrals over @xmath68 and @xmath69 we obtain that the gauge fixing term @xmath84 is effectively added to the classical action ( [ n=2_action ] ) . as usually , @xmath85 $ ] is presented as an integral over the faddeev popov ghost fields and gives the ghost action : @xmath86 = \int d\bar c\ , dc\ , d\bar c^+ dc^+ \exp\left(is_{\mbox{\scriptsize ghost}}\right)\ ] ] where @xmath87\ ] ] with @xmath88 + f_2 [ v , [ v , x ] ] + \ldots\ ] ] the ghost @xmath89 and the antighost @xmath90 are background - chiral ; the ghost @xmath91 and the antighost @xmath92 are background - antichiral . the ghost fields can be expressed in terms of the ( anti)chiral fields @xmath93 , @xmath94 , @xmath95 , and @xmath96 using equations similar to ( [ background_chiral_fields ] ) . thus , the generating functional can be written as @xmath97 where @xmath98 denotes the integration measure , @xmath99 is the original action of a @xmath0 supersymmetric theory ( which can also contain hypermultiplet superfields ) , @xmath32 is the regularizing action constructed in this paper , and @xmath100 is the action for the sources . the higher derivative term @xmath32 and the pauli villars determinants are invariant under the transformations of both supersymmetries . however , the gauge fixing term and the ghost action are invariant only under the transformations of the manifest supersymmetry . therefore , @xmath0 supersymmetry is broken only by the gauge fixing procedure . [ section_renormalization ] the results of the previous section show that all ingredients for proving the @xmath1 non - renormalization theorem take place in the considered theory . the theory is renormalizable and renormalization preserves the manifest @xmath1 supersymmetry . therefore , the divergences can be absorbed into the redefinitions of the coupling constant , fields , and masses : @xmath101 here we took into account that ( as we already mentioned above ) the effective action depends only on @xmath102 , and , therefore , it is not necessary to introduce two different renormalization constants for the background and quantum parts of this field . according to the standard prescription , the renormalization constants @xmath103 should be constructed so that to cancel the divergences , appearing in loop integrals . @xmath104 can be defined by the equation @xmath105 it is important that in eqs . ( [ renormalization ] ) and ( [ quantum_field_renormalization ] ) @xmath4 denotes the _ quantum _ gauge superfield . the _ background _ gauge superfield @xmath52 is not renormalized due to the unbroken background gauge invariance ( [ background_gauge_transformations ] ) . therefore , after the renormalization the total gauge superfield is renormalized as @xmath106 in our notation the renormalized fields in the adjoint representation are presented in the form @xmath107 where @xmath108 is the renormalized coupling constant , and @xmath109 denotes the generators of the fundamental representation , which are normalized by the condition @xmath110 the similar equations for the bare superfields have the form @xmath111 where @xmath15 is the bare coupling constant . therefore , in components @xmath112 where we have introduced the notation @xmath113 after substitution ( [ renormalization ] ) @xmath114 where @xmath115 is constructed from the total gauge superfield @xmath116 given by eq . ( [ total_gauge_superfield ] ) . the quantum gauge superfield is present in the action only in the combination @xmath117 . this quantum gauge superfield should be substituted by @xmath118 the ghost lagrangian is renormalized as @xmath119,\ ] ] where @xmath4 should be also substituted by @xmath120 according to eq . ( [ renormalized_v ] ) . using the @xmath1 nonrenormalization theorem @xcite according to which the superpotential @xmath121 is not renormalized , from eq . ( [ renormalized_n=2_action ] ) we immediately obtain @xmath122 certainly , finite renormalizations are possible , but in this paper we assume that for finite terms the renormalization constants are chosen equal to 1 . [ section_invariance ] naively , it is possible to suggest that due to existence of the second supersymmetry two first terms in eq . ( [ n=2_action ] ) are renormalized in the same way . as a consequence , the second supersymmetry would lead to some restrictions to the renormalization constants . however , this question is rather subtle . really , the regularized action proposed in this paper is invariant under both supersymmetries of the @xmath0 supersymmetric theory , but the gauge fixing term and the ghost actions are invariant only under the manifest supersymmetry . it is known @xcite that any symmetry of the classical action corresponds to a symmetry of the renormalized action and the renormalized effective action . nevertheless , we can not state that in the case under consideration the invariance corresponding to the second supersymmetry automatically fixes the renormalization constant @xmath123 . however , here we try to understand what is needed for fixing the renormalization constant @xmath123 . using the background field method we rewrite the transformations ( [ second_susy ] ) making the background quantum splitting for the gauge superfield and the chiral superfield @xmath3 : @xmath124 where @xmath116 is given by eq . ( [ total_gauge_superfield ] ) and @xmath125 these transformations can be obtained if we set @xmath126 for the background superfields and @xmath127 for the quantum superfields . we will show that if the effective action is invariant under the hidden supersymmetry transformations ( [ second_susy_with_background1 ] ) , then @xmath128 actually , this is a manifestation of the general statement that a symmetry can impose restrictions on the renormalization constants . in this sense eq . ( [ z_phi ] ) is similar , for example , to the equation @xmath129 which follows from the symmetry under transformations of the conformal group , see e.g. @xcite . note that the regularization proposed in this paper is important , because it ensures the invariance of the regularized action under the brst and @xmath0 supersymmetry transformations . now let us prove eq . ( [ z_phi ] ) assuming the invariance of the effective action under the transformations ( [ second_susy_with_background1 ] ) . if the sources for the hypermultiplet and quantum fields are set to 0 , then the invariance of the effective action under the transformations ( [ second_susy_with_background1 ] ) can be expressed by the equation @xmath130 where @xmath131 is obtained from the equation @xmath132 which follows from eq . ( [ second_susy_with_background1 ] ) . as a consequence , @xmath133 let us differentiate eq . ( [ st_identity_final ] ) with respect to @xmath134 and @xmath135 , where the subscripts denote points in the superspace . then , after setting all ( background ) fields to 0 we obtain @xmath136 because the effective action satisfies this relation , it is possible to choose such a subtraction scheme in which this relation is also valid for the renormalized action @xmath137 : @xmath138 where @xmath137 is constructed from the classical action by substituting the bare fields by the renormalized ones . in particular , a part of the renormalized action @xmath137 corresponding to the two - point functions of the background fields has the form @xmath139 where @xmath140 is the supersymmetric projection operator and @xmath141 denotes a matrix inverse to @xmath142 . substituting this expression into eq . ( [ two_point_functions ] ) we obtain @xmath143 integrating @xmath144 by parts and taking into account that the supersymmetry transformation parameter @xmath31 is a space - time constant , after some simple transformations involving the algebra of the covariant derivatives the integral in the right hand side of this equation can be rewritten as @xmath145 therefore , the renormalization constants @xmath146 and @xmath142 are related by the equation @xmath147 comparing this result with eq . ( [ zab ] ) , we see that the superfield @xmath55 is not renormalized , @xmath148 , and we prove eq . ( [ z_phi ] ) . as a result we get the criterium whether the effective action is invariant under the background hidden supersymmetry . this criterium is given by eq . ( [ z_phi ] ) . it means , in particular , that under this condition the renormalization group function @xmath149 [ section_finiteness_and_nsvz ] due to @xmath1 supersymmetry a @xmath2-function of sym theories is related with the anomalous dimension of the matter superfields by the nsvz relation @xcite . for our purposes it is convenient to write it in the following form @xcite : @xmath150 where @xmath151 is a bare coupling constant , @xmath152 is the anomalous dimension of the matter superfields , and the following notation is used : @xmath153 the renormalization group functions in eq . ( [ nsvz_general_beta ] ) are expressed in terms of the bare coupling constant @xmath154 . these functions are defined according to the following prescription : @xmath155 where @xmath156 denotes the renormalized coupling constant @xmath157 . the matter is that these are the functions for that the nsvz relation is obtained at least in the abelian case if the theory is regularized by higher derivatives @xcite . ) are scheme - independent , see e.g. @xcite . ] usually the renormalization group functions are defined by a different way , in terms of the renormalized coupling constant : @xmath158 it is well - known that the @xmath2-function and the anomalous dimension defined according to this prescription are scheme - dependent . however @xcite , if the boundary conditions @xmath159 are imposed on the renormalization constants in an arbitrary ( but fixed ) point @xmath160 , ) looks similar to the condition @xmath161 , which can be imposed for obtaining the conformal symmetry limit of a theory @xcite , there is a very important difference : in eq . ( [ nsvz_scheme ] ) @xmath129 only in a single ( but arbitrary ) point @xmath162 , while the conformal symmetry limit is obtained if @xmath129 for arbitrary values of @xmath163 . ] then the renormalization group functions ( [ rg_functions_renormalized ] ) coincide with the renormalization group functions ( [ rg_functions_bare ] ) : @xmath164 this implies that the boundary conditions ( [ nsvz_scheme ] ) at least in the abelian case give the nsvz scheme in all orders of the perturbation theory if the theory is regularized by higher derivatives . this statement was verified by the explicit three - loop calculations in refs . @xcite . all features of the higher covariant derivative regularization , in particular , factorization of integrals into integrals of ( double ) total derivatives @xcite , which gives the nsvz relation for the renormalization group functions ( [ rg_functions_bare ] ) in the abelian case , in the lowest loops also take place in the non - abelian case @xcite . although the all - loop derivation of the nsvz relation is not so far completed for sym theories , it seems reasonable to suggest that in the non - abelian case the nsvz relation is also obtained for the renormalization group functions defined in terms of the bare coupling constant with the higher covariant derivative regularization . in this section we prove that under this assumption the finiteness of @xmath0 sym theories beyond the one - loop approximation can be very easily derived from the nsvz relation if the effective action is invariant under the transformations ( [ second_susy_with_background1 ] ) . thus , the regularization proposed in this paper possibly allows not only to justify the non - renormalization theorems , but also to derive one of them in the easiest way . the main observation is that @xmath0 sym theories can be considered as a special case of @xmath1 supersymmetric yang mills theories , the representation @xmath165 for the matter superfield being reducible and equal to the direct sum @xmath166 here @xmath167 denotes the adjoint representation of the gauge group corresponding to the superfield @xmath3 . ( this superfield together with the gauge superfield @xmath4 forms the @xmath10 gauge supermultiplet . ) the fields in the representations @xmath20 and @xmath21 ( @xmath18 and @xmath19 , respectively ) form the hypermultiplet . let us find the constants @xmath168 and @xmath169 for the reducible representation ( [ r ] ) . for this purpose we note that the generators of the considered representation can be written in the form @xmath170 it is easy to see that for the adjoint representation @xmath171 . therefore , @xmath172 also we obtain @xmath173 we will prove that if the supersymmetric higher covariant derivative regularization is used , the anomalous dimension of the superfield @xmath174 ( defined in terms of the bare coupling constant ) is related with a @xmath2-function . this anomalous dimension is calculated according to the following prescription : @xmath175 where @xmath176 and the limit @xmath177 is assumed . then using eq . ( [ zab ] ) we obtain @xmath103 in eq . ( [ rg_functions_bare ] ) corresponds to @xmath178 . as a consequence , we obtain the factor @xmath179 in eq . ( [ gamma_for_phi ] ) . ] @xmath180 where @xmath181 is the anomalous dimension of the superfield @xmath3 defined in terms of the bare coupling constant . the anomalous dimension of the hypermultplet can be expressed through @xmath12 using eqs . ( [ z_hypermultiplet ] ) : @xmath182 therefore , the anomalous dimension can be written as @xmath183 substituting the expressions for @xmath169 , @xmath168 , and @xmath184 in eq . ( [ nsvz_general_beta ] ) we obtain @xmath185 solving this equation for @xmath186 , after some simple transformations we find that a @xmath2-function of the considered theory is @xmath187 thus , we see that in the theory under consideration the higher loop structure of the nsvz @xmath2-function is determined by the function @xmath188 . if @xmath189 , the expression ( [ exact_beta_for_n=2 ] ) contains only @xmath190 . therefore , the nsvz @xmath2-function for an arbitrary renormalizable @xmath0 supersymmetric yang mills theory does not vanish only in the one - loop approximation and coincides with conventional one - loop @xmath2-function : @xmath191 as the result , we can conclude that the @xmath0 non - renormalization theorem is equivalent to the statement @xmath192 . the equality @xmath189 follows from the invariance of the renormalized action under the background transformations of the hidden supersymmetry ( see section 4 ) . it is evident that this invariance takes place if both a regularization and a gauge fixing procedure are invariant under the complete @xmath0 supersymmetry . however , the considered gauge fixing term is invariant only under the manifest supersymmetry . nevertheless , we can present here some indirect arguments in favor of this equality . it is known that the @xmath2-function is gauge independent if the minimal substraction scheme is used for renormalization ( see e.g. @xcite ) . therefore , if there exists a completely @xmath0 supersymmetric gauge , then the regularized effective action will be invariant under the same amount of supersymmetries as the classical action and according to ( [ z_phi ] ) one gets @xmath189 and the @xmath2-function vanishes beyond one - loop . since the @xmath2-function is gauge invariant , the same result will be valid in any gauge , in particular in the gauge used in this paper . but the completely @xmath0 invariant gauge does actually exist @xcite . however , the @xmath0 supersymmetric gauge used for derivation of eq . ( [ z_phi ] ) was formulated in terms of @xmath0 superfields while the proposed regularization is formulated in terms of @xmath1 superfields and it is unclear whether these gauge and regularization are consistent one with another . thus , if we accept that @xmath189 , then up to a possibility of making finite renormalizations we obtain the following values of the renormalization constants ( exactly in all orders ) : @xmath193 where @xmath194 is a renormalization parameter . values of @xmath195 , @xmath196 and @xmath197 are not so far defined . in the next section we describe the one - loop calculation , which allows to find values of these renormalization constants in the one - loop approximation . [ section_one - loop ] in this section we calculate one - loop divergences using the version of the higher covariant derivative regularization constructed in this paper . ( 0,2 ) ( 4,0 ) ( 6.3,-0.1)@xmath4 ( 3.5,0.9)@xmath18 or @xmath19 ( 8,0 ) ( 10.3,-0.1)@xmath3 in the one - loop approximation the point green function of the hypermultiplet , @xmath198 , defined by the equation @xmath199 can be obtained by calculating two diagrams presented in fig . [ figure_hypermultiplet renormalization ] . it is easy to see that these diagrams cancel each other : @xmath200 this result is in a complete agreement with eq . ( [ renormalization_constants ] ) . ( for finite terms we always choose the renormalization constants equal to 1 . ) using exactly the same arguments we prove that the pauli villars fields @xmath37 , @xmath38 , @xmath47 , and @xmath201 are not renormalized in the one - loop approximation . ( 0,2.0 ) ( 1,0 ) in the one - loop approximation.,title="fig : " ] ( 1,0.9)@xmath3 ( 3.2,-0.2)@xmath4 ( 5,0.05 ) in the one - loop approximation.,title="fig : " ] ( 5.4,1.6)@xmath202 ( 9.4,0 ) in the one - loop approximation.,title="fig : " ] ( 9.6,1.6)@xmath3 ( 12.4,0 ) in the one - loop approximation.,title="fig : " ] ( 12.6,1.6)@xmath4 the two - point green function of the superfield @xmath3 in the one - loop approximation is determined by diagrams presented in fig . [ figure_phi renormalization ] . calculating these diagrams we obtain the function @xmath203 defined by the equation @xmath204 where @xmath205 and , for simplicity , we assume that the gauge group is simple . ( the background superfield @xmath55 is omitted , because the effective action depends only on @xmath56 . ) we are interested in the divergent part of the function @xmath203 . taking into account that in the one - loop approximation logarithmically divergent terms are proportional to @xmath206 , it can be found by differentiating the result for the function @xmath207 ( in the one - loop approximation this is equivalent to differentiating the function @xmath203 ) with respect to @xmath206 in the limit of the vanishing external momentum : @xmath208 here the first term ( proportional to @xmath209 ) is a contribution of the hypermultiplet and the corresponding pauli villars fields ( the second diagram in fig . [ figure_phi renormalization ] ) . the second term consists of the contributions of the first diagram in fig . [ figure_phi renormalization ] and the pauli villars fields @xmath37 and @xmath38 ( the second diagram in fig . [ figure_phi renormalization ] ) . the third and the fourth terms correspond to the third and the fourth diagrams in fig . [ figure_phi renormalization ] , respectively , and cancel each other . thus , we see that this expression is finite and can be easily calculated .. in eq . ( [ g_with_p=0 ] ) it is possible to take the limit @xmath210 due to the derivative with respect to @xmath206 . ] in order to do this we note that eq . ( [ g_with_p=0 ] ) can be rewritten as an integral over a double total derivative : @xmath211 taking into account that the pauli villars masses @xmath212 and @xmath40 are proportional to the parameter @xmath34 , we easily obtain ( setting @xmath213 ) @xmath214 as a consequence , @xmath215 evidently , the nielsen kallosh ghosts are not renormalized , because they interact only with the background gauge superfield . the one - loop renormalization of the faddeev popov ghosts can be found by calculating a diagram presented in fig . [ figure_ghost_renormalization ] . ( 0,2 ) ( 6.3,0.2 ) ( 6.2,1.1)@xmath216 ( 7.8,-0.3)@xmath4 ( 9.4,1.1)@xmath94 it is easy to see that contributions of the various ghosts fields cancel each other and this diagram is convergent and gives the vanishing contribution . therefore , in the one - loop approximation it is possible to choose @xmath217 . ( 0,4.5 ) ( 0.3,2.7 ) in the one - loop approximation . these diagrams without the ones with a loop of the nielsen kallosh ghosts @xmath77 also give a one - loop renormalization of the quantum gauge field.,title="fig : " ] ( -0.3,3.6 ) or @xmath4 ( 1.8,2.1)@xmath4 ( 4.3,2.7 ) in the one - loop approximation . these diagrams without the ones with a loop of the nielsen kallosh ghosts @xmath77 also give a one - loop renormalization of the quantum gauge field.,title="fig : " ] ( 5.9,2.1)@xmath3 ( 8.3,2.7 ) in the one - loop approximation . these diagrams without the ones with a loop of the nielsen kallosh ghosts @xmath77 also give a one - loop renormalization of the quantum gauge field.,title="fig : " ] ( 9.5,2.1)@xmath218 ( 12.3,2.7 ) in the one - loop approximation . these diagrams without the ones with a loop of the nielsen kallosh ghosts @xmath77 also give a one - loop renormalization of the quantum gauge field.,title="fig : " ] ( 12.8,2.1)@xmath202 ( 0.7,0 ) in the one - loop approximation . these diagrams without the ones with a loop of the nielsen kallosh ghosts @xmath77 also give a one - loop renormalization of the quantum gauge field.,title="fig : " ] ( 0.0,0.3 ) or @xmath4 ( 4.7,0 ) in the one - loop approximation . these diagrams without the ones with a loop of the nielsen kallosh ghosts @xmath77 also give a one - loop renormalization of the quantum gauge field.,title="fig : " ] ( 8.7,0 ) in the one - loop approximation . these diagrams without the ones with a loop of the nielsen kallosh ghosts @xmath77 also give a one - loop renormalization of the quantum gauge field.,title="fig : " ] ( 12.7,0 ) in the one - loop approximation . these diagrams without the ones with a loop of the nielsen kallosh ghosts @xmath77 also give a one - loop renormalization of the quantum gauge field.,title="fig : " ] renormalization of the coupling constant can be investigated by calculating the two - point green function of the background gauge superfield . due to the slavnov taylor identity @xcite this green function is transversal : @xmath219 in the one - loop approximation the function @xmath220 can be obtained by calculating the diagrams presented in fig . [ figure_alpha_renormalization ] . the result has the following form : @xmath221 the diagrams containing an internal loop of the quantum gauge superfield @xmath4 ( the first column in fig . [ figure_alpha_renormalization ] ) give a vanishing contribution in the limit @xmath210 . the diagrams with an internal loop of @xmath3 ( the second column in fig . [ figure_alpha_renormalization ] ) give the first term in eq . ( [ d-1_derivative ] ) . this term is exactly canceled by a contribution of the diagrams containing a loop of the nielsen kallosh ghosts @xmath77 , which is given by the second term in eq . ( [ d-1_derivative ] ) . the third term in eq . ( [ d-1_derivative ] ) corresponds to the contribution of the hypermultiplet @xmath222 and its pauli villars fields @xmath223 ( the fourth column in fig . [ figure_alpha_renormalization ] ) . the last term in eq . ( [ d-1_derivative ] ) consists of the faddeev popov ghosts ( @xmath224 ) contribution and the contribution of the pauli villars fields @xmath225 . taking into account that @xmath226 we obtain that a @xmath2-function of the considered theory is given by integrals of double total derivatives : @xmath227 in the considered approximation this result agrees with the exact expression ( [ exact_beta_for_n=2 ] ) . comparing it with eq . ( [ phi_anomalous_dimension ] ) we verify eq . ( [ exact_gamma ] ) in the considered ( one - loop ) approximation . due to the slavnov taylor identity the two - point green function of the quantum gauge superfield is also transversal : @xmath228 the function @xmath229 can be also found by calculating the diagrams presented in fig . [ figure_alpha_renormalization ] . the only difference is that the nielsen kallosh ghosts @xmath77 do no contribute to the renormalization of the quantum gauge superfield . the result has the following form : @xmath230 the expression in the right hand side of this equation is finite at finite values of @xmath34 and coincides with the corresponding expression in eq . ( [ d-1_derivative ] ) . the contributions of the superfield @xmath3 , the hypermultiplet ( with the corresponding pauli villars fields ) , and the pauli villars fields @xmath37 and @xmath38 are calculated exactly as earlier . however , contributions of the quantum gauge superfield and ghosts are different , if the external lines correspond to the quantum gauge superfield @xmath4 . as we have already mentioned above , the nielsen kallosh ghosts do not contribute to the renormalization of the quantum gauge superfield , because their action depends only on the background gauge superfield . the faddeev popov ghosts give only noninvariant terms proportional to @xmath231 , which exactly cancel similar terms coming from the diagrams with a loop of the quantum gauge superfield . the diagrams with a loop of the quantum gauge superfield also give invariant contribution , which is given by the second term in eq . ( [ dq-1_derivative ] ) . it is also expedient to compare eqs . ( [ g_derivative ] ) , ( [ d-1_derivative ] ) , and ( [ dq-1_derivative ] ) . for this purpose we write the one - loop divergences of the considered two point functions in the following form ( taking into account that the effective action depends on the superfield @xmath232 ) : @xmath233 from this equation we see that the regularization constructed in this paper in the considered approximation allows to obtain the manifestly @xmath0 supersymmetric effective action , although the gauge fixing procedure is not @xmath0 supersymmetric . from eq . ( [ divergences ] ) we also conclude that the superfield @xmath3 is not renormalized , @xmath234 , because all divergences are absorbed into the coupling constant renormalization . this completely agrees with eq . ( [ z_phi ] ) . certainly , it is possible to make a finite renormalization of the superfield @xmath3 . however , such a finite renormalization destroys @xmath0 supersymmetry and we will not make it . moreover , we see that the quantum field @xmath4 is not renormalized in the one - loop approximation , so that it is possible to choose @xmath235 . thus , we have verified that the proposed regularization does regularize the one - loop divergences and gives the correct values of the renormalization group functions in the one - loop approximation . in particular , we confirm eq . ( [ renormalization_constants ] ) by the explicit calculation in the one - loop approximation and also obtain @xmath236 we have proposed a new version of the higher covariant derivative regularization for general @xmath0 sym theories formulated in terms of @xmath1 superfields . at the classical level such theories are manifestly invariant under @xmath1 supersymmetry by construction , but these theories are also invariant under additional hidden on - shell supersymmetry . for calculation of quantum corrections it is convenient to define the effective action in the framework of the background field method and fix a gauge without breaking the background gauge invariance . in order to regularize the theory by higher covariant derivatives , we constructed the gauge invariant higher derivative functional which is invariant under the same amount of supersymmetries as the classical action . adding this functional to the classical action we regularize all divergences beyond the one - loop approximation in the gauge invariant and @xmath0 supersymmetric way . the remaining one - loop divergences are regularized by inserting appropriate pauli villars determinants into the generating functional . we show that these determinants preserve all supersymmetries of the classical action by construction . as a result , the hidden supersymmetry is broken only by the gauge fixing procedure . in this paper we have found that if the effective action is invariant under the background transformation of the hidden supersymmetry , the renormalization of the coupling constant is related with the renormalization of the superfields @xmath174 ( or , equivalently , the superfield @xmath237 is unrenormalized , @xmath234 ) . the exact nsvz @xmath2-function is naturally obtained with help of the higher derivative regularization . thus , it is possible to use the relation ( [ exact_beta_for_n=2 ] ) , which follows from the exact nsvz @xmath2-function . this , in turn , implies that the higher loop structure of exact nsvz @xmath2-function is determined by the anomalous dimension @xmath188 . if the function @xmath13 vanishes , the nsvz @xmath2-function is reduced to a purely one - loop expression . therefore , the equality @xmath192 discussed above can be considered as the exact criterium of finiteness of @xmath0 sym theories with matter beyond the one - loop approximation . we want to emphasize once more , that all previous proofs of the @xmath0 non - renormalization theorem were based on the assumption of existence of a regularization preserving all symmetries of the classical action in an arbitrary loop . however , all known regularizations do not satisfy this assumption . in this paper we actually presented such a regularization and showed how it works . also , we would like to point out that a completely off - shell @xmath0 supersymmetric regularization can in principle be developed within the harmonic superfield approach to @xmath0 supersymmetric theories @xcite . this approach allows to formulate @xmath0 sym theories in terms of off - shell @xmath0 superfields . moreover , the background field formalism and off - shell @xmath0 supersymmetric gauge fixing procedure are developed in the harmonic superfield approach @xcite . therefore , for constructing a manifestly @xmath0 supersymmetric regularization it is necessary to construct an appropriate gauge invariant higher derivative functional in terms of harmonic superfields . we plan to study this problem in a forthcoming work . * acknowledgements . * the authors are very grateful to a.l . kataev , a.a.slavnov , and i.v.tyutin for valuable discussions . the work of i.l.b was partially supported by rfbr grant , project no 12 - 02 - 00121 , rfbr - dfg grant , project no 13 - 02 - 91330 , grant for lrss , project no 88.2012.2 and grant of russian ministry of education and science , project tspu-122 . the work of k.v.s was supported by rfbr grant , project no 14 - 01 - 00695 . * appendix * [ appendix_noether ] in order to construct the action @xmath32 , given by eq . ( [ higher_derivative_term ] ) , it is convenient to use the noether method @xcite writing the supersymmetry transformations in terms of @xmath1 superfields @xcite . as a starting point we consider the action where @xmath34 is a regularization parameter . ( its dimension is equal to the dimension of a mass . ) in order to construct an action invariant under the transformations ( [ second_susy ] ) by the noether method , at the first step we calculate the variation of the action @xmath239 . the result is given by the following ( non - vanishing ) expression : @xmath240\nonumber\\ & & - 4\eta^ * \bar\nabla^2(e^{-\omega^+ } \bar w^{\dot a } e^{\omega^+ } ) \big [ e^{-\omega^+ } \bar w_{\dot a } e^{\omega^+ } , e^{\omega}\phi e^{-\omega}\big ] + \eta^ * e^{-\omega^+}\phi^+ e^{\omega^+ } \big[e^{\omega}\phi e^{-\omega } , \bar\nabla^2 \nabla^2 ( e^{\omega}\phi e^{-\omega})\big]\vphantom{\bigg(}\nonumber\\ & & - \eta^ * e^{-\omega^+}\phi^+ e^{\omega^+ } \bar\nabla^2 \big[e^{\omega}\phi e^{-\omega } , \nabla^2 ( e^{\omega}\phi e^{-\omega})\big ] + \eta e^{-\omega^+}\phi^+ e^{\omega^+ } \bar\nabla^2 \big[e^{-\omega^+}\phi^+ e^{\omega^+ } , \nabla^2(e^{\omega}\phi e^{-\omega})\big]\vphantom{\bigg(}\nonumber\\ & & - \eta e^{-\omega^+}\phi^+ e^{\omega^+ } \bar\nabla^2 \nabla^2 \big[e^{-\omega^+}\phi^+ e^{\omega^+ } , e^{\omega}\phi e^{-\omega}\big ] \bigg\}.\end{aligned}\ ] ] @xmath241 \vphantom{\frac{1}{2}}\nonumber\\ & & + ( e^{-\omega^+ } \bar w^{\dot a } e^{\omega^+ } ) \big[(e^\omega \phi e^{-\omega}),\bar\nabla_{\dot a}(e^{-\omega^+ } \phi^+ e^{\omega^+})\big ] \bigg)\end{aligned}\ ] ] @xmath243\nonumber\\ & & + \bar d_{\dot a } \eta^+ \big[e^{-\omega^+}\bar w^{\dot a } e^{\omega^+ } , e^\omega\phi e^{-\omega}\big]\bigg ) \big [ ( e^\omega \phi e^{-\omega } ) , ( e^{-\omega^+ } \phi^+ e^{\omega^+})\big].\end{aligned}\ ] ] m. t. grisaru and w. siegel , nucl . b * 201 * ( 1982 ) 292 [ erratum - ibid . b * 206 * ( 1982 ) 496 ] . s. mandelstam , nucl . b * 213 * ( 1983 ) 149 . l. brink , o. lindgren and b. e. w. nilsson , nucl . b * 212 * ( 1983 ) 401 . p. s. howe , k. s. stelle and p. k. townsend , nucl . b * 236 * ( 1984 ) 125 . p. s. howe , k. s. stelle and p. c. west , phys . b * 124 * ( 1983 ) 55 . m. t. grisaru , w. siegel and m. rocek , nucl . b * 159 * ( 1979 ) 429 . v. a. novikov , m. a. shifman , a. i. vainshtein and v. i. zakharov , nucl . b * 229 * ( 1983 ) 381 . d. r. t. jones , phys . b * 123 * ( 1983 ) 45 . v. a. novikov , m. a. shifman , a. i. vainshtein and v. i. zakharov , phys . b * 166 * ( 1986 ) 329 ; sov . j. nucl . * 43 * ( 1986 ) 294 ; [ yad . fiz . * 43 * ( 1986 ) 459 . ] m. a. shifman and a. i. vainshtein , nucl . b * 277 * ( 1986 ) 456 ; sov . jetp * 64 * ( 1986 ) 428 ; [ zh . . teor . * 91 * ( 1986 ) 723 . ] a. i. vainshtein , v. i. zakharov and m. a. shifman , jetp lett . * 42 * ( 1985 ) 224 [ pisma zh . fiz . * 42 * ( 1985 ) 182 ] . m. a. shifman , a. i. vainshtein and v. i. zakharov , phys . b * 166 * ( 1986 ) 334 . s. j. gates , m. t. grisaru , m. rocek and w. siegel , `` superspace or one thousand and one lessons in supersymmetry , '' front . * 58 * ( 1983 ) 1 . p. c. west , `` introduction to supersymmetry and supergravity , '' singapore , singapore : world scientific ( 1990 ) 425 p. i. jack and d. r. t. jones , `` regularization of supersymmetric theories , '' in * kane , g.l . ( ed . ) : perspectives on supersymmetry ii * 494 - 513 [ hep - ph/9707278 ] . g. t hooft and m. j. g. veltman , nucl . b * 44 * ( 1972 ) 189 . c. g. bollini and j. j. giambiagi , nuovo cim . b * 12 * ( 1972 ) 20 . j. f. ashmore , lett . nuovo cim . * 4 * ( 1972 ) 289 . g. m. cicuta and e. montaldi , lett . nuovo cim . * 4 * ( 1972 ) 329 . r. delbourgo and v. b. prasad , j. phys . g * 1 * ( 1975 ) 377 . w. siegel , phys . b * 84 * ( 1979 ) 193 . s. ferrara and b. zumino , nucl . b * 79 * ( 1974 ) 413 . d. r. t. jones , phys . b * 72 * ( 1977 ) 199 . e. c. poggio and h. n. pendleton , phys . b * 72 * ( 1977 ) 200 . l. v. avdeev , o. v. tarasov and a. a. vladimirov , phys . b * 96 * ( 1980 ) 94 . m. t. grisaru , m. rocek and w. siegel , phys . * 45 * ( 1980 ) 1063 . w. e. caswell and d. zanon , nucl . b * 182 * ( 1981 ) 125 . v. n. velizhanin , phys . b * 696 * ( 2011 ) 560 . l. v. avdeev and o. v. tarasov , phys . b * 112 * ( 1982 ) 356 . i. jack , d. r. t. jones and c. g. north , phys . b * 386 * ( 1996 ) 138 . d. r. t. jones , nucl . b * 87 * ( 1975 ) 127 . i. jack , d. r. t. jones and c. g. north , nucl . b * 486 * ( 1997 ) 479 . r. v. harlander , d. r. t. jones , p. kant , l. mihaila and m. steinhauser , jhep * 0612 * ( 2006 ) 024 . w. siegel , phys . b * 94 * ( 1980 ) 37 . l. v. avdeev , phys . b * 117 * ( 1982 ) 317 . l. v. avdeev and a. a. vladimirov , nucl . b * 219 * ( 1983 ) 262 . m. a. shifman and a. i. vainshtein , sov . j. nucl . * 44 * ( 1986 ) 321 [ yad . fiz . * 44 * ( 1986 ) 498 ] . j. mas , m. perez - victoria and c. seijas , jhep * 0203 * ( 2002 ) 049 . j. c. collins , `` renormalization . an introduction to renormalization , the renormalization group , and the operator product expansion , '' cambridge , uk : univ . ( 1984 ) 380p a. a. slavnov , phys . b * 518 * ( 2001 ) 195 . a. a. slavnov , theor . * 130 * ( 2002 ) 1 [ teor . mat . * 130 * ( 2002 ) 3 ] . a. a. slavnov and k. v. stepanyantz , theor . * 135 * ( 2003 ) 673 [ teor . mat . * 135 * ( 2003 ) 265 ] . a. a. slavnov and k. v. stepanyantz , theor . * 139 * ( 2004 ) 599 [ teor . mat . * 139 * ( 2004 ) 179 ] . a. a. slavnov , nucl . b * 31 * ( 1971 ) 301 . a. a. slavnov , theor.math.phys . * 13 * ( 1972 ) 1064 [ teor . mat . * 13 * ( 1972 ) 174 ] . v. k. krivoshchekov , theor . * 36 * ( 1978 ) 745 [ teor . mat . fiz . * 36 * ( 1978 ) 291 ] . p. c. west , nucl . b * 268 * ( 1986 ) 113 . a. a. slavnov , theor . * 23 * ( 1976 ) 305 . a. a. soloshenko and k. v. stepanyantz , theor . math . * 140 * ( 2004 ) 1264 [ teor . mat . * 140 * ( 2004 ) 430 ] . a. b. pimenov , e. s. shevtsova and k. v. stepanyantz , phys . b * 686 * ( 2010 ) 293 . k. v. stepanyantz , phys . part . * 8 * ( 2011 ) 321 . a. v. smilga and a. vainshtein , nucl . b * 704 * ( 2005 ) 445 . k. v. stepanyantz , `` factorization of integrals defining the two - loop @xmath2-function for the general renormalizable n=1 sym theory , regularized by the higher covariant derivatives , into integrals of double total derivatives , '' arxiv:1108.1491 [ hep - th ] . k. v. stepanyantz , j. phys . * 343 * ( 2012 ) 012115 . k. v. stepanyantz , j. phys . * 368 * ( 2012 ) 012052 . k. v. stepanyantz , nucl . b * 852 * ( 2011 ) 71 . k. v. stepanyantz , int . j. theor . phys . * 51 * ( 2012 ) 276 . k. stepanyantz , theor . * 146 * ( 2006 ) 321 [ teor . mat . * 146 * ( 2006 ) 385 ] . a. l. kataev and k. v. stepanyantz , nucl . b * 875 * ( 2013 ) 459 . a. l. kataev and k. v. stepanyantz , phys . b * 730 * ( 2014 ) 184 . v. k. krivoshchekov , phys . b * 149 * ( 1984 ) 128 . l. d. faddeev and a. a. slavnov , `` gauge fields . introduction to quantum theory , '' nauka , moscow , 1978 and front . * 50 * ( 1980 ) 1 [ front . phys . * 83 * ( 1990 ) 1 ] . a. a. slavnov , theor . phys . * 33 * ( 1977 ) 977 [ teor . mat . fiz . * 33 * ( 1977 ) 210 ] . j. c. taylor , nucl . phys . b * 33 * ( 1971 ) 436 . a. a. slavnov , theor . * 10 * ( 1972 ) 99 [ teor . mat . * 10 * ( 1972 ) 153 ] . grigorian and i. v. tyutin , theor . * 64 * ( 1985 ) 915 [ teor . mat . fiz . * 64 * ( 1985 ) 400 ] . a. l. kataev , jhep * 1402 * ( 2014 ) 092 . b. l. voronov and i. v. tyutin , yad . * 36 * ( 1982 ) 1041 . a. s. galperin , e. a. ivanov , v. i. ogievetsky , e. a. sokatchev , `` harmonic superspace , '' cambridge university press ( 2001 ) 306 p.

the single - impurity kondo model @xcite is unquestionably one of the most important models studied in condensed matter physics over the past decades . the investigation of its equilibrium properties has caused the development of important theoretical tools such as renormalization group methods or the bethe ansatz for impurity systems @xcite . more than 20 years ago it was also realized @xcite that the kondo model can be used to describe transport experiments through quantum dots . the developments in the ability to engineer devices on the nanoscale has led to the experimental realization of kondo physics in such systems @xcite . one particular advantage of these quantum dots is the almost full control over system parameters like temperature , bias and gate voltages , magnetic field , and exchange couplings . these possibilities have triggered a great interest in the theoretical study of quantum dots out of equilibrium . a wide range of theoretical methods has been applied in the past , including non - equilibrium perturbation theory @xcite , the flow - equation method @xcite , coulomb gas representations @xcite , real - time and functional renormalization group methods @xcite , a non - equilibrium extension of the numerical renormalization group ( nrg ) method @xcite , and time - dependent density matrix renormalization group ( dmrg ) techniques @xcite . these studies established the importance of relaxation and decoherence effects for the understanding of non - equilibrium physics . from an experimental point of view various measurable quantities like the steady - state current , the magnetization , and the static susceptibility have been calculated . beside the application of these techniques there were also attempts to employ the known integrability of certain impurity models in equilibrium , notably the anderson impurity model and the interacting resonant level model , to investigate their non - equilibrium properties . konik et al . @xcite calculated the differential conductance in the anderson impurity model by combining the well - known scattering states of the equilibrium system @xcite with a landauer - bttiker formalism . in this work the chemical potentials in the leads were coupled to dressed excitations rather than free electrons and the calculation was restricted to a subset of the excitations . in contrast , a different approach was recently put forward by metha and andrei @xcite to treat the interacting resonant level model . they constructed a new set of scattering states of bethe - ansatz form which share the quantum numbers of free electrons in the incoming channel , hence allowing the application of a finite voltage in the usual manner . however , questions concerning the existence of these scattering states and issues related to the used regularization scheme of the theory remain open . nevertheless , the interacting resonant level model has become one of the benchmark systems in the study of non - equilibrium physics @xcite . despite the large number of studies of impurity models out of equilibrium only few results are known for the dynamical correlation functions . the spin dynamics of a non - equilibrium quantum dot has been studied by using a majorana fermion representation @xcite , which yields the qualitative low - frequency properties of the correlation functions . the transverse susceptibility in a kondo model was studied by paaske et al . @xcite using non - equilibrium perturbation theory together with a pseudo - fermion representation of the kondo spin . they showed that the fourier transform of the transverse susceptibility possesses a peak if its frequency equals the value of the applied magnetic field , @xmath6 , and that the width of this peak is given by the transverse spin relaxation rate @xmath7 . their derivation was , however , restricted to either the regime @xmath8 or @xmath9 , where @xmath0 denotes the applied voltage . very recently , fritsch and kehrein @xcite applied the flow - equation method to study the longitudinal correlation function as well as the magnetization and t - matrix in a kondo model in and out of equilibrium . the numerical solution of the two - loop scaling equations allowed them to study the correlation function for all combinations of the parameters voltage , temperature , and magnetic field , provided the weak - coupling condition ( i.e. the presence of a large enough infrared cutoff ) was satisfied . in particular , these numerical solutions were used to compare the effects of an applied voltage and a finite temperature , revealing qualitative differences such as the appearance of kondo splitting in the non - equilibrium situation . in general , however , this method can not provide analytic expressions for the line shape . in this article we will generalize the real - time renormalization group method in frequency space @xcite to allow the calculation of dynamical correlation functions of arbitrary dot operators in systems describing spin and/or orbital fluctuations . in this setting the quantum dot is coupled to non - interacting leads which are held at different chemical potentials . the derived two - loop rg equations can be solved analytically in the weak - coupling regime . explicitly , we calculate the longitudinal and transverse spin - spin correlation and response functions in a two - lead kondo model in a magnetic field @xmath1 up to order @xmath10 . here @xmath11 denotes the effective coupling at the energy scale @xmath12 where the flow of the coupling constant is cut - off . in order to satisfy the weak - coupling condition @xmath13 either the applied voltage or the magnetic field have to be sufficiently large compared to the kondo temperature , @xmath14 , where @xmath3 is the scale where the system enters the strong - coupling regime . we note that the applied formalism does not rely on a fermionic representation of the kondo spin but rather deals with its matrix representation in liouville space directly . the longitudinal response function possesses a peak at the spin relaxation rate @xmath15 , which gets suppressed in a finite magnetic field . interestingly , in the case of a strong magnetic field , @xmath16 where @xmath17 denotes the renormalized magnetic field ( see eq . for the precise value ) , the longitudinal correlation and response function show `` kink - like '' structures at the frequencies @xmath18 , which were also observed using the flow - equation method @xcite . here we additionally provide the line shape close to these `` kinks '' and show that the real part of the response functions shows characteristic logarithmic features at @xmath18 . furthermore we study the longitudinal and transverse fluctuation - dissipation ratios . as expected these ratios show a revival of the fluctuation - dissipation theorem @xcite provided the applied voltage is small compared to the magnetic field or the considered oscillation frequency . this article is organized as follows . in the next two sections we will define the general set up we want to study and define the used notations . this will include the kondo model , the notion of liouville operators as well as the definition of the symmetrized correlation function and susceptibility . in sec . [ sec : pe ] we will then derive perturbative expansions for the kernels needed to calculate these correlation functions . this will be performed in liouville space ; the expansion is done in powers of the exchange coupling between the dot and the reservoirs . these perturbative expansions can be applied to any model describing spin and/or orbital fluctuations as well as correlation functions of arbitrary operators . in sec . [ sec : genrg ] we will use these results to derive the rg equations for the kernels of pure dot operators . in the following section we will further specialize to the two - lead kondo model , where we use the explicit expressions of the liouville operator and the coupling between the dot and the reservoirs to derive analytic results for the effective kernels appearing in the correlation functions . finally , these expressions for the kernels are used in secs . [ sec : cf ] and [ sec : tcf ] to calculate the longitudinal and transverse correlation and response functions . the real - time renormalization group in frequency space was applied in ref . to calculate various quantities including the spin relaxation and dephasing rates , the renormalized magnetic field , the magnetization and the current in the anisotropic kondo model in a finite magnetic field out of equilibrium . in this reference all notations which we will use in the following were originally set up . in order to increase the readability of the present manuscript we will briefly recall the basic formulas and notations . we consider a quantum dot with fixed charge in the coulomb blockade regime coupled to external reservoirs . as shown in detail in ref . , a standard schrieffer - wolff transformation leads to a hamiltonian of the form @xmath19 where @xmath20 is the reservoir part , @xmath21 characterizes the isolated quantum dot , and @xmath0 describes the coupling between reservoirs and quantum dot . they are given explicitly by @xmath22 here , @xmath23 are the fermionic creation ( @xmath24 ) and annihilation ( @xmath25 ) operators in the reservoirs and @xmath26 is an index characterizing all quantum numbers of the reservoir states , which contains the reservoir index @xmath27 and the spin quantum number @xmath28 . we measure the energy @xmath29 of the reservoir states relative to the chemical potential @xmath30 of reservoir @xmath31 . the eigenstates and eigenenergies of the isolated quantum dot are denoted by @xmath32 and @xmath33 . the interaction @xmath0 is quadratic in the reservoir field operators , which arises from second order processes of one electron hopping on and off the quantum dot coherently . this keeps the charge fixed and allows only spin / orbital fluctuations . the coupling vertex @xmath34 is an arbitrary operator acting on the dot states . it is written in its most general form , depending on the quantum numbers and energies of the reservoir states in an arbitrary way . as explained in ref . , the rg approach can be set up in its most convenient form if one assumes that the frequency dependence of the initial vertices is rather weak and varies on the scale of the band width @xmath35 of the reservoirs . for the model we have in mind , the isotropic spin-@xmath36 kondo model in a magnetic field , this is certainly satisfied . therefore , we will assume this in the following and introduce below [ see eq . ( [ contraction ] ) ] a convenient cutoff function into the free reservoir green s functions . to achieve a more compact notation for all indices , we write @xmath37 and sum ( integrate ) implicitly over all indices ( frequencies ) . the interaction is then written in the compact form @xmath38 @xmath39 denotes normal - ordering of the reservoir field operators , meaning that no contraction is allowed between reservoir field operators within the normal - ordering . within the normal - ordering of eq . ( [ spin_orbital ] ) , the field operators can be arranged in an arbitrary way ( up to a fermionic sign ) , therefore the coupling vertex can always be chosen such that antisymmetry holds : @xmath40 furthermore , due to the hermiticity of @xmath0 , the vertex has the property @xmath41 where @xmath42 . and @xmath43 involve exchange between the electron spins of the left / right reservoir and the local spin , @xmath44 transfers an electron from one reservoir to the other during the exchange process . we assume that the kondo model was derived from an anderson impurity model via a schrieffer - wolff transformation , which implies the relation @xmath45 . ] the specific model we want to study is the isotropic kondo model in an external magnetic field @xmath46 ( see fig . [ fig : kondomodel ] ) . in this case the above relations read explicitly @xmath47 -(j_{\alpha'\alpha})_0 s^i\sigma^i_{\sigma'\sigma } & \text{for $ \eta=-\eta'=-$ } \end{array } \right.\,,\end{aligned}\ ] ] where @xmath48 , @xmath49 is the @xmath50-component of the spin-@xmath36 operator of the quantum dot , @xmath51 is a pauli matrix , and @xmath52 are the initial exchange couplings . we will be interested in the antiferromagnetic model here , i.e. , we assume @xmath53 initially . if one derives the kondo model via a schrieffer - wolff transformation from an anderson impurity model ( see , e.g. , ref . ) , one further finds @xmath54 although the general formalism and many of the following formulas are also valid for an arbitrary number of reservoirs , we will consider the case of two reservoirs only with chemical potentials given by @xmath55 where @xmath0 is the applied voltage which we assume to be positive , @xmath56 . a contraction is defined with respect to a grand - canonical distribution of the reservoirs , given by @xmath57 @xmath58 is the fermi distribution function corresponding to temperature @xmath59 ( note that the chemical potential does not enter this formula since @xmath29 is measured relative to @xmath30 ) . furthermore , @xmath60 is the @xmath61-function in compact notation . furthermore , we have introduced the cutoff by the band width @xmath35 into the reservoir contraction via the density of states @xmath62 in order to calculate the dynamical spin - spin correlation functions we have to know the time evolution of the density matrix @xmath63 . formally , this follows from the solution of the von neumann equation @xmath64 where @xmath65}_-,\ ] ] is the liouvillian acting on usual operators in hilbert space via the commutator . the form of the hamiltonian yields a similar decomposition of the liouvillian , @xmath66 with @xmath67}_-$ ] , @xmath68}_-$ ] , and @xmath69}_-$ ] . we would like to note that the concept of liouville space and superoperators have been used in various contexts , for example in quantum statistical mechanics @xcite . initially , we assume that the density matrix is a product of an arbitrary dot part @xmath70 and a grandcanonical distribution @xmath71 for the reservoirs , @xmath72 furthermore , we introduce the laplace transform @xmath73 where we will frequently use the notation @xmath74 . the stationary density matrix is defined as @xmath75 which is understood in the sense @xmath76 for any local operator @xmath77 , and can be calculated using @xmath78 the existence of a stationary state was proven in ref . using non - equilibrium pertubation theory to all orders as well as in ref . using the rtrg - fs , which in particular clarified the generation of the relaxation and dephasing rates under the rg flow . the reduced density matrix of the dot is obtained by tracing out the reservoir degrees of freedom @xmath79 where @xmath80 denotes the effective liouvillian of the quantum dot formally defined in below . the stationary reduced density matrix can then be obtained similar to , @xmath81 the quantities of interest in this article are the two - point correlation function of two operators @xmath82 and @xmath83 as well as their dynamical susceptibility with respect to the steady state , @xmath84}_+\bigr\rangle_{st } , \qquad \label{eq : defphi}\\ \chi_{ab}(t)\!&=&\!{\text{i}}\,\theta(t)\bigl\langle { \left[a(t)_\mathrm{h},b(0)_\mathrm{h}\right]}_-\bigr\rangle_{st}. \label{eq : defchi}\end{aligned}\ ] ] where @xmath85 here the trace is taken over the dot states as well as the reservoir degrees of freedom , @xmath86 . the time - evolution of the operators in the heisenberg picture is given by @xmath87 instead of calculating and in real time we will study their respective fourier transforms @xmath88 where @xmath89 for @xmath90 ( @xmath91 ) . the susceptibility admits the standard decomposition @xmath92 . in order to calculate @xmath93 and @xmath94 we introduce the auxiliary correlation functions @xmath95}_\pm\bigr\rangle_{st } \label{eq : defbarcomega}\ ] ] with @xmath96 . its relations to the correlation functions are given by ( see app . [ sec : crel ] ) @xmath97 the static susceptibility is related to the dynamical susceptibility via @xmath98 where @xmath99 denotes the magnetization . some general properties of the correlation functions can be obtained by considering their spectral representations . let @xmath100 be a complete set of basis states of the full hamiltonian @xmath101 , i.e. @xmath102 . furthermore , the stationary density matrix @xmath103 satisfies @xmath104}_-=0 $ ] , thus the basis states @xmath105 can be chosen such that @xmath106 using this one easily verifies the spectral representations @xmath107 these relations imply @xmath108 as well as @xmath109 , @xmath110 , and @xmath111 . in equilibrium the matrix elements of the density matrix are given by @xmath112 with the partition sum @xmath113 , which implies the well - known fluctuation - dissipation theorem @xcite @xmath114 as well as @xmath115 . in this section we derive a perturbative expansion in liouville space for the auxiliary correlation functions @xmath116 , which will serve as the starting point for the derivation of the rg equations below . a similar perturbative expansion for the effective liouvillian of the quantum dot @xmath117 has been derived in refs . . we will generalize these results to @xmath116 while closely following the presentation of ref . . as starting point to set up the formalism we assume that the operators @xmath82 and @xmath118 admit a representation similar to , @xmath119 where we recall the short - hand notation @xmath120 and sum ( integrate ) implicitly over all indices ( frequencies ) . we further assume the operators @xmath82 and @xmath118 to be bosonic which implies @xmath121 and @xmath122 to be even . eventually we will be concerned with the correlation functions of the spin operators on the dot , i.e. @xmath123 . in this case the operators do not couple dot and reservoir degrees of freedom and hence only the terms with @xmath124 are non - vanishing . however , we will keep the general forms throughout this section , which for example include the case of current operators @xcite where @xmath125 . in order to set up the perturbative expansions in liouville space we define the operators @xmath126}_+,\quad l_b^\pm={\text{i}}{\left[b,.\right]}_\pm.\ ] ] then using together with we obtain after some algebra @xmath127 in the next step we use and and furthermore perform the laplace transform @xmath128 in to obtain @xmath129 here we have used @xmath96 to ensure convergence of the integral . the limit @xmath130 has to be taken before @xmath131 in order to reach the stationary state . the next step is to expand the expression in the interacting part @xmath132 of the liouvillian and to integrate out the reservoir part . this procedure was outlined for the reduced density matrix @xmath133 in detail in ref . ; we will generalize this to the case of the auxiliary correlation functions here . first , using @xmath134 with @xmath135 , can be formally expanded in @xmath132 , @xmath136 . \label{eq : cab3 } \end{split}\ ] ] second , in order to integrate out the reservoir degrees of freedom we write @xmath132 in the form @xmath137 where we implicitly sum ( integrate ) over @xmath138 as well as @xmath139 . @xmath140 is a quantum field superoperator in liouville space for the reservoirs , defined by ( @xmath141 is an arbitrary reservoir operator ) @xmath142 here @xmath143 serves as an auxiliary index which is similar to the keldysh index indicating whether the field operator is acting on the upper or the lower part of the keldysh contour . @xmath144 is a superoperator acting in liouville space of the quantum dot , and is defined by ( @xmath141 is an arbitrary operator on the quantum dot ) @xmath145 in the same way we define @xmath146 where the dot superoperators @xmath147 and @xmath148 act on arbitrary dot operators @xmath141 as @xmath149 for @xmath150 or @xmath151 we define @xmath152}_+$ ] and @xmath153}_\pm$ ] , respectively . as we consider only bosonic operators @xmath82 and @xmath118 which change the number of fermions by an even integer , the sign - superoperator is given by @xmath154 this operator was introduced to compensate additional signs due to interchanges of fermionic reservoir field operators as explained in detail in ref . . inserting the representations ( [ coupling_product ] ) as well as and into and shifting all reservoir field superoperators @xmath155 to the right using @xmath156 where we have introduced the short - hand notation @xmath157 , one can show@xcite that each term of perturbation theory can be written as a product of a dot part and an average over a sequence of field superoperators of the reservoirs with respect to @xmath71 . evaluating the latter with the help of wick s theorem , one can represent each term of the wick decomposition by a diagram ( see fig . [ fig : diagram_example ] for an example ) describing a certain process contributing to the auxiliary correlation function @xmath116 . each process consists of a sequence of interaction vertices @xmath158 between the dot and the reservoirs , and a free time propagation of the dot in between ( leading to resolvents in laplace space ) . since the reservoirs have been integrated out , the vertices are connected by reservoir contractions ( the green lines in fig . [ fig : diagram_example ] ) . this means that the various diagrams represent terms for the effective time evolution of the dot in the presence of dissipative reservoirs . . the time direction is to the left . each vertex @xmath159 is represented by two adjacent black dots indicating the two reservoir field operators associated with each vertex . the vertices @xmath160 and @xmath161 are represented in the same way by red dots . for this example we have chosen @xmath125 in as it would be the case for a current - current correlation function . the black horizontal lines connecting the vertices denote the free time propagation of the quantum system , leading to the resolvents @xmath162 in laplace space . the green lines are the reservoir contractions arising from the application of wick s theorem . the vertical blue lines between the vertices are auxiliary lines to determine the energy argument @xmath163 of the resolvents . ] each diagram for the auxiliary correlation function has the form @xmath164 where @xmath68}_-$ ] , @xmath165 indicates an interaction vertex , and @xmath166 is a contraction between the reservoir field superoperators , defined by @xmath167 we stress that only the initial reduced density matrix of the dot @xmath70 defined in appears in as we have already performed the trace over the reservoir degrees of freedom ( and hence @xmath71 ) to obtain the reservoir contractions @xmath168 . to factorize the wick decomposition , a fermionic sign has to be assigned to each permutation of reservoir field superoperators , indicated by the sign factor @xmath169 in ( [ value_diagram ] ) . for each pair of vertices connected by two reservoir lines , a combinatorial factor @xmath170 occurs , leading to the prefactor @xmath171 in ( [ value_diagram ] ) . the value of the frequencies @xmath163 in the resolvents between the interaction vertices is determined by the sum over all variables @xmath172 of those indices belonging to the reservoir lines which are crossed by a vertical line at the position of the resolvent ( the blue lines in fig . [ fig : diagram_example ] ) . thereby , the index of the left vertex has to be taken of the corresponding reservoir line . for example , the diagram shown in fig . [ fig : diagram_example ] is given by ( the obvious dependence on the keldysh indices has been omitted for simplicity , i.e. @xmath173 and @xmath174 ) @xmath175 where the resolvents are defined by @xmath176 with @xmath177 as well as @xmath178 as can be seen from the example ( [ diagram_example ] ) , each diagram consists of a sequence of irreducible blocks ( where a vertical line always cuts at least one reservoir line ) and free resolvents @xmath179 or @xmath180 in between . now there are two possibilities : ( i ) the vertices @xmath160 and @xmath161 do not belong to the same block ( see fig . [ fig : block1example ] for an example ) . ( ii ) the vertices @xmath160 and @xmath161 belong to the same block ( see fig . [ fig : block2example ] for an example ) . in the first case ( i ) one can formally resum those terms between the vertices @xmath160 and @xmath161 which are not connected to them similar to dyson equations with the result @xmath181 where @xmath182 here the kernel @xmath183 contains the sum over all irreducible diagrams , @xmath184 where the subindex @xmath185 indicates that only irreducible diagrams are allowed where any vertical line between the vertices cuts through at least one reservoir contraction . we further introduce irreducible blocks @xmath186 as well as @xmath187 which are given as the sum over all irreducible diagrams containing the vertices @xmath160 and @xmath188 , @xmath189 obviously , in the case of spin operators , @xmath190 , the vertex @xmath160 has no external legs and hence @xmath191 . in contrast , although the vertex @xmath188 does not possess external legs either , there exist irreducible diagrams containing @xmath188 and at least one vertex @xmath159 to the left and one to the right of @xmath188 . if we now proceed by resumming the irreducible blocks right to ( and not connected to ) the vertex @xmath188 similar to and perform the limit @xmath192 using , we deduce that all terms of type ( i ) contribute to @xmath193 . \label{eq : barcabterm1}\ ] ] in the second case ( ii ) we introduce a kernel similar to and which contains all irreducible diagrams containing both vertices @xmath160 and @xmath161 , @xmath194 in the case of spin operators , @xmath190 , there exist no irreducible diagrams connecting @xmath160 and @xmath188 , hence @xmath195 in this case . now using again for the sum of the irreducible blocks right to ( and not connected to ) the vertex @xmath188 we deduce that all terms of type ( ii ) contribute to @xmath196 . \label{eq : barcabterm2}\ ] ] hence , taking together ( i ) and ( ii ) we finally arrive at the main result of this section , @xmath197\nonumber\\ * & & \!\ ! -{\text{i}}\,\mbox{tr}_s\bigl[\sigma_{ab}^\pm(\omega,{\text{i}}0+)\,\rho_s^{st}\bigr ] , \label{eq : barcresult}\end{aligned}\ ] ] where the kernels are defined by , , , and , respectively . in addition we note that the diagrammatic series can be partially resummed by taking all closed sub - diagrams between two fixed vertices together which contain only contractions connecting vertices between the two fixed ones . this has the effect that the resolvents in ( [ value_kernel ] ) , , , and , are replaced by @xmath198 ( and similar for @xmath199 ) , i.e. , the full effective liouville operator occurs in the denominator . in this formulation the number of diagrams is reduced , i.e. diagrams containing closed sub - diagrams between two vertices are no longer allowed . in this section we will set up the generic rg equations for the kernels @xmath186 , @xmath187 , and @xmath200 , for spin operators on the dot in a model with spin / orbital fluctuations . hence we will assume the form for the coupling between the reservoirs and the quantum dot , but will keep the vertex @xmath158 arbitrary at this stage . the derivation will require some relations between the initial liouvillian @xmath201 , the vertex @xmath188 and the effective dot liouvillian @xmath117 which can be explicitly checked for the kondo model to be studied in the next section but have to be assumed here . these relations are , , , and . the generic rg equations for the vertex @xmath159 and the effective liouvillian @xmath117 have been derived and solved in ref . , we will quote these results without derivation when they are needed . furthermore , we will restrict ourselves to the calculation of the dynamical spin - spin correlations only , i.e. we will assume @xmath190 in what follows . this implies in particular , that the initial values of the vertices @xmath160 and @xmath188 defined in and do not possess any external lines , i.e. @xmath124 . explicitly , @xmath202}_+,\label{eq : avertexinitial}\\ \mathcal{b}_\pm&=&{\text{i}}{\left[b,.\right]}_\pm.\label{eq : bvertexinitial}\end{aligned}\ ] ] we will see below that this form of the vertex @xmath160 is conserved under the rg flow . in contrast , a new effective @xmath118-type vertex @xmath203 with two external lines will be generated . the fact that the initial vertex @xmath160 has no external lines directly implies the final results for the kernels and , namely @xmath204 hence , in the following we have to consider the kernel only . we note that the results of this section remain valid for any pure dot operators @xmath82 and @xmath118 , i.e. any operators with @xmath124 . the rg procedure is divided into two steps . in the first step we will integrate out the symmetric part of the reservoir contractions @xmath205 . the reason for this is as follows : the effective dot liouvillian @xmath80 can be diagonalized as @xmath206 where @xmath207 and @xmath208 denote the eigenvalues and corresponding projectors , respectively . this diagonalization implies for the resolvents @xmath209 now there exists a zero eigenvalue @xmath210 whose eigenstate for @xmath211 corresponds to the stationary state . the appearance of this zero eigenvalue can lead to infrared divergencies of the frequency integrations in the perturbative expansions for the vertex @xmath159 and the effective liouvillian @xmath117 as is elaborated on in detail in ref . . however , after the discrete rg step we can trivially sum over the keldysh indices @xmath212 and @xmath213 by introducing @xmath214 in the resulting rg equations only the symmetric vertex @xmath215 will appear . this vertex has the important property @xmath216 which is independent of the model specifics . hence , after the discrete rg step the zero eigenvalue can no longer appear in any resolvent standing left to @xmath215 ( i.e. in no resolvent except the one standing left to the vertex @xmath188 ) . this resolves the problem of infrared divergent internal frequency integrations , as the remaining eigenvalues @xmath207 have a strictly negative imaginary part ( see and ) . we would like to refer to ref . for a general discussion of this topic . in the second step we introduce a cut - off @xmath217 into the reservoir contractions via the fermi function . we then integrate out the reservoirs by sending @xmath218 , which results in a description of the system in terms of effective dot quantities like @xmath117 . this second , continuous , rg step is further divided into two substeps , first we integrate out the reservoir degrees of freedom down to an energy scale @xmath219 , and second we complete the flow from @xmath219 down to @xmath220 . in the first discrete rg step we integrate out the symmetric part @xmath221 of the fermi function in the contraction ( [ liouville_contraction ] ) . the discrete rg step for the kernel ( [ value_kernel ] ) and the vertex @xmath159 has been performed in ref . . here we will derive the analog results for the kernel and the vertex @xmath203 . this is achieved by decomposing the contraction ( [ liouville_contraction ] ) according to @xmath222,\end{aligned}\ ] ] with @xmath223 . using this decomposition in , one finds that each diagram decomposes into a series of blocks which are irreducible with respect to the symmetric part @xmath224 ( i.e. , any vertical line hits at least one symmetric contraction ) and connected to each other by antisymmetric contractions @xmath225 . the blocks which are irreducible with respect to @xmath224 can be formally resummed into an effective kernel @xmath226 and a newly generated effective vertex @xmath227 . the lowest order diagrams are shown in fig . [ fig : discrete_lg ] . ( left and middle diagram ) and the effective vertex @xmath227 ( right diagram ) when the symmetric part of the contraction is integrated out . @xmath228 ( @xmath229 ) denotes the symmetric ( antisymmetric ) contraction @xmath224 ( @xmath225 ) . ] using the diagrammatic rules together with ( [ contraction_decomposition ] ) and the convention ( [ e_definition ] ) , we obtain for the first two diagrams @xmath230 and for the third one ( including the interchange @xmath231 ) @xmath232 we use here the original perturbation series ( [ value_kernelb ] ) so that the unperturbed liouvillian @xmath201 occurs in the resolvents . performing the frequency integrations and assuming the band - width to be large , we obtain @xmath233 where we have performed the sum over the keldysh indices and used . we stress at this point that the kernel @xmath234 and the renormalized vertex @xmath235 without external lines are identical , @xmath236 nevertheless we will retain the distinction between the kernel and the vertex in the following , as the former appears in the final formulas for the correlation functions , whereas the latter appears in the diagrammatic expressions for the r.h.s . of the rg equations . we note that the frequency dependence in is generated during the flow as shown below . after integrating out the symmetric part of the fermi function in this way , we obtain a new diagrammatic series for the kernel analog to ( [ value_kernelb ] ) . the liouvillian and the vertices have to be replaced by the effective ones and the contractions between the effective vertices contain only the antisymmetric part @xmath225 . due to there occur no diagrams including the new vertex @xmath237 . furthermore , since the effective quantities have become energy dependent ( also the effective vertex @xmath238 becomes energy dependent in higher order perturbation theory ) , one has to replace @xmath239 ( with @xmath240 ) in ( [ value_kernelb ] ) . since the antisymmetric part of the contraction ( [ contraction_decomposition ] ) does not depend on the keldysh indices , only the effective vertex @xmath238 averaged over the keldysh indices occurs in the new perturbative series . in the second continuous rg procedure we deal with the remaining antisymmetric part of the fermi distribution function , where in each infinitesimal step a small energy shell is integrated out . instead of integrating out the energies on the real axis , it has turned out to be more efficient to integrate out the matsubara poles of the fermi distribution function on the imaginary axis @xcite . this is achieved by introducing a formal cutoff dependence into the antisymmetric part of the fermi distribution by @xmath241 where @xmath242 are the matsubara frequencies corresponding to the temperature of reservoir @xmath31 , and @xmath243 is a theta function smeared by temperature . for @xmath244 , ( [ fermi_cutoff ] ) yields the full antisymmetric part @xmath245 of the fermi distribution . in each rg step , one reduces the cutoff @xmath217 by @xmath246 , and integrates out the infinitesimal part @xmath247 of the fermi distribution . for example , the new effective liouvillian at scale @xmath248 @xmath249 and similarly the new effective vertices @xmath250 and @xmath251 as well as the kernel @xmath252 can be calculated technically in the same way as for the first discrete rg step . the only difference is that an infinitesimal small part is integrated out , so that the rg diagrams contain only one contraction involving @xmath253 . furthermore , since the diagrams have to be irreducible with respect to this part , this contraction must connect the first with the last vertex of the diagram . using this procedure the rg equations for the dot liouvillian @xmath254 and the vertex @xmath255 have been derived in ref . . here we will use this technique to obtain the rg equations for @xmath256 as well as @xmath257 . in @xmath258 . the slash indicates the contraction where the fermi function has to be replaced by @xmath259 . ] up to @xmath260 . ] the diagrams contributing to the rg equations for @xmath261 and @xmath262 are shown in fig . [ fig : rgb12 ] and fig . [ fig : rgsigmab ] , respectively . using the definition @xmath263 together with the convention @xmath264 we obtain the following rg equations : @xmath265 , \label{eq : bvertexcontrg}\ ] ] and @xmath266 \nonumber\\ & & -\frac{d\gamma^\lambda_{1}}{d\lambda}\,\gamma_2^\lambda\,\gamma_3^\lambda\ , \bar{g}_{12}(\omega)\,\pi_{12}(\omega ) \bigl [ \mathcal{b}_\pm(\omega_{12}+\bar{\omega}_{12},\xi_{12}+\bar{\omega}_{12})\ , \pi_{12}(\xi)\ , \bar{g}_{\bar{2}3}(\xi_{12}+\bar{\omega}_{12})\nonumber\\ * & & \hspace{20 mm } + \bar{g}_{\bar{2}3}(\omega_{12}+\bar{\omega}_{12})\,\pi_{13}(\omega)\ , \mathcal{b}_\pm(\omega_{13}+\bar{\omega}_{13},\xi_{13}+\bar{\omega}_{13})\ , \bigr]\pi_{13}(\xi)\ , \bar{g}_{\bar{3}\bar{1}}(\xi_{13}+\bar{\omega}_{13 } ) . \label{eq : bkernelcontrg}\end{aligned}\ ] ] we recall here that the kernel @xmath267 and the vertex without external lines @xmath188 equal each other , see , which yields a closed set of rg equations . we will further show in app . [ sec : b12 ] that the two - loop diagrams for @xmath203 as well as the one - loop diagrams containing @xmath203 itself do not contribute at second order in the coupling constant and hence can be neglected on the r.h.s . of . the initial conditions of the rg equations are given by ( [ eq : bkernela ] ) and ( [ eq : b12a ] ) . since @xmath268 , the solution at @xmath220 provides the result for the kernel @xmath269 from which the correlation functions can be calculated via . as the resolvents and the vertices on the r.h.s . of the rg equations are analytic functions in all frequencies @xmath270 in the upper half of the complex plane , all frequency integrations can be calculated analytically . the only poles contributing are the ones of the contractions and their derivatives given by with as well as @xmath271 here @xmath272 denotes the matsubara frequency @xmath273 which lies closest to the cutoff @xmath217 . after performing the integration we find@xcite that , due to the presence of the cutoff function @xmath274 , the r.h.s . of the rg equations gives a negligible contribution for @xmath275 . therefore , we can start the rg at @xmath276 and omit the cutoff function @xmath277 ( the precise ratio between @xmath278 and @xmath35 is determined such that no linear terms in @xmath35 in the effective liouvillian are generated @xcite ) . as a consequence , only the matsubara poles of the fermi function in the upper half plane contribute and all real frequencies are simply replaced by matsubara frequencies . from now on , we write the frequency dependence explicitly and define the analytic continuation of the liouvillian and the vertices in imaginary frequency space by @xmath279 where we keep the real and imaginary parts of the laplace variable @xmath74 and the external frequencies @xmath280 and @xmath281 separated from now on . furthermore , @xmath282 , @xmath283 correspond to matsubara frequencies and the compact indices @xmath284 and @xmath285 on the l.h.s . do no longer contain the frequencies @xmath286 . with the definition @xmath287 the rg equations and in matsubara space can be written as @xmath288 and @xmath289\nonumber\\ * & & \hspace{-25mm}\qquad\times \,\pi(\xi_{13},\xi'+\lambda_{t_{\alpha_1}}\!\!+\omega_3)\ , \bar{g}_{\bar{3}\bar{1}}(\xi_{13},\xi'+\lambda_{t_{\alpha_1}}\!\!+\omega_3 ; -\omega_3,-\lambda_{t_{\alpha_1 } } ) . \label{eq : bkernelmatsubara}\end{aligned}\ ] ] in these equations we implicitly sum over all indices and matsubara frequencies on the r.h.s . of the rg equations which do not occur on the l.h.s . only positive matsubara frequencies smaller than the cutoff @xmath217 are allowed and each sum has to be written as @xmath290 which reduces to an integral @xmath291 for zero temperature . in the next two subsections we will solve the rg equations ( [ eq : bvertexmatsubara ] ) and ( [ eq : bkernelmatsubara ] ) analytically in the weak coupling regime up to @xmath258 . weak coupling is defined by the condition that the renormalized vertices @xmath292 stay small compared to one throughout the rg flow , so that the expansion in powers of @xmath159 on the r.h.s . of the rg equations is well defined . this condition is fulfilled if the various cutoff scales occurring in the resolvents are much larger than the kondo temperature @xmath3 at which the vertices would diverge in the absence of any cutoff scales . as is discussed in detail in refs . there exists a characteristic energy scale @xmath294 where @xmath295 is the renormalized magnetic field . for @xmath296 the cutoff scales @xmath297 , @xmath298 and @xmath5 can be neglected in the rg equation for the vertex @xmath215 ( see below ) . this leads to a reference solution @xmath299 which serves as the starting point for a systematic expansion in powers of the coupling constant @xmath300 , where @xmath301 ( see ( [ g_vertex_liouville ] ) together with ) . this yields a perturbative solution of the rg equations in the regime @xmath296 . these results serve as initial values for the flow in the second regime @xmath302 . here the renormalization of the vertex @xmath303 is at least of order @xmath10 , where @xmath304 . provided the weak - coupling condition @xmath305 is satisfied all quantities can be calculated perturbatively . this fact crucially relies on the appearance of some relaxation / dephasing rate in the resolvents , which is guaranteed by ( as then in all resolvents standing left to a vertex @xmath215 the zero eigenvalue of the liouvillian can not contribute ) . this analysis has been performed for the liouvillian and the current kernel in the anisotropic kondo model in ref . . we note that the zero eigenvalue may appear in the resolvent left to the vertex @xmath188 in . we will show below that this does not lead to any problems in the calculation of the spin - spin correlation functions in the kondo model up to order @xmath10 . regarding the appearance of the external frequency @xmath306 as one of the cut - off parameters in we see from the perturbative expansion that @xmath307 does not appear in the resolvents and as a cut - off parameter for all vertices right to @xmath188 . this fact , however , will only affect the results in the regime @xmath308 . we will therefore use as unique cut - off for all vertices appearing in the derivation of the kernel @xmath309 . in sec . [ sec : expl ] we will show that for the spin operator in the kondo model the difference yields a correction @xmath310 and can thus be neglected . nevertheless we stress that all vertices appearing in the stationary reduced density matrix @xmath311 in do not possess @xmath307 as cut - off parameter . we thus deduce that in order to stay in the perturbative regime we can not rely on the external frequency @xmath307 but have to require @xmath312 . we finally note that temperature serves as a unique cutoff for all terms on the r.h.s . of the rg equations as for @xmath313 the matsubara sums are reduced to one term and the cutoff @xmath314 becomes independent of @xmath217 . this trivial cutoff is set to zero in the following , i.e. we will set @xmath315 . after these preliminary remarks let us turn to the evaluation of the rg equations . the one - loop rg equation for the vertex @xmath215 is at zero temperature given by @xcite @xmath316 the rg equation for the reference solution @xmath317 is obtained by assuming @xmath217 to be much larger than any other term appearing in the resolvent , which gives @xmath318 . \label{eq : poorman}\ ] ] the initial condition for @xmath319 at @xmath320 is the bare vertex @xmath321 defined in . the leading order solution is proportional to the coupling constant @xmath322 . we stress that the term on the r.h.s . , which is @xmath323 , contributes to the change of the vertex at order @xmath324 . this is a general feature of the rg above @xmath219 . in order to calculate the change of a quantity at order @xmath325 one has to analyze those terms @xmath326 , where @xmath327 is some energy scale , and @xmath328 on the r.h.s . of the corresponding rg equation . in contrast , terms @xmath329 @xmath330 do not contribute to the change at order @xmath325 . the rg equation for the vertex @xmath331 is given by with the replacement @xcite @xmath332 . using this leading order solution we can formally expand all quantities in powers of @xmath324 , i.e. @xmath333 here @xmath68}_-$ ] is the bare dot liouvillian . we recall , which implies @xmath334 in all orders in @xmath324 . furthermore , we have already indicated which terms will depend on the matsubara frequencies , external frequencies @xmath335 and @xmath336 and the laplace variable @xmath337 . the vertex @xmath215 and the liouvillian were calculated in ref . . we will here state those results needed for the calculation of and . the second order vertex @xmath338 is further decomposed as @xmath339 here the vertex @xmath340 is given by @xcite @xmath341.\ ] ] for the kondo model the vertex @xmath342 turns out to have the same matrix structure as the leading order solution @xmath299 . this implies that both can be put together by redefining @xmath343 , which amounts to a two - loop renormalization of the kondo temperature @xcite . furthermore , the vertex @xmath344 is generically given by @xmath345 the zeroth - order liouvillian is given by the initial condition @xmath68}_-$ ] with while the first - order liouvillian is further decomposed as @xmath346 where @xmath347 and @xmath348 do not depend on the laplace variable . the second - order liouvillian was calculated in ref . , however , we will not need it for the solution of ( [ eq : bvertexmatsubara ] ) and ( [ eq : bkernelmatsubara ] ) in the regime @xmath296 . let us now turn to the calculation of and . the zeroth - order term of the kernel is just given by the initial condition , i.e. @xmath349}_\pm.\ ] ] for the derivation of an rg equation for @xmath350 we have to keep all terms @xmath351 on the r.h.s . of ( [ eq : bkernelmatsubara ] ) . taking the zero - temperature limit and keeping only the order @xmath352 in the resolvents we obtain @xmath353 we evaluate this integral by assuming @xmath354}_-=\kappa\,\mathcal{b}_\pm^{(0 ) } , \label{eq : defkappa}\ ] ] which has to be checked for the specific model at hand . in the kondo model we will find @xmath355 for @xmath356 and @xmath357 for @xmath358 ( see sec . [ sec : expl ] below ) . can be used to shift @xmath359 to the right and evaluate the remaining integral by a partial fraction expansion @xmath360\bar{g}_{\bar{2}\bar{1}}^{(1)},\quad \label{eq : rgb1}\end{aligned}\ ] ] where we have defined @xmath361 the leading term in is extracted by treating the terms @xmath362 separately , @xmath363 where @xmath364 can be integrated by @xmath365 with @xmath366 which has the asymptotic behavior @xmath367 as @xmath368 . using in we obtain the frequency independent result @xmath369 the initial condition is given by , i.e. @xmath370 . as mentioned above we identify this with the vertex in first order , @xmath371 . using the result for @xmath372 we derive in app . [ sec : b12 ] the vertex , which is given by @xmath373 , \label{eq : b12result}\end{aligned}\ ] ] where @xmath374 ( see sec . [ sec : expl ] below ) . we note that @xmath375 . for large @xmath217 and using the result we can derive the rg equation for the kernel in second order ( see app . [ sec : bkernel ] ) @xmath376\bar{g}_{\bar{2}\bar{1}}^{(1)}\nonumber\\ * & & \hspace{-20mm}-\frac{{\text{i}}}{2\lambda}\bigl [ \bar{g}_{12}^{(2a_1)}\,\mathcal{b}_\pm^{(0)}\,\bar{g}_{\bar{2}\bar{1}}^{(1 ) } + \bar{g}_{12}^{(1)}\,\mathcal{b}_\pm^{(0)}\,\bar{g}_{\bar{2}\bar{1}}^{(2a_1 ) } \bigr]-\frac{1}{2\lambda}\,\bar{g}_{12}^{(1)}\,\mathcal{b}_\pm^{(1)}\ , \bar{g}_{\bar{2}\bar{1}}^{(1 ) } + \frac{1}{2\lambda}\,\bar{g}_{12}^{(1)}\ , { \left[z^{(1)},\mathcal{b}_\pm^{(0)}\right]}_+\ , \bar{g}_{\bar{2}\bar{1}}^{(1 ) } , \label{eq : b2kernelrg}\end{aligned}\ ] ] where @xmath377 . the initial condition is given by . at this point it turns out to be useful to decompose the second - order kernel as @xmath378 here @xmath379 is given by @xmath380\bar{g}_{\bar{2}\bar{1}}^{(1)},\label{eq : sigmab2aabovelc}\end{aligned}\ ] ] which satisfies the initial condition @xmath381 the remaining terms in satisfy @xmath382,\quad\label{eq : sigmab2babovelc}\\ & & \frac{d}{d\lambda}\sigma_b^{\pm,(2c)}= -\frac{1}{2\lambda}\,\bar{g}_{12}^{(1)}\,\mathcal{b}_\pm^{(1)}\ , \bar{g}_{\bar{2}\bar{1}}^{(1)}\nonumber\\ * & & \quad+\frac{1}{2\lambda}\,\bar{g}_{12}^{(1)}\ , { \left[z^{(1)},\mathcal{b}_\pm^{(0)}\right]}_+\ , \bar{g}_{\bar{2}\bar{1}}^{(1)},\label{eq : sigmab2cabovelc}\end{aligned}\ ] ] with initial condition ( according to ( [ eq : bkernela ] ) and ( [ eq : sigmab2a_initial ] ) ) @xmath383 the rg equations in the regime @xmath296 derived in this section for a model describing spin / orbital fluctuations will be specialized and solved for the isotropic kondo model in sec . [ sec : kondolargel ] below . in the next section we will first study the effect of the rg flow in the regime @xmath302 . as explained in ref . the rg above @xmath219 has resummed all leading and subleading logarithmic contributions in @xmath384 into the renormalized vertices . at @xmath385 , the bare coupling constant is replaced by a renormalized one @xmath11 , all logarithmic contributions are eliminated , and a simple power series in @xmath11 remains . thus the rg equations can be solved perturbatively provided @xmath386 . in addition , the liouvillian in the resolvents is replaced by the full effective liouvillian @xmath80 . when calculating diagrams containing the resolvent an obvious problem is that the frequency dependence of the effective liouvillian @xmath80 is not known explicitly . to circumnavigate this complication we use the following approximation for the resolvents , @xmath387 here the poles @xmath388 of the resolvent follow from the self - consistency equation @xmath389 the residue satisfy @xcite @xmath390 and hence can be set to one in the following . starting with the one - loop rg equation for the vertex @xmath215 we observe that the terms on the r.h.s . are already of @xmath391 . therefore , the renormalization of the leading - order vertex @xmath299 stops at @xmath385 and we have to use its value @xmath392 in all calculations from now on . note that we indicate the use of the coupling constant at @xmath219 by the additional superscript @xmath393 . furthermore , as we are eventually interested in the kernel @xmath394 up to @xmath391 we deduce that the higher - order vertices @xmath340 and @xmath344 will not be needed below @xmath219 . using the replacement @xmath395 in we obtain up to @xmath391 @xmath396 where we have already used that the flow of the kernel in order @xmath324 also stops at @xmath219 . the second - order kernel satisfies @xmath397 where we have used the approximation and only kept terms @xmath398 . the initial condition is given by the solution at @xmath385 . using the decomposition we obtain @xmath399 where the flow of @xmath400 below @xmath219 is governed by . the initial value at @xmath385 is given by , which we can rewrite as @xmath401\nonumber\\ * & & \hspace{-30mm}\times \bar{g}^{(1)c}_{12}\,p_i(z_i)\,\mathcal{b}_\pm^{(0)}\,p_j(z_j)\ , \bar{g}^{(1)c}_{\bar{2}\bar{1}}. \label{eq : sigmab2ainitiallambdac}\end{aligned}\ ] ] in doing so we have assumed @xmath402 and neglected terms of order @xmath403 . we will show in app . [ sec : ass ] that these assumptions are fulfilled for the kondo model in a magnetic field . we will further show that @xmath404 for all pairs @xmath405 for which the last line in is non - zero . when applying the results of this section to other models of spin / orbital fluctuations one has to ensure the validity of the assumptions made above . the initial value problem with is readily solved using @xmath406 with @xmath407 the result for the kernel at @xmath220 is then obtained using @xmath408 : @xmath409 . \label{eq : sigmab2aresult}\end{aligned}\ ] ] together with @xmath410 and @xmath411 determined in the rg procedure above @xmath219 this yields the final result for the kernel in second order in @xmath11 . it is applicable to any operator @xmath118 which does not couple the dot and reservoir degrees of freedom , i.e. whose initial value satisfies @xmath151 in . furthermore the calculations were done for a generic model describing spin or orbital fluctuations as the initial vertex @xmath159 was assumed to have two external legs . the only assumptions we have made regarding the model specifics are the commutation relations and , eqs . and as well as the specific relation between the parameter @xmath412 and the poles @xmath388 and @xmath413 , which have to be determined from the self - consistency equation . in the next section we will apply the results derived above to the spin - spin correlation functions in the isotropic kondo model . in particular , we will show that the assumptions discussed above are justified in this model . finally we note that similar results for the effective liouvillian have been derived in refs . . in this section we will specialize the generic results derived above to the case of the spin - spin correlation functions in the isotropic , antiferromagnetic kondo model in a magnetic field . the hamiltonian was presented in sec . [ sec : kondo ] ; in particular , the dot hamiltonian and the coupling to the leads as given in and . the first step is to represent the initial vertex and hamiltonian in liouville space , @xmath414}_-=h_0\,(l^{+z}+l^{-z})=h_0\,l^h,\end{aligned}\ ] ] where the spin superoperators @xmath415 are defined by their action on an arbitrary operator @xmath82 on the dot hilbert space via @xmath416 an explicit matrix representation for the superoperators @xmath417 is given in app . [ sec : appl ] , where also further superoperators are defined . the leading - order vertex @xmath299 was derived in refs . . it can be parametrized for @xmath418 as @xmath419 where @xmath420 is the vector formed by the pauli matrices . the vertex for @xmath421 is obtained using the antisymmetry @xmath422 . inserting into the rg equation and using the antisymmetry @xmath423 , , and @xmath424 we obtain @xmath425 if we assume the form , i.e. @xmath426 with @xmath427 , we obtain the usual poor - man scaling equation @xmath428 with the solution @xmath429 eq . explicitly shows that the term @xmath323 on the r.h.s . contributes to the renormalization in order @xmath324 . similarly , the renormalization of a quantity in order @xmath325 is determined by the terms @xmath430 as well as @xmath328 . in contrast , a term @xmath431 with @xmath432 does not contribute at order @xmath325 , as can be seen from @xmath433 the vertex @xmath331 is given by @xmath434 beside the leading - order vertex and the zero - order liouvillian we also need explicit expressions for the vertex @xmath340 as well as the liouvillian in first order . these are given by @xcite @xmath435 where @xmath436 is the coupling of the leading - order vertex @xmath299 and the trace tr is taken in the reservoir indices , @xmath437 . furthermore we have taken the scaling limit @xmath438 , @xmath439 such that the kondo temperature @xmath3 remains constant . as the next step it is straightforward to derive the following results for the initial vertex @xmath359 : @xmath440 where @xmath228 takes the values @xmath441 for @xmath442 and we have set @xmath443 , @xmath444 . we stress that in this way the spin operators are directly represented by their matrices in liouville space . hence , we do not have to use a pseudo - fermion representation of the kondo spin . the vertex @xmath160 defined in is just given by @xmath445 . furthermore , we recall that the kernels in zeroth order are just given by @xmath191 and @xmath446 , respectively . now the rg equation for @xmath350 reads @xmath447 where we have used and . the additional factor of two is due to the implicit summation over @xmath448 . hence the solution in the scaling limit is given by @xmath449 in second order we will here determine only the terms @xmath450 and @xmath451 as their flow is cut - off at @xmath385 . the remaining term @xmath400 will be derived in the next section . hence we have to solve the rg equation , which using and reads @xmath452 analogously we obtain for using , and @xmath453 the solutions read @xmath454 where we have already taken the scaling limit so that the contributions from the initial condition are negligible . as we have already explained above the rg flow of the leading order solution @xmath299 stops at the scale @xmath219 which is given by the maximal one of the external parameters , @xmath455 the value of the vertex at @xmath219 is given by with @xmath456 where due to the couplings satisfy @xmath457 with @xmath458 and @xmath459 the asymmetry ratio is defined as @xmath460 . as already mentioned the flow of all vertices right to @xmath188 does not stop at @xmath219 as defined in but rather at @xmath461 . this affects the result for the kernel @xmath309 only in the regime @xmath462 . we will discuss the changes in this case separately at the end of this section . we stress , however , that the flow of all vertices @xmath215 needed for the derivation of the stationary reduced density matrix @xmath311 is cut off by @xmath461 . hence in order to stay in the weak - coupling regime we need @xmath463 as the definition of the scale @xmath219 is to some extent arbitrary as long as it remains of the order of the external energy scales in the problem , it is necessary to study the effect of a redefinition @xmath464 with @xmath465 . this will induce a redefinition of the coupling as @xmath466 as we will show below , the redefinition @xmath464 does not change the final results for the liouvillian or the kernel @xmath309 up to order @xmath10 . the liouvillian up to @xmath391 can be parametrized as @xcite @xmath467 the liouvillian can be diagonalized ( see ) using the eigenvalues ( @xmath74 ) @xmath468 and the projectors @xmath469 the eigenvalues can now be used to determine the poles @xmath388 of the resolvent defined in . solving the self - consistency equation one finds @xmath470 , @xmath471 , and @xmath472 , where the spin relaxation and dephasing rates and the renormalized magnetic field are given up to @xmath391 by @xmath473 the higher order terms @xmath474 for the rates were obtained in ref . . the voltage was defined in and we always assume @xmath475 . as these rates are obtained from the liouvillian at @xmath476 and @xmath477 , respectively , the external frequency @xmath307 does not appear as a cut - off in the definition of @xmath219 . furthermore , we have defined the auxiliary functions @xmath478 which arises from terms like @xmath479 by neglecting the imaginary part of @xmath413 , which is proportional to @xmath480 , in the prefactor and takeing the real and imaginary parts . we note that as @xmath472 the renormalized magnetic field automatically appears in the logarithm . we have therefore also kept @xmath5 in the linear prefactor . the deviation of @xmath481 from @xmath482 is only important for @xmath483 and will be neglected otherwise . furthermore , we have omitted the imaginary part of the laplace variable , @xmath29 , since for the correlation functions calculated below we only need the liouvillian on the real axis . for @xmath484 and using , the functions in the parametrization are given up to order @xmath10 by @xcite @xmath485 in the last line we have replaced the bare magnetic field by the renormalized field , such that the latter appears consistently in all functions . the change is of @xmath486 . we further note that a redefinition @xmath464 yields the same result for the liouvillian with the replacement @xmath487 . naively , the linear terms in @xmath488 and @xmath489 yield additional contributions in order @xmath490 . these are , however , exactly cancelled by terms appearing from the logarithms using @xmath491 , where @xmath492 is given by with @xmath464 . the facts presented above allow us to justify the assumptions and made in sec . [ sec : genericrgbelowlambdac ] above for the specific case of the kondo model ( see app . [ sec : ass ] ) . after the recall of the liouvillian in second order we will finally evaluate the kernel @xmath394 in the kondo model . the flow of the kernel also stops at @xmath219 except for one term , namely @xmath400 . hence we find @xmath493 where the first term is given by or depending on the operator studied , and the second , fourth , and fifth is obtained from , as well as using the replacement @xmath494 . for the evaluation of the remaining contribution @xmath495 from let us first consider the operator @xmath358 . using @xmath496 for @xmath497 and @xmath498 ( in all other cases the l.h.s . vanishes ) we find @xmath499 we note that in the zero eigenvalue of the effective liouvillian appears in the resolvent left to @xmath500 . as we will show in the next section the resulting term does not contribute to the correlation functions . for the evaluation of the kernel for the calculation of the susceptibility we use @xmath501 0 , & \text{otherwise } , \end{array}\right . \label{eq : gpb - pg}\ ] ] which results in @xmath502\nonumber\\ * & & \hspace{-30 mm } -\frac{{\text{i}}}{2}\frac{j_rj_l}{\omega-\xi+{\text{i}}(\delta-\xi ' ) } \biggl[\left(\mathcal{h}_2(\omega+v+\tilde{h } ) + \mathcal{h}_2(\omega - v+\tilde{h})-\mathcal{h}_2(\omega+v-\tilde{h } ) -\mathcal{h}_2(\omega - v-\tilde{h})\right.\nonumber\\ * & & \left.\quad -\mathcal{h}_2(\xi+v+\tilde{h})-\mathcal{h}_2(\xi - v+\tilde{h } ) + \mathcal{h}_2(\xi+v-\tilde{h})+\mathcal{h}_2(\xi - v-\tilde{h})\right ) \left(l^a-\tfrac{1}{2}\,l^c\right)\nonumber\\ * & & -\frac{1}{2}\left(\mathcal{h}_2(\omega+v+\tilde{h } ) + \mathcal{h}_2(\omega - v+\tilde{h})+\mathcal{h}_2(\omega+v-\tilde{h } ) + \mathcal{h}_2(\omega - v-\tilde{h})\right.\nonumber\\ * & & \left.\quad -\mathcal{h}_2(\xi+v+\tilde{h})-\mathcal{h}_2(\xi - v+\tilde{h } ) -\mathcal{h}_2(\xi+v-\tilde{h})-\mathcal{h}_2(\xi - v-\tilde{h})\right)l^h \biggr ] . \label{eq : sigmab2aresult2}\end{aligned}\ ] ] next we consider the case @xmath503 and start with the evalution of the kernel @xmath504 from by using @xmath505 \mp\frac{{\text{i}}}{4}\frac{\gamma^{3z}(z_1)}{\gamma^a(z_1)}l^5_\pm , & i=1,j=-,\\[2 mm ] 0 , & \text{otherwise}. \end{array}\right.\end{aligned}\ ] ] hence the double sum in reduces to a sum over @xmath498 , where the two terms have opposite signs and otherwise equal each other up to the appearance of the rate @xmath15 in the second term . as can be easily shown this sum vanishes in second order , i.e. @xmath506 similarly , the kernel @xmath507 is evaluated using @xmath508 \!\!-\frac{{\text{i}}}{4}l^4_\pm , & i=\pm , j=1,\\[2 mm ] 0 , & \text{otherwise } , \end{array}\right.\ ] ] which results in @xmath509\nonumber\\ * & & \hspace{-30 mm } -\frac{{\text{i}}}{2}\frac{j_rj_l}{\omega-\xi-\tilde{h}+{\text{i}}(\delta-\xi ' ) } \biggl[\left(\mathcal{h}_1(\omega+v)+\mathcal{h}_1(\omega - v ) -\mathcal{h}_2(\xi+v+\tilde{h})-\mathcal{h}_2(\xi - v+\tilde{h})\right ) l^5_\pm\nonumber\\ * & & + \left(\mathcal{h}_2(\omega+v-\tilde{h})+\mathcal{h}_2(\omega - v-\tilde{h } ) -\mathcal{h}_1(\xi+v)-\mathcal{h}_1(\xi - v)\right ) l^4_\pm\biggr ] . \label{eq : sigmab2aresult2pm}\end{aligned}\ ] ] a redefinition @xmath464 yields the same result for the kernels @xmath309 with the replacement @xmath487 as can be easily shown in the same way as for the liouvillian . finally , let us consider the kernel in first order in the regime @xmath510 . starting from we obtain using @xmath511 ( @xmath512 ) @xmath513 here the second vertex still depends on @xmath217 . when integrating this from @xmath461 to @xmath514 we obtain a contribution @xmath310 . hence we deduce @xmath515 a similar analysis shows that the difference between @xmath516 and @xmath517 is at least @xmath518 . in this section we will use the results for the liouvillian and the kernel to calculate the correlation functions and . we first calculate the auxiliary correlation functions using . we start with the parametrization of the liouvillian , which implies the form for the resolvent in @xmath519 . we further deduce from the previous section that the kernels @xmath520 admit the parametrizations @xmath521 where for example @xmath522 . we stress that in contrast to the parametrization of the liouvillian we have not introduced additional factors of @xmath523 here . the stationary reduced density matrix has the form @xmath524 with @xmath525 . the stationary reduced density matrix can be determined using . without magnetic field one simply finds @xmath526 . furthermore , the rates and are given by @xmath527 we now rewrite the resolvent using the projectors and use @xmath528&=&-\frac{1}{2}\,\delta_{i1 } , \quad\label{eq : bla1}\\ \mbox{tr}_s\bigl[\sigma_{s^z}(\omega)\,p_i(\omega)\ , \sigma_{s^z}^-(\omega,{\text{i}}0+)\,\rho_s^{st}\bigr]&=&\nonumber\\ * & & \hspace{-20mm}{\text{i}}\frac{\pi}{4}(j_r+j_l)^2\,\delta_{i1},\end{aligned}\ ] ] where we have applied as well as . we note in particular that the term @xmath529 does not contribute to . this yields with and @xmath530 : @xmath531 since @xmath532 we can neglect @xmath533 in . on the other hand , the real part of @xmath534 in the denominator has to be kept as it becomes large compared to @xmath307 in the small - frequency limit . hence we arrive at @xmath535 we note that the leading term of the susceptibility is of order @xmath10 . the correlation functions are plotted in figs . [ fig : plot1][fig : plot3 ] . we observe very good agreement with the results obtained by fritsch and kehrein using the flow - equation method @xcite . in the symmetric kondo model ( @xmath536 ) for various values of the applied voltage @xmath0 . ] in the symmetric kondo model ( @xmath536 ) for various values of the applied voltage @xmath0 . ] for @xmath537 and various values of the asymmetry ratio @xmath460 . ] let us further study the behavior of the correlation functions analytically . for small values of the frequency @xmath538 has the lorentzian form @xmath539 where we have used @xmath540 . this result for the small - frequency regime agrees with conclusions drawn from a mapping of the spin correlators to the one - particle green s function of majorana fermions @xcite . on the other hand , in the limit of large frequencies we find @xmath541 in agreement with the flow - equation method @xcite . we note that the @xmath542 appearing in the correlation function have their origin in the resolvent @xmath543 , hence the external frequency @xmath307 serves as a cut - off parameter in @xmath219 . this results in the logarithmic corrections at large frequencies . the susceptibility in the limiting regimes reads @xmath544 the first result shows a dependence of the gradient at small @xmath307 on the asymmetry ratio @xmath460 , while the second result indicates the revival of the fluctuation - dissipation theorem for @xmath545 . we note that the derivation of relies on the fact that the coupling constants @xmath11 appearing in the kernel @xmath546 are cut - off by the external frequency @xmath307 , as it was discussed at the end of sec . [ sec : expl ] . furthermore , the susceptibility @xmath547 has a maximum at @xmath548 , where it takes the value @xmath549 this behavior was also deduced using the flow - equation method @xcite . for various values of the asymmetry @xmath550 and applied voltage @xmath0 . in order to get a smooth behavior at @xmath551 we have kept the @xmath552 in the definition of @xmath553 in this region . the dotted line is a guide to the eye . ] in order to investigate the revival of the fluctuation - dissipation theorem we introduce the longitudinal fluctuation - dissipation ratio @xcite @xmath554 which is in equilibrium simply given by @xmath555 ( @xmath556 ) . using our results and we obtain @xmath557 which is plotted in fig . [ fig : plot4 ] . we find @xmath558 , i.e. the equilibrium result , whereas for small frequencies we get @xmath559 in agreement with refs . . we note that @xmath560 increases with increasing asymmetry @xmath550 as the coupling of the voltage to the dot becomes less effective . in the symmetric kondo model ( @xmath536 ) for @xmath537 and various values of the applied magnetic field @xmath1 . for @xmath562 we already have @xmath16 , which implies @xmath563 in order @xmath10 ( see sec . [ sec : longlargeh ] ) . ] in the presence of an external magnetic field the stationary reduced density matrix is given by @xmath564 , @xmath565 , @xmath566 , with the magnetization in leading order @xcite ( see also refs . ) @xmath567 to evaluate the correlation function we now use @xmath568\\ & \qquad\qquad= \left\{\begin{array}{ll}m\frac{\gamma^{3z}(\omega)}{\gamma^a(\omega)},&i=0,\\ -\frac{1}{2}-m\frac{\gamma^{3z}(\omega)}{\gamma^a(\omega)},&i=1,\\ 0 , & i=\pm , \end{array}\right . \end{split}\ ] ] where the term @xmath529 again does not contribute . from this a straightforward calculation using yields the correlation function up to @xmath391 @xmath569 where the zero - frequency @xmath61-peak does not appear because of our definition . the suppression of the correlation function by the finite magnetic field is shown in fig . [ fig : plot5 ] , which agrees very well with similar plots obtained using the flow - equation method @xcite . in the zero - frequency limit we find @xmath570 while the leading term @xmath518 in the large frequency regime is given by ( including the logarithmic corrections in the coupling constants ) . furthermore we observe a weak feature at @xmath571 which has for @xmath572 the line shape @xmath573.\quad \label{eq : featurev - h}\end{aligned}\ ] ] similar features appear at @xmath574 . in the symmetric kondo model ( @xmath536 ) for @xmath537 and various values of the applied magnetic field @xmath1 . for @xmath575 we have @xmath16 , which implies @xmath576 in order @xmath10 ( see sec . [ sec : longlargeh ] ) . ] for the calculation of the susceptibility we need @xmath577\\ & \qquad\qquad=\left[\frac{{\text{i}}}{2}\,\gamma_{s^z}^{-,3z}(\omega ) + { \text{i}}\,m\,\gamma_{s^z}^{-,a}(\omega)\right]\delta_{i1 } , \end{split}\ ] ] which directly yields @xmath578 , \quad\label{eq : chipresultfiniteh}\\ \chi_{s^zs^z}''(\omega)&= & \left[\frac{\pi}{4}(j_r+j_l)^2+m\,re\,\gamma_{s^z}^{-,a}(\omega)\right ] \nonumber\\ * & & \qquad\times\frac{\omega}{\omega^2+re\,\gamma^a(\omega)^2}. \label{eq : chippresultfiniteh}\end{aligned}\ ] ] in we have kept the terms in the second line , which are of @xmath579 , as due to @xmath580 ( @xmath581 ) they become dominant in the small - frequency limit . for larger frequencies these terms have to be neglected . thus the static susceptibility is given in leading order by @xmath582 in agreement with the literature @xcite . for larger frequencies the real part of the susceptibility possesses logarithmic features at @xmath583 due to the term @xmath584 . for example , @xmath585 where the terms represented by the dots do not contain any logarithmic features at @xmath586 . the imaginary part of the susceptibility is plotted in fig . [ fig : plot6 ] . it has a finite gradient at @xmath587 given by @xmath588 as well as a maximum at @xmath548 , where it takes the value @xmath589 in the large - frequency limit @xmath590 coincides with the correlation function . furthermore , the imaginary part of the susceptibility has features at @xmath583 which have their origin in the function @xmath591 contained in @xmath592 and hence have a line shape similar to . for @xmath593 and different values of the asymmetry @xmath550 and magnetic field @xmath1 . in order to get a smooth behavior at @xmath594 we have kept the @xmath552 in the definition of @xmath553 . for @xmath16 the fluctuation - dissipation ratio is independently of @xmath307 given by @xmath595 . the dotted line is a guide to the eye . ] the fluctuation - dissipation ratio @xmath596 defined in reads in the presence of a magnetic field @xmath597 which is plotted in fig . [ fig : plot7 ] . larger values of the magnetic field push the system closer to its equilibrium behavior , as only those lead electrons in the energy interval @xmath598 can couple to the dot and thus induce the non - equilibrium behavior . we note , however , that the equilibrium result @xmath595 is only reached for @xmath599 . furthermore , increasing the asymmetry @xmath550 drives the system towards the equilibrium situation as the coupling of the voltage to the dot becomes less effective . this effect is suppressed by increasing the magnetic field as overall less electrons couple to the dot . for small frequencies we obtain @xmath600 in the symmetric kondo model ( @xmath536 ) for @xmath602 and various values of the applied voltage @xmath0 . the correlation function is given by @xmath603 . for @xmath604 the susceptibility vanishes in order @xmath10 . the line shape close to @xmath586 is given by . ] in the symmetric kondo model ( @xmath536 ) for @xmath602 and various values of the applied voltage @xmath0 . we observe characteristic logarithmic features at @xmath18 . ] in the case of a strong magnetic field , @xmath16 , the correlation functions up to quadratic order in the coupling are still given by , and , respectively , where the magnetization is simply @xmath605 . one can easily show using @xmath606 that @xmath607 , which implies the equilibrium result @xmath608 . furthermore , the correlation function vanishes identically in order @xmath10 for @xmath604 . physically the kondo spin is in its ground state @xmath609 and the energy difference to the state @xmath610 due to the external magnetic field is given by @xmath5 . hence one has to apply at least the frequency @xmath611 to obtain any response from the spin , where the energy @xmath0 is provided by the applied voltage . this has to be contrasted with the result for the susceptibility in the equilibrium kondo model derived by garst et al . they used a relation between the inelastic electron scattering and the correlation function to show that the susceptibility in equilibrium has the small - frequency behavior @xmath612 , i.e. it is non - zero for @xmath613 . this linear behavior was also observed by costi and kieffer @xcite as well as hewson @xcite using a numerical renormalization group calculation . in analogy , we expect the non - equilibrium correlation functions to be nonzero for @xmath604 in higher order in @xmath11 . the consistent calculation of terms @xmath614 in the real - time rg procedure applied here would involve , however , 5-loop diagrams and is hence beyond the scope of this work . the correlation function in the regime @xmath16 is plotted in fig . [ fig : plot8 ] . we find excellent agreement with numerical results recently obtained by fritsch and kehrein using the flow - equation method @xcite . in particular , we observe a splitting of the sharp edge at @xmath586 due to the applied voltage , which leads to characteristic features at @xmath18 . using our result we can derive analytic expressions for the line shape close to these frequencies . for example , at @xmath4 we find @xmath615 the first term shows that the gradient of @xmath547 will become negative for @xmath613 if the applied voltage is large enough , i.e. @xmath616 . in the vicinity of @xmath617 the correlation function shows similar kink - like behavior . the physical origin of these kinks lies in the fact that at each of the energies @xmath18 a new process sets in , which involves a spin - flip on the dot costing the zeeman energy @xmath5 as well as the virtual hopping of an electron on and off the dot gaining or costing the energy @xmath618 , @xmath619 , or @xmath0 , respectively . the real part @xmath620 of the susceptibility shows logarihmic features at @xmath18 as is shown in fig . [ fig : plot8a].we stress that the splitting of the sharp edge at @xmath586 is a true non - equilibrium effect . finally let us discuss the transverse correlation functions in the presence of a magnetic field . we note that by virtue of we can restrict ourselves to the @xmath621-correlations . the corresponding kernels up to second order in @xmath11 were calculated in , , , , as well as . we will first discuss the susceptibility and present the results for the correlation function afterwards . in the symmetric kondo model ( @xmath536 ) for @xmath602 and various values of the applied voltage @xmath0 . ] for @xmath622 , @xmath602 , and various values of the asymmetry ratio @xmath460 . the result is invariant under @xmath623 . ] in order to derive the susceptibility we start with the parametrization @xmath624 where for example @xmath625 . note that we already indicated that the explicit frequency dependence in order @xmath10 appears in @xmath626 exclusively . ( there is of course an implicit frequency dependence of @xmath627 and @xmath628 through @xmath11 . ) now using @xmath629\nonumber\\*[1 mm ] & & \qquad={\text{i}}\left(\gamma_{s^+}^{-,3}+\bigl(\gamma_{s^+}^{-,2 } + 2\gamma_{s^+}^{-,5}(\omega)\bigr)m\right)\delta_{i+}\qquad\end{aligned}\ ] ] we obtain @xmath630 where we have introduced the short - hand notation @xmath631 . for @xmath632 we find @xmath633 . the transverse susceptibility has a peak at the solution of @xmath634 which is up to first order solved by @xmath635 in a finite magnetic field the spin on the dot will be in its ground state @xmath609 . the energy difference to the excited state @xmath610 is given by @xmath5 , leading to an enhanced response of the system at this frequency . at the peak the imaginary part of the susceptibility takes the value @xmath636 as is shown in figs . [ fig : plot9 ] and [ fig : plot10 ] . the peak is suppressed by increasing the voltage , since this reduces the probability for the kondo spin to be in its ground state . on the other hand , for a fixed value of the voltage the peak increases with increasing asymmetry ratio as the coupling of the voltage to the dot becomes less effective . the width of the peak is up to order @xmath10 given by @xmath637 with the limiting cases @xmath638 in the equilibrium limit , @xmath639 , this corresponds to the result obtained in ref . . the real part of the transverse susceptibility possesses logarithmic features similar to at @xmath18 . using a pseudo - fermion representation of the kondo spin together with non - equilibrium perturbation theory paaske et al . @xcite previously obtained the transverse susceptibility . in order to compare these results to we first make the approximations @xmath640 and @xmath641 . in the limit @xmath642 we then obtain using @xmath643 @xmath644 in the regime @xmath645 we have @xmath646 , thus we can neglect @xmath628 as well as @xmath626 in the numerator in , which results in @xmath647 these approximations agree with the results obtained in ref . ( we have to replace @xmath648 due to a different sign in the definition of the bare dot hamiltonian @xmath21 ) . furthermore , in the regime @xmath649 we can use @xmath650 where we have replaced @xmath651 in the real part of @xmath626 , to obtain @xmath652 this confirms a conjecture by paaske et al . we would like to stress , however , that our result goes beyond the approximation . in analogy to the susceptibility one finds for the correlation function @xmath653 where we have neglected all terms of order @xmath403 in the numerator . this allows the calculation of the transverse fluctuation - dissipation ratio @xmath654 which is plotted in fig . [ fig : plot11 ] . for negative frequencies @xmath655 the fluctuation - dissipation ratio takes the value @xmath656 , whereas for frequencies @xmath657 we find @xmath658 , thus recovering the equilibrium situation in these limits . as for the longitudinal fluctuation - dissipation ration we observe that increasing the magnetic field or the asymmetry ratio @xmath550 drives the system towards the equilibrium situation . for @xmath593 and various values of the asymmetry ration @xmath550 and the applied magnetic field @xmath1 . the plot shows good agreement with similar results obtained in ref . . the dotted line is a guide to the eye . ] in this article we have generalized the real - time renormalization group method in frequency space to allow the calculation of dynamical correlation functions of arbitrary dot operators in systems describing spin and/or orbital fluctuations . we applied this to the two - lead kondo model in a magnetic field , where we calcualted the longitudinal and transverse spin - spin correlation and response functions up to second order in the exchange coupling . we wish to stress that within this formalism the kondo spin is directly represented by matrices in liouville space , hence there is no need to apply a pseudo - fermion representation . specifically , we derived the two - loop rg equations for the dot operators and solved them analytically up to order @xmath10 in the weak - coupling regime . here @xmath11 denotes the effective coupling at the energy scale @xmath12 which has to satisfy @xmath14 . our results show several features attributed to the non - equilibrium situation , e.g. the splitting of the edge at @xmath586 of the longitudinal correlation function in a strong magnetic field or the suppression of the peak in the transverse susceptibility by a finite applied voltage . furthermore , we find very good agreement with results for the longitudinal correlation function recently obtained by fritsch and kehrein using the flow - equation method @xcite . a particular advantage of our approach is the possibility to obtain analytic expressions for all correlation functions in the weak coupling limit . we have calculated the spin - spin correlation functions for the nonequilibrium kondo model in the weak coupling regime @xmath14 . the regime of strong coupling , @xmath659 , is still an open problem . in this case , the exchange couplings @xmath11 become of order @xmath660 and a controlled truncation of the rg equations is no longer possible . within the present rtrg - fs method it was shown in ref . that the relaxation / dephasing rates saturate to the kondo temperature in the strong coupling regime . as an effect the coupling constants do not diverge as in poor man scaling methods but remain finite . however , the numerical solution of the rg equations in lowest order showed an instability against an exponentionally small change in the initial condition for the relaxation / dephasing rates . although it was possible to find excellent agreement for the temperature dependence of the linear conductance with nrg calculations , it was necessary to fine tune the initial condition for the rates . therefore , up to now , it is not yet clear whether a controlled solution of the strong coupling regime is possible by using rtrg - fs . the nonequilibrium kondo model describes the spin fluctuation ( or coulomb blockade ) regime of the more general nonequilibrium anderson impurity model ( for a systematic derivation of the kondo model from the anderson model using a schrieffer - wolff transformation , see e.g. ref . ) . in this model the single - particle spectral function is of most interest , which has recently been studied within nrg in a scattering wave basis @xcite . the present rtrg - fs method can also be applied to this model but , as explained in detail in ref . , in the charge fluctuation regime it is not yet clear whether a well - defined weak coupling regime exists at zero temperature . at resonance ( i.e. when the renormalized single - particle level is identical to one of the chemical potentials of the leads ) there is no energy scale except the broadening of the level itself and the expansion parameter is of order @xmath660 . nevertheless the results obtained in ref . ( using a previous version of the real - time rg method ) after an a priori uncontrolled truncation of the rg equations were in excellent agreement with the bethe ansatz solution @xcite for the equilibrium occupation of the local level . these and related topics are of high interest and will be the subject of forthcoming research . we would like to thank benjamin doyon , markus garst , stefan kehrein , verena krting , mikhail pletyukhov , and frank reininghaus for valuable discussions . this work was supported the dfg - forschergruppe 723 `` functional renormalization group in correlated fermion systems '' and the dfg - forschergruppe 912 `` coherence and relaxation properties of electron spins '' . obviously the time - dependent correlation function can be written as @xmath661}_+\bigr\rangle_{st}- \bigl\langle a\bigr\rangle_{st}\bigl\langle b\bigr\rangle_{st}.\ ] ] now , applying the time - translational invariance @xmath662 for all @xmath663 fixed and finite as well as the relation @xmath664}_\pm\bigr\rangle_{st}^*= \bigl\langle{\left[a(t)_\mathrm{h},b(0)_\mathrm{h}\right]}_\pm\bigr\rangle_{st},\ ] ] which can be verified by a straightforward calculation using @xmath83 , we obtain . in the same way can be used to derive the relation @xmath665 . in this appendix we will calculate the vertex @xmath667 . starting from ( [ eq : bvertexmatsubara ] ) we first take the zero - temperature limit . furthermore , we can expand the resolvents up to @xmath668 as@xcite @xmath669 with @xmath670}_+.\ ] ] using this we obtain together with the expansion @xmath671 in ( [ eq : bvertexmatsubara ] ) @xmath672}_+\right ) \frac{1}{\xi_{12}+{\text{i}}\xi'+{\text{i}}\lambda+{\text{i}}\omega_1-l_s^{(0)}-\tilde{l}_s^{(1 ) } } \left(1-\frac{z^{(1)}}{2}\right)\bar{g}_{\bar{2}1 ' } -(1\leftrightarrow 1'),\end{aligned}\ ] ] where we have omitted the arguments of the vertices @xmath215 for simplicity . using @xmath673}_-= \tilde{\kappa}\,\mathcal{b}_\pm^{(0)},\quad { \left[l_s^{(0)}+\tilde{l}_s^{(1)},\mathcal{b}_\pm^{(1)}\right]}_-= \tilde{\kappa}\,\mathcal{b}_\pm^{(1)},\quad { \left[l_s^{(0)}+\tilde{l}_s^{(1)},{\left[z^{(1)},\mathcal{b}_\pm^{(0)}\right]}_+\right]}_- = \tilde{\kappa}\,{\left[z^{(1)},\mathcal{b}_\pm^{(0)}\right]}_+ , \label{eq : commapp}\ ] ] with @xmath674 for @xmath356 and @xmath675 for @xmath358 ( see sec . [ sec : expl ] ) we obtain after a partial fraction expansion @xmath676}_+\right)\nonumber\\ * & & \,\times \left [ \frac{1}{\omega_{12}-\tilde{\kappa}+{\text{i}}\delta+{\text{i}}\lambda+{\text{i}}\omega_1 -l_s^{(0)}-\tilde{l}_s^{(1)}}- \frac{1}{\xi_{12}+{\text{i}}\xi'+{\text{i}}\lambda+{\text{i}}\omega_1-l_s^{(0)}-\tilde{l}_s^{(1 ) } } \right]\left(1-\frac{z^{(1)}}{2}\right)\bar{g}_{\bar{2}1 ' } + ( 1\leftrightarrow 1').\qquad\end{aligned}\ ] ] we see that no terms @xmath677 can occur on the r.h.s . and , hence , in order to determine the rg equation for @xmath667 we have to replace the vertices @xmath215 by the leading order ones @xmath299 and omit all terms containing @xmath372 , @xmath348 and @xmath678 . this yields using @xmath374 as well as @xmath679 : @xmath680\bar{g}_{\bar{2}1'}^{(1)}+(1\leftrightarrow 1 ' ) . \label{eq : b12rgapp}\end{aligned}\ ] ] using the rg equation for the leading - order vertex we can integrate up to higher - order corrections and obtain the solution , which in particular satisfies the initial condition @xmath681 as given by from the discrete rg step . . ] which contain @xmath203 itself . ] the two - loop diagrams for the vertex @xmath203 are shown in fig . [ fig : twoloopb12 ] . as each diagram contains three vertices @xmath682 and we are interested in the vertex up to second order , we have to extract the terms @xmath683 . this is done by expanding the resolvents in lowest order @xmath684 where @xmath685 ( @xmath29 ) stands for either @xmath307 ( @xmath61 ) or @xmath686 ( @xmath687 ) , respectively . using this we immediately see that the first two diagrams are proportional to @xmath688 the fourth and fifth to @xmath689 and the third and sixth to @xmath690 thus the two - loop diagrams behave as @xmath691 and hence do not contribute to the renormalization of the second - order vertex @xmath667 . finally we have to study the one - loop diagrams which contain @xmath203 itself . these diagrams are shown in fig . [ fig : rgoneloopb12withb12 ] . as we can easily see from , the leading - order result for @xmath203 behaves for large @xmath217 as @xmath692 now the one - loop diagrams contain an additional vertex @xmath215 as well as a resolvent @xmath693 . hence we deduce that the diagrams yield terms proportional to @xmath694 which do not contribute to the renormalization of the second - order vertex @xmath667 . in this appendix we will derive the rg equation for the second - order kernel @xmath696 . for this we have to evaluate all terms on the r.h.s . of . we start with the first line . using ( [ eq : restildel ] ) and expanding @xmath215 in powers of @xmath324 yields @xmath697}_+\right)\nonumber\\ * & & \qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\times \frac{1}{\xi_{12}+{\text{i}}\xi'+{\text{i}}\lambda+{\text{i}}\omega_2-l_s^{(0)}-\tilde{l}_s^{(1 ) } } \left(1-\frac{z^{(1)}}{2}\right)\bar{g}_{\bar{2}\bar{1}}^{(1 ) } \label{eq : appc1}\\ & & + \int_0^\lambda d\omega_2\ , \bigl({\text{i}}\,\bar{g}_{12}^{(2a_1)}+ \bar{g}_{12}^{(2b)}(\omega,\delta;\lambda,\omega_2)\bigr)\ , \frac{1}{\omega_{12}+{\text{i}}\delta+{\text{i}}\lambda+{\text{i}}\omega_2-l_s^{(0)}}\ , \mathcal{b}_\pm^{(0)}\ , \frac{1}{\xi_{12}+{\text{i}}\xi'+{\text{i}}\lambda+{\text{i}}\omega_2-l_s^{(0)}}\ , \bar{g}_{\bar{2}\bar{1}}^{(1)}\label{eq : appc2}\\ & & + \int_0^\lambda d\omega_2\,\bar{g}_{12}^{(1)}\ , \frac{1}{\omega_{12}+{\text{i}}\delta+{\text{i}}\lambda+{\text{i}}\omega_2-l_s^{(0)}}\ , \mathcal{b}_\pm^{(0)}\ , \frac{1}{\xi_{12}+{\text{i}}\xi'+{\text{i}}\lambda+{\text{i}}\omega_2-l_s^{(0)}}\nonumber\\ * & & \qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\times \bigl({\text{i}}\,\bar{g}_{\bar{2}\bar{1}}^{(2a_1)}+ \bar{g}_{\bar{2}\bar{1}}^{(2b)}(\xi_{12},\xi'+\lambda+\omega_2 ; -\omega_2,-\lambda)\bigr),\label{eq : appc3}\end{aligned}\ ] ] where we have already neglected the terms @xmath348 and @xmath678 in the second and third line as they lead only to higher - order corrections . using the commutators the first line can be treated similarly to @xmath350 derived in sec . [ sec : wkabove ] . in the term @xmath698 we only keep the term containing @xmath364 ( the term @xmath699 was already used to calculate @xmath350 ) , while in the other two terms we have to extract the term @xmath677 . thus we arrive at @xmath700\bar{g}_{\bar{2}\bar{1}}^{(1)}\\ & \qquad-\frac{1}{2\lambda}\,\bar{g}_{12}^{(1)}\,\mathcal{b}_\pm^{(1)}\ , \bar{g}_{\bar{2}\bar{1}}^{(1 ) } + \frac{1}{2\lambda}\,\bar{g}_{12}^{(1)}\ , { \left[z^{(1)},\mathcal{b}_\pm^{(0)}\right]}_+\,\bar{g}_{\bar{2}\bar{1}}^{(1)}. \end{split } \label{eq : appbkernelresult1a}\ ] ] using the same steps for the terms containing @xmath340 in and one finds @xmath701 . \label{eq : appbkernelresult1b}\ ] ] for the terms of and containing @xmath344 we use in the form @xmath702 when inserted into and the first terms do not depend on the integration variable @xmath703 . the remaining integral can be done as usual by a partial fraction expansion . this yields ( @xmath704 for @xmath356 and @xmath357 for @xmath358 ) @xmath705\ , \bar{g}_{\bar{2}\bar{1}}^{(1)}\nonumber\\ * & & \hspace{-8mm}+ \frac{{\text{i}}}{\omega\!-\!\xi\!-\!\kappa\!+\!{\text{i}}(\delta\!-\!\xi')}\ , \bar{g}_{12}^{(1)}\,\left [ \mathcal{k}_\lambda(\omega_{12}+{\text{i}}\delta - l_s^{(0 ) } ) \mathcal{b}_\pm^{(0)}-\mathcal{b}_\pm^{(0)}\ , \mathcal{k}_\lambda(\xi_{12}+{\text{i}}\xi'-l_s^{(0)})\right]\ , \bar{g}^{(1)}_{\bar{2}3 } \ln\frac{2\lambda-{\text{i}}(\xi_{13}+{\text{i}}\xi'-l_s^{(0)})}{\lambda } \,\bar{g}^{(1)}_{\bar{3}\bar{1}}.\nonumber\end{aligned}\ ] ] if we now expand @xmath706 and the logarithm for large @xmath217 , @xmath707 and @xmath708 , and keep only the terms proportional to @xmath709 we arrive at @xmath710 \bar{g}_{\bar{3}\bar{1}}^{(1)}. \label{eq : bg2b1stterm}\ ] ] in contrast , the second terms of and do depend on the integration variable @xmath703 . the evaluations is , however , straightforward . we use a partial fraction expansion for the resolvents left and right to the vertex @xmath359 as well as @xmath711 this leads to integrals of the form @xmath712 where we are only interested in the @xmath683 terms . now using the asymmetry @xmath423 we find @xmath713 \bar{g}_{\bar{3}\bar{1}}^{(1)}. \label{eq : bg2b2ndterm}\ ] ] which has to be combined with for the full result from the terms containing @xmath344 . the second and third line of containing the vertex @xmath667 can be treated using the same steps as were used to evaluate the @xmath344-dependent parts of and . the result reads after some tedious but straightforward algebra @xmath714 \bar{g}_{\bar{3}\bar{1}}^{(1)}. \label{eq : appbkernelresult2}\ ] ] finally , to extract the leading term @xmath715 of the fourth line of one can simply replace the resolvents @xmath716 by @xmath717 . this yields @xmath718 \bar{g}_{\bar{3}\bar{1}}^{(1)}. \label{eq : appbkernelresult3}\ ] ] hence , the result for the rg equation of the kernel @xmath696 is obtained by summing , , , , and , using @xmath365 with in as well as @xmath719 \bar{g}_{\bar{3}\bar{1}}^{(1)}= \frac{d}{d\lambda}\,\bar{g}_{12}^{(1)}\,\mathcal{b}_\pm^{(0)}\ , \bar{g}_{\bar{2}\bar{1}}^{(1)}.\ ] ] this yields . consider an operator @xmath82 acting on the dot hilbert space having matrix elements @xmath720 with respect to the basis @xmath721 . if @xmath722 denotes the superoperator acting on dot operators via @xmath723}_\pm$ ] then for an arbitrary dot operator @xmath118 we have @xmath724 furthermore , we represent superoperators in the matrix representation @xmath725 if @xmath726 is the product of two superoperators , then @xmath727 and the matrix of @xmath77 is simply given by the matrix product of the matrices of @xmath728 and @xmath729 . a basis for the operators in the liouville space of the kondo dot can be built up by the spin superoperators @xmath730 and @xmath731 defined in . an explicit representation in the basis is provided by @xmath732 @xmath733 @xmath734 furthermore we define the operators @xmath735 as well as @xmath736 we note that @xmath737 as well as @xmath738 . in the spin sector we will use frequently @xmath739 as well as @xmath740 the liouville operators satisfy ( the sums are over @xmath741 while the index @xmath212 takes the values @xmath143 ) @xmath742}_+\,l^{2i}&= & \delta_{p-}\,\frac{{\text{i}}}{2}\,\text{tr}\,j\,l^h.\label{eq : appe5}\end{aligned}\ ] ] + the results for the liouvillian presented in sec . [ sec : resultsbelowlc ] allow us to obtain @xmath743 where we have used @xmath470 , @xmath744 and @xmath745 . furthermore , we can expand the zero - eigenvalue projector as @xmath746 where @xmath747 . this directly yields . to prove the final statement , @xmath404 for all pairs @xmath405 for which @xmath748 is non - vanishing , we note that for @xmath358 these pairs are given by ( see and ) @xmath749 for @xmath500 and @xmath750 for @xmath751 , respectively . on the other hand we have @xmath357 in both cases , which implies @xmath752 for @xmath358 . the same analysis can be performed for @xmath356 where the relevant pairs are given by @xmath753 for @xmath500 and @xmath754 for @xmath751 .

during the past ten years , the evolution of structural relaxation in glass - forming liquids has been intensively studied using neutron - scattering spectroscopy , various light - scattering techniques , dielectric - loss spectroscopy , and molecular - dynamics simulation . results of this work have also been used to test the mode - coupling theory ( mct ) , which interpretes the structural relaxation as precursor of the glass - transition . originally , the mct was proposed as an approximation approach for the cage effect in liquids @xcite . in its simplest version , the mct equations of motion describe an ideal liquid - to - glass transition , i.e. a bifurcation from ergodic to nonergodic dynamics , if control parameters like temperature @xmath3 or packing fraction @xmath4 cross critical values @xmath5 or @xmath6 , respectively . this bifurcation is connected with the evolution of a two - step relaxation scenario entirely determined by the regularly changing equilibrium structure . two divergent time scales appear , closely connected to two power - law decay processes . a detailed description of these results can be found in ref . @xcite and references therein . comparisons of the theoretical results for simple model systems with experiments done on colloids @xcite , and with computer - simulation studies @xcite demonstrate the validity of the microscopic mct approach . for the solutions of the mct equations , a variety of results has been derived by asymptotic expansions , using as a small parameter the distance from the critical point , @xmath7 , or @xmath8 , respectively . the leading - order results of this expansion establish universality features of the mct dynamics . assessments of the theory have been reached by comparing spectra in the ghz regime or relaxation curves within the pico - second window with the universal results . the outcome of this work , which is reviewed in ref . @xcite , leads to the conclusion that mct properly describes some essential features of structural relaxation even for some complicated molecular liquids . the mct for simple systems has been extended recently to liquids of nonspherical molecules @xcite . but so far , only the bifurcation equation for the so - called nonergodicity parameters resulting within the new theory could be solved . comparing these results with the findings of molecular - dynamics simulations for a liquid of linear molecules @xcite and for water @xcite indicates that the mct for molecular liquids is promising . it was also predicted that there can be two states of nonergodic motion for nonspherical molecules . these states are connected by a type - a transition if the molecules exhibit a top - down symmetry @xcite . such transitions are generic possibilities in mct , provided there is some symmetry in the problem rendering certain mode - coupling coefficients zero @xcite . at a type - a transition , the nonergodicity parameters change continuously , whereas at the conventional mct transition , referred to in this context as a type - b transition , a discontinuity occurs @xcite . in this paper , correlation functions and susceptibility spectra shall be discussed , which deal with the glassy dynamics of the orientational degrees of freedom of nonspherical molecules . the results are obtained as solutions of the equations of motion derived previously @xcite for the dynamics of a linear probe molecule immersed in a simple liquid . a top - down symmetric dumbbell of two fused hard spheres will be considered as the molecule , and as the solute , a hard - sphere system is chosen . this model deals with the simplest problem concerning glassy rotational dynamics , namely the influence of the cages formed by the neighbors of the molecule on the molecule s reorientational motion as it is caused by steric hindrance . the dynamics will be exemplified for two cases : a molecule with a large elongation and a molecule with a small elongation . it will be shown that large elongations lead to strong coupling of the rotational degrees of freedom to the density fluctuations of the solute , such that the glassy dynamics of the latter enforces the validity of all the universal mct laws for the solvent . moreover , the corrections to the leading - order - asymptotic laws show the same qualitative trends as studied for simple liquids @xcite . a motivation of the present study is the explanation of three general properties of the @xmath0-relaxation in molecular liquids , which are exhibited in fig . [ ersteabb ] . in this figure , experimental susceptibility spectra for the van - der - waals liquid propylene carbonate ( pc ) are reproduced for four temperatures . one set of data deals with the response for angular - momentum index @xmath1=1 ; it was obtained by dielectric - loss spectroscopy @xcite . the other set was measured by depolarized - light - scattering spectroscopy @xcite and deals with the ( @xmath1=2 ) reorientational dynamics . the data show for @xmath9k and @xmath10k @xmath0-relaxation peaks at @xmath11ghz ( @xmath1=1 ) and @xmath12ghz ( @xmath1=2 ) respectively . these temperatures exceed the melting temperature @xmath13k of pc by more than @xmath14k . lowering @xmath3 to @xmath15k , the @xmath0-peaks of the spectra are shifted down by about two orders of magnitude . the shape of the @xmath0-peak is temperature independent , and the ratio of the @xmath0-process - time scales , characterizing the @xmath0-peak - maximum positions for the two values of @xmath1 , is also @xmath3-independent . these are two features which mct predicts to be universal . the first nonuniversal feature to be understood is , that the @xmath0-peak intensity , taken relative to that of the band of microscopic excitations at around 1thz , is larger for the ( @xmath1=1 ) response than for the ( @xmath1=2 ) case : the former exceeds the latter by about a factor of @xmath16 . second , the ( @xmath1=1 ) response is slower than the response for @xmath1=2 : the ratio of the @xmath0-peak positions is about @xmath17 . third , the @xmath0-peak of the ( @xmath1=1 ) response is less stretched than the peak for @xmath1=2 , i.e. the halfwidth of the ( @xmath1=1 ) peak is smaller than that of the ( @xmath1=2 ) peak . if one describes these peaks by the spectra of the kohlrausch law , @xmath18 , the stretching exponent @xmath2 for @xmath1=1 , @xmath19 @xcite , is larger than the one for @xmath1=2 , @xmath20 @xcite . the same three @xmath0-peak features are noticed , if one compares the depolarized - light - scattering spectra of glycerol @xcite with the corresponding dielectric - loss spectra @xcite . a fourth general feature to be explained is the large ratio of the @xmath0-relaxation - time scale found by depolarized - light - scattering spectroscopy and the one found for the longitudinal elastic modulus by brillouin - scattering spectroscopy . for salol , a ratio of about 10 was reported @xcite , while for pc , a factor of about 5 was found @xcite . the small elongation of concern in this paper is chosen so that it exceeds the critical value for the above - mentioned type - a transition by about @xmath12% . the theory for the corrections to the leading - order asymptotic laws @xcite implies that these diverge at a type - a transition . therefore , the range of validity of the universality features of the standard mct bifurcation shrinks upon approaching the type - a transition . it will be shown that in our example the standard results are not exhibited any more for reasonable choices of the distance parameter @xmath21 . in particular , it is impossible to identify a two - step scenario for the odd-@xmath22 correlators , nor is @xmath0-relaxation scaling observed . the paper is organized as follows . in sec . [ themodelsystem ] , the model is defined , and the mct equations are noted . after an overview of the general scenario for the evolution of the glassy relaxation of the reorientational correlators ( sec . iiia ) , the differences between the relaxation patterns for the ( @xmath1=1 ) and ( @xmath1=2 ) response are described for strong ( sec . iiib ) and weak ( sec . iiic ) steric hindrance . in sec . iiid it is demonstrated how the @xmath2-relaxation is described by the first scaling law , and in sec . iiie it is discussed how the @xmath0-relaxation - scaling - law description emerges . the concluding sec . iv summarizes the results . a system of @xmath23 spherical particles shall be considered as the solvent . the basic variables describing the structure are the density fluctuations for the wave vector @xmath24 : @xmath25 . here @xmath26 , labels the centers of the particles . the structure factor @xmath27 provides the simplest information on the equilibrium distribution of the particles ; here , @xmath28 denotes canonical averaging . because of rotational symmetry , @xmath29 depends on the wave - vector modulus @xmath30 only . the structure factor can be expressed through the direct correlation function @xmath31 via the ornstein - zernicke equation @xmath32 ; where @xmath33 denotes the particle density @xcite . the simplest quantities , characterizing the structural dynamics in a statistical manner , are the normalized auto - correlation functions for the density fluctuations , called the density correlators @xmath34 . the evolution with increasing time @xmath35 is given by the canonical equations of motion . we will also need fourier - laplace transforms for complex frequency @xmath36 , @xmath37 , @xmath38 , using the convention : @xmath39 . for real frequency @xmath40 , one gets with @xmath41 : @xmath42 . the imaginary part @xmath43 is called the fluctuation spectrum , and @xmath44 is the susceptibility spectrum @xcite . the basic version of mct consists of two equations @xcite . the first one is exact and derived within the zwanzig - mori formalism : [ eins ] @xmath45 here , @xmath46 , with @xmath47 denoting the thermal velocity . the relaxation kernel @xmath48 is a fluctuating - force correlator . the equation has to be solved with the initial condition @xmath49 @xcite . equation ( [ einsa ] ) is equivalent to the double fraction @xmath50 the second mct equation is obtained by writing the kernel as a sum of a regular term and a contribution describing the cage effect . the latter is treated by kawasaki s factorization approximation for the force correlations . it is found to be a quadratic functional of the density fluctuations : @xmath51 . for the sake of simplicity , the regular term shall be neglected in the following . furthermore , the wave - vector modulus will be discretized to @xmath52 values with equal spacing @xmath53 . thus , @xmath54 , @xmath55 , @xmath56 can be considered as labels running from @xmath57 to @xmath52 . as a result , the kernel is given as a quadratic mode - coupling polynomial @xmath58 of the @xmath52 correlators @xmath59 , @xmath60 : @xmath61=\sum_{kp } v_{qkp}\,\phi_{k}(t)\,\phi_{p}(t ) . \label{zwei}\ ] ] the positive coupling coefficients @xmath62 are given by @xmath29 and @xmath31 @xcite . anticipating these equilibrium quantities to be known , eqs . ( [ einsa ] ) and ( [ zwei ] ) are closed . equations ( [ eins ] ) and ( [ zwei ] ) exhibit a transition from liquid - state dynamics in the regime @xmath63 or @xmath64 to glass - state dynamics for @xmath65 or @xmath66 . in the former regime the density fluctuations decay to zero for long times , @xmath67 . the ideal - glass states exhibit a nontrivial long - time limit , which is called the nonergodicity parameter , @xmath68 . it is the debye - waller factor of the glass . at the transition , this long - time limit is discontinuous , and the jump is called critical nonergodicity parameter or plateau , @xmath69 . at the critical point , the correlators decay algebraically : @xmath70 . the exponent @xmath71 , @xmath72 , is called the critical exponent , and @xmath73 is denoted as the critical amplitude . @xmath74 marks the time scale of the transient from the microscopic motion to the relaxation dynamics of the mct . the mct @xmath0-process is defined as the dynamics for those times , where the correlators of the liquid decay from the plateau @xmath75 to zero . the mct @xmath2-process deals with the dynamics , where the correlators are near the plateau , i.e. @xmath76 . the first relaxation step of the anomalous dynamics is given by the initial part of the @xmath2-process ; it deals with the decay towards the plateau for times outside the transient : @xmath77 , @xmath78 . the second step is the @xmath0-decay in the liquid . its initial part is identical with the final part of the @xmath2-process , and it follows von schweidler s law @xmath79 . the exponent @xmath80 , @xmath81 , is called the von - schweidler exponent . in a leading - order expansion in the small parameter @xmath82 one finds the universal results for the @xmath2-process . there holds the factorization theorem @xmath83 the dependence on time and on control parameters is given by the @xmath54-independent function @xmath84 , which is called the @xmath2-correlator . it is determined by the equation @xmath85 to be solved with the initial condition @xmath86 . the number @xmath87 , @xmath88 , is referred to as the exponent parameter . @xmath89 is a smooth function of the control parameters and is called the separation parameter . its zero defines the critical point . expanding in leading order in the distance @xmath21 , one can write @xmath90 . from eq . ( [ viera ] ) , one derives the first scaling law @xmath91 here , @xmath92 denotes the amplitude scale , and @xmath93 abbreviates the first characteristic time scale of the mct - transition scenario : [ sechs ] @xmath94 the master functions @xmath95 are determined by @xmath87 as solutions of eq . ( [ viera ] ) for @xmath96 , respectively . they interpolate monotonously between @xmath97 and @xmath98 or @xmath99 . von schweidler s law is obtained as the long - time limit on the scale @xmath93 in the form : @xmath100 . here @xmath101 abbreviates the second characteristic scale of the theory : @xmath102 the @xmath0-process obeys for @xmath103 the second scaling - law , called the superposition principle : @xmath104 the control - parameter independent master function @xmath105 exhibits the initial decay @xmath106 . the parameters @xmath75 , @xmath107 and @xmath87 are determined by @xmath58 from eq . ( [ zwei ] ) for control parameters at the critical point . the same holds for the function @xmath105 . the constant @xmath108 is determined by the first taylor coefficient in @xmath21 of the deviations of @xmath58 from its value at the critical point . formulas for these quantities can be found in ref . @xcite , where also the original work is cited . the theory for the leading corrections to the quoted results has been worked out in ref . @xcite . the calculations in this paper will be done for the hard - sphere system ( hss ) . the temperature does not enter the structure , but determines the time scale via the thermal velocity only : @xmath109 . the relevant control parameter is the packing fraction : @xmath110 , where @xmath111 is the particle diameter . the structure factor will be calculated within the percus - yevick theory @xcite . the discretization will be done for @xmath112 wave - vector values with step size @xmath113 . for this model , all the mentioned mct quantities have been reported in ref . in particular it was found : @xmath114 , @xmath115 , @xmath116 the results for the glass - transition of the hss are documented comprehensively in refs . @xcite , albeit for a brownian microscopic dynamics . the bifurcation scenario for the model with newtonian dynamics as defined by eqs . ( [ eins ] ) and ( [ zwei ] ) is demonstrated for the wave - vector @xmath117 in ref . @xcite , where the transient time - scale was determined : @xmath118 . for the presentation of our results in the following figures , the units of length and time will be chosen so that @xmath119 and @xmath120 . the control parameters @xmath4 shall be cited by the logarithm @xmath121 of the distance parameter @xmath21 : @xmath122 as in the previous work @xcite , the mct equations are solved in the time domain . the solutions are then laplace - transformed to get @xmath123 . similarly , the transformed kernel @xmath124 is calculated from @xmath48 in eq . ( [ zwei ] ) . these results are used to compare the left - hand side of eq . ( [ einsb ] ) with the right - hand side . thereby a verification of the numerical solutions is obtained . as a model for a dilute solution of molecules we shall consider a single linear molecule immersed in a simple system . the position of this molecule is described by the tensor - density fluctuations @xmath125 . here , @xmath126 denotes the center - of - mass position and @xmath127 abbreviates the axis of the molecule . the @xmath128 are related to the spherical harmonics by : @xmath129 . the solute - solvent equilibrium correlations are described by the generalized structure factors @xmath130 , where @xmath131 . the proper generalization of the density correlators for simple systems are tensor - density correlators for the molecule , @xmath132 . the mct for these quantities shall be simplified by restricting the correlators to the diagonal elements @xmath133 correlation functions for wave vectors @xmath24 different from @xmath134 can be obtained from the specified ones by elementary transformations @xcite . the first equation of the mct for the molecule dynamics reads @xcite : @xmath135 here , @xmath136 is the characteristic frequency for the translational motion of a tagged particle . @xmath137 is the analogue for the rotational dynamics , where @xmath138 plays here and in the following a similar role as @xmath139 for the translational motion . the frequency @xmath140 denotes the thermal velocity for the rotation . the relaxation kernels @xmath141 and @xmath142 are approximated along the same lines as indicated above for simple systems . they are obtained as a functional of the density correlators of the solvent , multiplied by the tensor - density correlators of the solute @xcite . let us discretize the wave vector to , say , @xmath143 values with equal spacing @xmath144 . let us also restrict the angular - momentum index by some upper cutoff value @xmath145 . one obtains the kernels as mode - coupling polynomials @xmath146\\ \label{vierzehn } & = & \sum_{kpl\nu } v_{\alpha qj\mu}(kpl\nu ) ~ \phi(kl\nu , t)~ \phi_{p}(t)\ , , \qquad \alpha={\rm r , t } .\end{aligned}\ ] ] the positive coefficients @xmath147 , @xmath148 are given in ref . @xcite as specialization of the results in ref . they are expressed in terms of @xmath29 and @xmath149 for @xmath150 . anticipating @xmath29 , @xmath151 and @xmath59 as known , eq . ( [ dreizehn ] ) and ( [ vierzehn ] ) are closed equations for the determination of the @xmath152 correlators @xmath153 . the quantities of main interest for a statistical description of the rotation of the molecule are the reorientational correlators , defined with the legendre polynomials @xmath154 : [ fuenfzehnneu ] @xmath155 they are the long - wave - length limits of the general correlators ; @xmath156 @xcite . one gets from eq . ( [ dreizehn ] ) the fraction representation in analogy to eq . ( [ einsb ] ) : @xmath157 here the kernel @xmath158 is the @xmath159 limit of @xmath160 . carrying out the limit in the general formula for @xmath160 @xcite and discretizing the wave - vector integral afterwards , one finds : @xmath161 = \sum_{j~kl\nu}~ v_{kl\nu}^{j}~ \phi(kl\nu , t)\phi_{k}(t).\ ] ] the positive coupling coefficients @xmath162 are listed in ref after evaluation of @xmath163 and @xmath164 the correlators for the @xmath143 values of @xmath55 , eq . ( [ siebzehn ] ) yields the kernel @xmath165 . fourier - laplace transformation gives @xmath166 and eq . ( [ sechzehn ] ) provides @xmath167 . fourier - cosine transformation of the spectrum @xmath168 leads to @xmath169 . the theory shall be applied for a dumbbell consisting of two equal fused hard spheres of diameter @xmath111 and distance @xmath170 between the centers . thus , besides the packing fraction @xmath4 , there is the elongation parameter @xmath171 as the second control parameter specifying the structure . the structure factor @xmath149 and the corresponding pair correlation functions @xmath172 are evaluated within the percus - yevick theory @xcite . figure [ zweiteabb ] exhibits the probability distribution to find a solvent particle in the plane through the symmetry axis of the dumbbell . the upper panel , calculated for @xmath173 , shows a pronounced quadrupolar pattern extending over several shells . for the small elongation @xmath174 , the lower panel shows that anisotropy is almost lost from the third shell onwards . the calculations of the dynamics will be done for such moment of inertia that @xmath175 . the discretization will be done with @xmath176 wave vectors with spacing @xmath177 . the cutoff for the angular - momentum index is chosen as @xmath178 for @xmath173 and @xmath179 for @xmath174 . the equation of motion ( [ dreizehn ] ) is transformed to an integro - differential equation in analogy to eq . ( [ einsa ] ) and then solved by an algorithm similar to that used for the standard mct problem @xcite . figure [ dritteabb ] demonstrates the transition scenario for the solute correlators for two representative wave vectors @xmath54 and three values of the angular momentum index @xmath1 . the calculated correlators exhibit a very weak dependence on the helicity index @xmath180 , and therefore only the solutions for @xmath181 are shown . the wave vector @xmath182 is close to the structure - factor - peak position , and @xmath183 is near the first minimum of @xmath29 . the correlator for @xmath184 is the probability distribution of the molecule s center - of - mass position , i.e. the analogue of the incoherent - intermediate scattering function for simple liquids : @xmath185 . results for @xmath1=1 and @xmath1=2 deal with the propagation of the dipole- and quadrupole - density fluctuations , respectively . the critical - decay curves , i.e. the solutions for @xmath186 , organize the bifurcation pattern . they deal with the stretched decay towards the plateaus @xmath187 . if @xmath4 increases above @xmath6 , the long - time limits @xmath188 increase above the plateau because the molecule gets more tightly localized in the frozen solvent . the @xmath189-versus-@xmath54 curves are bell - shaped , since the molecules are localized with a nearly gaussian probability distribution @xcite . for @xmath64 , the correlators exhibit a long - time decay from the plateau to zero , and this is the @xmath0-process . the @xmath0-decay time is the larger the smaller the wave vector @xmath54 , while the @xmath0-relaxation stretching increases with increasing @xmath54 . the @xmath54-dependence of the relaxation features are similar as observed and explained previously for the tagged - particle correlator @xmath190 for simple liquids . therefore , the following discussion shall be restricted to the ( @xmath191 ) limit , i.e. to the reorientational correlators @xmath169 . figure [ vierteabb ] exhibits representative decay curves @xmath169 for the liquid state for two separation parameters , and fig . [ fuenfteabb ] exhibits an extended set of susceptibility spectra @xmath192 . the plateaus and critical amplitudes shall be denoted by @xmath193 and @xmath194 , respectively . they have been calculated from the mode - coupling functionals @xcite , and some examples are listed in tab . [ erstetabelle ] . these parameters specify the leading - order asymptotic results for the @xmath2-relaxation process as explained in sec . [ eqnsolv ] for the solvent . the factorization theorem holds in analogy to eq . ( [ drei ] ) @xcite : @xmath195 the @xmath2-correlator @xmath196 is the same function as explained in connection with eqs . ( [ viera])([sechs ] ) for the solvent . this implies for the critical correlator the asymptotic law [ neunzehn ] @xmath197 the nonergodicity parameter of the glass state , @xmath198 , exhibits the @xmath199-singularity @xmath200 and the @xmath0-process initial decay is given by von schweidler s law for @xmath201 and @xmath202 : @xmath203~ , \qquad \tilde{\tau}_\alpha^j=(f_j^c / h_j)^{1/b } ~ t_\sigma^{\prime } ~ .\ ] ] let us introduce two ad hoc time scales for the description of the liquid relaxation outside the transient regime . the center of the @xmath2-relaxation process , @xmath204 , shall be defined as the time , where the correlator has decayed to the plateau : @xmath205 . the center of the @xmath0-process shall be defined as the time , where the correlator has decayed to 50% of the plateau : @xmath206 . some values are listed in tab . [ zweitetabelle ] , and open squares and circles mark these @xmath0- and @xmath2-relaxation times , respectively , in fig . [ vierteabb ] . the slowing down of the dynamics upon approaching the glass - transition point is reflected by the increase of the time scales with decreasing distance parameter @xmath207 . the two step scenario emerges , because the ratio of the scales @xmath208 increases as well . the @xmath0-decay leads to the @xmath0-peaks of the susceptibility spectrum , which are separated from the microscopic excitation peaks by a susceptibility minimum , as is demonstrated in fig . [ fuenfteabb ] . there , the @xmath0-peak - maximum positions , @xmath209 , and the minimum positions , @xmath210 , decrease for @xmath211 . the open squares and circles in fig . [ fuenfteabb ] demonstrate , that @xmath212 and @xmath213 as @xmath214 . the two - step scenario implies that the ratio @xmath215 also decreases upon approaching the glass - transition point . thus , for @xmath211 , the @xmath0-peak gets more and more separated from the rest of the spectrum . in this limit , the plateau height is the relative area under the @xmath216-versus-@xmath217 curve @xcite : @xmath218 figure [ vierteabb ] demonstrates that for @xmath219 the dynamics deals with oscillatory motion , i.e. with rotations and librations which are influenced by steric hindrance affects . if these effects would lead to some fast decay towards the correlator s long - time limit , one would find a white - noise low - frequency fluctuation spectrum : @xmath220 . equivalently , one would obtain a regular low - frequency susceptibility spectrum varying linearly with @xmath40 , @xmath221 , as is indicated schematically by the straight dashed - dotted line in the upper left panel of fig . [ fuenfteabb ] . a linear susceptibility spectrum is obtained for the glass - spectra for @xmath222 , since the correlators approach the limit @xmath223 exponentially for @xmath224 . this is shown by the ( @xmath225 ) spectra in fig . [ fuenfteabb ] . such regular spectra are also found for the low - frequency wings of the @xmath0-peaks , since the liquid correlators approach zero exponentially for @xmath226 . at the bifurcation point , however , the critical decay leads to a power law spectrum which , according to eq . ( [ neunzehna ] ) , reads @xmath227 for @xmath228 , the correlators follow the critical decay decay if @xmath229 is small . therefore the spectra are approaching the asymptotic @xmath230-law for @xmath231 , as is demonstrated for the ( @xmath232 ) results in fig . [ fuenfteabb ] . the stretching of the first relaxation step leads to the strong enhancement of the intensity of the spectral minimum @xmath233 relative to any possible estimation of a white - noise - background spectrum . this enhancement also is exhibited by the experimental data reproduced in fig . [ ersteabb ] . let us consider the probability density @xmath234 for the molecule s axis @xmath235 to have the projection @xmath236 on its initial direction @xmath127 : @xmath237 . since @xmath238 , one gets [ zweiundzwanzig ] @xmath239 thus , knowledge of the set of @xmath169 , @xmath240 , is equivalent to knowing @xmath241 . if the summation over @xmath1 is understood with the cutoff @xmath145 , eq . ( [ zweiundzwanziga ] ) describes the evolution of the distribution with the initial value @xmath242 figure [ sechsteabb ] exhibits results for the small distance parameter @xmath243 corresponding to @xmath244 . the dotted lines exhibit @xmath245 , calculated with @xmath246 for @xmath173 , and @xmath247 for @xmath174 , respectively . within the dynamical window , where the leading - order result for the @xmath2-relaxation , eq . ( [ achtzehn ] ) , applies , one gets [ dreiundzwanzig ] @xmath248 @xmath249 thus the distribution relaxes towards the distribution @xmath250 , which is frozen for @xmath186 . the relaxation does not exhibit any correlation between changes in time described by @xmath84 , and variations with angle described by @xmath251 . this is the scenario expected for relaxation due to dephasing in the random distribution of sizes and shapes of the cages producing steric hindrance for the rotations . for @xmath173 , the @xmath2 regime extends from @xmath252 to about @xmath253 as shown in fig . [ vierteabb ] , and the upper panel of fig . [ sechsteabb ] exhibits the described phenomena for @xmath254 and @xmath255 . the @xmath2-relaxation window is somewhat smaller for @xmath174 , as will be discussed in quantitative detail below in connection with fig . [ siebteabb ] . the dephasing relaxation for this case is demonstrated in fig . [ sechsteabb ] for @xmath254 and @xmath256 . the beginning of the @xmath0-relaxation process follows von schweidler s law , eq . ( [ neunzehnc ] ) . it is identical with the end of the @xmath2-process , and thus it is described within the scenario based on eqs . ( [ dreiundzwanzig ] ) . the most drastic difference between large- and small - elongation relaxation shows up for the @xmath0-process outside the von - schweidler regime . for @xmath173 , the probability decreases monotonically if the angle @xmath257 of the axis increases from its initial value @xmath258 to @xmath259 . this is shown in the upper panel of fig . [ sechsteabb ] for @xmath260 . as time increases , the probability for @xmath261 decreases , while it increases for @xmath262 . thus , the relaxation towards the equilibrium distribution @xmath263 is similar to what one would expect for diffusion on a sphere . for @xmath174 , the correlators for odd @xmath1 decay faster than the corresponding correlators with the even index @xmath264 this is demonstrated in fig . [ vierteabb ] and by the numbers @xmath265 in tab . [ zweitetabelle ] . therefore the @xmath0-process consists of an intermediate time step leading to a probability distribution which is nearly symmetric with respect to the equator @xmath266 . only at later times , the symmetric distribution relaxes to the equilibrium one . figure [ sechsteabb ] shows that , already for the rather short time @xmath254 , @xmath241 exhibits a minimum . for @xmath267 , there is an overshooting effect of the probability for @xmath268 : @xmath269 ; and this effect increases if the time increases to @xmath270 . thus , the relaxation pattern is that expected for a random process of large - angle flips of the molecule s axis . the equations for the nonergodicity parameters @xcite imply that the @xmath223 increase towards unity if the coupling coefficients in eq . ( [ vierzehn ] ) are increased towards infinity . for this strong - coupling limit , one derives from eq . ( [ zweiundzwanziga ] ) that @xmath271 . because of continuity , for strong steric hindrance and for @xmath272 , @xmath241 is a narrowly - peaked distribution centered around @xmath261 . thus , one expects the expansion coefficients @xmath223 for not too large values of @xmath1 to form a smoothly decreasing sequence of @xmath1 : @xmath273 , @xmath274 . table [ erstetabelle ] demonstrates this result quantitatively for @xmath173 and @xmath186 : [ vierundzwanzig ] @xmath275 in particular , the ratio @xmath276 of the relative strengths of the @xmath0-peaks for the dipole relaxation , @xmath277 , and for the quadrupole relaxation , @xmath278 , is larger than unity . one can not conclude quantitatively from @xmath279 the ratio @xmath280 of the @xmath0-peak heights , since the shapes of the spectra depend on @xmath1 . however , fig . [ fuenfteabb ] demonstrates that the two ratios are close to each other . one can also characterize the @xmath0-peak height relative to the microscopic - peak height , @xmath281 , or relative to the minimum intensity , @xmath282 . from fig . [ fuenfteabb ] , one infers @xmath283 , and @xmath284 , i.e. the ( @xmath1=1)-versus-(@xmath1=2 ) enhancement effect appears even more pronounced . according to eq . ( [ neunzehnb ] ) , the nonergodicity parameters increase with increasing @xmath285 . on the other hand , @xmath286 . therefore , @xmath194 must decrease if @xmath193 increases , so that the strongly coupled parameters @xmath223 leave the asymptotic regime for eq . ( [ neunzehnb ] ) for similar magnitudes of @xmath89 . table [ erstetabelle ] quantifies this result for @xmath173 . in particular @xmath287 the reasoning assumes @xmath223 to be large , and thus it can not be applied for too large @xmath1 . there is some @xmath288 , so that @xmath194 decreases with increasing @xmath1 for @xmath289 . within the frequency window , where the leading - order asymptotic law for the critical decay is valid , eq . ( [ einundzwanzig ] ) , one derives an enhancement of the ( @xmath1=2 ) spectrum relative to the ( @xmath1=1 ) spectrum , since : @xmath290 the dotted lines @xmath291 and @xmath292 in the upper panels of fig . [ fuenfteabb ] demonstrate this result . for a strongly near-(@xmath293=1)-peaked probability distribution @xmath241 , one can approximately replace averages of functions of @xmath293 by the functions of the average @xmath294 . thus , lebon et al . concluded : @xmath295 @xcite . specializing to @xmath186 , one quantifies the sequence of @xmath193 in terms of its first value @xmath277 : [ sechsundzwnzig ] @xmath296 substituting into eq . ( [ neunzehnb ] ) and specializing to @xmath297 , one can also quantify the sequence of @xmath194 by the first term @xmath298 : @xmath299 from tab . [ erstetabelle ] one infers , that for @xmath173 the error of eq . ( [ sechsundzwnziga ] ) for @xmath300 is as small as @xmath301 , and eq . ( [ sechsundzwanzigb ] ) is obeyed for @xmath300 within @xmath302 . the strong nonlinear couplings of the structural - relaxation modes require that all correlators enter the first relaxation step , the second relaxation step , and the equilibrium state nearly at the same respective time . this is demonstrated in fig . [ dritteabb ] for @xmath173 . the most striking manifestation of the coupling effect occurs at the center of the @xmath2-relaxation window for @xmath303 . in this case , the factorization theorem , eq . ( [ achtzehn ] ) , is valid . all correlators cross their plateau at the same time , say @xmath304 , where @xmath304 is the zero of the @xmath2 correlator @xmath84 . because of the scaling law , eq . ( 5 ) , one gets the result @xmath305 , i.e. @xmath306 here , @xmath307 is the scale from eq . ( [ sechsa ] ) , and @xmath308 is the zero of the master function : @xmath309 . for the hss it reads @xmath310 @xcite . the open and full circles in fig . [ vierteabb ] show that the asymptotic eq . ( [ siebenundzwanzig ] ) is obeyed very well for @xmath173 . since the @xmath0-processes of @xmath311 and @xmath312 start at the same time @xmath313 and reach zero nearly at the same time , one expects from @xmath314 that the decay time for @xmath315 is larger than that for @xmath316 : @xmath317 furthermore , the @xmath318-versus-@xmath319 plot is somewhat steeper than the corresponding graph for @xmath1=2 . this means that the stretching is larger for the ( @xmath1=2 ) @xmath0 process than for the ( @xmath1=1 ) @xmath0 process . if one interpolates the decay functions by a kohlrausch law , @xmath320 $ ] , the stretching exponent for @xmath1=1 is larger than that for @xmath1=2 : @xmath321 stretching can also be quantified by the width @xmath322 at half height of the @xmath0-peak of the susceptibility spectrum . for @xmath173 our model yields for @xmath323 @xmath324 decades . the kohlrausch processes leading to the same @xmath322 require stretching exponents @xmath325 . the derivation of the inequality for the time scales can be put on a quantitative level by combining eq . ( [ neunzehnc ] ) with the two inequalities in eqs . ( [ vierundzwanzig ] ) . one gets in analogy to eq . ( [ achtundzwanzig ] ) : @xmath326 . the @xmath0-relaxation law for the @xmath169 holds in analogy to eq . ( [ sieben ] ) : @xmath327 . if the shape function @xmath328 would be independent of @xmath1 , the ratio @xmath329 would be equal to the ratio @xmath330^{1/b}$ ] . but the latter is about @xmath331 times larger than @xmath329 . there are two universal phenomena which are relevant for a discussion of the dynamics for weak steric hindrance . the first one concerns the limit @xmath332 of the center - of - mass correlator @xmath333 , which is identical to the tagged - particle - density correlator of the larger of the two spheres forming the dumbbell . if the radius of this sphere , say @xmath334 , is of the same order or larger than the radius @xmath111 of the solvent spheres , the steric hindrance is very effective . in this case , @xmath190 exhibits the canonical bifurcation scenario if @xmath4 crosses @xmath6 , as was discussed comprehensively in ref . this implies that for @xmath335 the ( @xmath184 ) correlators exhibit only a smooth @xmath171-dependence for @xmath171 decreasing to zero . a side remark shall be added to this conclusion . if the ratio of the diameters @xmath336 of a sphere moving in a glass of hard spheres decreases towards zero , there occurs a percolation transition at some critical value @xmath337 . this is a type - a transition , i.e. a bifurcation where the lamb - mssbauer factor decreases continuously to zero for @xmath338 approaching zero from above @xcite . because of continuity , it is obvious that for a dumbbell built of sufficiently small spheres , @xmath339 , there will be a type - a transition if the elongation @xmath171 decreases to some critical value @xmath340 . if @xmath171 crosses @xmath341 , the dynamics changes from one dealing with molecules localized in the hard - sphere glass to one dealing with delocalized molecular motion . this small-@xmath171 phenomenon for small molecules is not considered in this paper . the second universal phenomenon deals with a type - a transition resulting from the fact , that for top - down symmetrical molecules the mct equations of motion of the even-@xmath1 correlators decouple from the odd-@xmath1 ones @xcite . the even-@xmath1 correlators couple to the function @xmath190 , and thus the conventional transition scenario of this correlator enforces the same for all other correlators with even @xmath1 . however , such coupling does not exist for odd @xmath1 . for large @xmath171 , this results in no considerable effect . but all coupling coefficients in the equations of motion approach zero for odd @xmath1 if @xmath171 tends to zero . consequently , for all @xmath342 there is some critical elongation @xmath343 for a type - a transition . for the studied model @xmath344 @xcite . choosing @xmath171 sufficiently close to @xmath343 , it can happen that for odd @xmath1 : @xmath345 or even @xmath346 @xcite . the transition at @xmath343 shall not be studied in this paper . for the demonstration of the small-(@xmath347 ) phenomena , the value @xmath174 has been chosen so large that the canonical sequence for the plateau values in the ( @xmath191 ) limit , eq . ( [ vierundzwanziga ] ) , is not violated , as is quantified in tab . [ erstetabelle ] . but it is chosen so small , that the precursor effects of the type - a transition seriously influence the results for the dumbbell dynamics . thereby , the results are also representative for such cases , where the type - a transition singularity is avoided @xcite due to a weak breaking of the top - down symmetry of the solvent - solute interaction . the even-@xmath1 correlators show the conventional behavior . therefore , the discussion of their trends with decreasing @xmath171 for fixed @xmath1 can be held analogously to that given in sec . iiib for the trends with increasing @xmath1 for fixed large @xmath171 . thus one understands that the ( @xmath1=2 ) @xmath0-process for @xmath174 is weaker , faster , and more stretched than that for @xmath173 , as it is demonstrated in figs . [ vierteabb ] and [ fuenfteabb ] . the halfwidth of the @xmath0-peak for @xmath348 is @xmath349 decades as for a kohlrausch process with exponent @xmath350 . notice in particular from fig . [ vierteabb ] , that the @xmath2-relaxation scale @xmath351 for the @xmath244-result is close to the @xmath171- and @xmath1-independent number @xmath313 from eq . ( [ siebenundzwanzig ] ) . for @xmath352 , the asymptotic formula is obeyed reasonably , but the preasymptotic corrections are larger for @xmath174 than for @xmath173 . the most obvious precursor of the type - a transition is the suppression of the plateau values @xmath223 for odd @xmath1 . this leads to a violation of the rule @xmath353 , as is quantified in tab . [ erstetabelle ] . the general qualitative reasoning from sec . iiib explains , that the suppression of @xmath354 is connected with an enhancement of @xmath298 : @xmath355 . the amplitude @xmath298 is given by the resolvent of the so - called stability matrix , and at a type - a transition the resolvent exhibits a pole @xcite . hence @xmath356 , and the regular trend , eq . ( [ vierundzwanzigb ] ) , is reversed : @xmath357 for our example one infers from tab . [ erstetabelle ] that @xmath358 . according to eq . ( [ fuenfundzwanzig ] ) , the critical spectrum for the dipole relaxation is considerably larger than that for the quadrupole relaxation , as is demonstrated by the dotted lines in the lower two panels of fig . [ fuenfteabb ] . combining eq . ( [ einunddreissig ] ) with von schweidler s law , eq . ( [ neunzehnc ] ) , one concludes that the @xmath311-versus-@xmath319 curve crosses its plateau @xmath277 much steeper than the @xmath312-versus-@xmath319 curve . this is illustrated in the lower panel of fig . [ vierteabb ] . hence the @xmath0-relaxation of the ( @xmath1=1 ) response is faster than the one of the ( @xmath1=2 ) response : @xmath359 again the order for large @xmath171 , eq . ( [ achtundzwanzig ] ) , is reversed . from tab . [ zweitetabelle ] , one infers for @xmath244 : @xmath360 . accordingly , the @xmath0-peak positions for the ( @xmath244 ) spectra for @xmath1=1 and @xmath1=2 in the lower panels of fig . [ fuenfteabb ] differ by about one order of magnitude . for the ratio of the von schweidler scales in eq . ( [ neunzehnc ] ) , one gets @xmath361^{1/b}\to 0 $ ] for @xmath362 , and this identifies the smallness of the ratio @xmath363 as a precursor of the type - a transition . the preceding discussion is valid more generally and explains that all the odd-@xmath1 correlators decay faster than the nearby even-@xmath1 ones . as a result , the probability distribution @xmath241 approaches first an even function of @xmath293 , before the equilibrium value @xmath364 is obtained , as is demonstrated in the lower panel of fig . [ sechsteabb ] . the factorization theorem for the @xmath2-relaxation , eq . ( [ achtzehn ] ) , means that the rescaled correlators @xmath365 are given independently from @xmath1 by the @xmath2-correlator @xmath84 of the solvent . the latter obeys the scaling law , specified by eqs . ( [ viera])-([sechsa ] ) . for fixed rescaled time @xmath366 , the cited formulas deal with the results correctly up to order @xmath367 @xcite . the leading corrections are of order @xmath229 , and they explain the range of validity of the leading results for separations @xmath21 @xcite . figure [ siebteabb ] demonstrates these statements . on a @xmath368-accuracy level the leading - order results describe @xmath369 of the decay of the correlators around the plateau for @xmath173 , @xmath370 . these decay intervals are indicated in fig . [ vierteabb ] by vertical lines . for @xmath371 , the corresponding dynamical window extends from about @xmath252 to about @xmath372 , while it extends from about @xmath373 to about @xmath374 for @xmath375 . this discussion requires a reservation : the corrections to the scaling results can lead to such a violation of eq . ( [ siebenundzwanzig ] ) , which appears as an offset of the plateau @xcite . this offset can be noted in the lower panel of fig . [ siebteabb ] for the odd-@xmath1 results . the good description of the @xmath2-decay of the ( @xmath174 ) results for odd @xmath1 holds only after a correction of the offset . for @xmath376 , the correction effects cause the @xmath377 for @xmath173 to differ from @xmath84 ; one infers from fig . [ siebteabb ] that the @xmath377 increase with increasing @xmath1 . the general results for the theory of the corrections imply , that then also the @xmath377 increase with @xmath1 for @xmath378 @xcite . the @xmath377-versus-@xmath379 curves do not intersect for @xmath313 but they touch each other as is demonstrated in the upper panel of fig . [ siebteabb ] . corresponding results also hold for @xmath174 after the mentioned offset is eliminated . equations ( [ fuenf ] ) and ( [ achtzehn ] ) lead to the scaling law for the susceptibility spectra : @xmath380 . the master spectra @xmath381 are given by the fourier - cosine transform @xmath382 of the master functions @xmath95 . the master spectrum for the glass state describes the crossover from a regular spectrum for small rescaled frequencies , @xmath383 , to the critical spectrum at large rescaled frequencies , @xmath384 . it deals with the knee exhibited by the spectra for @xmath225 and @xmath385 in fig . [ fuenfteabb ] . the master spectrum for the liquid describes the crossover from the von - schweidler high - frequency tail of the @xmath0-peak , @xmath386 , to the critical decay for large rescaled frequencies , @xmath387 . the results describe in the small-@xmath89 limit the @xmath2-relaxation minimum as it can be seen in fig . [ fuenfteabb ] for the @xmath244 and @xmath232 results . in particular , the factorization theorem explains why the spectral minima @xmath388 are located at the same position independently of @xmath1 and @xmath171 . the leading - order formulas imply @xmath389 , where @xmath390 denotes the minimum of the master spectrum @xmath391 . for the hard - sphere system , one gets @xmath392 @xcite . obviously , the @xmath2-relaxation scaling laws can describe the susceptibility minimum only for such small distance parameters , for which @xmath388 is located in that frequency window where the ( @xmath393 ) spectrum exhibits the asymptotic @xmath394-law , eq . ( [ einundzwanzig ] ) . figure [ fuenfteabb ] shows that for the model under study this window is restricted to @xmath395 . this means , that @xmath388 has to be located about three decades below the peak of the microscopic susceptibility spectrum . for @xmath396 , the critical spectrum is modified by crossover effects to the transient dynamics . the susceptibility minimum with @xmath397 is due to the crossover of the @xmath0-peak tail to the microscopic excitation spectrum ; it can not be discussed by the universal asymptotic laws for the mct bifurcation . one concludes from fig . [ fuenfteabb ] , that @xmath398 needs to be satisfied in order to apply the @xmath2-scaling laws for the model under study . the @xmath0-relaxation scaling law reads for the reorientational correlators in analogy to eq . ( [ sieben ] ) : @xmath399 the @xmath21-independent master function @xmath400 obeys as initial condition the von schweidler law : @xmath401 . the superposition principle for the susceptibility spectra reads @xmath402 with @xmath403 denoting the rescaled frequency . the @xmath21-independent master spectrum is given by the fourier - cosine transform of the master correlators @xmath404 . consequently , the above defined @xmath0-relaxation time scales @xmath405 and susceptibility - maximum positions @xmath209 read @xmath406 where @xmath407 is defined by @xmath408 and @xmath409 denotes the peak frequency of @xmath410 . the scaling law implies that a representation of @xmath169 as a function of the rescaled time @xmath411 should superimpose correlators for different distance parameters @xmath21 on the common curve @xmath412 . asymptotic validity means that the @xmath413 interval , where the scaling law is obeyed , expands to arbitrary size for @xmath103 . a corresponding statement holds for the representation of the susceptibility peaks as functions of the rescaled frequency . the corrections to the leading - order asymptotic laws are the larger , the larger the critical amplitude @xmath194 is @xcite . figures [ fuenfteabb ] and [ achteabb ] demonstrate , that the described scenario for the evolution of the @xmath0-process is valid for @xmath173 , and also for @xmath174 in the case @xmath1=2 . for strong steric hindrance , the @xmath0-scaling law works for larger values of @xmath414 , than the @xmath2-scaling law . this is so , because the leading corrections to the @xmath0-scaling law are of relative size @xmath415 , while they are of relative size @xmath416 for the @xmath2-scaling law @xcite . figure [ achteabb ] demonstrates a drastic ( @xmath1=1)-versus-(@xmath1=2 ) effect of the @xmath0-scaling for @xmath174 . the dipole correlators do not exhibit the superposition principle for @xmath417 , nor do the correlators for the other odd values of @xmath1 . for @xmath418 the plateau emerges only for extremely small values of the distance parameter @xmath419 . the @xmath0-peak heights of the dipole spectra decrease with decreasing @xmath207 in fig . [ fuenfteabb ] in contradiction to the scaling - law prediction . this anomaly is caused by the large size of the critical amplitude @xmath298 , which was explained in connection with eq . ( [ einunddreissig ] ) . more precisely , it is caused by the large percentage of the decay of @xmath311 described by the @xmath2-scaling law as is indicated by the vertical lines in fig . [ vierteabb ] . to formulate this observation quantitatively , let us remember that the decay of the correlator near the plateau is described by eqs . ( [ fuenf ] ) , ( [ sechs ] ) and ( [ achtzehn ] ) : @xmath420 . the master function @xmath421 for small positive values and all negative ones is well approximated by : @xmath422 . here , @xmath423 is determined by the exponent parameter @xmath87 and for our solvent model reads @xmath424 @xcite . thus one gets for @xmath425 within the window for the validity of the @xmath2-relaxation scaling law : [ vierunddreissig ] @xmath426 \}~ .\ ] ] the leading corrections to this formula can explain the possible offset of @xmath193 or , equivalently , of the scales @xmath307 @xcite , which was noticed above in connection with fig . [ siebteabb ] for @xmath174 . equation ( [ vierunddreissiga ] ) can be rewritten as @xmath427 + h_j |\sigma| b_1/\tilde{t}^b$ ] . here , the bracket is the @xmath0-scaling - law description of the initial part of the @xmath0-process , and the term proportional to @xmath423 is the leading correction . the correction term to the @xmath0-scaling law deals with that part of the @xmath2-process below the plateau , which is not given by the von schweidler s large-@xmath428 asymptote . therefore , one can write the for the @xmath0-process for not too large values of rescaled time @xmath429 : @xmath430 the analogue of this formula was shown in ref . @xcite to describe the evolution of the @xmath0-process of the density correlators of the hss perfectly for @xmath431 . it was also shown , that the corresponding spectrum describes the susceptibility peak to increase above the scaling - law constant @xmath432 if the separation @xmath207 increases from @xmath433 to @xmath434 . in fig . [ neunteabb ] , the evolution of the ( @xmath1=1 ) @xmath0-process for small steric hindrance is reexamined . instead of rescaling the time with the ad - hoc scale @xmath435 , the theoretically motivated scale @xmath436 is chosen . one recognizes , that the found scenario does not exhibit any qualitative peculiarity anymore , compared to what is presented in fig . [ achteabb ] for @xmath174 and @xmath1=2 . the ( @xmath1=1)-versus-(@xmath1=2 ) anomaly is identified as an anomaly of the size of the corrections only . in the case of the small elongation , the distance parameter @xmath207 has to be taken almost two orders of magnitude smaller for @xmath1=1 in order to render the corrections to the @xmath0 scaling as small as found for @xmath1=2 . for @xmath437 , even including the leading corrections to von schweidler s law , one can explain the relaxation from the plateau only up to some offset in the time scale . this is demonstrated in fig . [ neunteabb ] by the dotted lines for @xmath438 . two remarks concerning tests of the second scaling law shall be added . the definition of the time scale @xmath265 used in fig . [ achteabb ] was arbitrary . let us consider more general definitions to be parameterized by a number @xmath439 and denoted as @xmath440 . the subscripts @xmath0 and @xmath1 shall be dropped for the sake of simplicity , and the definition shall be : @xmath441 . if the scaling law is valid , one finds in analogy to eq . ( [ dreiunddreissig ] ) : @xmath442 . here , @xmath443 is defined by @xmath444 . in this case , the choice of @xmath55 is irrelevant , since the ratio of two different scales is @xmath21-independent : @xmath445 . however , if preasymptotic corrections are present , the scales are not equivalent . the range of validity of the superposition principle expands from large to small rescaled times . this follows from eq . ( [ vierunddreissigc ] ) and is demonstrated in fig . [ neunteabb ] . one gets for @xmath446 : @xmath447 for a detection of the superposition principle for an as large as possible value of @xmath207 , one should therefore choose an as large as possible value of @xmath55 for the rescaling procedure . thereby , the artificial crossing point of the rescaled curves at @xmath448 is suppressed as much as possible . otherwise , one introduces a time scale @xmath440 for the characterization of a decay process which can not be characterized by a single scale . the outcome of this ill - defined procedure is demonstrated in the lower left panel in fig . [ achteabb ] . in this case , @xmath449 is a parameter extracted from the correlator which , according to fig . [ siebteabb ] , is adequately to be specified by the two scales @xmath450 and @xmath307 of the @xmath2-relaxation scaling law . the dashed line in the inset of fig . [ neunteabb ] demonstrates explicitly , that the scale @xmath435 does not exhibit the asymptotic behavior for @xmath451 . the asymptotic law @xmath452 is followed only for @xmath453 . the second remark concerns the determination of the exponent @xmath454 entering the power - law behavior for the @xmath0-relaxation time scale , as specified by eqs . ( [ sechsb ] ) and ( [ dreiunddreissig ] ) . these results are based on the validity of the scaling law @xcite . therefore , one can not appeal to mct if one fits power laws for scaling times for cases where the scaling law is violated . the dashed line in the inset in fig . [ neunteabb ] demonstrates , that the scale @xmath435 for @xmath455 can be fitted well by a power law for a two - decade variation of the distance parameter @xmath207 . the identified effective exponent @xmath456 describes the variation of @xmath457 over three orders of magnitude ; but nevertheless @xmath458 has no well defined meaning for the discussion of our model . solving the mct equations of motion for the dynamics of a hard - sphere dumbbell moving in a hard - sphere liquid , first - principle results have been obtained for the evolution of the glassy dynamics of the reorientational degrees of freedom of a molecule . it was found that one has to distinguish between two scenarios , namely between strong steric hindrance as found for large elongations @xmath171 of the dumbbell , and weak steric hindrance as found for small elongations . for strong steric hindrance , the mode - coupling coefficients for the reorientational degrees of freedom in eq . ( [ vierzehn ] ) are of the same order as the ones entering eq . ( [ zwei ] ) for the description of the translational degrees of freedom of the solvent . the dependence of the various parameters on the angular - momentum index @xmath1 is similar to the dependence on the wave vector @xmath54 . one has to view @xmath138 as the analogue of @xmath139 . while the @xmath54-dependence reflects the decomposition of the direct solute - solvent correlations in plane waves the @xmath1-dependence reflects the decomposition in spherical harmonics . hence one finds , that all reported results on the @xmath1-dependence of the reorientational correlators @xmath169 are similar and can be explained in a similar manner as known from the previous work on the tagged - particle - density correlators @xmath190 in simple liquids @xcite . in particular , it was shown that with increasing @xmath1 the @xmath0-peak - strength parameters @xmath193 , eq . ( [ vierundzwanziga ] ) , the @xmath0-relaxation - time scales @xmath405 , eq . ( [ achtundzwanzig ] ) , and the stretching exponents @xmath459 , eq . ( [ neunundzwanzig ] ) , decrease . these findings reproduce qualitatively the three general differences between dielectric - loss and depolarized - light - scattering spectra , which were discussed in sec . i in connection with fig . [ ersteabb ] . because of eq . ( [ einsa ] ) , the relaxation of the correlator @xmath460 follows that of the kernel @xmath461 . therefore , the @xmath0-relaxation time scale of the ( @xmath159 ) density fluctuations , say @xmath462 , is larger than the corresponding scale of the longitudinal elastic modulus @xmath463 , say @xmath464 . for strong steric hindrance , the decay of the cage is the prerequisite for the reorientation of the molecule , and therefore @xmath465 . thus one expects the fourth general feature of the @xmath0 relaxation listed in the introduction : @xmath466 . for our model one gets for @xmath375 : @xmath467 , @xmath468 , @xmath469 . the ratio @xmath470 is of the same order as cited in sec . i for pc and salol . the mode - coupling coefficients in eq . ( [ vierzehn ] ) decrease to zero if @xmath1 tends to infinity . thus the solutions for large @xmath1 are sums of many small terms , which are not strongly correlated . each term exhibits the short - time von schweidler - law behavior for the @xmath0-relaxation : @xmath471 . therefore , one expects for @xmath169 the characteristic function of the stable lvy distribution , @xmath472 $ ] @xcite . for the density correlators of the solvent , fuchs has worked out the limit behavior for the @xmath0-relaxation master function for @xmath473 and showed how the kohlrausch law with @xmath474 arises @xcite . we suspect that a similar derivation can be done for the reorientational correlators . therefore , we conjecture that the sequence of kohlrausch exponents @xmath475 converges towards the von schweidler exponent @xmath80 . molecular - dynamics - simulation data for a model of water have been interpreted consistently within the standard mct scenario @xcite . in particular , the @xmath169 exhibit the conventional behavior @xcite . one concludes that water exhibits strong - steric - hindrance effects . therefore it is reinsuring that the sequence of the first five kohlrausch exponents @xmath459 decreases with increasing @xmath1 monotonously towards the von schweidler exponent @xcite . a further general result , namely the increase in the initial part of the series of critical amplitudes , eq . ( [ vierundzwanzigb ] ) , is also found in the simulation data for @xmath476 @xcite . figure [ ersteabb ] exhibits as full lines the ( @xmath1=1 ) and ( @xmath1=2 ) spectra calculated for @xmath173 . the lines for @xmath477 and @xmath352 are the ones discussed in fig . [ fuenfteabb ] , and the other two refer to @xmath478 and @xmath479 , respectively . in order to transfer the mct results , which are calculated with ad hoc units specified in sec . ii , to the units used by the experimentalists , one has to introduce three scales . the first and second scale transfer the calculated dimensionless normalized spectra @xmath480 for @xmath1=1 and @xmath1=2 to the units used by the experimentalists for their dielectric - loss and depolarized - light - scattering spectra , respectively . the third scale shifts our frequency scale to the ghz scale . in the double - logarithmic representation , the first two scales define an overall vertical shift of the diagrams in fig . [ fuenfteabb ] , while the third scale defines a horizontal shift of the figures . intending to compare data for pc for different temperatures with the mct results for different packing fractions @xmath4 , one gets a mapping of the @xmath3-scale onto the @xmath4-scale via eq . ( [ elf ] ) . the result is shown as an inset in fig . [ ersteabb ] . the inset also includes the point with coordinates of the critical packing fraction of the solvent @xmath6 and the critical temperature @xmath481 . this value for @xmath5 was determined for pc by analyzing neutron scattering data @xcite , and has recently been corroborated in an mct analysis of various other pc experiments @xcite . our results in fig . [ ersteabb ] describe the evolution of the two types of pc spectra semi - quantitatively . in particular , the extrapolation of the @xmath3-@xmath4-relation yields a reasonable estimation of the critical temperature for that system , which is demonstrated through the dashed line in the inset . there is no obvious reason why the studied dilute solution of hard symmetric dumbbells in a hard - sphere solvent should produce spectra , which are similar to the data for pc . we consider the found similarities to a large extend as accidental . the theoretical curves are added in fig . [ ersteabb ] with the mere intention to justify the conclusion : the model studied in this paper and our choice of parameters are relevant for achieving an understanding of experiments in glass - forming molecular liquids . in order to further corroborate the preceding conclusion , let us consider fig . [ zehnteabb ] . the data points exhibit a susceptibility spectrum of pc measured by incoherent - neutron - scattering spectroscopy for the wave vector @xmath482 @xcite . a remarkable feature of the @xmath0-peak spectrum compared to the spectra shown in fig . [ ersteabb ] is that it is less pronounced relative to the spectrum of the microscopic excitation band and that is is more stretched . the two dashed lines in fig . [ zehnteabb ] exhibit the spectra for the center - of - mass correlator @xmath483 for @xmath484 in order to emphasize , that this leading approximation for the scattering function can not easily explain the experimental findings . the shown @xmath0-peaks of @xmath485 have a half width of @xmath486 decades as produced by a kohlrausch process with exponent @xmath487 . the scattering function @xmath488 is a sum over the contributions of the molecule s constituents and hence it is a superposition of the density correlators for all angular momentum indices @xmath1 . for the symmetric dumbbell one gets up to some normalization constant @xcite @xmath489 where @xmath490 are the spherical bessel functions . the full lines in fig . 10 exhibit the spectra for @xmath488 for @xmath484 . the @xmath0-peaks have a halfwidth of @xmath491 decades as produced by a kohlrausch law with stretching exponent @xmath492 . the frequency was rescaled as explained in connection with fig . [ ersteabb ] and the scale for the spectra was adjusted to meet the one of the data . comparison of the full line with the dashed one for @xmath352 shows the features distinguishing the @xmath0-processes of @xmath493 from that of @xmath494 . the result calculated for @xmath477 shows that the finding for our model semiquantitatively accounts for the @xmath0-peak data . some side remarks considering the comparisons in figs . [ ersteabb ] and [ zehnteabb ] might be useful . a schematic - model analysis of the pc data gave the exponent parameter @xmath495 @xcite , in good agreement with the values found from analyses of the susceptibility minima with the @xmath2-relaxation scaling laws @xcite . the value is close to the result @xmath496 for the hard - sphere system , eq . ( [ neun ] ) . this accident ensures that the master function for the susceptibility minimum and the values of all anomalous exponents of pc agree within the experimental uncertainties with the corresponding quantities of the model studied in this paper ; and this is a prerequisite of a successful fit . accidently , the ratio of the @xmath0-relaxation times @xmath497 noted in tab . [ zweitetabelle ] for @xmath498 is only a bit larger than the ratio of the @xmath0-peak - maximum positions of pc , @xmath499 , exhibited in fig . 1 ; and this is another request for a reasonable fit . since the ratio decreases with decreasing @xmath171 , some @xmath500 could be chosen to reproduce the specified ( @xmath1=1)-versus-(@xmath1=2 ) effect quantitatively . nevertheless , it is remarkable that the fit in fig . [ ersteabb ] reproduces the ratios of @xmath0-peak maximum intensity to @xmath2-minimum intensity @xmath501 reasonably well for both values of @xmath1 . neither it is trivial , that the model reproduces reasonably the ( @xmath1=1)-versus-(@xmath1=2 ) effect for the stretching . a new liquid - glass - transition scenario is predicted which is referred to as the regime of weak steric hindrance for reorientational motion . it is characterized by ( @xmath1=1)-versus-(@xmath1=2 ) effects , more generally by odd-@xmath1-versus - even-@xmath1 effects , which are qualitatively different from the results described above as strong - steric - hindrance results . the new scenario occurs , if precursor phenomena of a type - a - transition between two nonergodic states strongly influence the asymptotic results for the conventional mct bifurcation . the scenario appears if the particle interactions deviate not too strongly from spherical symmetry , e.g. , if a linear molecule exhibits only small deviations from a top - down symmetry and if there are not too large elongations . six features characterize the weak - steric - hindrance scenario . first ( i ) , the plateaus @xmath193 for the reorientational correlators for odd @xmath1 are suppressed in comparison to what one would expect by interpolating or extrapolating the values for nearby even-@xmath1 plateaus ( tab . [ erstetabelle ] ) . most importantly ( ii ) , the critical amplitude @xmath298 is larger than @xmath502 , eq . ( [ einunddreissig ] ) , so that the canonical ordering of the @xmath194 for small @xmath1 , eq . ( [ vierundzwanzigb ] ) , is reversed . third ( iii ) , the percentage of the decay of the reorientational correlators @xmath169 , which can be explained by the leading- plus next - to - leading - order asymptotic formulas for the @xmath2-relaxation is larger for @xmath1=1 than for @xmath1=2 ; as is indicated by the vertical lines in fig . the structural relaxation of the reorientations is dominated by large - angle flips ( iv ) , as shown in fig . [ sechsteabb ] for the dumbbell with elongation @xmath174 . the @xmath0-relaxation - time scale for @xmath1=1 is smaller than for @xmath1=2 , eq . ( [ zweiunddreissig ] ) , ( v ) so that the canonical order of the @xmath0-relaxation scales , eq . ( [ achtundzwanzig ] ) , is reversed . this can cause the @xmath169-versus-@xmath319 graphs for @xmath1=1 to cross the graphs for @xmath1=2 , as is shown for the @xmath174 results in figs . [ dritteabb ] and [ vierteabb ] . finally ( vi ) , for distance parameters @xmath455 , where the conventional @xmath169-versus-@xmath503 plot exhibits the @xmath0-relaxation scaling law for @xmath1=2 , the correlators for @xmath1=1 do not show the validity of the superposition principle , as is demonstrated in fig . [ achteabb ] . nor does the scale @xmath435 , defined as the time for a 50% decay of the @xmath0-relaxation correlator , exhibit the power law behavior with the correct exponent @xmath454 as is shown in the inset of fig . [ neunteabb ] . a side remark concerning a molecular - dynamics study of the evolution of glassy dynamics in a lennard - jones - dumbbell liquid by kmmerer et al . @xcite might be in order . it was reported that the correlators dealing with translational degrees of freedom and also for the ones for the reorientational dynamics for angular index @xmath504 could be interpreted qualitatively within the universal asymptotic mct formulas . however , the evolution of the dipole correlators did not fit into the standard mct pattern . it was found that @xmath505 and @xmath506 . a drastic violation of the @xmath0-scaling law was noted quite similar to what is exhibited in the lower left panel of fig . [ achteabb ] . the scale @xmath435 exhibited a deviation from the asymptotic law @xmath507 , but a fit by @xmath508 with @xmath458 as discussed in the inset of fig . [ neunteabb ] was possible . these simulation results for @xmath1=1 differ from those for water simulations @xcite as well as from the experimental findings for propylene carbonate quoted in fig . [ ersteabb ] . however , they agree with the features ( ii ) , ( v ) , and ( vi ) specified in the preceding paragraph . moreover , the property ( iv ) concerning the large angle flips is also obtained in ref . accidently , the @xmath352 results in the lower panel of fig . [ fuenfteabb ] show that the minimum position of the @xmath1=1 spectrum exceeds that of the @xmath1=2 spectrum by nearly one order of magnitude , in agreement with the corresponding finding in ref . @xcite . furthermore , the @xmath0-peak variation with @xmath121 shown in the lower left panel of fig . [ fuenfteabb ] is in qualitative agreement with that reported in ref . @xcite . in view of these observations it does not seem impossible , that the scenario studied in refs . @xcite fits into the framework of the ideal mct . however , it is not clear , whether or not the results of ref . @xcite can be explained by our theory for type - a precursors of a dilute solution of molecules . first , the simulation results for the dumbbell liquid do not exhibit a particular decrease of @xmath277 relative to @xmath278 . second , the @xmath2-relaxation scaling has not been documented for the dumbbell liquid and so it is unclear , whether or not the feature ( iii ) holds for that case . summarizing , it shall be emphasized that all qualitative features for the evolution of the structural relaxation studied in this paper have been explained by means of the formulas for the leading - asymptotic expansions and their leading - order - correction formulas for the bifurcation scenario . in this sense , these asymptotic formulas can be considered as the essence of mct . however , in order to explain the characteristic ( @xmath1=1)-versus-(@xmath1=2 ) differences for the relaxation patterns , it is necessary to also understand the general trends of the nonuniversal parameters with variations of wave - vector @xmath54 and angular - momentum index @xmath1 . and this requires to use mct as a microscopic theory based on the knowledge of the equilibrium structure . we thank the authors of refs . @xcite for the permission to use their data files and m. fuchs for many stimulating discussions . we kindly thank f. sciortino for the permission to cite his unpublished simulation results for the reorientational dynamics of water . we acknowledge gratefully helpful critique and suggestions for improvements of the manuscript written to us by h.z . cummins , r. pick , and r. schilling . this work was supported by verbundprojekt bmbf 03-g05tum .

the central engines of powerful type i active galactic nuclei ( agns ) contain gas that is moving at extreme velocities , apparent from their broad emission lines ( bels ) and , in the case of broad absorption line ( bal ) quasars , in blueshifted absorption troughs . these spectral features arise from compact gas flows close to the accretion disk , apparent from the presence of high ionization lines and size scales from reverberation mapping @xcite . these compact flows are probably the origin of larger - scale outflows ( e.g. chelouche & netzer 2005 ; crenshaw & kraemer 2007 ; storchi - bergmann 2010 ) and result in feedback that is believed to be important on many scales , regulating supermassive black hole ( smbh ) fueling , quenching star formation , and so driving galaxy - smbh coevolution ( e.g. silk and rees 1998 ; fabian 1999 ; sommerville et al . 2008 ; zubovas & king 2012 ) as well as influencing the intergalactic and intracluster medium and affecting galaxy and large - scale structure formation ( e.g. scannapieco & oh 2004 ; rafferty , mcnamara & nulsen 2006 ; sommerville et al . 2008 ; fabjan et al . 2010 ; wang 2015 ) . yet despite the importance of these processes , the connection between compact and large - scale outflows is not understood . in the case of the broad emission line region ( belr ) , even the basic kinematics are unclear ; independent reverberation mapping results support outflow , inflow , and orbital motion @xcite , and microlensing results have been consistent with both virial belrs @xcite and with more complex , multi - component flows @xcite . bal quasars offer perhaps the clearest evidence of nuclear outflows in agns . approximately 15% of optically selected quasars exhibit high velocity , blueshifted absorption features @xcite . the outflows responsible for these features may be common to most quasars if this 15% represents a special line of sight to outflows with a similar covering factor @xcite , or a particular phase in the lifecycle of quasars . an accurate measurement of the absorption profile of a bal quasar would go a long way to revealing the kinematics of these outflows , however this measurement is complicated by the fact that their intrinsic absorption profile can not be recovered without knowing the shape of the underlying bel that it absorbs . these bel profiles can vary dramatically between quasars , and in the case of bal quasars may be strongly affected by a biased viewing angle . the gravitationally lensed bal quasar h1413 + 117 ( the cloverleaf ; magain et al . 1988 ) offers a unique opportunity to simultaneously study both bel and bal flows in the same source . the four lensed images of this @xmath1 quasar exhibit deep bal features in high - ionization lines , as well as low - ionization absorption in aliii , making it a lobal quasar . significant microlensing has been observed in h1413 + 117 over the past 15 years @xcite . in microlensing , substructure within the foreground lensing galaxy causes changes in the magnification of individual lensed images . the degree of microlensing depends on size of lensed region , and in the case of h1413 + 117 this has resulted enhanced magnification in the continuum in lensed image d compared to the broad emission lines in this component . given a model of the microlensing magnification of the lens system , these differential magnifications can be used to derive size constraints of emission regions . this approach has been used to contrain accretion disk size and temperature gradient in several lensed quasars ( e.g. kochanek et al . 2004 ; bate et al . 2008 ; floyd , bate & webster 2009 ; blackburne et al . 2011 ; jimnez - vicente et al . 2014 ; blackburne et al . 2015 ) , including in h1413 + 117 by @xcite ( hereafter h10 ) . it has also provided constraints on belr size @xcite and kinematics @xcite in a number of quasars , however no such constraints have been derived for the belr of h1413 + 117 . another novel use of microlensing takes advantage of differential magnifications between lensed images to algebraically decompose the spectral contributions of different physical regions within the quasar @xcite . @xcite ( hereafter a90 ) and h10 apply this technique to h1413 + 117 , shedding light on the true shape of the broad absorption profile , and in the case of h10 decomposing into microlensed and non - microlensed components which reveal substructure in each component and may point to a two - component outflow comprising a polar outflow and equatorial disk - wind . in this paper we present new gemini north gmos integral field spectroscopy of h1413 + 117 , and based on the strong microlensing observed in lensed image d we derive new statistical size constraints on the belr . we also present a new spectral decomposition , deriving belr and continuum components , as well as the intrinsic bal profiles for nv , siiv , civ , and aliii . in section [ obs ] we describe the observations and extraction of lensed image spectra . in section [ sigs ] we analyse the microlensing signatures in this new epoch of data . in section [ microlensingbehavior ] we use lensing simulations to explore the size dependence of the observed microlensing signatures and deriving size constraints for the belr . in section [ algebra ] we decompose the spectra in terms of the broad emission line region and continuum , emphasizing the physical interpretation of the decomposed regions , and we derive the intrinsic bal absorption profile . we discuss the interpretation of these decomposed components in the context of disk - wind outflow models in section [ disc ] . in section [ conc ] we summarize our conclusions . 117.,width=302 ] h1413@xmath2117 was observed on the night of april 5th , 2006 with the gmos integral field unit ( ifu ) @xcite on the gemini north telescope ( program i d : gn-2006a - q-15 ) . 3@xmath320 minute exposures were taken in one - slit mode using the b600 grating with a spectral range of 3961 to 6845 and a resolution of @xmath4 . at the redshift of h1413@xmath2117 , @xmath1 , this covers the ly@xmath5 through ciii ] broad emission lines . seeing conditions were good , with a psf fwhm of 08 . the data was reduced using the gemini iraf package v1.10 under iraf 2.14.1 . we used calibration files taken as part of the standard gemini facility calibration unit ( gcal ) suite . mostly , the standard gmos ifu reduction procedure was followed for overscan subtraction and trimming , bias subtraction , flat fielding and response correction , arc calibration , fiber extraction , sky subtraction , flux calibration , cube reconstruction with atmospheric dispersion correction . additionally , we corrected the gcal flat and the twilight flat for scattered light by fitting a surface to residual light detected between the fiber bundles . that correction was not necessary for the science observations , the signal being too weak to generate any significant scattered light . the cosmic rays were corrected before extraction with la - cosmic @xcite , followed , after extraction , by a floor and ceiling masking of remaining extreme hits . the calibrated cubes for each science observation were aligned and combined with a customized version of the pyfu software package . figure [ 2dimage ] shows the reduced data cube collapsed into a 2-d image . all lensed components are distinct and well - resolved . the lensed images are well resolved in these observations , allowing for a good extraction of their spectra . we perform this extraction using the method described in @xcite . we refer the reader to that paper for a detailed description , however in short : relative lensed image fluxes at each wavelength bin are determined by least - squares fitting of reconstructed point spread functions ( psfs ) . a separate psf is determined for each 100 - wide spectral bin . each given psf is initially estimated using carefully - masked fragments of the ifu frame ( stacked over 100 ) , and are iteratively improved with a series of least - squares minimizations of each frame . figure [ spec ] shows the extracted spectra of the four quasar images . we see distinct broad emission lines for ly@xmath5 , nv , oi , cii , siiv , civ , and ciii ] as well as prominent iron emisison features at @xmath01600 - 1800 . emission from heii and aliii is also apparent , blended with iron and ciii ] respectively . clear redshifted broad absorption troughs are seen for ly@xmath5/nv , siiv , civ , and aliii . gravitational lensing is achromatic , and so in the absence of microlensing , differential extinction , or time - dependent spectral variation , the spectra of lensed images of a multiply - imaged quasar are identical modulo a scaling factor . microlensing is typified by a difference in the magnification of one or more lensed images compared to the time - averaged magnification for that lensed image ( which we will refer to as the macrolens magnification , or just macro - magnification ) . _ differential microlensing _ is characterized by magnifications that vary between different emission regions , resulting in a change in the shape of the spectrum of the affected lensed image . differential microlensing is sensitive to emission region size and projected location , is frequently seen between bels and continua , and also often shows wavelength dependence within these features . spectral variability coupled with different light travel times can result in spectral differences between lensed images . in the case of h1413 + 117 the time delay is potentially significant , estimated , for example at 23.4@xmath64 days between images a and d @xcite . this variability can potentially lead to differences in four key observables between lensed images : bal and bel profile shapes , continuum - to - bel flux ratio , and the spectral index of the continuum . while it is difficult to extricate the differences due to intrinsic variability from those due to microlensing , we can comment on the effect of this on our analysis . variation in broad absorption and emission line profile shapes would be very difficult to discriminate from microlensing . fortunately the effect is probably minimal . as we show in section [ fluxratios ] the bel profiles are identical between lensed images , precluding both differential microlensing and intrinsic variability in the belr spectrum . it is not so easy to compare bal profiles , however we note that , while this profile has been observed to vary , this variation occurs over months and years , not days ( h10 ) . as we will see in section [ microlensingbehavior ] we estimate the size of the belr ( and hence also the bal wind ) at @xmath7100 light days , and so their intrinsic variation timescales should be significantly longer than the time delay . variation in the continuum - to - bel ratio between lensed images can not be ruled out , and we do see significant variation in this ratio , which we attribute to microlensing because it is consistent with the microlensing behavior of the lens in recent years . in section [ algebra ] we use this difference in continuum - to - bel ratios to decompose the spectrum into intrinsic components . it s important to note that for the purpose of the decomposition of spectral components , it does not matter why this ratio varies between images whether by microlensing or intrinsic variability the decomposition is still valid as long as the bels do nt show wavelength - dependent variations . variation in the slope of the continuum could also be due to intrinsic variability , and we do indeed observe a difference in slope ( sect . [ fluxratios ] ) , however we do nt investigate this in detail in this paper . differential extinction between lensed images will not affect the continuum - to - bel ratios as our measurement of the continuum strength is made at the relevant bel wavelength range . nor will it significantly affect broad line profiles , which appear identical between lensed images anyway . differential extinction only affects the continuum slope and ratios between bels , which are not a focus of his work . to inspect differential microlensing properties of h1413 + 117 we first obtain an estimate of the unabsorbed continuum spectrum for each lensed image by performing a linear fit to regions of the spectrum least affected by emission or absorption ( fig . [ spec ] shows fits for images a & d ) . prominent iron emission features force us to choose small windows at rest frame 14201460 , 16851692 , and 17951805 . while the continuum is unlikely to be linear over the full range of this spectrum , we find that the fit gives a very plausible estimate , particularly redward of the nv line . subtraction of the fitted continua allow us to measure the average flux ratios across each bel for each lensed image with respect to image a. these ratios and the average flux ratios for the continuum at each line center are given in table [ fluxratiotable ] . for the purpose of inspection of the microlensing properties , we find that changing the continuum level and slope by allowing variation of @xmath65% ( the width of the pale gray region surrounding fits in fig . [ spec ] ) does not significantly affect results . lcccccccc & & & & & + & line & continuum & & line & continuum & & line & continuum + nv & 0.78 @xmath6 0.02 & 0.92 @xmath6 0.02 & & 0.93 @xmath6 0.03 & 0.81 @xmath6 0.02 & & 0.45 @xmath6 0.03 & 0.77 @xmath6 0.03 + siiv line & 0.82 @xmath6 0.02 & 0.91 @xmath6 0.01 & & 0.93 @xmath6 0.02 & 0.79 @xmath6 0.01 & & 0.38 @xmath6 0.02 & 0.75 @xmath6 0.02 + civ & 0.81 @xmath6 0.01 & 0.91 @xmath6 0.01 & & 0.85 @xmath6 0.01 & 0.77 @xmath6 0.01 & & 0.41 @xmath6 0.01 & 0.73 @xmath6 0.01 + ciii ] & 0.84 @xmath6 0.02 & 0.88 @xmath6 0.02 & & 0.87 @xmath6 0.03 & 0.71 @xmath6 0.02 & & 0.38 @xmath6 0.03 & 0.68 @xmath6 0.02 + fe complex & & & & & + mean bel & & & & & + mean cont . & & & & & + figure [ linesratios ] ( upper panels ) compares the spectra of all lensed components around the broad emission lines . to highlight the effect of differential microlensing we scale the spectra of lensed images b , c , & d so that the average continuum emission at each line ( from the continuum fits ) matches that of image a. we see significant differential microlensing between broad emission lines and continuum in lensed image d versus all other images , with the emission lines at lower magnification in d relative to its continuum . other lensed images show weak differential magnification between lines and continuum . due to an unknown lens redshift the lens model is not well established and so the macro - magnifications are not well known . for this reason we do nt know which , if any of a , b , or c are free of microlensing . nonetheless , we can conclude that image d is undergoing significant differential microlensing between belr and accretion disk . these results are consistent with observations over the past decade ( angonin et al . 1990 , stensen et al . 1997 , monier et al . 1998 , anguita et al . 2008 , h10 ) , which show strong anomalous line - continuum flux ratios in image d compared to a , b , and c. our observations show wavelength - dependent differential microlensing of the continuum in the d / a and c / a ratios . the d / a ratio tends to more extreme values ( deviating from the average bel ratio ) towards shorter continuum wavelength , as was also observed by h10 . it is tempting to interpret this as a simple inverse correlation between emission region size and microlensing intensity , however we note that the observed c / a ratios have the opposite behaviour , tending towards the average c / a bel ratio towards _ longer _ wavelengths . in section [ microlensingbehavior ] we show that we can not expect a monatonic trend of magnification ratio with size across multiple orders of magnitude in emission region size . proper analysis of differential microlensing across the continuum requires careful lensing simulation , however we do not perform that analysis in this work . to investigate the microlensing state of the bels we determine the wavelength - dependent flux ratios for the bel seds for lensed images a and d. these are just the ratios of the spectra of images a and d after subtracting the continuum estimates described above . in figure [ linesratios ] ( lower panels ) we can see that the flux ratios for the nv , siiv , civ , and ciii ] broad emission lines are flat over the wavelength range redward of bal absorption . ly@xmath5 is omitted due to its overlapping the nv absorption trough , which makes continuum subtraction impossible . this flatness indicates that there is no differential microlensing across the belr velocity structure redward of the bal absorption , and very little difference between these emission species . the mean and individual bel flux ratios are given in table [ fluxratiotable ] . error intervals include standard error of the mean and uncertainty in the continuum fits . variation of the level of continuum subtraction up to its estimated uncertainty at the location of each emission line does not change the observed flatness of these flux ratios . in @xcite we analyzed microlensing signatures in the civ and ciii ] bels of q2237 + 0305 , the einstein cross . in that case there was clear differential microlensing across the velocity structure of these broad lines , with magnification ratios varying by @xmath025 - 30% between line center and wings . the bel magnification ratio approached the highly anomalous continuum magnification ratio at the highest velocities . our simple model indicated that this trend was consistent with the broad line flow being a keplerian field , and not supporting a pure radiation - driven outflow without a substantial gravitational component . this differential microlensing signature sharply contrasts the wavelength - independent bel magnification ratios seen in h1413 + 117 ( sect . [ sigs ] ) . this wavelength independence suggests that the belr in h1413 + 117 is large enough to average over fluctuations in the magnification surface . given the very low redshift of the lensing galaxy in q2237 + 0305 ( @xmath8 ) and its commensurately large fluctuation scale - length , differential microlensing of its belr is not surprising . to explore the implications of this difference in more detail we compare the expected microlensing behavior of these two quasars as a function of source size . although the lensing galaxy of h1413 + 117 does not have a confirmed redshift , it is possible to explore plausible microlensing scenarios using estimates from literature : z@xmath9 = 0.94 based on the probability distribution from lens parameters @xcite and supported by the z@xmath10 0.9 photometric redshift of overdensities in the field @xcite ; and z@xmath9 = 1.4 @xcite , that is also comparable to the z @xmath0 1.5 absorption systems in the quasar spectrum @xcite . for the h1413 + 117 lens macro - model we use the convergence and shear estimates of @xcite ( @xmath11 case ) , which are nearly identical to those of @xcite . we assume an 80% smooth matter percentage , consistent with the values estimated in other lenses where the line of sight does not pass through the galactic core @xcite . for q2237 + 0305 we use the macro - model of @xcite , and trott & wayth ( priv . comm . ) . in this case lensed images are clustered around the star - dominated galactic core and microlensing is consistent with a 0% smooth matter percentage @xcite , which we employ here . we obtained simulated maps of microlensing magnifications from the gerlumph database @xcite based on these models . we simulate the expected microlensing behavior of these lenses , using monte carlo simulations . emission region surface brightness distributions are simulated with gaussian profiles whose half - light radii represent emission region sizes . note that except in the case of very strong microlensing events , half - light radius rather than the detailed surface brightness distribution governs microlensing properties @xcite . a pair of random locations on gaussian - convolved magnification maps ( one map for each of a pair of lensed images ) yields a possible magnification ratio between those lensed images . by varying half - light radius and location we can then explore microlensing behavior as a function of emission region size . we determine d / a ratios for h1413 + 117 at 50,000 location pairs and b / a ratios for q2237 + 0305 at 25,000 location pairs . we explore sizes from @xmath12 m to @xmath13 m , constrained by the resolution and size of the gerlumph maps . figure [ ratiovssize ] ( top 3 panels ) shows magnification ratio versus emission region size for a random selection of 26 of these locations . magnification ratios are normalized to unity at the macro - model magnification ratios ( the large - scale average ratios ) . these figures , which we will refer to as _ convergence plots _ , illustrate two important points : \1 ) as expected , magnification ratios will converge on the macro - model ratios above a certain emission region size ( the _ convergence size _ ) . we see that for h1413 + 117 ( top 2 panels ) the d / a convergence size is @xmath14 m ( @xmath020 light days ) for both the @xmath15 and @xmath16 cases . if the wavelength independence that we observe in the bel magnification ratios ( sect . [ fluxratios ] ) is due to such convergence , then this gives us an approximate lower limit on the size of observed belrs . in the case of q2237 + 0305 ( panel 3 ) the convergence of the b / a magnification ratio does not occur until significantly greater than @xmath17 m , beyond the limit of our simulated magnification maps , and so the observation of differential microlensing across the full velocity width of its belr is not surprising . \2 ) we expect tracks on the convergence plot to be flat for sizes larger than the convergence size , however we also see many flat tracks for very small emission regions . this independence of magnification ratio with size often occurs in cases where the magnification ratio deviates significantly from the macro - model value ( i.e. where there is strong microlensing ) . for h1413 + 117 , most simulated tracks are relatively flat below @xmath18 m. there is an intermediate range over which all tracks converge on the macro - model ratios the _ convergence range_. the convergence range spans @xmath18 m to @xmath14 m for h1413 + 117 , and in it we see strong size dependence of magnification ratios , and here the only flat tracks are those already close to the macro - model ratio . the interpretation of this behavior is straightforward : the small - scale flatness seen the convergence plots occurs when a small emission region falls in relatively featureless regions between strong magnification gradients in the caustic network . as size increases into the convergence range , our larger source is beginning to overlap high or low magnification structures in the caustic network , resulting in size - dependent , differential microlensing . when the emission region is large enough to sample many such structures , tracks again flatten at the macro - model magnification ratio . the magnification ratios of the civ and ciii ] bels in h1413 + 117 vary by less than 5% across the unabsorbed component of their velocity structure ( sect . [ fluxratios ] ) , and so must reside on a `` flat '' region of a convergence plot track . we can place more formal constraints on the possible sizes of these emission regions by determining the number of tracks that show similar flatness over a given size range . we conservatively assume that the broad line gas flow spans a size range of 0.5 dex across its entire velocity structure . in fact the belr probably spans a larger range , but choosing a smaller range yields more tracks within our flatness constraint , and so produces weaker ( more conservative ) size constraints . figure [ ratiovssize ] ( bottom ) shows the percentage of tracks consistent with @xmath195% variation in the magnification ratio over 0.5 dex in emission region size as a function of the maximum size of the emission region , for both h1413 + 117 ( @xmath15 ) and q2237 + 0305 . in the case of q2237 + 0305 figure [ ratiovssize ] shows a reasonable likelihood of observing size - independent microlensing over this full range of emission region sizes . for h1413 + 117 , size - independent microlensing is unlikely ( @xmath20 5% ) for belrs spanning 0.5 dex with maximum size in the range @xmath18 m to @xmath21 m ( @xmath20 1% for @xmath22 m to @xmath23 m ) . at larger sizes , size independence will become more probable , and extrapolating our results we estimate that size independence will become likely ( @xmath750% ) at belr sizes of 2 - 3@xmath24 m , or around 100 light days . it is important to note that h10 also observe this size - independent microlensing of the belr at a different epoch , greatly decreasing the probability that the belr is in the @xmath18 m to @xmath21 m range . for a reasonable probability of observing size independent magnification ratios in the belr of h1413 + 117 in two independent epochs , the belr probably needs to be larger than @xmath17 m ( @xmath040 light days ) , and any significantly luminous component likely arises well above the convergence size of @xmath14 m ( @xmath020 light days ) . these constraints are for the @xmath15 case ; larger values of @xmath25 push us to larger belr sizes . we can also consider the possibility that the entire belr is emitting at sizes smaller than @xmath18 m ( @xmath04 light days ) . we think this unlikely for two related reasons : 1 ) our observations and those of h10 show wavelength - dependent differential microlensing in the continuum d / a ratio , which precludes flat tracks at small size needed for size independence in the bel magnification ratios ; 2 ) for a high likelihood of size independence at small scales at two epochs , the emission region must be closer to a few times @xmath26 m , which is approaching the scale of the accretion disk ( see sect . [ sizes ] ) , and yet the d / a ratios indicate very strong differential microlensing between continuum and bels at multiple epochs ( see sect . [ fluxratios ] ) . from the reverberation mapping belr size - luminosity relation for luminous quasars of @xcite , h1413 + 117 ( with its @xmath27 f@xmath28(1350)@xmath29 erg s@xmath30 ) is expected to have a civ belr in the range of 2590 light days . from their microlensing study of of 16 lensed quasars ( not including h1413 + 117 ) , @xcite find a belr size range for high - ionization lines in their luminous subsample of @xmath31 light days . both of these results support our estimate of a belr that is signficantly larger than 20 light days , and very unlikely to be @xmath204 light days . we estimate the black hole mass in h1413 + 117 at @xmath32 m@xmath33 based on the civ width and using the @xcite m@xmath34fwhm relation . we measure the civ fwhm of 3,400 km / s based on the unabsorbed , red side . note that the intrinsic civ bel is either symmetric or shows a blue excess , as discussed in section [ belrcomp ] , and so the correct fwhm and corresponding black hole mass may be a larger . in order to obtain an upper limit estimate on the size of the accretion disk in h1413 + 117 we assume a shakura - sunyaev thin disk @xcite with a @xmath35 m@xmath33 non - rotating black hole accreting at the eddington limit , and observed at 150 nm . this yields a source half - light radius of @xmath36 m , shown in figure [ ratiovssize ] . for our black hole mass estimate of @xmath32 m@xmath33 and the same accretion parameters the 150 nm source size is @xmath37 m. based on differential microlensing across the continuum , h10 estimate a continuum source size of @xmath38 m. the accretion disk for both quasars is much smaller than the convergence range . for h1413 + 113 both size - dependent and size - independent magnification ratios are reasonably common at small radii , and it s not surprising that the continuum shows some differential magnification in both our observations and those of h10 . this is also true for q2237 + 0305 , however size dependence occurs less frequently at these small size scales , in @xmath020% of cases . the d / a magnification ratio in the h1413 + 117 continuum tends towards the average d / a bel ratio towards longer continuum wavelengths ( although never getting close to the bel ratio ) , and so it s tempting to interpret this as the accretion disk becoming large enough to start averaging over causic structures and hence approach the macro - model magnifications . however figure [ ratiovssize ] shows that even a large accretion disk is much smaller than the convergence range . in this regime , magnfication ratio can have a size dependence that either approaches the macro - model value or moves away from the macro - model value , even if all tracks ultimately converge on the same value . in short , we can not interpret a trend towards the macro - model magnifications as a reliable proxy for increasing emission region size _ except _ in the convergence range . it is therefore not surprising that the c / a continuum ratios tend in the opposite direction to the d / a ratios , with shorter wavelengths approaching the c / a bel ratio . proper interpretation of the differential microlensing in the continuum requires higher resolution magnification maps and a careful statistical analysis , which we leave to a later study . we can consider the spectra of the lensed images in h1413 + 117 to be linear combinations of intrinsic spectral components from different emission regions , each subject to absorption and lensing that may vary between lensed images and emission regions . if the relative strengths of the emission components vary between one lensed image and another due to differential microlensing or intrinsic variability , then with some assumptions it can be possible to solve for these components . @xcite first apply this approach to h1413 + 117 to derive an intrinsic bal absorption profile for siiv , civ , and aliii . they describe the spectrum as composed of physical components : a continuum + bel . h10 explore the approach in more detail , applying the method of @xcite to h1413 + 117 to extract ly@xmath5+nv , siiv , civ bal troughs . in the @xmath39 method , the quasar spectrum is explicitly described in terms of a component that is only macrolensed and one that is both micro- and macrolensed . both approaches lead to similar results , and this is because their assumptions are similar : that the continuum is microlensed but the belr is not ; as a result the micro+macrolensed component of h10 is essentially the absorbed , lensed continuum . here we derive a new spectral decomposition of h1413 + 117 . our approach is mathematically similar to both angonin et al . and h10 , and we discuss their connection in [ h10comparison ] . we choose to formulate the decomposition in terms of physical emission regions rather than by their degree of microlensing . we do this because : 1 ) our observation of almost constant magnification ratios across the emission line velocity structures ( see sect . [ sigs ] ) ensures a straightforward extraction of the bel component , and 2 ) we include in our formulation a separate bal absorption term for both the continuum and belr , which we feel adds clarity to the interpretation of the derived components , and 3 ) any source of variation microlensing or intrinsic variability coupled with time delay enables a decomposition , and yet the possible presence of the latter means that results can not be easily interpreted as microlensed / non - microlensed components . we describe the spectrum * * _ f_**@xmath40 of lensed image x as a linear combination of two components : 1 ) @xmath41 : the intrinsic continuum spectrum * * _ f_**@xmath42 multiplied by the bal absorption profile affecting the continuum * * _ a_**@xmath42 and the ( potentially wavelength - dependent ) magnification * * _ m_**@xmath43 affecting the continuum in image x , and 2 ) @xmath44 : a broad emission line spectrum * * _ f_**@xmath9 with its own , potentially different absorption profile * * _ a_**@xmath9 and lensing magnification * * _ m_**@xmath45 . the * _ f _ * and * _ a _ * components are intrinsic to the quasar and identical for all lensed images ( modulo quasar variability , but see below ) . the lines of sight to different lensed images do not diverge significantly within the bal wind ( see sect . [ lineofsight ] ) and so all lensed images see the same absorption from this outflow . however separate absorption terms are required becaues the accretion disk and belr are viewed through different regions of the bal wind . the magnification factor is different for each lensed image and in the presence of microlensing this factor may differ between the continuum and bels , and each of these may have wavelength dependence . we have neglected to include an explicit term for reddening or to account for intrinsic variability in the quasar . both of these effects cause wavelength - dependent multiplicative changes to component spectra , and so in terms of the decomposition we can consider these to be folded into the magnification terms ; the origin of these multiplicative effects does not affect the decomposition . for lensed images 1 and 2 we have : @xmath46 @xmath47 the only known quantities in these equations are the observed spectra , * * _ f_**@xmath48 and * * _ f_**@xmath49 . although we do nt know the intrinsic magnification factors of each component , it is possible to measure their ratios . in the absence of differential absorption or reddening these are just the wavelength - dependent flux ratios for each component . we set @xmath50 and @xmath51 = * * _ m_**@xmath52**_f_**@xmath53 : the magnification ratios in continuum and emission line . we can then solve for @xmath54 and @xmath55 for one lensed ] @xmath56 we can think of these equations as representing a scaled subtraction of the spectrum of one lensed image ( * * _ f_**@xmath49 ) from another ( * * _ f_**@xmath48 ) , with the scaling factor @xmath57 giving complete subtraction of one spectral component . in eq . [ cont ] , for example , * * _ f_**@xmath49 is scaled so that its emission line flux matches that of * * _ f_**@xmath48 . as long as the line - to - continuum ratio is different between * * _ f_**@xmath48 and * * _ f_**@xmath49 then * * _ f_**@xmath58 * * _ f_**@xmath49 eliminates the bel spectrum leaving a representation of the absorbed continuum spectrum with a slope somewhere between that of lensed images 1 and 2 . the scaling factor @xmath59 then recovers the normalization and slope of the continuum spectrum seen in image 1 . a similar interpretation of eq . [ line ] eliminates the continuum component to give the lensed , absorbed bel spectrum for image 1 . for our decomposition of h1413 + 117 we define * * _ f_**@xmath48 to be the spectrum of image d and * * _ f_**@xmath49 to be that of image a. we take @xmath60 to be the wavelength - dependent continuum ratio from the fits described in section [ sigs ] . as seen in sect.[sigs ] , the bel magnification ratios are not wavelength dependent over each line and vary relatively little between siiv , civ , and ciii ] . to reconstruct spectral components over the full wavelength range of the spectra we take @xmath61 to be a constant : @xmath62 , the mean continuum - subtracted broad line flux ratio from table [ fluxratiotable ] . for individual bel decompositions we use their individual magnification ratios ( see below ) . figure [ decomp ] shows the resulting decomposition over the full spectrum . the upper panel shows @xmath63 : the lensed , absorbed quasar continuum ( eq . [ cont ] ) , with the continuum magnification factor of image d. we overplot the continuum fit for the image d spectrum . the lower panel shows @xmath64 : the lensed , absorbed bel spectrum ( eq . [ line ] ) , again as magnified in image d. we overplot the observed spectrum of image d minus the continuum fit and also an iron emission spectrum from the templates of @xcite , scaled to roughly match the iron features in the spectrum . figure [ civdecomp ] shows a nearly identical decomposition as fig . [ decomp ] over the range of the civ line , with the only difference being that we take @xmath65 , the continuum - subtracted flux ratio over the red wing of the civ line . [ civdecomp ] also shows the range of uncertainty for these decompositions ( red lines ) : the maximum and minimum decomposed bel and continuum components obtained by allowing the continuum flux to vary by @xmath66% ( see sect . [ sigs ] ) and @xmath67 to vary over its uncertainty range . it can be seen that the general shape of the decomposed components does not change significantly over the range of possible decompositions . lens image d is experiencing by far the strongest microlensing between the continuum source and belr . however we do see some differences in broad line to continuum flux ratios in a , b and c. figure [ comparedecomp ] shows these same decompositions over the siiv and civ lines for image pairs a & d and for b & c. in these cases we take the a / d and b / c emission line flux ratios to be the average for the siiv and civ lines for each image pair . we see very consistent results , albeit with significantly more noise in the b & c decomposition due to the smaller differences between these spectra . if we divide @xmath68 , the lensed , _ absorbed _ continuum spectrum , by our estimate of the lensed , unabsorbed continuum * * _ m_**@xmath69 * * _ f_**@xmath42 we recover an estimate of the intrinsic absorption profile affecting the continuum . figure [ pcygni ] shows this for the civ , siiv , nv , and aliii absorption troughs . . velocity zero - point is set at line center or at lower wavelength line center for doublets . vertical dotted lines show line centers . letters are designations for absorption trough features discussed in section [ features ] . [ pcygni],width=453 ] the derived spectral components are similar to those found by h10 , and the derived absorption profile is similar to that of @xcite . we discuss briefly similarities and differences in these works . decompose in terms of macrolensed - only ( their @xmath70 ) and microlensed + macrolensed ( @xmath71 ) components , and use the combined spectra of lensed images a and b ( @xmath72 ) as well as that of image d ( @xmath73 ) . the spectral subtraction scaling factors used by h10 are @xmath74 , the magnification ratio of the macrolensed - only component , and @xmath75 , the magnification ratio of the microlensed + macrolensed component . @xmath75 is taken as the continuum ratio in regions unaffected by broad line flux , and so is similar to our @xmath60 . they measure a single value of @xmath75 for each emission line from the average continuum bounding that line . this accounts for changes in the continuum magnification ratio from one line to the next , but not for variation across the wavelength range of the lines , however the latter is of order 1 - 2% and so is not a large effect . @xmath74 is determined under the assumption that at least part of each broad emission line is not microlensed , so that within the wavelength range of each line there is a region of the reconstructed micro+macrolensed component ( @xmath71 ) that is equivalent to the continuum spectrum . this can be interpreted as subtraction of @xmath72 from @xmath73 , with @xmath72 scaled to eliminate at least part of the emission line ( the part of the line closest to the continuum magnification ) . in practice , because the emission line magnifications are constant across their wavelength range ( see sect . [ sigs ] ) , this should mean almost complete subtraction of emission lines and produces a result that is very similar to our equation 2 . @xcite ( a90 ) describe their decomposition in less detail than h10 , however their approach is mathematically similar to both ours and that of h10 . they express the spectrum of lensed image d to be that of the ( purportedly ) unmicrolensed a+b+c composite plus a microlensed , absorbed continuum component . after deriving a continuum spectrum from the a+b+c composite they fit parameters to yield the absorption profile . this appears to be equivalent to performing a scaled subtraction of a+b+c from d to eliminate the bel component and at the same time scaling out the wavelength - dependent continuum , reddening , lensing magnifications . in summary , although a90 and h10 formulate decompositions differently and make some different assumptions , all three approaches are mathematically similar and yield similar decompositions . it is worth noting that in our formulation the interpretation of the decomposition does not depend on their being any macrolensed - only component , only that the continuum - belr ratio differs between the two lensed images . while both h10 and a90 assume that the belr is not microlensed , any such wavelength - independent microlensing or variability - related differences in the belr do not invalidate the decompositions . we compare the details of these decomposition results in section [ disc ] . the continuum spectral component ( fig . [ decomp ] & [ civdecomp ] , upper panels ) and the absorption profile that we derive from it ( fig . [ pcygni ] ) provide a picture of the bal absorption along the line - of - sight column to the accretion disk , and the shape of these profiles allows us to draw specific conclusions about the structure and kinematics of the outflow . we focus on interpretations in the context of rotating disk - wind models , in which the material responsible for both bal and bel features arises from the accretion disk and is driven out by radiation and possibly gas pressure @xcite . in the case of civ the continuum is experiencing essentially no absorption until @xmath01,500 km / s blueward of the emission line central wavelength , at which point transmission drops sharply over @xmath0600 km / s to opacity @xmath76 . this near - black region extends to @xmath07,500 km / s before transmission ramps roughly linearly to the end of the absorption feature at @xmath015,000 km / s blueward of line center . two narrow emission peaks are seen in this latter region . in the case of siiv , absorption also has a sharp onset blueward of the siiv @xmath771394 line center , in this case at @xmath03000 km / s . the deep region of its absorption trough ends much more sharply than in the case of civ at @xmath07,500 km / s . a single clear emission peak is observed in the deep part of the trough . a weaker absorption trough appears redward of the deeper trough onset and may result from siiv @xmath771403 . in any case , all absorption is blueward of the siiv doublet . as with civ , the nv trough begins at a blueshifted @xmath01,500 km / s , while the aliii trough appears at @xmath04,500 - 5,000 km / s , though may show some absorption from @xmath02,000 km / s . of particular interest is the sharp , blueshifted onset of this absorption , which is especially striking in civ . this sharpness is not ambiguous , and is apparent over the full range of uncertainty in the decomposition ( fig . [ civdecomp ] , lower ) . while bals with highly blueshifted absorption onsets are sometimes observed ( so - called detached troughs , blueshifted at several thousand km / s ; e.g. turnshek 1987 ) , our decomposition of h1413 + 117 provides the first unambiguous description of a sudden , blueshifted absorption onset that is still within the velocity range of the bel . the steepness and blueshift of this onset tells us two things : 1 ) the outflow has already accelerated to some degree before either it enters our sightline or it enters the region of efficient absorption within our sightline ; and 2 ) the kinematics of absorbing gas in our sightline is strongly dominated by outflow , not orbital motion . @xcite intepret highly detached absorption troughs as resulting from sightlines that graze the upper layer of an equitorial outflow , hence missing the vertical base of their wind model where the flow has a small radial component . @xcite ( h13 ) apply radiative transfer code to their disk - wind model to predict civ bal profiles , also finding that these same grazing sightlines produce a very sharp , blueshifted onset . in these equitorial disk - winds we expect any deeper line of sight those passing near the launch region of the wind to show absorption at low velocities . not only do they encounter launching gas prior to significant outward acceleration , but they also encounter gas with very high rotational velocities in the vicinity of the accretion disk . this latter effect is important , as shown by h13 ; at launch , the bal flow shares the keplerian velocity of the accretion disk and so is dominated by rotational motion . if the inner disk is occluded by this gas along our line of sight , and if this gas is significantly absorbing in the vicinity of the accretion disk , then some fraction of the continuum will be absorbed by the component of the orbiting flow that is receding along our sightline , producing partial absorption of the continuum at lower and/or redshifted projected velocities . on the other hand , even large rotational motion far from the accretion disk will only have a transverse velocity component within our line - of - sight column to the disk . this argument against a rotational contribution to the line - of - sight velocity is supported by @xcite , who show that , for a spherically expanding outflow , a sharp absorption onset is only expected in cases with small rotational components . any significant rotation results in a gradual absorption onset . this same grazing sightline in the h13 model also appears to reproduce the ramped blue edge of the civ absorption trough that we observe in h1413 + 117 . the models of @xcite show that a ramped increase in transmission at the blue edge indicates a large variation in both velocity and density in the absorbing region along the line of sight , which appears to occur when this line of sight is close to tracing a streamline in the flow . this is consistent with a line of sight along the upper surface of the wind . the interpretation of @xcite contradicts this . they classify sdss bal quasars according to the appearance of siiv and aliii troughs alongside the ubiquitous civ trough , and interpret those containing all three ( h1413 + 117 does , making it a lobal ) as resulting from lines of sight on the underside of the outflow , intersecting the base of the wind . the argument of the grazing sightline does not explain the different velocity shifts we see in the absorption onset for different line species , which should be the same if caused by our line of sight exiting the outflow . while ion species with the highest ionization potentials , civ ( 64.5 ev ) and nv ( 97.9 ev ) , have similar @xmath01,500 km / s absorption onsets , the onset for siiv ( 45.1 ev ) is @xmath03,000 km / s and the deeper part of the aliii ( 29.4 ev ) trough starts at @xmath04,500 - 5,000 km / s ( though may show weaker absorption from @xmath02,000 km / s ) . this increase in absorption onset with diminishing ionization potential may suggest that for siiv and aliii we are in fact seeing very sudden jumps in transition rate as ionizing radiation and/or gas density reach optimal levels as a function of radius and hence outflow velocity . for example , this may be the radius where it is first possible for `` locally optimized clouds '' @xcite to form due to the diminishing ionization parameter . a combination of the above may be needed to explain the onset of all absorption troughs ; we propose that for civ and the higher ionization nv we are seeing our line of sight exit the outflow near its launch point , while for lower ionization lines we are seeing the onset of favorable absorption conditions further out in the flow . in the ionization / radiative transfer calculations of h13 , high civ fractions are expected at the highest ionization parameters in their luminous benchmark quasar ( m@xmath78 m@xmath33 , eddington ratio = 0.2 ) , and so it is expected that civ absorption should be seen at its inner - most radii . figures [ decomp ] & [ civdecomp ] ( lower ) shows a plausible bel spectrum , however it should be remembered that this component includes the bal absorption of this broad line emission . it is not possible to extricate the true absorption profile @xmath79 from the line profiles @xmath80 . in the case of the civ line ( fig . [ civdecomp ] , lower ) we see a region of near - zero emission in the rest frame 1515 - 1525 range ( @xmath045006400 km / s ) where we might otherwise expect to see broad - line emission either civ or feii , given the presence of one or both surrounding this region , and the significant iron emission in this wavelength range of the @xcite template . we take this as evidence of absorption of broad - line emission by the bal wind . if this black region is due to absorption then we are seeing very high column densities _ and _ very high covering factors for absorption of the civ and/or feii emission at these velocities . this is very interesting due to the presumably much larger extent of the source of this broad - line emission compared to the continuum the line - emitting gas is completely enveloped by absorbing gas across the entire belr at these projected velocities . further evidence of absorption of bel light by the bal wind comes from the significant absorption of ly@xmath5 by the nv trough in this source ( see also turnshek 1987 ) . while the symmetry of the civ emission in this component resembles that of an unabsorbed bel , it should be remembered that civ very frequently shows a strong blue excess which may be absorbed in this case . the siiv line shows a sharp drop in flux in its blue wing that also suggests absorption of bel light . the iron emission template of @xcite shown in these figures provides a good match to several features in the emission line spectrum , and an imperfect match to others . the complexity and potential variation in iron emission makes a perfect match unlikely , however the scale and shape of these features match the template adequately . as well as natural variation , it is also possible that we are seeing differential microlensing acrosss the iron emission spectrum . the strong presence of iron features in component @xmath81 implies that some of this emission arises on size scales similar to other bels . the fact that the iron template provides a reasonable match to the iron emission across the full wavelength range of the belr component ( fig . [ decomp ] ) gives us confidence in our choice of continuum , and that we avoided most of the iron emission ; if not then we would expect a significant and wavelength - dependent mismatch between the decomposed bel spectrum and the template . low - velocity - width emission peaks are seen within the region of the civ bal absorption trough , marked in figure [ civdecomp ] : in the continuum component centered at blueshifts of @xmath08,000 km / s ( _ a _ ) ) and @xmath011,000 km / s ( _ b _ ) and in the bel component at @xmath04,000 km / s ( _ c _ ) , @xmath07,500 km / s ( _ d _ ) , and @xmath013,000 km / s ( _ e _ ) . an emission peak is also observed blueward of siiv @xmath771393.76 at @xmath04,500 km / s ( which we designate feature _ f _ , fig . [ pcygni ] ) . h10 observe a similar emission peak in their decomposition ; it appears in all epochs in @xmath06,5009,500 km / s blueward of civ in their @xmath82 ( bel ) component ( from the decomposition of lensed image d with respect to a & b ) . this appears to correspond to our feature _ d_. also notable is a feature in their @xmath83 ( continuum ) component from the decomposition of image c which corresponds to our feature _ a_. a similar peak was also observed in the civ absorption profile derived by a90 at @xmath05,500 km / s . in all of these previous decompositions , signal quality makes it is difficult to rule out the presence of any of the other features that we note in the current work . the features observed in the civ and siiv troughs of the continuum component ( _ a _ , _ b _ & _ f _ ) indicate regions of reduced absorption in the bal outflow . this may be due to radial fluctuations in overall outflow density and/or relative ion species density due to variations in ionization parameter . this interpretation reflects that of h10 , who propose a strongly absorbing cloud embedded within the outflow but with lower velocity width that results in a double - peaked civ line . in the context of a sightline grazing the upper layer of a fast , equitorial outflow , @xcite show that , in the hydrodynamic model of @xcite , `` blobs '' of failed outflow can cause clumpiness in this layer as they fall from the over - ionized high - inclination regions . interpretation of the features in the bel component is more complicated . d _ and _ e _ may result from clumpy absorption of a strong blue excess in civ and/or feii emission ( which may or may not be absorbed ) . as discussed in section [ belrcomp ] , the near - black region between features _ c _ and _ d _ does seem to indicate absorption of bel light by the outflow . whether this is also true of the dark region between features _ d _ and _ e _ is unclear , as feature _ e _ may be isolated feii emission . c _ is probably iron emission . it is clear that features _ a _ and _ d _ correspond to the same peak in the original spectrum , and so to the same region of diminished absorption . b _ in the continuum component has no obvious counterpart in the bel component . this may be due to intrinsically low bel flux in this region or a difference in the bel vs. continuum absorption profiles . the differences observed in the bal profiles between lens images can not be due to differences in the line of sight . the bal absorption profile affecting the continuum is determined by the region of the bal flow sampled by the column along the line of sight to the accretion disk , and so different lensed images will only experience different absorption if their line - of - sight columns diverge significantly within this outflow . the image separation between images a and d is 118 , which corresponds to a divergence angle of the a and d lines of sight from the source of 066 and 20 for @xmath15 and 1.4 respectively . assuming @xmath16 and a conservatively small accretion disk with half - light radius @xmath26 m ( ss disk with @xmath84 and @xmath85 ) , lines of sight have diverged by one half - light radius at @xmath0100 light years . this is probably much larger than the strongly - absorbing region of the bal flow . choosing a smaller @xmath25 gives an even smaller divergence . on the @xmath201000 light day size scale of a large belr @xcite , line - of - sight columns to the accretion disk still have @xmath86% overlap . line - of - sight columns to the belr , assuming 100 light day cross - section , diverge completely at @xmath028,000 ly , and so may be subject to different line - of - sight absorption features both within the host galaxy and in the igm . this is also true of the continuum , however the continuum will be subject to differences in absorption features from a greater depth within the host galaxy . these line - of - sight differences do account for the many differences in narrow absorption features observed in our spectra . @xcite show that these features arise both in the igm and potentially the lensing galaxy , but probably not the host galaxy which has a higher redshift than observed features . we do nt investigate the absorption features in detail in this paper , however we note that the @xmath87 feii lines ( @xmath88 2344.2 , 2374.5 , 2382.8 ) follow the same behavior in our spectra as observed by @xcite ; it is present in lensed images a and b and absent in c and d. several other features exhibit similar behavior , and we will investigate these in a separate study . we have presented new spectroscopy of the gravitationally lensed broad absorption line quasar h1413 + 117 using gemini north s gmos ifu . extracted spectra of the four lensed images reveal that differential microlensing strongly affects the relative magnification of the broad emission lines compared to the continuum in lensed image d , with the continuum in d appearing anomalously bright compared to all other lensed images . lensed images a , b , and c also show small differences in their continuum - to - broad line flux ratios relative to each other . subtracting a fit to the continuum emission outside the broad emission and absorption regions allows us to determine the wavelength - dependent d / a flux ratios of the broad emission lines . these flux ratios are flat , indicating no differential microlensing across the velocity structure of the belr . monte carlo simulations were performed to analyze the dependence of the flatness of broad line flux ratios on the the size range spanned by the belr velocity structure . it was found that belrs with an upper size limit in the @xmath89@xmath17 m range , and spanning 0.5 dex across their velocity structure are very unlikely to produce the observed flat flux ratios . smaller belrs are also unlikely due to the strong differential magnification observed between broad lines and continuum . therefore , it is inferred that the belr in h1413 + 117 is significantly larger than @xmath17 m , and produces no significant emission from regions smaller than @xmath14 m , or 20 light days . we take advantage of the differential magnification between lensed images a and d to algebraically decompose the quasar spectrum into two components : the absorbed broad line spectrum and the absorbed continuum spectrum , and we use the latter to derive intrinsic absorption profiles . our decomposition leads us to these conclusions : @xmath90 a sharp , blue - shifted onset of absorption is observed in all prominent bal features in the continuum component . in the case of civ , nv , and ly@xmath5 this onset is at @xmath01,500 km / s , which may represent the inner edge of a radial outflow . in the context of typical disk - wind models this would place the onset of bal absorption near its launch point , and so in the vicinity of the accretion disk . in the case of siiv and aliii , which have a larger absorption onsets , we may be seeing increases in absorption efficiency of lower ionization species with distance from the ionizing source . this sharp onset also indicates that the line - of - sight velocity is dominated by outflow , not rotational motion . @xmath90 overall the derived absorption profile for civ resembles that of an accelerating outflow that is either spherical or viewed relatively close to a streamline , albeit with a @xmath01,500km / s initial velocity . @xmath90 the civ and/or feii belr appears to be strongly absorbed in the range @xmath04,5006,400 km / s . it is unclear whether emission features blueward of this trough are civ or feii , however the complete absorption of either of these features suggests a very high covering factor for the absorbing material at large radii . combined with the apparent strong absorption near the outflow launch point , this suggests that gas responsible for bal features maintains a high - covering - factor over at least two orders of magnitude in size . angonin m .- c . , vanderriest c. , remy m. , surdej j. , 1990 , a&a , 233 , l5 allington - smith , j. , graham , m. , content , r. , dodsworth , g. , davies , r. , miller , b. w. , jorgensen , i. , hook , i. , crampton , d. , murowinski , r. , 2002 , pasp , 114 , 892 anguita , t. , faure , c. , yonehara , a. , wambsganss , j. , kneib , j .- p . , covone , g. , alloin , d. , 2008 , a&a , 481 , 615 baldwin , j. , ferland , g. , korista k. , verner , d. , 1995 , apj , 455 , l119 bate , n. f. , floyd , d. j. e. , webster , r. l. , wyithe , j. s. b. , 2008 , mnras , 391 , 1955 bate , n. f. , floyd , d. j. e. , webster , r. l. , wyithe , j. s. b. , 2011 , apj , 731 , 71 bentz , m. c. , walsh , j. l. , barth , a. j. , et al . , 2009 , apj , 705 , 199 blackburne , j. a. , pooley , d. , rappaport , s. , schechter , p. l. , 2011 , apj , 729 , 34 blackburne , j. a. , kochanek , c. s. , chen , b. , dai , x. , chartas , g. , 2014 , apj , 798 , 95 blandford , r. d. , mckee , c. f. , 1982 , apj , 255 , 419b borguet , b. , hutsemkers , d. , 2010 , a&a , 515 , 22 chartas , g. , kochanek , c. s. , dai , x. , poindexter , s. , garmire , g. , 2009 , apj , 693 , 174 chelouche , d. , netzer , h. , 2005 , apj , 625 , 95c crenshaw , d. m. , kraemer , s. b. , 2007 , apj , 659 , 250 dai , x. , kochanek , c. s. , chartas , g. , kozlowski , s. , morgan , c. w. , garmire , g. , agol , e. , 2010 , apj , 709 , 278 denney , k. d. , peterson , b. m. , pogge , r. w. , et al . , 2009 , apj , 704 , l80 doroshenko , v. t. , sergeev , s. g. , klimanov , s. a. , pronik , v. i. , efimov , y. s. , 2012 , mnras , 426 , 416 eigenbrod , a. , courbin , f. , sluse , d. , meylan , g. , agol , e. , 2008 a&a , 480 , 647 elvis , m. , 2000 , apj , 545 , 63 fabian , a. , 1999 , mnras , 308 , l39 fabjan , d. , borgani , s. , tornatore , l. , saro , a. , murante , g. , dolag , k. , 2010 , mnras , 401 , 1670 faure , c. , alloin , d. , kneib , j-.p . , courbin , f. , 2004 , a&a 428 , 741 filiz ak , n. , brandt , w. n. , hall , p. b. , et al . , 2014 , apj , 791 , 88 floyd , d. j. e. , bate , n. f. , webster , r. l. , 2009 , mnras , 398 , 233 gallagher , s. c. , everett , j. e. , 2007 , _ the central engine of active galactic nuclei _ , asp conference series , vol . l. c. ho and j .- m . wang gallagher , s. c. , hines , d. c. , blaylock , myra , priddey , r. s. , brandt , w. n. , egami , e. e. , 2007 , apj , 556 , 619 giustini , m. , proga , d. , 2012 , apj , 758 , 70 goicoechea , l. j. , shalyapin , v. n. 2010 , apj , 708 , 1001 guerras , e. , mediavilla , e. , jimenez - vicente , j. , kochanek , c. s. , muoz , j. a. , falco , e. , motta , v. , 2013 , apj , 764 , 160 hall , p. b. , brandt , w. n. , petitjean , p. , et al . , 2013 , mnras , 434 , 222 higginbottom , n. , knigge , c. , long , k. s. , sim , s. a. , matthews , j. h. , 2013 , mnras , 436 , 1390 hook , i. , jorgensen , i. , allington - smith , j. r. , davies , r. l. , metcalfe , n. , murowinski , r. g. , crampton , d. , 2004 , pasp , 116 , 425 jimnez - vicente , j. , mediavilla , e. , kochanek , c. s. , muoz , j. a. , motta , v. , falco , e. , mosquera , a. m. , 2014 , apj , 783 , 47 kaspi , s. , smith , p. s. ; netzer , h. ; maoz , d. , jannuzi , b. t. , giveon , u. , 2000 , apj , 533 , 631 kaspi , s. , brandt , w. n. , maoz , d. , netzer , h. , schneider , d. p. , shemmer , o. , 2007 , apj , 659 , 997 keeton , c. r. , burles , s. , schechter , p. l. , wambsganss , j. , 2006 , apj , 639 , 1 kneib , j .- p . , alloin , d. , mellier , y. , et al . , 1998 , a&a , 329 , 827 kochanek , c. s. , 2004 , apj , 605 , 58 lamy , h. , hutsemkers , d. , 2004 , a&a , 427 , 107 magain , p. , surdej , j. , swings , j .- p . , borgeest , u. , kayser , r. 1988 , nature , 334 , 325 mediavilla , e. , muoz , j. a. , falco , e. , motta , v. , guerras , e. , canovas , h. , jean , c. , oscoz , a. , mosquera , a. m. , 2009 , apj , 706 , 1451 monier , e. m. , turnshek , d. a. , lupie , o. l. , 1999 , apj , 496 , 177 mortonson , j. m. , schecter , p. l. , & wambsganss , j. , 2005 , apj , 628 , 594 mosquera , a. m. & kochanek , c. s. , 2011 , apj , 738 , 96 murray , n. , chiang , j. , 1995 , apj , 454 , l105 murray , n. , chiang , j. , grossman , s. a. , voit , g. m. , 1995 , apj , 451 , 498 monier , e. m. , turnshek , d. a. , lupie , o. l. 1998 , apj , 496 , 177 odowd , m. , bate , n. f. , webster , r. l. , wayth , r. , labrie , k. , 2011 , mnras , 415 , 1985 ostensen , r. , remy , m. , lindblad , p. o. , refsdal , s. , et al . , 1997 , a&as , 126 , 393 peterson , b. m. , 1993 , pasp , 105 , 247 pooley , d. , rappaport , s. , blackburne , j. , schechter , p. l. , schwab , j. , wambsganss , j. , 2009 , apj , 697 , 1892 proga , d , stone , j. m. , kallman , t. r. , 2000 , apj , 543 , 686 proga , d , kallman , t. r. , 2004 , apj , 616 , 688 rafferty , d.a . , mcnamara , b.r . , & nulsen , p.e.j . , 2008 , apj , 687 , 899 reichard , t. a. , richards , g. t. , hall , r. b. , et al . , 2003 , aj , 126 , 2594 richards , g. t. , keeton , c. r. , pindor , b. , et al . , 2004 , apj , 610 , 679 scannapieco , e. , oh , s. p. , 2004 , apj , 608 , 62 shakura , n. i. , sunyaev , r. a. , 1973 , a&a , 24 , 337 silk , j. , rees , m. , 1993 , a&a , 331 , l1 sim , s.a . , long , k. s. , miller , l. , turner , t. j. , 2008 , mnras , 388 , 611 sluse , d. , claeskens , j .- f . , hutsemkers , d. , surdej , j. , 2007 , a&a , 468 , 885 sluse , d. , schmidt , r. , courbin , f. , hutsemkers , d. , meylan , g. , eigenbrod , a. , anguita , t. , agol , e. , wambsganss , j. , 2011 , a&a , 528 . 100 storchi - bergmann , t. , 2010 , _ proceedings of the iau symposium no . 267 _ , eds . peterson , r.s . somerville and t. storchi - bergmann somerville , r. s. , hopkins , p. f. , cox , t. j. , robertson , b. e. , hernquist , l. , 2008 , mnras , 391 , 481 surdej , j. , fraipont - caro , d. , gosset , d. , refsdal , s. , remy , m. , 1993 , proc . 31st lige international colloq . ( ligel univ . lige ) , eds . j. surdej , d. fraipont - caro , e. gosset , s. refsdal , m. remy trott c. m. , treu t. , koopmans l. v. e. , webster r. l. , 2010 , mnras , 401 , 1540 turnshek , d. a. , 1987 , in _ qso absorption lines : probing the universe _ , ed . j. c. blades , d. turnshek , & c. norman , ( cambridge : cambridge university press ) , 17 van dokkum , p. g. , 2001 , pasp 113 , 1420 vernardos , g. , fluke , c. j. , bate , n. , f. , croton , d. , 2014 , apj , 211 , 16 vestergaard , m. & wilkes , b. j. , 2001 , apjss , 134 , 1 vestergaard , m. , peterson , b. m. , 2006 , apj , 641 , 689 wang , j. , 2015 , newa , 37 , 15 wayth , r. b. , odowd , m. j. , webster , r. l. , 2005 , mnras , 359 , 561 weymann , r. j. , morris , s. l. , foltz , c. b. , hewett , p. c. , 1991 , apj , 373 , 23 witt , h. , j. , mao , s. , schechter , p. , l. , 1995 , apj , 443 , 18 zubovas , k. , king , a. r. , 2012 , mnras , 426 , 2751

excitonic molecules - or biexcitons - are known and well studied since several decades in conventional semiconductors . their formations due to the coupling of a pair of exciton and their photophysics properties have been well established both theoretically and experimentally @xcite . the outstanding facts are related to the giant oscillator strength associated with the one exciton - biexciton transition ; we may cite , for instance , the giant two photon absorption , shown theoretically by hanamura @xcite and observed first on cucl by gale and mysyrowicz @xcite . there , five or six orders of magnitude more than for typical two - photon interband transitions are observed . in organic compounds , the interest for excitonic bound states - or excitonic complexes or last , to use a terminology coming from the bethe ansatz technique , excitonic strings - is rather new since the main studies started about five or six years ago . they concern three types of compounds : some organic charge transfer solids ( ct ) , _ j _ or _ h_-aggregates ( _ j - h_a ) and conjugated polymers ( cp ) . the organic ct solids considered here , consist of planar aromatic donor ( d ) and acceptor ( a ) molecules alternately arranged along a one - dimensional stack . a large distance between molecules of the same stack ( more than 3 @xmath0 ) gives a rather small hopping integral compared to the characteristic coulomb repulsion . moreover , a very large distance between neighbouring chains ensures the quasi - one dimensionality of these systems . a prototype is given by the anthracene - pmda ( pyromellitic acid dianhydride ) ; for this compound , by comparison between theoretical calculations on small clusters and experimental results of differential transmission spectroscopies , the existence of biexcitons - and even triexcitons - has been demonstrated recently @xcite . this result stays , for the moment , the only clear evidence for excitonic strings in organic compounds . _ j _ and _ h_-aggregates are also , in many cases , one dimensional stacks of organic molecules . but , on the contrary to the previous class of compounds , they have no ionic character and the attractive exciton - exciton interaction is smaller . they are typical organic solids in the sense that their low - lying excitations can be understood in terms of frenkel excitons . these compounds are claimed to be the ideal ones to get biexcitons but , without convincing demonstration for the moment - possibly be due to a too small binding energy @xcite . conjugated polymers are very complex compounds . as far as we are concerned with low - lying excitations , we may consider them , at first approximation , as an one dimensional interacting electron gas , which strongly interacts with the lattice too @xcite . moreover , for a more realistic description , in many cases , three dimensional effects and inter- and intra - chain disorder should be considered . because of these difficulties , despite a large amount of work , even the nature of the primary excitations remains subject of intense controversies @xcite . anyway , some photoinduced absorption ( pa ) experiments @xcite and two photon absorption ( tpa ) experiments @xcite show some features which are possibly due to biexcitons . only a few theoretical investigations of excitonic strings in organic compounds have been done @xcite . most of them are based on exact diagonalizations of small clusters using different kinds of model hamiltonians depending on the system under consideration @xcite . from these results and based on physical reasoning , it has been argued that , when biexcitonic states exist , the intensity of the optical one exciton - biexciton transition may decrease when the binding energy of the biexciton increases ; moreover , this intensity is shown to be less important than the one of the one exciton - two excitons ( two free excitons ) transitions and , even , is supposed to be independent of the system size for sufficiently large binding energies @xcite . therefore , if the tendencies observed on finite clusters remain the same for the infinite system , on the contrary to the situation observed in some inorganic semiconductors @xcite , it may be difficult to create biexcitons by tpa . a more appropriated experimental technique to observe excitonic strings would then be the differential transmission spectroscopy used by the way in @xcite . throughout this manuscript , we often refer to these numerical works for comparisons , especially concerning the intensity of the transition from one exciton to biexciton states . in this paper , we propose a model for studying excitonic complexes in organic compounds . based on physical arguments , we think , this model is relevant for the three classes of compounds enumerated above . for organic crystals , as the @xmath1 and @xmath2 aggregates , our model is equivalent , in the low energy sector , to an effective _ xxz _ spin one - half heisenberg model already used to study this problem @xcite . for the ct solids and cp , some extensions are necessary because the proper excitons are not of frenkel type . they are realized by introducing effective hard core boson particles with a finite extension on a lattice , which mimics the exciton extension . next , we propose an effective hamiltonian in terms of these particles which captures the essential physical ingredients to describe excitonic strings in organic compounds . this model is considerably easier to study than the traditionally employed models and therefore some analytical investigations are possible . in this paper , we study the proposed model for two and three particles in a finite and infinite lattice . our main interests concern the behaviours of the binding energies of the excitonic strings with the parameters of the model and of the one exciton - biexciton transition oscillator strengths ; are they giant as in inorganic semiconductors or , on the contrary , small as suggested by exact diagonalizations on small clusters @xcite ? this question is crucial in order to clarify the role of hypothetical biexcitons - or even more extended excitonic strings - in the photophysics of organic compounds . the paper is organized as follow . in section ii , we introduce our hard core boson hamiltonian . in section iii , iv and v , we study this hamiltonian for one , two and three particles respectively . first , for each class of systems mentioned in this study , we emphasize the main characteristics which are important for our purpose . _ j _ and _ h _ aggregates are typical organic crystals . the ground state of these systems is the tensorial product of the molecular ground states . the low - lying excitations are frenkel - type excitons where the hole and the electron of a monoexcitation are located on the same molecular site @xcite . we can draw a picture of these excitations in the following way : if we represent symbolically the ground state of such a quasi - one dimensional system as @xmath3 , where @xmath4 stands for a molecule in its ground state , the frenkel excitons are the bloch states of local configurations such as @xmath5 where the superscripts @xmath6 and @xmath7 are for the electron and the hole , respectively ; @xmath8 is then a monoexcited molecule . organic ct solids are stacks of donor ( d ) and acceptor ( a ) molecules . we already mentioned that they are , with a good approximation , quasi - one dimensional systems with a nearest - neighbour hopping term small compared to the coulombic term . the strong coupling regime is then a good starting point to consider such systems . moreover , if one consider only a short range coulombic term , their ground state would be a charge density wave depicted roughly by the unique configuration @xmath9 where d molecules get two electrons and a molecules none . the low - lying excitations would then be the excitons where an electron is transfered from a d to an a molecule ; they would be bloch states of excited configurations such as @xmath10 ( more details could be found in ref @xcite ) . we get a so - called charge transfer exciton ( cte ) . for a more realistic coulomb potential , the picture proposed above will remain reasonable , at first approximation . conjugated polymers are quite different from the other systems just considered since , obviously , they are not molecular crystals . on the contrary , they are usually described as an one dimensional electron gas governed by the pariser - parr - pople ( ppp ) hamiltonian which includes both long range coulomb interaction and , semi - classically , electron - phonon interaction @xcite . numerical studies of this hamiltonian give , for a proper choice of parameters , excitonic states of small radius , typically a few monomeric units @xcite . on the other hand , the ground state may be analyzed in terms of basis sets completely localized on the monomers . then , the ground state of finite clusters is shown to contain mainly intermonomer charge fluctuations of very short range - especially the nearest - neighbours ones @xcite . according to these specificities , an effective model has been proposed very recently @xcite . starting from the ppp hamiltonian , this model reaches a kind of molecular - crystal description for the low - lying excitations in conjugated polymers . in this new approach the vacuum ( or ground state ) is still not the simple molecular - type of ground state , proper to the simpson related models @xcite , but a matrix - product like state composed by local configurations extended at most , over two monomers . excitons are then composed mainly by local configurations where either the electron and the hole are localized on the same monomer or distant by only one monomer ; the weight of the configurations where the electron and the hole are more separated decreases exponentially in the wave functions . therefore , the radius of the exciton , @xmath11 , defined as an average distance between the hole and the electron is rather small . if we assume @xmath12 , the following pictorial representation can be done : if we represent the ground state as @xmath3 , where @xmath4 is a monomer , the excitons will be the bloch states of such configurations @xmath13 . because of the electron - hole symmetry , the charge transfer of the electron on the right side and on the left side must be combined either symmetrically or antisymmetrically . in contrast to the two previous cases , excitons have here , a structure . in conclusion , in these three cases , excitons are bloch states of very localized excitations . the vacuum and the nature of these local excitations vary , of course , following the system under consideration : crystal molecular ground state and frenkel excitons for @xmath1 and @xmath2 aggregates , cdw and cte for ct solids and finally , matrix - product state and some kind of cte but including the two possible arrangements of the charges , for cp . however , all these very localized excitations can be thought in terms of effective hard core bosons ( hcb ) extended over a few bonds - with an extension of the order of the exciton s radius . we adopt such a view in this work , and describe the dynamics of these excitations within the same model which we introduce now . in this study , we consider only one dimensional lattices with n sites , a site being a molecular - site in the case of molecular crystals@xcite and a monomer - site in the case of cp @xcite . as we already mentioned , this is a good approximation for a large class of compounds studied here . however , the generalization to higher dimensions should be possible and the subject of further studies . excitons are extended over @xmath11 sites , the radius of the excitons , which varies from zero to a few units in the cases considered here . in this study , to mimic the excitons , we introduce hcb particles extended over one bond ( @xmath12 ) . this minimal extension is adopted more by a pictorial preoccupation but , also , in accordance with the work developed in ref @xcite . indeed , at the thermodynamic limit where we will derive all the analytical formulae , this extension plays obviously no role , since it remains small compared to the system size . moreover , for finite size cluster calculations , the qualitative behaviours shown in this paper could also be obtained by choosing @xmath14 , @xmath15 or even other small @xmath11 . the restriction to @xmath12 does not affect the conclusions but , if necessary , more ( or less ) extended particles could also be considered without any difficulties . working with extended particles , instead of particles localized on a site , does not imply technical complications . the coulomb interaction between excitons must be included in order to describe excitonic strings . for that purpose , the relative position between the electron and the hole , the two elementary constituents of an exciton , becomes important . we introduce then two species of hcb extended over one bond : * the so - called right - bosons ( @xmath16 ) which mimic nearest neighbour electron - hole pairs with the electron on the right ( see figure [ rlbosons].a ) . * the so - called left - bosons ( @xmath17 ) which also mimic nearest neighbour electron - hole pairs but with the electron on the left ( see figure [ rlbosons].b ) . the @xmath16 and @xmath17 particles obey the following hamiltonian @xmath18 @xmath19 and @xmath20 are the destruction ( creation ) operators of the @xmath16 and @xmath17 bosons at site @xmath21 , @xmath22 , the operators number of hcb @xmath16 ( @xmath17 ) at site @xmath21 . @xmath23 is the excitonic energy ( @xmath24 ) . the @xmath1 term is the hopping term of the hcb ( @xmath25 ) ; it is directly related to the exciton band width given by @xmath26 . @xmath27 is an effective term for virtual interactions which couple locally @xmath16 and @xmath17 bosons ( @xmath28 ) . the @xmath16 and @xmath17 particles are effective representations of more complex particles , invoking a more sophisticated level of description ; these `` high level '' particles possess a structure which can be very complicated as it is the case for most of the cp . the interaction @xmath27 reflects the existence of virtual processes exchanging the relative position of the electron and the hole of an exciton ; this can be due , for instance , to kinetic terms . indeed , by a second order process involving a n.n . hopping term , it is readily possible to exchange the relative position of the two building particles of an exciton . the last term of ( [ hamiltonian ] ) , @xmath29 ( @xmath30 ) , is the interaction between excitons restricted to nearest - neighbours . we have to distinguish between two situations : for two excitons of the same species , @xmath31 or @xmath32 , the nearest charges are of opposite sign so that the resulting interaction must be attractive ; the situation is reversed for two excitons of opposite species , @xmath33 or @xmath34 , where the resulting interaction is repulsive . for structureless excitons , frenkel or ct - excitons , the definition of two species of degenerate hcb makes no sense . to study strings of such excitons with the general hamiltonian ( [ hamiltonian ] ) , it is sufficient to : ( i ) identify the @xmath16 and @xmath17 particles and ( ii ) to take the limit @xmath35 . to be more precise , @xmath16 and @xmath17 particles may exist in the case of ct solids , but with a large energy difference which results , in our model , in a very small @xmath27 ; therefore , it is a good approximation to simply neglect this term in this case . in the following , we will study the hamiltonian ( [ hamiltonian ] ) for one , two and three particles , which are the cases of experimental interest . excitons appear as bloch states of symmetric or antisymmetric combinations of @xmath16 and @xmath17 bosons , corresponding to two different symmetry classes @xmath36 where @xmath37 is the state without any boson . the separation in energy between this two states is given by @xmath38 , the @xmath39 state being lower . in the case of _ j _ ( _ h _ ) aggregates and ct solids , we identify @xmath16 and @xmath17 bosons and get only one excitonic state @xmath40 in all cases , exciton states result in a very intense peak in the linear absorption spectrum for @xmath41 , while the electron - hole continuum is almost not visible . this very unusual feature , if we think about conventional semiconductors , is an important characteristic of these one dimensional compounds @xcite . this is easily understandable since the radius of the exciton is very small and the ground state contains mainly short range charge fluctuations @xcite . now , we write down the elementary transition moments due to a localized hcb , @xmath42 , in the case of cp and @xmath43 , in the case of _ j _ ( _ h _ ) aggregates and ct solids ( @xmath44 is for organic crystal ) ; these quantities will serve in the following to express the transition moments between one exciton and two exciton states @xmath45 where @xmath46 is the charge of the electron and @xmath47 the position operator . the intensity of the ground state - one exciton transition at @xmath41 is proportional to @xmath48 or @xmath49 , depending on the system under consideration @xcite . in the case of cp , only the excitonic state @xmath50 is observable in linear absorption . however , the state @xmath51 could become important , as it is the case in electroabsorption experiments for instance . the wave functions with two excitons are written as @xmath52where @xmath53 and @xmath54 are @xmath16 or @xmath17 bosons , @xmath55 the ket relative to the situation where the particle @xmath53 is in @xmath56 and @xmath54 in @xmath57 with @xmath58 , @xmath59 is the corresponding amplitude . as usual , we treat separately the movement of the center of mass and of the relative position of the two particles . moreover , we have to specify the nature , @xmath16 or @xmath17 , of the particles located on @xmath56 and @xmath57 , it is then convenient to use matrix formalism . we write @xmath60_{n}\mid n_{1},n_{2}>\]]with @xmath61 ( @xmath62 ) and @xmath63_{n}=\left ( \begin{array}{l}\varphi_{rr}(n)\\ \varphi_{rl}(n)\\ \varphi_{ll}(n)\\ \varphi_{lr}(n ) \end{array}\right)\ ] ] where the small left ( right ) indices refer to the nature of the left ( right ) particle , @xmath16 or @xmath17 bosons . we apply periodic boundary conditions ( pbc ) to the wave functions which give the following constraints @xmath64_{-n+n}=[\varphi]_{n } \quad \mbox{where } \quad n = n_{2}-n_{1 } \end{array}\right.\ ] ] moreover , we work with hard core bosons extended over one link so that we must preserve @xmath63_{0}=[\varphi]_{1}=[0]\]]where @xmath65 $ ] is the zeroth four component vector . with the wave function ( [ wf ] ) , the eigenvalue equation is written @xmath66_{n}=-2j\cos q[i]\{[\varphi]_{n-1}+[\varphi]_{n+1}\}-\alpha[\sigma][\varphi]_{n}-\delta_{n,2}v[\lambda][\varphi]_{n}\]]where @xmath67 $ ] is the 4 by 4 identity matrix , and @xmath68=\left ( \begin{array}{c } 0 \quad 1 \quad 0 \quad 1 \\ 1 \quad 0 \quad 1 \quad 0 \\ 0 \quad 1 \quad 0 \quad 1 \\ 1 \quad 0 \quad 1 \quad 0 \end{array}\right ) \quad , \quad [ \lambda]=\left ( \begin{array}{c } 1 \quad \quad \quad \quad \quad \quad \\ \quad \quad -1 \quad \quad \quad \quad\\ \quad \quad \quad \quad 1 \quad \quad \\ \quad \quad \quad \quad \quad \quad -1 \end{array}\right)\ ] ] with the equation([vp ] ) , we have reached an impurity like problem where free particles , belonging to four different continuum , interact with impurities pinned at @xmath69 . first , by applying a local unitary transformation , we solve ( [ vp ] ) away from the impurities . @xmath70_{n}=[u][\varphi]_{n}\\ \quad \\ \left(\begin{array}{c}\varphi_{0}(n)\\ \varphi_{\bar{0}}(n)\\ \varphi_{+}(n)\\ \varphi_{-}(n)\end{array}\right ) = \left(\begin{array}{crlrlrl } \frac{1}{\sqrt{2 } } & & 0 & - & \frac{1}{\sqrt{2 } } & & 0 \\ 0 & & \frac{1}{\sqrt{2 } } & & 0 & - & \frac{1}{\sqrt{2 } } \\ \frac{1}{2 } & & \frac{1}{2 } & & \frac{1}{2 } & & \frac{1}{2 } \\ \frac{1}{2 } & - & \frac{1}{2 } & & \frac{1}{2 } & - & \frac{1}{2 } \end{array}\right ) \left(\begin{array}{c}\varphi_{rr}(n)\\ \varphi_{rl}(n)\\ \varphi_{ll}(n)\\ \varphi_{lr}(n ) \end{array}\right ) \end{array}\ ] ] we can now rewrite equation ( [ vp ] ) in terms of @xmath71_{n}$ ] ; we find @xmath72_{n}=-2j\cos q[i]\{[\bar{\varphi}]_{n-1}+[\bar{\varphi}]_{n+1}\}-2\alpha[\sigma_{2}][\bar{\varphi}]_{n}-\delta_{n,2}v\{[\sigma_{1}][\bar{\varphi}]_{n}+[\sigma_{3}][\bar{\varphi}]_{n}\}\]]where @xmath73=\left(\begin{array}{c}1 \quad \quad \quad \quad \quad \quad \\ \quad \quad -1 \quad \quad \quad \quad \\ \quad \quad \quad \quad 0 \quad \quad \\ \quad \quad \quad \quad \quad \quad 0\end{array}\right ) , \quad [ \sigma_{2}]=\left(\begin{array}{c } 0 \quad \quad \quad \quad \quad \quad \\ \quad \quad 0 \quad \quad \quad \quad \\ \quad \quad \quad \quad 1 \quad \quad \\ \quad \quad \quad \quad \quad \quad -1\end{array}\right ) , \quad [ \sigma_{3}]=\left(\begin{array}{c}0 \quad \quad \quad \quad \quad \quad \\ \quad \quad 0 \quad \quad \quad \quad \\ \quad \quad \quad \quad 0 \quad 1 \\ \quad \quad \quad \quad 1 \quad 0\end{array}\right)\ ] ] without the impurity like states at @xmath69 , there is no mixing between @xmath74 , @xmath75 , @xmath76 and @xmath77 components ; with the scattering of particles at @xmath69 , only the @xmath76 and @xmath77 components are mixed . from equations ( [ moment ] ) and ( [ unitary ] ) , it is clear that only the @xmath78 and the @xmath79 components of @xmath71_{n}$ ] are relevant to study the optical responses of two exciton states . since we are mainly interested in spectroscopic properties of biexcitons , we consider only these two components in the following . the @xmath80 and @xmath81 components have no physical meaning for structureless excitons , but could be important for cp ; their characteristics could be calculated , following exactly the same way as the one presented here . we write , following h. bethe @xcite , @xmath82which give the following expressions for the energy where @xmath83 and where we choose @xmath84 as reference @xmath85 @xmath86 , @xmath87 , @xmath88 and @xmath89 , @xmath90 and @xmath91 are the constants to be determined . note that , if in ( [ fonctions ] ) @xmath90 and @xmath91 are purely imaginary numbers , we get free excitons . on the contrary , if they are purely real numbers , we get bound - states , the biexcitons . the ansatz ( [ fonctions ] ) gives obviously the exact solution of ( [ vp2 ] ) without impurities ( @xmath92 ) ; with impurities , ( [ fonctions ] ) must fulfill the following constraints in order to remain solution of ( [ vp2 ] ) @xcite @xmath93 moreover , the pbc imply @xmath94 in the following , we give the solutions of the above mentioned problem for different choices of parameters . in this trivial case , the two excitons are free ; we get two degenerated continuum for @xmath78 and @xmath79 within the energy range @xmath95 $ ] . it is easy to show that a very intense linear absorption takes place for @xmath96 and @xmath97 only ; then the transition moment of the one exciton - two free exciton transition at the edge of the continuum behaves as @xmath98 plus corrections in @xmath99 . for @xmath100 and/or @xmath101 , this strong absorption remains but slightly shifted toward the high energies . this transition is similar to the ground state - one exciton transition ; indeed , the excitation energies and the behaviour of the transition moments are the same in these two cases . this characteristic was already used in @xcite to explain the lack of bleaching signal in pa experiments for charge transfer and frenkel exciton systems . this case concerns ct solids and _ j _ or _ h _ aggregates for instance . without the @xmath27 term , our model is equivalent to a standard interacting frenkel exciton hamiltonian already studied numerically @xcite and analytically @xcite . our results , obtained with a different technique , are in accordance with the conclusions reach by these various studies . since @xmath102 , bound states - or biexcitons - may exist ; in this subsection we consider the biexcitonic states only and then @xmath90 and @xmath91 of equations ( [ fonctions ] ) , as real quantities . when @xmath35 , equations ( [ energie ] ) imply that @xmath103 . from equations ( [ impurities ] ) together with the pbc ( [ pbc ] ) , we get at the thermodynamic limit , the following set of equations @xmath104where @xmath105 . additionally to the continuum , if @xmath106 , i.e @xmath107 , we get with the second equation of ( [ set ] ) and equations ( [ energie ] ) , two bound states with energies given by the following simple expressions @xmath108 by using a two - particle green function approach @xcite , these equations have already been obtained for @xmath96 in @xcite and for every @xmath109 in @xcite together with the biexciton condition @xmath110 . the first equation gives a bound state below the continuum of two free exciton states , this is the biexciton of interest for us ; the second equation gives another bound state on top of the continuum . in the following , we consider only the first state and calculate its wave function and the transition moment associated with the one exciton - biexciton transition . first , we write @xmath111 and @xmath112 . with the pbc ( [ pbc ] ) and equations ( [ set ] ) , we get the expression for @xmath113 @xmath114then , with equations ( [ fonctions ] ) and ( [ set ] ) , we get @xmath115)\]]valid in the interval @xmath116 . @xmath117 is the normalization constant given by @xmath118^{-\frac{1}{2}}\ ] ] the first equation of ( [ set ] ) shows that the wave function of the lowest bound state is a symmetric mixture of the @xmath76 and @xmath77 components resulting in configurations with two like particles @xmath31 or @xmath32 . on the contrary , the highest bound state contains only configurations with two unlike particles @xmath33 or @xmath34 . we recall that for this case @xmath16 and @xmath17 particles are identified so that the distinction between the case with @xmath31 or @xmath32 bosons and @xmath33 or @xmath34 bosons is then a way to treat the case of attractive and repulsive interaction between bosons at the same time . in case of attractive interaction ( two bosons of same species ) , we get a bound state below the two exciton continuum ; the situation is reversed in the case of repulsive interaction ( two bosons of opposite species ) , where a bound state appears above the two exciton continuum @xcite . for large systems , the transition moment of the one exciton - biexciton transition is non - zero only for @xmath96 and @xmath41 ( @xmath109 , the center of mass of the biexciton , @xmath119 , the exciton momentum ) ; its expression is then given with the help of the elementary transition moment ( [ moment ] ) by ( more precisely , the square of the transition moment is shown here ) @xmath120going to the thermodynamic limit , we found the simple result @xmath121which is exactly the result found in @xcite by using a two particle green function approach . we recall that the bound states ( biexcitons ) exist only if @xmath122 and that the biexciton binding energy increases when @xmath123 approaches @xmath124 . with this formula , the behaviour of the absorption from the one exciton to biexciton state appears very clearly . first les us consider its behavior when @xmath123 approaches @xmath124 or @xmath125 , at the thermodynamic limit ( equation ( [ momentthermo ] ) ) : for @xmath126 , the transition moment reaches saturation to @xmath127 which is easily understandable since , in this case , the two excitons are strongly bounded together in two n.n . sites ; for @xmath128 , the transition moment diverges following the asymptotic behaviour in @xmath129 . next , we can make some conclusions by studying the equation ( [ momentn ] ) . * at large and fixed @xmath130 , with decreasing @xmath123 ( which means , with increasing the biexciton binding energy ) , the intensity of the transition decreases . * at fixed @xmath123 and sufficiently large @xmath130 , @xmath131 is almost independent of the system size . on the contrary , the intensity of the transition between one exciton and two free excitons increases linearly with the system size ( cf . . [ momentfree ] ) . however , as expected , when @xmath123 approaches @xmath125 , @xmath132 recovers progressively a linear dependence in @xmath130 . both of these behaviours depict inverse mechanisms than the ones theoretically predicted for the ground state - one exciton transition @xcite , where the intensity of the excitonic peak is proportional to the size of the system and increases with the binding energy . these behaviours are illustrated in figure [ intensity ] where the equation ( [ momentn ] ) is evaluated for @xmath133 . there , we see that the intensity of the one exciton - biexciton transition decreases dramatically when @xmath123 decreases ( which means , when the binding energy increases ) reaching saturation for small @xmath123 . then , this intensity is independent of @xmath130 which gives in pa and tpa spectroscopy an intensity proportional to @xmath134 instead of @xmath130 for the transition to the continuum of two free excitons ; this last behaviour can be seen by comparison with equation ( [ momentthermo ] ) evaluated at the large @xmath130 limit ( dashed line ) . consequently , at the thermodynamic limit , biexcitons may be observable by spectroscopy experiments , but with the need of very clean compounds , otherwise we may expect at first analyse , that the intense transition to two free excitons would not permit to detect possible biexcitons . within our model , no two - photon giant resonance @xcite can be expected for @xmath1 , @xmath2 aggregates and organic charge transfer solids . the very same conclusions were already pointed out in ref @xcite from numerical calculations and in @xcite from analytical studies . this case concerns cp . as in the previous subsection , we look for bound states only so that @xmath90 and @xmath91 are again assumed to be real quantities . the question we want to address more specifically here is how the @xmath27 term will affect the binding energy , the wave function and the transition moment related to the biexcitons . for @xmath136 , from equations ( [ impurities ] ) and ( [ pbc ] ) , we get at the thermodynamic limit the following set of equations @xmath137 from equations ( [ energie ] ) and the second equation of ( [ set2 ] ) , we get @xmath138where @xmath105 and @xmath139 . this expression together with equations ( [ energie ] ) , gives the energy of the lowest biexciton . the binding energy of the biexciton , @xmath140 , is then given by the following equation @xmath141\]]the study of this expression shows that the biexciton binding energy decreases when @xmath27 increases ( see figure [ bindingenergy ] ) . this behaviour is not surprising since the @xmath27 term mixed together @xmath16 and @xmath17 particles ; by doing so , configurations with two unlike particles , @xmath33 or @xmath34 , appear in the wave function which increase its energy and then reduce the binding energy with respect to the case for @xmath142 . the critical value @xmath143 for which the binding energy becomes zero is given by the solution of the equation @xmath144 , @xmath145 the critical value , @xmath143 , increases when @xmath123 decreases , to diverge for @xmath146 : for a very large binding energy , @xmath27 has no effect on the bound states anymore . another interesting quantity along this critical line , is the critical value @xmath147 @xmath148 from this equation , we can see without any surprise , than @xmath147 - the critical value above which one gets biexcitonic states - continuously increases when @xmath149 increases . the attraction between excitons must be stronger when the mixing between @xmath16 and @xmath17 bosons becomes more important . we analyze now the wave function and the resulting transition moment associated with the one exciton - biexciton transition at @xmath96 . we proceed as in the previous case ; we write @xmath150 , @xmath151 and @xmath152 , @xmath153 and get with equations ( [ pbc ] ) and ( [ set2 ] ) @xmath154)+(-1)^{\delta_{\sigma_{1},\sigma_{2}}}e^{(\mu+\ln\lambda)(n-3)}\cosh([\mu+2\ln\lambda][n-\frac{n}{2}])\right)\ ] ] where again @xmath155 . @xmath156 is the normalization constant which could be easily calculated if necessary . the first term of ( [ wf2 ] ) comes from the @xmath76 component , the second term from the @xmath77 component . for @xmath35 , @xmath157 and the previous result is recovered . for @xmath158 , the weight of the @xmath77 component decreases smoothly in the wave function . consequently , the a priori unfavourable configurations @xmath33 or @xmath34 appear in the biexciton wave function , on the contrary to the previous case without the @xmath27 interaction . the transition moment of the one exciton - biexciton transition at @xmath96 can be easily calculated with the expression of the wave function ( [ wf2 ] ) , but contains no new information compared to what was already pointed out for @xmath35 . the results obtained in the previous subsection remain valid in this case . the only difference comes from the fact that the @xmath77 component is non active in linear absorption . consequently , for the same binding energy , if we assume @xmath159 , the biexcitonic peak would be a bit more intense for @xmath1 ( @xmath2 ) aggregates or ct solids than for cp . for instance , for @xmath35 , we get @xmath160 ; for @xmath158 , the @xmath76 becomes more important than the @xmath77 component in the wave function : the intensity of the one exciton - biexciton transition then increases . now , let us take reasonable parameters for conjugated polymers . extracting from experimental data ( we take roughly the binding energy of the @xmath161 exciton state observed in polydiacetylene compounds ) , we get @xmath162 @xcite . by calculating the coulomb interaction between two hcb using the mataga potential ( for instance ) @xcite and for a reasonable lattice constant ( @xmath163 ) , we get @xmath164 the @xmath27 term may be evaluated from the energy difference between the @xmath161 exciton and the @xmath165 `` ionic '' exciton . these states are represented here , in an effective way , by @xmath50 and @xmath51 respectively ( see eq . [ 1exciton ] ) . this quantity could be estimated , for instance , from exact resolutions of the ppp hamiltonian for small clusters or , in a better way , from electroabsorption experiments . there , the state @xmath166 shows a red - shift with no change of the spectral line and a quadratic dependence on the applied field . this quadratic stark effect of the exciton can be reasonably well reproduced with the state @xmath51 only , neglecting , in particular , the continuum effects @xcite . within this assumption , it is then very easy to estimate @xmath27 . we take here a reasonable value , @xmath167 . with this choice of parameters , we get a binding energy of @xmath168 for the biexcitons which is in accordance with experimental results @xcite . by the way , it is easy within our model to get such a value for the binding energy for other values of @xmath27 or / and @xmath1 by changing @xmath29 , which is more flexible . in figure [ finitesize ] , a calculation for a finite cluster of ten sites is shown . the parameters are @xmath169 and @xmath170 which give a typical spectrum . be is for the peak associated with the biexciton and , fe , for the peak associated with the more intense two free exciton state ( even if such concepts are not well defined for finite clusters ) . the be peak is a bit smaller than the fe one , which is in accordance with the results of ref . @xcite from exact calculations using the extended hubbard hamiltonian , a short version of the ppp hamiltonian . at the thermodynamics limit , only the fe peaks will survive , slightly shifted toward the low energies . to conclude with the two particle case , we may summarize our main results . for @xmath35 , biexcitons exist if @xmath122 ; the transition moment of the one exciton - biexciton transition shows inverse behaviour compared to the ground - state - one exciton transition : it decreases when the biexciton binding energy increases and , for sufficiently large @xmath130 , is independent of the system size . for @xmath171 , the conclusions aforementioned remain valid at the condition that @xmath172 for @xmath1 and @xmath29 fixed . above this critical value , biexcitons fall in the continuum . finite size cluster calculations give qualitatively similar results as other calculations performed with more complicated hamiltonians . for more than two particles , the question about the integrability of the hamiltonian ( [ hamiltonian ] ) becomes important . to answer this question , it is sufficient to analyze carefully the case with only two particles @xcite . for that purpose , let us first rewrite the hamiltonian ( [ hamiltonian ] ) in a more convenient way by introducing two new operators @xmath173 then , the hamiltonian becomes diagonal in @xmath27 @xmath174to illustrate this new picture , note that the states with one exciton ( [ 1exciton ] ) are the bloch functions of one @xmath175 or one @xmath176 particle . let us now consider the case with two like particles ( @xmath177 or @xmath178 ) located in @xmath56 and @xmath57 ( @xmath58 ) . because the last term of ( [ hamiltonian2 ] ) exchanges the color ( @xmath175 or @xmath176 ) of the two particles , the corresponding amplitude in the wave function is written as @xmath179the constants @xmath180 being complex numbers determined in the previous section by solving the schrdinger equation . the system is invariant by translation , hence the total momentum is conserved : @xmath181 . we then write @xmath182 , @xmath183 and @xmath184 , @xmath185 . now , let us do a scattering experiment. starting , for instance , with two @xmath175 particles , after the scattering processes we get two @xmath176 particles . the collision being elastic , the energy ( [ energie ] ) must be conserved @xmath186 depending on the @xmath27 value , we can make some conclusions . * if @xmath100 , @xmath187 ( see eq . [ wf2 ] for instance ) . hence , there is a change in the momenta during the scattering processes . in other words , diffractive processes are involved during the scattering which is sufficient to conclude that the model defined by ( [ hamiltonian ] ) ( or ( [ hamiltonian2 ] ) ) is non - integrable @xcite . * if @xmath35 , @xmath188 . there is no diffractive processes anymore ; then , as we look for the lowest bound - states , we will see that the model ( 1 ) may be expressed , in that case , as an effective spin-1/2 _ xxz _ hamiltonian which is well known to be integrable @xcite . rigorously , to conclude about the integrability of this model in the general case , the study of the two particles @xmath189 matrix has to be done @xcite . we will not attempt such study here . in the following two subsections , we will study both cases , starting with the simplest one where @xmath35 . for @xmath35 , by interacting with each other , particles do not exchange their color ; the interaction is reduced to an exchange of momenta alone . in such case , it is better to consider the hamiltonian ( [ hamiltonian ] ) expressed in terms of @xmath16 and @xmath17 particles . since @xmath16 and @xmath17 particles are hcb , an _ n_-electron configuration is subdivided into _ n _ separated regions . each particle stays in its own area of the chain ; within optical terminology , only reflections are allowed . the hamiltonian is studied independently for each configuration of color which corresponds to a configuration of pair - interactions , attractive or repulsive , depending on the nature of the neighbouring particles . in the general case , the problem is very complex . however , the interesting case for us is given by the configurations where all the particles have the same color . in this case , the model can be written in the form of a _ xxz _ ferromagnetic hamiltonian for spin-1/2 which has been solved exactly by using the bethe - ansatz @xcite . first , our goal is to show the equivalence of our model ( [ hamiltonian ] ) in the low energy sector and the _ xxz _ model . for simplicity , we consider here particles without extension ( @xmath14 ) . of course , it is possible to work with our defined @xmath16 and @xmath17 particles , but the comparison with the _ xxz _ model would then claimed some additional notations unnecessary for our purpose . by the way , in the second subsection of this section , we will show more formally the equivalence between the two models working with the extended particles . second , since we consider the simplest case with _ n_-particles like - for instance , the states with _ n _ _ r _ particles - the hamiltonian of the system may be expressed in terms of spin - one - half operators : @xmath190 being for a site @xmath21 without any particle and @xmath191 , for a site occupied by a @xmath16 particle ( an exciton ) . the model becomes then equivalent to the spin-1/2 @xmath192 model with an attractive anisotropic term @xcite @xmath193 the ground state of ( [ xxz ] ) is the state with only up - spins , @xmath194 . @xmath195 creates a down - spin ( a @xmath16 particle or exciton ) localized at site @xmath21 , @xmath196 . we introduce here , as usual , the parameter of anisotropy @xmath197 @xcite . this model has been subject of intense studies since several decades , and many results are known about it @xcite . concerning the bound states , or more precisely , the so - called `` string - states '' , their properties depend on the anisotropic parameter @xmath198 . within the `` string hypothesis '' , the following results were found@xcite . * if @xmath199 , the bound states of @xmath21-spins , here @xmath21-excitons , exist without any restriction on @xmath21 @xcite . the energy of a @xmath21-excitonic string is then given by the following formula first appeared in @xcite @xmath200with @xmath201 and @xmath109 being the wave number of the n - string center of mass . for @xmath69 , the energy ( [ energiebiexciton ] ) is of course recovered . * if @xmath202 , the @xmath21-excitonic strings may exist but with a strong restriction on @xmath21 @xmath203with @xmath204 , @xmath205 . moreover , these excitonic strings fall in the continuum spectrum . we have seen in the previous section , that biexcitons exist if @xmath199 . in this case , from the above conclusions about the _ xxz_-model , we can say that @xmath21-excitonic strings will also exist and that their binding energies will increase with @xmath21 ( [ strings ] ) ; this result is in agreement with numerical studies on small clusters @xcite . however , as we have shown in detail for the biexcitonic case , it is not easy to observe such states experimentally , since the particles are tightly bounded inside the complex , resulting in a lack of oscillator strength . for @xmath206 , excitonic strings may still exist but with an energy falling in the continuum . in this limit , the excitons are less bounded and , therefore , it could be interesting to study the signatures of these states under the influence of an electric field . we leave this question to further work . for @xmath100 , the interactions between particles involve both exchange of momenta and exchange of color . by a simple argument , showing the appearance of diffractive terms in the two particles scattering case , the non - integrability of the model ( [ hamiltonian2 ] ) has been shown . therefore , to study _ n_-excitonic strings with @xmath207 one needs to do some approximations . in the following , we propose a trial wave function for the three particle case build from the bethe ansatz solution of the _ xxz_-heisenberg model within the string hypothesis . let us consider the expression ( [ hamiltonian2 ] ) of our model expressed in terms of @xmath176 and @xmath175 particles . for the three particle case , eight configurations have to be distinguished depending on the color of the particles and on their relative positions ; we denote these configurations as @xmath208 , where @xmath209 is for the color ( @xmath176 or @xmath175 ) . these configurations are separated into two disconnected channels : on one hand , we have @xmath210 , @xmath211 , @xmath212 and @xmath213 ; on the other hand , @xmath214 , @xmath215 , @xmath216 and @xmath217 . we consider only the latter channel which is the lowest in energy . the wave function for the triexciton is expressed as a linear combination @xmath218where @xmath219 stands for the @xmath214 , @xmath215 , @xmath216 and @xmath217 states respectively ( @xmath220 for _ right _ , _ center _ and _ left _ refer to the position of the @xmath175 particle in the three particles complex ) . @xmath221 are the coefficients to determine . every component @xmath222 is assumed to have the bethe ansatz structure @xmath223where , more precisely , the particles are again separated by one bond , at least , and with @xmath224the @xmath225 are the quasi - momenta and the @xmath226 the phase terms @xcite . the summation runs over the @xmath227 permutations of the indices . we apply periodic boundary conditions to the wave functions , @xmath228 , and we get the well known equations @xcite @xmath229 since we are looking for excitonic complexes , we are interested in complex solutions of these equations . following ovchinnikov @xcite , we assume that the only relevant phase terms are @xmath230 and @xmath231 , then @xmath232where @xmath233 are integers between @xmath124 and @xmath234 . it follows @xmath235 next , we assume @xmath236 and @xmath237 without loss of generality ; then , at the thermodynamic limit , among the summation ( [ ba ] ) only one term remains with an exponential accuracy in @xmath130 @xmath238 we may notice , that the used assumption about the phase terms @xmath226 , is rigorously correct in the case without @xmath27 @xcite . the wave function ( [ wf3 ] ) with the simplification ( [ wfstring ] ) , must satisfy the following constraints due to the @xmath29 term , in order to be an eigenfunction of ( [ hamiltonian2 ] ) : * for @xmath239 , ( @xmath240 ) @xmath241 * for @xmath242 , ( @xmath243 ) @xmath244 moreover , by symmetry , we have @xmath245 , @xmath246 and @xmath247 . then , the previous set of equations together with the expression ( [ wfstring ] ) gives @xmath248-ve^{ik_{2}^{0}}\frac{v}{j}\frac{e^{ik_{2}^{l}}}{e^{i(k_{1}^{l}+k_{2}^{l})}+1}=0\\ j[e^{i(k_{2}^{0}+k_{3}^{0})}+1]-ve^{ik_{3}^{0}}\frac{v}{j}\frac{e^{ik_{3}^{l}}}{e^{i(k_{2}^{l}+k_{3}^{l})}+1}=0\\ j[e^{i(k_{1}^{l}+k_{2}^{l})}+1]-ve^{ik_{2}^{l}}\frac{v}{j}\frac{e^{ik_{2}^{c}}}{e^{i(k_{1}^{c}+k_{2}^{c})}+1}=0\\ j[e^{i(k_{2}^{l}+k_{3}^{l})}+1]-ve^{ik_{3}^{l}}\frac{v}{j}\frac{e^{ik_{3}^{c}}}{e^{i(k_{2}^{c}+k_{3}^{c})}+1}=0 \end{array } \right.\]]to which we add the momentum conservation law @xmath249@xmath109 being the momentum associated with the motion of the center of mass of the excitonic complex . this system of equations is obviously non - soluble in the general case . hence , we add an additional simplification to the trial wave function . if @xmath35 , @xmath250 , and therefore @xmath251 ; it is easy to verify that the system ( [ system ] ) becomes equivalent , in that case , to the corresponding equations for the _ xxz _ heisenberg model @xcite as we already pointed out in the previous subsection . then the system ( [ system ] ) leads to the energy ( [ strings ] ) , for @xmath199 . within the string hypothesis , @xmath252 and @xmath253 are complex conjugate , @xmath254 , a purely real quantity . enlightened by the exact results for @xmath35 @xcite , we assume for @xmath100 @xmath255where @xmath256 , a real number , is given by the following equation obtained from the energy conservation law @xmath257the left - hand - side is the energy of the @xmath214 component , the right - hand - side , the energy of the three other components . with this assumption on the momenta @xmath225 , the solutions stay very close to the bethe ansatz form ; the binding between the two bordered particles of the complex ( with momenta @xmath258 and @xmath259 ) , ensures the cohesion of the three particle bound states . the set of equations composed by the two first equations of ( [ system ] ) and equations ( [ q ] ) , ( [ ansatz ] ) and ( [ epsilon ] ) are then solved numerically . it is important to realize that our approximation is variational . this important statement could be checked in the following way . one starts with the trial wave function given by ( [ wf3 ] ) where the components @xmath222 are assumed to be of the form defined by the equations ( [ amplitude ] ) , ( [ wfstring ] ) and ( [ ansatz ] ) together with the restrictions ( [ q ] ) and ( [ epsilon ] ) . the coefficients @xmath221 of ( [ wf3 ] ) , and their complex conjugates , are determined variationally by minimizing the functional @xmath260where @xmath261 is the lagrange parameter for the normalization constraint of the wave function . we get @xmath262and similar equations for all the other minimizations . at the end of the day , we obtain nothing else than the usual secular equations and one way to solve them , following a. bethe @xcite , is to impose the constraints given by the sets of equations ( [ constraints1 ] ) and ( [ constraints2 ] ) . our procedure is then variational and the results found are an upper bound for the problem . typical results are shown in figure ( [ bindingenergy ] ) . the binding energy of the biexciton got in the previous section and the one of the triexciton , obtained with our variational calculation , are represented for @xmath96 and @xmath169 as a function of @xmath263 . the triexciton binding energy is defined with respect to the continuum of one biexciton plus one free exciton which is the lowest in energy . the following remarks can be made . ( i ) the triexciton binding energy for @xmath35 is the exact one . ( ii ) the binding energy of the triexciton is larger than the one of the biexciton . ( iii ) a critical value of @xmath27 exists also for the triexciton ; it is larger than @xmath143 ( [ alphac ] ) for the biexciton . it follows that triexcitons may exist without biexcitons ( in the sense that @xmath264 ) . we may suspect the same behaviour for the binding energy of n - string for any @xmath21 : the binding energy and the critical @xmath27 may be larger for a n - string than for a ( n-1)-string in accordance with the results for the _ xxz _ model . ( iv ) since our result is variational , we get an upper bound for the triexciton binding energy so that our conclusions are qualitatively correct . to illustrate the behaviour of the binding energy of a _ n_-string at strong @xmath27 , we study this limit perturbatively . for @xmath265 , the hilbert space of ( [ hamiltonian2 ] ) is naturally separated in two subspaces : the first , lower in energy , with configurations without @xmath176 particles and , the second , higher in energy , with configurations with at least one @xmath176 particle . the first one is associated with the projector @xmath266 , the second one with the projector @xmath267 ( @xmath268 ) . the hamiltonian may be written as @xmath269 by a schrieffer - wolf type of canonical transformation @xcite , we want to derive an effective hamiltonian @xmath270 , with @xmath271 , @xmath272 being the effective term to be determined . the study of the n - excitonic strings will then be performed by diagonalizing the projection of this model into the space without @xmath176 particle , @xmath273 . introducing the unitary transformation @xmath274 ( with @xmath275 ) , and following the method described in detail in @xcite , we found the simple effective hamiltonian valid in the strong coupling limit where @xmath265 , at second order in @xmath276 @xmath277 we have recovered , once again , the spin-1/2 _ xxz _ hamiltonian ( [ xxz ] ) . however , the anisotropic parameter is now given by @xmath278 , instead of @xmath279 , and the interpretation of the spin operators is different : @xmath190 is still for a site without any particle but , @xmath191 , is now for a site occupied by a @xmath175 particle instead of a @xmath16 particle . in the effective model , the attractive interaction is reduced by the @xmath27 term and , for the strong limit ( @xmath280 ) where this derivation is valid , one reaches a _ xx_-heisenberg hamiltonian which is well known to be equivalent to a spinless free fermion system @xcite . hence , in this limit , the excitonic strings do not exist which is consistent with the results of our variational calculation for @xmath281 . the states with free @xmath175 particles are then preferred . in this paper , we propose a simplified model to describe excitonic strings in quasi - one dimensional organic compounds including organic charge transfer solids , _ j_- and _ h_-aggregates and conjugated polymers . for all these compounds , the excitonic states are characterized by small radius . hence , our model starts with the definition of hard core boson particles extended over one bond which describe , in an effective way , the excitons in one dimensional organic compounds . there are two possibilities in one dimension to place the electron and the hole , the two elementary constituents of an exciton . either the electron is on the right or on the left of the hole ; we introduce two kinds of hcb which illustrate these two situations : the _ right_-bosons and the _ left_-bosons . the proposed model in terms of these effective particles contains four terms : the excitonic energy ( @xmath23 ) , which separate in energy the space with @xmath21 particles from the space with @xmath282 particles ; the kinetic energy of the hcb ( @xmath1 ) ; the local interaction between the two species of bosons ( @xmath27 ) , which illustrates the fact that there exist effective interactions which can exchange the relative position of the hole and the electron of an exciton ; and last , the interaction between two excitons which are attractive between particles of the same species and repulsive between particles of opposite species ( interaction of intensity @xmath29 ) . the kinetic energy and the @xmath27-interaction combined with the repulsive two body interaction act against the attractive two body interaction - they tend to delocalize the excitons on the contrary to the two particle attraction which tends to create bound states . the cases with two and three particles are mainly studied in this work . we summarize briefly our results considering the case where @xmath109 , the momentum of the center of mass of the excitonic complexes , is zero . the case with two particles is solved exactly . for @xmath35 , biexcitonic states exist if @xmath283 . then , the intensity of the transition from the one exciton state decreases dramatically with the increase of the binding energy , to saturate and become independent of the system size . this behaviour makes the biexcitonic state difficult to observe at the thermodynamic limit . for @xmath100 , a critical value @xmath143 is found , above which biexcitonic states do not exist otherwise , the behaviour of the oscillator strength remains qualitatively unchanged . for @xmath35 , the @xmath21 particle case can be solved in the low energy sector . then , the model is equivalent to the spin one half _ xxz _ heisenberg model which is exactly solvable in one dimension . for @xmath283 , it is well known that @xmath21-excitonic strings exist with a binding energy which increases with @xmath21 @xcite . for @xmath284 , the model is non - integrable . the three particle case is then studied using a variational ansatz , which may be extended to a more general case with @xmath21 particles . again , a 3-string is found with a binding energy larger than the one for the biexciton . we also found a critical value for @xmath27 , larger than the @xmath143 valid for the 2-strings . from numerical studies on small clusters @xcite , it has been argued that _ n_-strings may exist for any value of _ n _ in organic compounds . our work gives some confidence to this statement . indeed , with our hcb hamiltonian , if biexcitons exist , triexcitons will exist with a larger binding energy ; moreover , if @xmath35 , _ n_-strings will also exist with binding energies which increase with _ n _ and , we believe , from our present results , that it will also be the case for @xmath100 . the next important question concerning the _ n_-string peaks in pa and tpa experiments , is fully clarified for the biexciton ( the most important case ) and confirms some conclusions of previous works @xcite . the intensity of the transition from the one exciton to the biexciton decreases with the binding energy and , for sufficient binding energy , becomes independent of the system size . hence , the pa intensity to the biexciton is proportional to @xmath134 instead of @xmath130 for the two free exciton case . this property comes from the fact that the two excitons are tightly bounded within the complex , rendering the 2-string states difficult to observe for infinite systems ( clean samples are needed ) and more accessible experimentally for small oligomers @xcite . in the case of conjugated polymers , we think some additional studies are needed going in two different directions . first , in the context of pa experiments , excitations extended over two molecules have been invoked @xcite ; extension of our work to two dimensions or , at least , for two coupled polymers would be suitable . second , conjugated polymers are disordered systems with several possible sources of disorder which result , in practice , in the vague definition of the so - called conjugation length @xcite . as we have already mentioned , since the transitions to biexcitons are much less intense than the ones involving two free excitons , we may expect at first analyse - and it was implicitly our point of view all along this manuscript - that a sufficient disorder will render the biexcitons not observable . however , some relatively recent results about the two interacting particle problem in a disordered medium @xcite show very unexpected behaviours which may have dramatic consequences for our problem , and possibly in a way reverse to our intuition . in the light of the intriguing results of @xcite , the effects of disorder on transition moments should be clarified for a correct interpretation of pa and tpa results . we leave these considerations to further works . i would like to thank prof . ovchinnikov for his support and illuminating discussions . this work would certainly not be developed in the same way without his help . this work has been supported by the european commission through the tmr network contract erbfnrx - ct96 - 0079 ( qucex ) . m. ueta , h. kanzaki , h. kobayashi , y. toyozawa and e. hanamura , excitonic processes in solids ( springer , berlin , 1986 ) . e. hanamura , solid state comm . * 12 * , 951 ( 1973 ) . g.m . gale and a. mysyrowicz , lett . * 32 * , 727 ( 1974 ) . s. mazumdar , f. guo , k. meissner , b. fluegel , n. peyghambarian , m. kuwata - gonokami , y. sato , k. ema , r. shimano , t. tokihiro , h. ezaki and e. hanamura , j. chem . phys * 104 * , 9283 ( 1996 ) . h. ezaki , t. tokihiro and e. hanamura , phys . rev . * b 50 * , 10 506 ( 1994 ) . a. chakrabarti , a. schmidt , v. valencia , b. fluegel , s. mazumdar , n. armstrong and n. peyghambarian , phys . rev . * b 57 * , r4206 , ( 1998 ) . d. baeriswyl , d.k . campbell and s. mazumdar , in _ conjugated conducting polymers _ , edited by h. kiess ( springer - verlag , heidelberg , 1992 ) , pp7 - 133 . , edited by n.s . sariciftci ( world scientific publishing , singapore , 1997 ) . sinclair , d. mcbranch , t.w . hagler and a.j . heeger , synth . met . * 49 - 50 * , 593 ( 1992 ) . klimov , d.w . mcbranch , n. barashkov and j. ferraris , phys . rev . * 58 * , 7654 ( 1998 ) . townsend , w.s . fann , s. etemad , g.l . baker , z.g . soos and p.c.m . mcwilliams , chem . 180 * , 485 ( 1991 ) . v.a . shakin and s. abe , phys . * b 50 * , 4306 ( 1994 ) . f. guo , m. chandross and s. mazumdar , phys . lett . * 74 * , 2086 ( 1995 ) . g. vektaris , j. chem . phys . * 101 * , 3031 ( 1994 ) . f.c . spano , chem . * 234 * , 29 ( 1995 ) . spano and e. manas , j. chem . phys . * 103 * , 5939 ( 1995 ) . wu , x. sun and k. nasu , phys . rev . * b52 * , 4849 ( 1995 ) . f.b . gallagher and f.c . spano , phys . b 53 * , 3790 ( 1996 ) . davydov , theory of molecular excitons , new york , 1971 . m. chandross , y. shimoi and s. mazumdar , phys . b 59 * , 4822 ( 1999 ) . s. pleutin and j.l . fave , j. phys . * 10 * , 3941 ( 1998 ) . s. pleutin , in preparation . simpson , j. am . soc . * 77 * , 6164 ( 1955 ) . m.j . rice and y.n . gartstein , phys . lett . * 73 * , 2504 ( 1994 ) . a. horvath , g. weiser , c. lapersonne - meyer , m. schott and s. spagnoli , phys b 53 * , 13507 ( 1996 ) . h. bethe , z. physik * 71 * , 205 ( 1931 ) . b. sutherland , synth . met . * 84 * , 11 ( 1997 ) . y.a izyumov and y.n . skryabin , `` statistical mechanics of magnetically ordered systems '' , consultants bureau , new - york and london ( 1988 ) . m. takahashi , in `` conformal field theories and integrable models '' , springer - verlag , 1997 ( berlin ) . m. gaudin , `` la fonction donde de bethe '' , masson ( 1983 ) . ovchinnikov , jetp lett . * 5 * , 38 ( 1967 ) . schrieffer and p.a . wolf , phys . rev . * 149 * , 491 ( 1966 ) k.a . chao , j. spalek and a.m. oles , j. phys c : solid state phys * 10 * , l271 ( 1977 ) . e. lieb , t. schultz and d. mattis , annals of phys . * 16 * , 407 ( 1961 ) . hsu , m. yan , t.m . jedju , l.j . rothberg and b.r . hsieh , phys . * b 49 * , 712 ( 1994 ) . dorokhov , jetp * 71 * , 360 ( 1990 ) ; d.l . shepelyansky , phys . lett . * 73 * , 2607 ( 1994 ) .

instabilities in plastic flow has been an object of attention for a long time in metallurgical literature . conceptually simplest form of this [ lubahn & felgar , 1961 ; hall , 1970 ] manifests when the material is subjected to a creep test wherein a force is applied and the response in the form of elongation of the material is measured , which under normal conditions , is a smooth strain - time curve . however , under certain metallurgical conditions , one sees steps on creep curve suggesting a form of instability [ lubahn & felgar , 1961 ; hall , 1970 ; da silveira & monteiro , 1979 ] . a better known form of the instability having the same physical origin , but conceptually more difficult , appears when the material is deformed under tensile deformation [ bodner & rosen , 1967 ; brindley & worthington , 1970 ; penning , 1972 ] . here , the material is subjected to a predetermined response namely a constant rate of deformation , and the force or the stress developed in the sample is sought to be measured . even in this case , one finds a smooth stress - strain curve under normal conditions . however , when the system is in the regime of instability , for some values of material parameters the stress - strain curve exhibits multiple load drops . each of the load drops is related to the formation and propagation of dislocation bands [ chihab _ et al _ , 1987 ] . it is in the latter type of testing where plastic instability manifests much more easily than in the former . the phenomenon is referred to as the portevin - le chatelier ( plc ) effect or the jerky flow and is seen in several metals such as commercial aluminium , brass , alloys of aluminium and magnesium [ brindley & worthington , 1970 ] . it is observed only in a window of strain rates and temperature . in contrast , the phenomenon of steps on creep curve , which is the subject of the present discussion , is seen in few instances [ da silveira & monteiro , 1979 ; zagarukuyko _ et al _ , 1977 ; stejskalova _ et al _ , 1981 ] . the reason attributed to this is that it is difficult to obtain proper control on metallurgical parameters wherein this form of instability can be observed . there is one more form of instability which manifests under constant stress test where similar strain jumps are seen as a function of time . the origin of instability in all these three modes of testing is known to be the same . ( for a current status of both experiments and theory see kubin _ et al _ , 1993 ) . eventhough steps on creep curve are seen in a limited number of experiments , it is often straight forward to translate many of the experimental results of constant strain rate test into that of creep test . it is generally agreed that the microscopic origin of the instabilities arises due to the interaction of dislocations with mobile point defects and is referred to as dynamic strain ageing . this leads to negative strain rate characteristic . the basic idea was formulated by cottrell [ 1953 ] few decades ago . the early phenomenological models are all static since they do not deal with time development . in contrast , methodology of dynamical systems addresses precisely this aspect . until about a decade and a half ago , there were no models which looked at the problem from the point of view of bifurcation theory . an attempt to understand the problem in the above perspective was first made by our group several years ago [ ananthakrishna & sahoo , 1981b ; ananthakrishna & valsakumar , 1982 , 1983 ; valsakumar & ananthakrishna , 1983 ] . in a series of papers starting from an extended fokker - planck equation for the distribution function of the velocity of dislocation segments , and then splitting this into a mobile and an immobile component , we arrived at a model which consisted of three types of dislocations and some transformations between them [ ananthakrishna & sahoo , 1981a , 1981b ; sahoo & ananthakrishna , 1982 ; valsakumar & ananthakrishna , 1983 ] . the basic idea could be summarized by stating that the phenomena is due to a hopf bifurcation resulting from nonlinear interactions between three different types of dislocations , suggesting a new mathematical mechanism for the instability . eventhough the spatial inhomogeneous structure was ignored and only the temporal oscillatory state was sought to be described , the model and its extensions to the case of constant strain rate test , proved to be very successful in that it could explain most of the experimentally observed features ( for critical reviews see kubin & lepinoux , 1988 and kubin _ et al _ , 1993 ) . _ for instance , the existence of negative strain rate sensitivity in a regime of strain rates , comes out naturally as a consequence of the hopf bifurcation _ [ ananthakrishna & valsakumar , 1982 ] . it must be emphasized , that this feature has been measured in most experiments on the plc effect , and is _ assumed _ in all other theoretical models [ cottrell , 1953 ; bodner & rosen , 1967 ; penning , 1972 ; kubin and estrin 1990 ; see also kubin & martin , 1988 ; martin & kubin , 1992 ; kubin _ et al _ 1993 ] . other experimentally observed features such as the existence of bounds on strain rate for the plc effect to occur , the existence of critical strain and its dependence on applied strain rate , the dependence of the amplitude on the strain etc . , also follow . _ one other important prediction which is direct consequence of the dynamical basis of the model is the existence of chaotic stress drops in a window of strain rates _ [ ananthakrishna & valsakumar , 1983 ; ananthakrishna & john , 1990 ] . _ recently this prediction has been verified which in turn implies that only a few degrees of freedom are required for a dynamical description of the phenomenon . this offers justification for the use of only a few degrees of freedom for the description of the temporal aspect , eventhough the system is spatially extended . _ [ ananthakrishna _ et al_. , 1995 ; ananthakrishna & noronha , 1995 ; quaouire & fressengeas , 1995 ; venkadesan _ et al _ 1995 ; noronha _ et al _ 1996 & 1997 ] . ( note that a spatially extended system implies infinite degrees of freedom . ) description of the phenomenon which includes the initiation and propagation of the bands during the plc effect has also been recently attempted [ ananthakrishna , 1993 ] . since the introduction of bifurcation theory into this field several years ago by our group , there has been a resurgence of interest in plastic instabilities in the light of introduction of new methodology borrowed from the theory of dynamical systems . this has further helped to obtain new insights hitherto not possible [ kubin & martin , 1988 ; estrin & kubin , 1989 ; kubin & estrin , 1990 ; see also kubin & martin , 1988 ; martin & kubin , 1992 ; kubin , 1993 ; ananthakrishna _ et al_. , 1995a & b and references therein ] . one of the aims of such theories is to be able to relate the microscopic dislocation mechanisms to the measurable macroscopic quantities . however , in the process , we feel that finer aspects of dynamical systems have been glossed over in this field . for instance , one often finds that casual remarks are made about fast and slow modes without actually going through the procedure of demonstrating the existence of such modes and eliminating the fast modes in favour of the slow ones [ aifantis , 1988 ; hahner & kubin , 1992 ; hahner , 1993 ] . in addition , under the adiabatic elimination , the slow modes are complicated functions of the original modes . yet , hand waiving arguments have been used in building models which we believe are technically suspect . the purpose of this article is at least three fold . first , we wish to illustrate the power and utility of dynamical methods to the study of dynamical aspects of the model . we demonstrate the technical aspects of adiabatic elimination ( largely addressed to metallurgists working in the area of plastic instabilities ) and then derive the equation for the slow modes having the form of time - dependent ginzburg - landau ( tdgl ) equation . second , we wish to compare the results of the dependence of the amplitude and the period of the strain jumps on stress and temperature with experiments on creep curve as also with the results translated from the constant strain rate case to the creep case . this will allow us to relate the theoretically introduced parameters to the macroscopically measured quantities . third , we wish to compare these results and the detailed numerical results via power series expansion with the approximate solutions for the limit cycles obtained earlier [ ananthakrishna & sahoo , 1981b ; valsakumar & ananthakrishna , 1983 ] . the principal results of this work are as follows . we show that the derived quintic tdgl equation for the slow modes is valid in the mid - range of the instability domain . within this range , there is a narrow sub - range of the parameter where the system exhibits a supercritical bifurcation , beyond which subcritical bifurcation is exhibited . the results of this calculations on the amplitude and period of the strain jumps are compared with experiments . in addition , analytical work based on the tdgl equation and a detailed comparative numerical study via the power series solution shows that the model exhibits unusual properties in a certain domain of the parameter . the study further shows that very high order nonlinearities actually govern the nature of the subcritical bifurcation beyond the domain handled by the quintic tdgl equation . the plan of the paper is as follows . in what follows ( section 2 ) we present a brief summary of the model . in section 3 , the technique of extracting the tdgl equation using reductive perturbative approach is applied to the model . in section 4 , approximate limit cycle solution obtained through the tdgl equation is compared with experiments and with the numerical solution of the model as well as earlier studies on the model . section 5 contains summary and discussion . we start with a brief summary of the model . the basic idea of the model is that instabilities in plastic flow are a consequence of nonlinear interactions between different types of dislocation populations . spatial dependence is ignored and only temporal aspects are described , the idea being that the few degrees of freedom used for describing the phenomenon ( dislocation populations ) correspond to the collective modes of dislocations in the spatially extended system . as mentioned in the introduction , this finds support from the recent experimental verification of chaos in the plc effect . the details of the model can be found in the original references [ ananthakrishna & sahoo , 1981b ; valsakumar & ananthakrishna , 1983 ] . the model consists of mobile dislocations @xmath0 and immobile dislocations @xmath1 and another type which mimics the cottrell s type @xmath2 , which are dislocations with clouds of solute atoms . let the corresponding densities be @xmath3 , @xmath4 and @xmath5 , respectively . the basic dislocation mechanisms included are the following : ( a ) production of dislocations by cross glide ( @xmath6 , @xmath7 is the cross glide coefficient , @xmath8 is the velocity of mobile dislocations ; @xmath9 [ alexander , 1986 ] where @xmath10 is the applied stress ) , ( b ) immobilization of two mobile dislocations ( @xmath11 , @xmath12 is the rate at which two mobile dislocations get immobilized ) and annihilation of two mobile dislocations ( @xmath13 ) , and annihilation of a mobile with an immobile dislocation ( @xmath14 ) , ( c ) once a mobile dislocation is immobilized , it can be reactivated by athermal or thermal means ( @xmath15 ) , ( d ) lastly , a mobile dislocation can acquire a cloud of solute atoms and move with them . we consider such dislocations as distinct from both the mobile as well as the immobile and denote them @xmath2 . this process is represented by @xmath16 , @xmath17 being a function of the concentration of solute atoms . however , as the solute atoms gather progressively around dislocations , these dislocations eventually will be immobilized ( @xmath18 ) . ( note that in principle , @xmath12 , @xmath19 and @xmath20 are different . ) these mechanisms lead to the rate equations for the densities of dislocations : @xmath21 ( for the sake of simplicity , we take @xmath22 and @xmath23 . ) one can give a more transparent interpretation for eq . ( 3 ) which represents whole process of acquiring a cloud of solute atoms by a mobile dislocation and eventual immobilization . this can be seen by defining @xmath24 , where the kernel @xmath25 which represents the fact that solute atoms are arriving at a certain rate has been chosen to be @xmath26 . it can be easily checked that eq . ( 3 ) is the differential form of the definition of @xmath5 . one last remark about some of the terms appearing in these equations . the first two terms plus the last term in eq.(1 ) , and equivalent terms in eqs.(2 ) , were derived starting from a velocity distribution function of the dislocation segments [ ananthakrishna & sahoo 1981a ; sahoo & ananthakrishna , 1982 ] . thus , this model is an improved model to explain the instability in plastic flow . equations ( 1 - 3 ) can be cast into a dimensionless form by using scaled variables : @xmath27 to get @xmath28 where @xmath29 . the dot represents differentiation with respect to @xmath30 . equations ( 5 - 7 ) are coupled set of nonlinear equations which support limit cycle solutions for a range of parameters @xmath31 and @xmath32 , that are physically relevant . @xmath33 refers to the concentration of the solute atoms , @xmath34 refers to the reactivation of immobile dislocations and @xmath35 to the time scales over which the slowing down occurs . we have demonstrated the existence of limit cycle solutions and also obtained approximate closed form solutions for the limit cycles [ ananthakrishna & sahoo , 1981b ; valsakumar & ananthakrishna , 1983 ] . in addition , the model has been studied numerically . using the orowan equation which relates the rate of change of strain(@xmath36 ) to mobile dislocation density and the mean velocity : @xmath37 , with @xmath38 as the burger s vector , steps on the creep curve follow automatically since the densities of dislocations are oscillatory . several experimental results are reproduced [ ananthakrishna & sahoo , 1981b ; valsakumar & ananthakrishna , 1983 ] . for the sake of completeness , we outline the reductive perturbative approach . near the point of hopf bifurcation of the system ( eqs.5 - 7 ) , corresponding to a critical value of the drive parameter , a pair of complex conjugate eigenvalues and another real negative eigenvalue exist for the linearized system of equations around the steady state . as we approach the critical value from the stable side , the real parts of the pair of complex conjugate eigenvalues approach zero and hence the corresponding eigenvectors get slower and slower . in contrast , the effect of the change in the drive parameter on the real negative eigenvalue is negligible . thus , _ the eigenvector corresponding to the real negative eigenvalue is the fast mode_. for this reason , the slow modes are the ones which determine the formation of new states of order . _ the reductive perturbative method is a method where the asymptotic equation is extracted in a systematic way _ [ taniuti & wei , 1968 ; newell & whitehead , 1969 ; kuramoto & tsuzuki , 1974 ; mashiyama _ et al_. , 1975 ; richter _ et al_. , 1981 ] . the method involves in first finding the eigenvectors corresponding to the slow modes and looking for a solution in the subspace spanned by these vectors . the effect of the nonlinearity is handled progressively as a perturbation of the linear solution in power series of the deviation from the critical value of the drive parameter . it may be worth emphasizing that this method is essentially the same as reduction to center manifold . indeed , the equivalence of the center manifold theory [ carr , 1981 ; guckenheimer & holmes , 1983 ; troger & steindl , 1991 ] with the reductive perturbation has been established [ chen _ et al _ , 1996 ] . other techniques of extracting amplitude equations have been devised whose end results are basically the same . for instance , perturbative renormalization group method [ goldenfeld _ et al _ 1989 ; chen _ et al _ , 1996 ] and its recent extension on the basis of envelope theory [ kunihiro , 1995 & 1996 ] has also been developed as a tool for global asymptotic analysis which can be used to extract the amplitude equations . in the present model , there is one simplification arising from the fact that @xmath39 can be identified as the fast ( stable ) mode , and hence this will be eliminated adiabatically . then we derive the equation for the slow dynamics through a power series expansion . although this adiabatic elimination is approximate , the result is found to be in a reasonably good agreement with the accurate method ( which will be reported elsewhere ) except for finer details . consider the eqs.(5 - 7 ) . it can be shown that there is a domain of instability for the parameters @xmath33 , @xmath34 and @xmath35 . the range of values of @xmath34 is @xmath40 , that of @xmath35 is @xmath41 while @xmath42 . thus , @xmath43 and @xmath44 @xmath45 . thus , there are three time scales that are in principle different from each other . physically , they correspond to the mobile ( fastest @xmath33 ) , the cottrell type ( @xmath35 ) and the immobile ( slowest @xmath34 ) . under these conditions we can eliminate one full equation adiabatically . this aspect becomes transparent if the above equations are written in terms of variables which are deviations from the steady state . there is only one steady state defined by : @xmath46^{1/2}}\over{4b_0 } } , { \rm and},y_{a}=1/2.\ ] ] defining new variables which are deviations from the steady state @xmath47 eqns . ( 5 - 7 ) take the form @xmath48 where @xmath49 now , we can rescale the time - like variable by @xmath50 and get @xmath51 it is clear that in the limit of small @xmath34 , @xmath52 , unless the left hand side vanishes identically . thus , we can eliminate @xmath53 in favour of the other two variables and obtain : @xmath54 , \\ \mbox{and}\hspace{.5 cm } { { d{z}}\over{d\tau } } & = & c\left[x -z\right],\end{aligned}\ ] ] where @xmath55^{\frac{1}{2 } } \right).\ ] ] ( the other root for @xmath53 is unphysical since it corresponds to negative dislocation density . ) it must be emphasized that this adiabatic elimination becomes more exact as the value of the parameter @xmath34 gets smaller . equations ( 17 - 18 ) will be solved reductive perturbatively . writing these equations as a matrix equation where the nonlinear part appears separately from the linear part , we get @xmath56 where @xmath57 @xmath58 with @xmath59 and the nonlinear part , @xmath60 , is given by @xmath61 the coefficients @xmath62 and @xmath63 appearing in the last expression for @xmath60 are functions of @xmath64 , @xmath65 and @xmath35 whose expressions for the required first few are given in eqs.(a8-a15 ) of appendix a. consider stability of the fixed point as a function of the parameter @xmath35 . the eigenvalues , @xmath66 , of the matrix @xmath67 are @xmath68 the fixed point becomes unstable when @xmath35 is less than @xmath69 ( @xmath35 is non - negative ) and the discriminant of eq.(24 ) is negative giving a pair of complex conjugate eigenvalues . this holds when the inequality @xmath70 is satisfied . the instability region in the @xmath71 versus @xmath33 plane for a fixed value of @xmath72 is shown in fig . it extends approximately between @xmath73 and @xmath74 . to get approximate analytical solution of eq.(20 ) , we follow the reductive perturbative approach similar to that used by mashiyama _ et al_. [ 1975 ] and richter _ et al_. [ 1981 ] . we choose @xmath75 with @xmath76 and write the matrix @xmath67 as a sum of two matrices , @xmath77 , where @xmath78 is the matrix @xmath67 evaluated for @xmath79 and @xmath80 the eigenvalues of @xmath78 are @xmath81 , where @xmath82 . considering that the solution for @xmath83 grows continuously out of the critical eigenmodes , we can express it as @xmath84 where @xmath85 and @xmath86 are time - dependent complex amplitudes while @xmath87 are right eigenvectors defined by @xmath88 . similarly , we introduce left eigenvectors , @xmath89 defined by @xmath90 . ( note that @xmath91 . t refers to transpose operation of the matrix concerned ) . substituting this expression for @xmath83 in eq.(@xmath92 ) and multiplying it by @xmath93 from the left - hand side gives the equation governing @xmath85 : @xmath94 @xmath95 and @xmath96 are complex coefficients which are functions of @xmath33 , @xmath34 and @xmath35 . their expressions are , respectively , given in eqs . ( a1 ) and ( a4 ) in appendix a. we express @xmath85 in power series of @xmath97 : @xmath98 and introduce slowly varying multiple time scales @xmath99 , @xmath100 , ... replacing @xmath101 by @xmath102 . with this substitution in eq.(@xmath103 ) , it can be solved successively by equating terms of equal power in @xmath104 on both sides of the equation . at first , @xmath105(@xmath97 ) terms give rise to the equation @xmath106 which means that , on the time scale of @xmath30 , @xmath107 is constant . the next higher order , @xmath108 , terms give the equation @xmath109 which can be integrated to get @xmath110 . @xmath105(@xmath111 ) terms give the equation @xmath112 + p_3 ( \psi_1 e^{i\omega\tau } + c.c . ) ^3 \right ] e^{-i\omega\tau},\ ] ] from which we extract the slow dynamics : @xmath113 since to @xmath114 @xmath115 , eq.(@xmath116 ) gives the dynamics for the complex amplitude @xmath85 : @xmath117 the expression for the complex coefficient @xmath118 is given in eq . ( a2 ) in appendix a. equation ( @xmath119 ) is a _ cubic _ tdgl equation where @xmath85 is the complex amplitude for the slow mode which is also referred to as complex order parameter ( of the new state of temporal order ) . its steady state solution gives the amplitude squared as @xmath120 and the associated frequency , @xmath121 , ( with @xmath122 ) as @xmath123 note that @xmath85 describes the limit cycle oscillation . this solution exists provided @xmath124 is negative since @xmath125 is positive ( see expression of @xmath126 in eq.(a1 ) in appendix a ) . @xmath127 is found to be negative in a narrow region ( @xmath128 ) as can be seen in fig . 2 . in this case , since the amplitude of the slow modes grows continuously in proportion to @xmath97 ( see eq.(@xmath129 ) ) the transition is supercritical bifurcation ( continuous or ` second order ' ) for other values of @xmath33 ( within the instability range ) , @xmath124 is positive implying that the transition is subcritical bifurcation ( discontinuous or ` first order ' ) . one has to then go to quintic or even higher terms in the tdgl equation for obtaining an expression for the slow modes ( order parameter ) . we have carried out the reductive perturbative method further to derive the _ quintic _ tdgl equation : @xmath130 the steady state solution of this equation gives the amplitude squared as @xmath131 and the frequency as @xmath132 we found @xmath133 to be negative in the range @xmath134 ( see fig . _ this means a narrow region : @xmath135 within the subcritical bifurcation exists over which the quintic tdgl equation has a solution _ ( see fig . to start with , consider the supercritical regime where expressions for the amplitude and period of the limit cycle are simple . using the steady state solution of the cubic tdgl equation , eqs.(@xmath129 ) & ( @xmath136 ) , the dependence of @xmath137 and period , @xmath138 , of the limit cycle on @xmath33 are , to the leading order , found to be @xmath139 and @xmath140 , respectively . the parameter @xmath141 is a function of the applied stress ( @xmath10 ) and temperature ( t ) . as remarked earlier , @xmath17 is proportional to the concentration of the solute atoms , and therefore @xmath142 . using the standard expression [ alexander 1986 ] @xmath143 ( with @xmath144 ) , we get @xmath145 from this we see that the amplitude of the limit cycle has a strong increasing dependence on stress and an increasing dependence on temperature if we assume that @xmath146 . in contrast , the period of the limit cycle has a weak decreasing dependence on stress ( as the leading contribution is constant ) and a decreasing dependence on temperature . since stress and temperature are measurable quantities , our predictions can be compared with experimental results . the amplitude and period of the limit cycle are related , respectively , to the amount of strain jumps and the period of the jumps on the creep curve through the orowan equation . there are very few experiments in this mode of testing as mentioned in the introduction . the only experiment where this dependence on stress and temperature has been measured is that by zagoruyko _ et al _ [ 1977 ] . however , it is possible to translate the results from experiments in constant strain rate case to the creep case and compare them with the results of this calculation . according to zagoruyko _ et al _ [ 1977 ] , the amplitude of the strain jumps increases with stress while its period has a weak but decreasing dependence on stress . experiments from constant strain rate case also exhibit the same trend when the results are translated in terms of constant stress experiments . it is well known that the amplitude of the stress drops decreases with applied strain rate . in fact , even the present model predicts this behaviour for the constant strain rate mode [ ananthakrishna & valsakumar 1982 ] . this implies that the dependence of the amplitude of strain jumps on stress should increase [ kubin _ et al _ 1993 ] . [ this relation can be seen as follows . in constant strain rate case , the deformation rate is fixed and the stress developed in the sample is measured . when the contribution to the plastic strain rate increases due to increased dislocation motion ( for whatever reasons ) , the stress has to fall in order to keep the applied strain rate constant . thus , the relation between strain rate and stress is opposite . ] clearly , the general trend is consistent with the experimental results . 3 shows the actual dependence of the amplitude and the period on @xmath33 in the domain of supercritical bifurcation ( not the leading one as given by the above expressions ) . zagarukuyko @xmath147 [ 1977 ] also report that the amplitude of the strain jumps increases while its period decreases with temperature , which is consistent with our result ( provided @xmath148 ) . in the case of subcritical bifurcation , the dependence of the amplitude of the strain jumps and the period on @xmath33 have similar behaviour as in the case of the supercritical bifurcation . eventhough , in this case , leading order dependence on @xmath33 is difficult to obtain we have their actual dependence plotted in fig . it may be noted that they have qualitatively the same behaviour as in fig . 3 with the degree of dependence more pronounced in this case . experiments in constant strain rate case show that the stress drops are seen to arise both abruptly as well as continuously [ kubin _ et al _ , 1988 ] . translating this result to the constant stress case , it implies that the strain jumps can arise both abruptly and continuously . this feature again follows from our calculations . having derived the tdgl equation , we would first like to compare its result with the numerical solutions obtained via eqs . ( 17 - 18 ) and via eq.(@xmath92 ) . secondly , due to the fact that the quintic tdgl equation ( eq . @xmath149 ) is valid in a limited domain , one suspects that higher order nonlinearities are controlling the subcritical bifurcation , which is quite unusual . therefore , we would like to investigate the numerical solution obtained by solving eq.(@xmath92 ) keeping successive leading powers in @xmath150 . this analysis confirms the above suspicion that higher order nonlinearities are important in this model outside the domain of validity of the quintic tdgl equation . using the steady state solution of the quintic tdgl equation , eq.(@xmath149 ) , in eq.(@xmath151 ) , we get an approximate expression for the limit cycle which is usually called the secular motion . the equations governing the secular motion are given by eqs . ( b5 ) and ( b6 ) ( hereafter called secular equations ) in appendix b. this can be compared with the numerical solutions obtained using eqs . ( 17 - 18 ) ( hereafter called the reduced model ) and with that obtained using eq.(@xmath92 ) _ where _ the expansion in @xmath150 of eq . ( 23 ) is limited up to a certain power ( hereafter called the @xmath152-truncated model , n signifying the highest power of @xmath150 after truncation ) . _ at the outset we state that the reduced model has bounded solution over the entire instability domain_. in the case of the numerical solutions obtained from eq.(@xmath92 ) , we have found that while truncating the @xmath150-expansion in eq . ( 23 ) at @xmath153 is sufficient to give a bounded solution for the range of supercritical bifurcation ( @xmath154 ) , it fails beyond this . on the other hand , truncating the @xmath150-expansion at @xmath155 does extend the domain of bounded solution ( @xmath156 ) which matches with the domain where the quintic tdgl equation works . figure 5 shows the plots of the solutions to the secular equations ( eqs . b5 and b6 ) , the reduced model ( eqs . 17 - 18 ) and the @xmath153-truncated model ( eq . @xmath92 ) for @xmath157 and @xmath158 . all of them are numerically almost indistinguishable . we have verified that the solutions obtained by these three methods show that the amplitude of the limit cycle scales with @xmath97 . figure 6 shows the solutions obtained by the three ways for @xmath159 and @xmath158 where , this time , the truncation is at @xmath155 . the agreement between the three solutions is reasonably good . part of the discrepancy between the secular motion and the two numerical solutions can be attributed to the range of values of the various terms ( @xmath160 s , @xmath161 s , @xmath0 s and @xmath162 s of appendix c ) contributing to @xmath163 and @xmath164 . ( they range from @xmath165 to @xmath166 ) . we have also verified that the amplitude of the limit cycle has a finite jump for this case . for @xmath167 , where the quintic tdgl equation is inadequate , @xmath155-truncated model is also inadequate . however , @xmath168-truncated model has a bounded solution for the region @xmath169 . in particular , fig . 7 shows the plots of the solutions obtained by the @xmath168-truncated model and that by the reduced model for @xmath170 and @xmath158 . _ this implies that seventh order tdgl equation need to be considered for this range of @xmath33_. we have verified that @xmath171 term extends the domain of bounded solutions upto @xmath172 . from this trend , it appears that very high powers in @xmath150 need to be retained to cover the entire domain upto @xmath173 . the region @xmath174 is even more interesting . while the quintic tdgl equation is inadequate to give a bounded solution , we found that retaining even upto the ninth power in @xmath150 does not give a bounded solution for the entire range of values @xmath175 ( even for those close to @xmath176 ) . this suggests that this region is one where the nonlinearity is strong . on the other hand , as pointed out earlier the reduced model has limit cycle solution in this range as well . in particular , fig . 8 shows such a plot for @xmath177 and @xmath158 . thus , the results of the tdgl equation , the @xmath152-truncated model and the reduced model are consistent with each other . further , the numerical solutions via the power series in @xmath150 throws light on the nature of nonlinearities governing the solution in different regions of instability . we have carried out the reductive perturbative approach to the problem of steps on creep curve and shown that the dynamics of the system is described by a tdgl equation for the amplitude of the slow modes ( complex order parameter ) @xmath85 in the neighbourhood of the bifurcation point . since the above derivation is valid only in the neighbourhood of the critical value , the expression for @xmath85 is valid only for small @xmath104 . this has been exemplified by the reasonable agreement between the secular motion and the numerical solution of the reduced model within the domain of the validity of the quintic tdgl equation . we have shown that both subcritical as well as supercritical bifurcations are seen in the instability range of the parameter @xmath33 . _ while the subcritical bifurcation ( abrupt or ` first order ' transition ) seen over most of the values of @xmath33 is consistent with our earlier calculation _ [ ananthakrishna & sahoo , 1981b ; valsakumar & ananthakrishna , 1983 ] , _ we see a supercritical bifurcation ( continuous or ` second order ' transition ) over a narrow mid - range of _ @xmath33 . the existence of the supercritical bifurcation was missed by us earlier . _ the existence of both supercritical and subcritical bifurcations is consistent with experiments . in addition , the dependence of the amplitude and the periodicity of the strain jumps on stress and temperature are consistent with experiments_. from this point of view , even though the derived tdgl equation is valid in a limited domain , the present calculation allows a direct mapping of these quantities . this calculation also helps us to demonstrate the complicated dependence of the slow modes on the original modes . this will serve as a warning to those using hand waiving arguments for declaring certain modes as fast modes and others as slow modes in modelling of such of these problems . looked at from the point of view of the properties of the model , there are some interesting features . for the region outside the validity of the quintic tdgl equation , on the basis of a careful numerical analysis , we found that high order nonlinearities govern the nature of the subcritical bifurcation . it must be emphasized that this feature is unusual and this insight would not have been possible but for the numerical solutions obtained by keeping successive higher powers in @xmath150 , which in itself is the basis for the derivation of the tdgl equation . this unusual feature of the model interpreted in the language of phase transitions give better insight and could also find applicability . for conventional models , the ` free energy ' is described by an expansion in power series of the order parameter . even in the case of dynamic transition such as the present one , it is possible to associate a ` free energy ' like function such that @xmath178 @xmath179 , where @xmath180 $ ] is the ` free energy ' like function . while the ` free energy ' for ` second order ' phase transitions is described by retaining up to fourth power in the order parameter , up to sixth power is conventionally required for ` first order ' phase transitions ( with the appropriate signs for the coefficients in the expansion ) . in the present case , however , we need to go to as high as twelfth power ( or more ) of the order parameter to cover the interval @xmath181 ( or equivalently twelfth power or higher in @xmath150 ) . in this case , including successively higher powers in @xmath150 increases the domain of description . on the other hand , in the interval of @xmath175 , even retaining upto tenth power in @xmath150 ( which is the highest power we have checked ) does not work . this must be contrasted with the existence of bounded solution for the reduced model over the entire interval . _ this feature is rather unusual and , to the best of our knowledge , we are not aware of any other model which exhibits this property_. one of us ( m.b . ) would like to thank the international program in physical sciences , uppsala university ( sweden ) for offering a fellowship to study at indian institute of science . part of this work has been supported by ifcpar grant no . 1108 - 1 . * references * aifantis , e. c. [ 1988 ] `` on the problem of dislocation patterning and persistent slip bands '' in _ non linear phenomena in materials science i _ , eds . kubin , l. p. & martin , g. ( trans tech , switzerland ) , pp . 397 - 406 . alexander , h. [ 1986 ] `` dislocations in covalent crystals '' in _ dislocations in solids _ ed . nabarro , f. r. n. ( north - holland ) , 151 . ananthakrishna , g. [ 1993 ] `` formation , propagation of bands and chaos in jerky flow '' , _ scripta . _ * 29 * , 1183 - 1188 . ananthakrishna , g. & john , t. m. [ 1990 ] `` instabilities and chaos in plastic flow '' , in _ directions in chaos _ vol . 3 , ed . hao bai - lin ( world scientific , singapore ) , pp . 133 - 148 . ananthakrishna , g. , fressengeas c. , grosbras , m. , vergnol , j. , engelke , c. , plessing , j. , neuhaeuser , h. , bouchaud , e. , planes , j. & kubin , l. p. [ 1995a ] `` on the existence of chaos in jerky flow , '' _ scripta metall . _ * 32 * , 1731 - 1737 . ananthakrishna , g. , kubin , l. p. & martin , g. ( eds . ) [ 1995b ] _ solid state phenomena _ vol . 42&43 ( scitec , switzerland ) . ananthakrishna , g. & noronha , s. j. [ 1995 ] `` chaos in jerky flow : theory and experiment '' in _ solid state phenomena _ vol . 42&43 , eds . ananthakrishna , g. , kubin , l. p. & martin , g. ( scitec , switzerland ) , pp . 277 - 286 . ananthakrishna g. & sahoo , d. [ 1981a ] `` a statistical dislocation dynamics with application to lif '' , _ d _ * 14 * ) , 699 - 713 . ananthakrishna , g. & sahoo , d. [ 1981b ] `` a model based on nonlinear oscillations to explain jumps on creep curves '' , _ d _ * 14 * , 2081 - 2090 . ananthakrishna , g. & valsakumar , m. c. [ 1982 ] `` repeated yield drop phenomena : a temporal dissipative structure '' , _ j. phys . d _ * 15 * , l171-l175 . ananthakrishna , g. & valsakumar , m. c. [ 1983 ] `` chaotic flow in a model for repeated yielding '' , _ phys . a _ * 95 * , 69 - 71 . ananthakrishna , g. , fressengeas c. , grosbras , m. , vergnol , j. , engelke , c. , plessing , j. , neuhaeuser , h. , bouchaud , e. , planes , j. & kubin , l. p. [ 1995 ] `` on the existence of chaos in jerky flow , '' _ scripta metall . _ * 32 * , 1731 - 1737 . bodner , s. r. & rosen , a. [ 1967 ] `` discontinuous yielding of commercially - pure aluminium '' , _ j. mech . phys . solids _ * 15 * , 63 - 77 . brindley , b. j. & worthington , p. j. [ 1970 ] `` yield - point phenomena in substitutional alloys '' , _ metall . reviews _ * 145 * , 101 - 114 . carr , j. [ 1981 ] _ applications of center manifold theory _ ( springer - verlag , berlin ) . chen , l. y. , goldenfeld , n. & oono , y. , [ 1996 ] `` the renormalization group and singular perturbations : multiple- scales , boundary layers and reductive perturbation theory '' , _ phys . e _ * 54 * , 376 - 394 . chihab , k. , estrin , y. , kubin , l. p. & vergnol , j. [ 1987 ] `` the kinetics of the portevin - le chatelier effect in an al-5 at% mg alloy '' , _ scripta . _ * 21 * , 203 - 208 . cottrell , a. h. [ 1953 ] `` a note on the portevin - le chatelier effect '' , _ phil . mag . _ * 44 * , 829 - 832 . da silveira , t. l. & s. n. monteiro , s. n. [ 1979 ] `` jumps in the creep curve of austentic stainless steels '' , _ met . a _ * 10 * , 1795 - 1796 . estrin , y. & kubin , l. p. [ 1989 ] `` collective dislocation behavior in dilute alloys and the portevin - le chatelier effect '' , _ j. mech . behaviour of materials _ * 2 * , 255 - 292 . goldenfeld , n. , martin , o. & oono , y. [ 1989 ] `` intermediate asymptotics and renormalization group theory '' , _ comp . _ * 4 * , 355 . guckenheimer , j. & holmes , p. [ 1983 ] _ nonlinear oscillations , dynamical systems , and bifurcations of vector fields _ vol . 42 of _ applied math . ( springer - verlag , berlin ) , 123 - 145 . hahner , p. & kubin , l. p. [ 1992 ] `` coherent propagative structures in plastic deformation : a theory of ludlers bands in polycrystals '' in _ solid state phenomena _ vol . 3&4 , eds . kubin , l. p. & martin , g. ( trans tech , switzerland ) , pp . 385 - 402 . hahner , p. [ 1993 ] `` modelling the spatiotemporal aspects of the portevin - le chatelier effect '' , _ mater . sci . and eng . _ * 164 * , 23 - 34 . hall , e. o. [ 1970 ] _ yield point phenomena in metals and alloys _ ( macmillan , london ) . kubin , l. p. , chihab , k. & estrin , y. [ 1988 ] `` the rate dependence of the portevin - le chatelier effect '' , _ acta metall . _ * 36 * , 2707 - 2718 . kubin , l. p. & estrin , y. [ 1990 ] `` evolution of dislocation densities and the critical conditions for the portevin - le chatelier effect '' , _ acta metall . _ * 38 * , 697 - 708 . kubin , l. p. & estrin , y. ( organizers ) [ 1993 ] `` viewpoint set no . 21 : propagative instabilities '' , scripta metall . * 29 * , 1147 - 1188 . kubin , l. p. & lepinoux , j. [ 1988 ] `` the dynamic organization of dislocation structures '' , _ strength of metals and alloys _ ( icsma 8) , vol . kettunen , p. o. , lepisto , t. k. , & lehtonen , m. e. , ( pergamon press , oxford ) , pp . 35 - 59 . kubin , l. p. & martin , g. ( eds . ) [ 1988 ] _ solid state phenomena _ vol . 3&4 ( trans tech , switzerland ) . kunihiro , t. [ 1995 ] `` a geometrical formulation of the renormalization group method for global analysis '' , _ prog . phys . _ * 94 * 503 - 514 . kunihiro , t. [ 1996 ] `` the renormalization group method applied to asymptotic analysis of vector fields '' , in hep - th/9609045 . kuramoto , y. & tsuzuki , t. [ 1974 ] `` reductive perturbative approach to chemical instabilities '' , _ prog . _ * 52 * , 1399 - 1401 . lubahn , j. d. & felgar , r. p. [ 1961 ] _ plasticity and creep of metals _ ( john wiley , new york ) . martin , g. & kubin , l. p. ( eds . ) [ 1992 ] _ solid state phenomena _ vol.23&24 ( trans tech , switzerland ) . mashiyama , h. , ito , a. & ohta , t. `` fluctuations and phase transitions far from equilibrium '' , _ prog . * 54 * , 1050 - 1066 . newell , a. c. & whitehead , j. a. [ 1968 ] `` finite bandwidth , finite amplitude convection '' , _ j. fluid mech . _ * 38 * , 279 - 303 . noronha , s. j. , ananthakrishna , g. , quaouire , l. & fressengeas , c. [ 1997 ] `` chaos in jerky flow- experimental verification of a theoretical prediction '' , _ pramana _ , in press . noronha , s. j. , ananthakrishna , g. , quaouire , l. , fressengeas , c. & kubin , l. p. [ 1996 ] `` chaos in the portevin - le chatelier effect '' , submitted to int . j. bif . & chaos . penning , p. [ 1972 ] `` mathematics of the portevin - le chatelier effect '' , _ acta metall . _ * 20 * , 1169 - 1175 . quaouire , l. & fressengeas , c. [ 1995 ] `` dynamical analysis of the portevin - le chatelier effect '' in _ solid state phenomena _ vol . 3&4 ( trans tech , switzerland ) , eds . ananthakrishna , g. , kubin , l. p. & martin , g. ( scitec , switzerland ) , pp . 293 - 302 . richter , p. h. , procaccia , i. & ross , j. [ 1981 ] `` chemical instabilities '' in _ advances in chemical physics _ prigogine , i , & rice , s. ( wiley - interscience , new york ) , pp . 217 - 268 . sahoo , d. & ananthakrishna , g. [ 1982 ] `` a phenomenological dislocation transformation model for the mobile fraction in simple systems '' , _ j. phys . d _ * 15 * , 1439 - 1449 . stejskalova , v. , hammersky , m. , luckac , p. , vostry , p. & spursil , b. [ 1981 ] _ czech . * b31 * , 195 . taniuti , t. & wei , c. c. [ 1968 ] `` reductive perturbative method in nonlinear wave propagation i '' , _ j. phys . japan _ * 24 * , 941 - 946 . troger , h. & steindl , a. [ 1991 ] _ nonlinear stability and bifurcation theory _ ( springer - verlag , wien , new york ) , pp . 52 - 68 . valsakumar , m. c. & ananthakrishna , g. [ 1983 ] `` a model based on nonlinear oscillations to explain jumps on creep curves : ii . approximate solutions '' , _ j. phys . d _ * 16 * , 1055 - 1068 . venkadesan , s. , murthy , k. p. n. , & valsakumar , m. c. [ 1995 ] `` chaotic flow in al - mg alloy '' in _ solid state phenomena _ vol . 42&43 , eds . ananthakrishna , g. , kubin , l. p. & martin , g. ( scitec , switzerland ) , pp . 287 - 292 . zagoruyko , l. n. , osetskiy a. i. & soldatov , v. p. [ 1977 ] `` discontinuous deformation of zinc single crystals under creep conditions '' , _ phys . met . metallogr . _ * 43*(5 ) , 156 - 164 . in this appendix we give the expressions for the coefficients appearing in the tdgl equation up to the quintic term , i.e. the expressions for @xmath126 , @xmath163 and @xmath164 in the equation @xmath182 . they are given by @xmath183 @xmath184 and @xmath185 where @xmath186,\ ] ] and @xmath187 @xmath188 @xmath189 the expressions for @xmath190 s and @xmath191 s used in eq.(a5 ) are given by @xmath192,\ ] ] @xmath193 @xmath194,\ ] ] @xmath195,\ ] ] and @xmath196 @xmath197 @xmath198 @xmath199 where @xmath200 expressions for @xmath160 s , @xmath161 s , @xmath0 s and @xmath162 s are given in appendix c. in this appendix the equations describing the secular motion are derived for the case when the order parameter , @xmath85 , is found from the quintic tdgl equation ( eq . 44 ) . from eq . ( 39 ) , @xmath201 and @xmath202 where the complex amplitude is @xmath203 with @xmath204 given in terms of functions of @xmath107 and phase factors ( see appendix c ) . on the other hand , to @xmath114 , @xmath205 which , upon inverting , gives @xmath107 in terms of @xmath85 : @xmath206 substituting the @xmath207-expressions from appendix c in eqs . ( b1 ) and ( b2 ) and , then , using eq . ( b4 ) leads to obtaining the secular equations given by : @xmath208 and @xmath209 where @xmath210 with @xmath211 where @xmath212 stands for either @xmath213 , @xmath214 , @xmath215 or @xmath216 , while @xmath217 stands for either @xmath160 , @xmath161 , @xmath0 , or @xmath162 correspondingly . expressions for @xmath213 , @xmath214 , @xmath215 and for @xmath216 can be obtained from @xmath160 s , @xmath161 s , @xmath0 s and @xmath162 s which are given in appendix c. expressions for @xmath110 , @xmath218 , @xmath219 , @xmath220 ( with a factor @xmath221 ) are @xmath222 where @xmath223,\\ \ell_{3 } & \equiv & \frac{i}{2\omega}\,\left[2p_{2}(k_{1}^{\star}+k_{2}^{\star}+k_{2 } + k_{3})+3p_{3}\right],\\ \ell_{4 } & \equiv & \frac{i}{4\omega}\,\left[2p_{2}(k_{1}^{\star}+k_{3})+p_{3}\right],\\ m_{1 } & \equiv & -\frac{i}{\omega}\left[-\mu\ , k_{1 } - \mu^{\star}\ , k_{3}^{\star } + 2 p_{2 } \ , \ell_{1}^{\star } \right],\\ m_{2 } & \equiv & \frac{i}{\omega}\left[-\mu^{\star}\,(k_{2}+ k_{2}^{\star})+ 2 p_{2 } ( \ell_{1}+\ell_{1}^{\star } ) \right],\\ m_{3 } & \equiv & \frac{i}{3\omega}\left[\mu\ , k_{3 } - \mu^{\star}(k_{1}^\star+ 2k_{3 } ) + 2p_2 \ell_{1}\right],\\ m_{4 } & \equiv & - \frac{i}{3\omega } \left\{p_{2}\left[(k_{1}+k_{3}^{\star})^{2}+ 2 ( \ell_{2}+\ell_{4}^{\star})\right]+ 3p_{3 } ( k_{1}+k_{3}^{\star } ) + p_{4 } \right\},\\ m_{5 } & \equiv & - \frac{i}{\omega}\left\{2p_{2}\left[(k_{1}+k_{3}^{\star})(k_{2}+k_{2}^{\star } ) + \ell_{2}+\ell_{3}^{\star } + \ell_{4}^{\star}\right]+ 3p_{3 } \left[2(k_{1}+k_{3}^{\star } ) + k_{2 } + k_{2}^{\star}\right]\right . \nonumber \\ & & + 4p_{4}- 2k_{1}\,\eta \left . \right\},\\ m_{6 } & \equiv & \frac{i}{\omega}\left\{p_{2}\left[2(k_{1}+k_{3}^{\star } ) ( k_{1}^\star+k_{3})+(k_{2}+k_{2}^{\star})^{2 } + 2(\ell_{3 } + \ell_{3}^{\star})\right ] \right . \nonumber\\ & & + 3p_{3}\left[k_{1}+k_{1}^{\star } + 2(k_{2 } + k_{2}^{\star})+k_{3}+k_{3}^{\star}\right ] + 6p_{4}-2k_{2}\eta_1 \left . \right\},\\ m_{7 } & \equiv & \frac{i}{3\omega}\left\{2p_{2}\left[(k_{1}^{\star}+k_{3})(k_{2}^{\star}+k_{2 } ) + \ell_{2}^{\star}+\ell_{3 } + \ell_{4}\right]+ 3p_{3 } \left[2(k_{1}^{\star}+k_{3 } ) + k_{2 } + k_{2}^{\star}\right]\right . \nonumber \\ & & + 4p_{4}-2k_{3 } \eta^{\star}\left . \right\},\\ m_{8 } & \equiv & \frac{i}{5\omega}\left\{p_{2}\left[(k_{1}^{\star}+k_{3})^{2 } + 2(\ell_{2}^{\star } + \ell_{4})\right ] + 3p_{3}(k_{1}^{\star } + k_{3})+p_{4 } \right\},\\ n_{1 } & \equiv & \frac{i}{2 \omega}\left(\mu - \mu^{\star}\right ) \ell_{1},\\ n_{2 } & \equiv & -\frac{i}{2 \omega } \left\ { 2 p_{2}\left[\ell_{1}^{\star } ( k_{1}+ k_{3}^{\star } ) + m_{1 } + m_{3}^{\star}\right ] + 3p_{3}\ , \ell_{1}^{\star } -2\mu\ell_{2 } - \mu^{\star } \ell^{\star}_{4 } \right\},\\ n_{3 } & \equiv & \frac{i}{2\omega } \left\ { 2 p_{2}\left[\ell_{1}(k_{2}+ k_{2}^{\star } ) + \ell^{\star}_{1}(k_{1}^{\star } + k_{3 } ) + m_{1}^{\star } + m_{2}^{\star } + m_{2}+ m_{3}\right]\right . \nonumber\\ & & + 3p_{3}(\ell_{1}^{\star}+2\ell_{1})\left.-\eta^{\star } \ , \ell_{1 } - 2 \mu^{\star}\,\ell_{3 } \right\},\\ n_{4 } & \equiv & \frac{i}{4\omega } \left\ { 2 p_{2}\left[\ell_{1}(k_{1}^{\star}+ k_{3 } ) + m_{1}^{\star } + m_{3}\right ] + 3p_{3}\ , \ell_{1}-\mu^{\star } \ell_{2}^{\star } + ( \mu - 3\mu^{\star})\ell_{4 } \right\},\\ n_{5 } & \equiv & - \frac{i}{4\omega}\left\{2p_{2}\left[m_{4}+m_{8}^{\star } + ( k_{1}+k_{3}^{\star } ) ( \ell_{2}+\ell_{4}^{\star})\right ] + 3p_{3 } \left[(k_{1}+k_{3}^{\star})^2 + \ell_{2 } + \ell_{4}^{\star}\right ] \right.\nonumber \\ & & + 4p_{4}(k_{1}+k_{3}^{\star } ) + p_{5}\left . \right\},\\ n_{6 } & \equiv & -\frac{i}{2\omega}\left\ { 2p_{2}\left[m_{4 } + m_{5 } + m_{7}^{\star } + m_{8}^{\star } + \ell_{3}^{\star}(k_{1}+k_{3}^{\star } ) + ( \ell_{2 } + \ell_{4}^{\star } ) ( k_{2 } + k_{2}^{\star } ) \right ] \right . \nonumber \\ & & + 3p_{3 } \left[\ell_{3}^{\star } + 2(\ell_{2 } + \ell_{4}^{\star } ) + ( k_{1 } + k_{3}^{\star})^{2 } + 2(k_{1}+k_{3}^{\star})(k_{2}+k_{2}^{\star } ) \right]\nonumber \\ & & + 4p_{4 } \left [ k_{2 } + k_{2}^{\star } + 3(k_{1}+k_{3}^{\star } ) \right ] + 5 p_{5}\left.-3\eta \ell_{2 } \right\},\\ n_{7 } & \equiv & \frac{i}{2\omega}\left\ { 2p_{2}\left[m_{5}^{\star } + m_{6}^{\star}+ m_{6 } + m_{7 } + ( \ell_{2}^{\star}+\ell_{4})(k_{1}+k_{3}^{\star } ) + \ell_{3 } ( k_{2 } + k_{2}^{\star})+\ell_{3}^{\star}(k_{1}^{\star}+k_{3 } ) \right ] \right . \nonumber\\ & & + 3p_{3 } \left [ ( k_{2 } + k_{2}^{\star})^{2 } + 2(k_{1}^{\star}+k_{3})(k_{1}+k_{2}+k_{2}^{\star}+k_{3}^{\star})+\ell_{2}^{\star } + \ell_{3}^{\star } + 2\ell_{3}+\ell_{4}\right ] \nonumber\\ & & + 4p_{4 } \left [ k_{1 } + k_{3}^{\star } + 3(k_{1}^{\star}+k_{2}^{\star}+k_{2 } + k_{3})\right ] + 10 p_{5}\left.-(\eta + 2\eta^{\star})\ell_{3 } \right\},\\ n_{8 } & \equiv & \frac{i}{4\omega}\left\{2p_{2}\left[m_{4}^{\star } + m_{5}^{\star } + m_{7 } + m_{8 } + \ell_{3}(k_{1}^{\star}+k_{3 } ) + ( \ell_{2}^{\star } + \ell_{4 } ) ( k_{2 } + k_{2}^{\star } ) \right ] \right . \nonumber \\ & & + 3p_{3}\left [ ( k_{1}^{\star } + k_{3})^{2 } + 2(k_{2}+k_{2}^{\star})(k_{1}^\star+k_{3})+2 ( \ell_{2}^{\star } + \ell_{4 } ) + \ell_{3}\right ] \nonumber \\ & & + 4p_{4 } \left [ k_{2 } + k_{2}^{\star } + 3(k_{1}^{\star}+k_{3 } ) \right ] + 5 p_{5}\left.-3\eta^{\star}\ell_{4 } \right\},\\ n_{9 } & \equiv & \frac{i}{6\omega}\left\{2p_{2}\left[m_{4}^{\star } + m_{8 } + ( \ell_{2}^{\star } + \ell_{4 } ) ( k_{1}^{\star } + k_{3}\right ] + 3p_{3 } \left [ ( k_{1}^{\star } + k_{3})^{2 } + \ell_{4}^{\star } + \ell_{4}\right ] \right.\nonumber \\ & & + 4p_{4 } ( k_{1}^{\star}+k_{3 } ) + p_{5}\left . \right\}.\end{aligned}\ ] ] note that expressions for @xmath224, ... ,@xmath225 are given in eq . ( a5 ) in appendix a. fig . 1 : : plot of @xmath226 versus @xmath33 with @xmath227 . the hatched portion shows the instability region which extends between @xmath73 and @xmath74 approximately . fig . 2 : : plots of @xmath127 and @xmath228 versus @xmath33 ( @xmath227 ) . note that @xmath127 is negative ( but finite ) in the interval @xmath229 . similarly , @xmath228 is negative ( but finite ) in the interval @xmath134 . fig . 3 : : plots of @xmath230 and @xmath231 versus @xmath33 in the region of supercritical bifurcation . fig . 4 : : plots of @xmath230 and @xmath231 versus @xmath33 in a subregion of subcritical bifurcation , i. e. the region where quintic tdgl equation holds . fig . 5 : : plots of the ( numerical ) limit cycle solutions obtained from the secular equations ( dotted line ) , the reduced model ( dashed line ) and the @xmath153-truncated model ( crossed line ) for @xmath157 and @xmath158 . fig . 6 : : plots of the ( numerical ) limit cycle solutions obtained from the secular equations ( dotted line ) , the reduced model ( dashed line ) and the @xmath155-truncated model ( crossed line ) for @xmath159 and @xmath158 . fig . 7 : : plots of the ( numerical ) limit cycle solutions obtained from the reduced model ( dashed line ) and the @xmath168-truncated model ( crossed line ) for @xmath170 and @xmath158 . : : plot of the ( numerical ) limit cycle solution obtained from the reduced model for @xmath177 and @xmath158 .

the intuitive feeling that hadron matter should be maximally perturbed in the high energy extremely inelastic collisions was a main reason of our effort to consider such processes . it was a hope to observe new dynamical phenomena , or new degrees of freedom , unattainable in other ordinary hadron reactions . this paper presents a first attempt to describe the particularity of considered processes , to give a review of existing models prediction and , at the end , we will offer the field - theoretical formalism for hadron inelastic processes . thus , considering hadrons mean multiplicity @xmath3 as a natural scale of the produced hadrons multiplicity @xmath4 at given cm energy @xmath5 , we would assume that n>>. 0.1at the same time we wish to have n<<n_max=/m , 0.2where @xmath6 gev is the characteristic hadron mass . the last restriction is introduced to weaken the unphysical constraints from the finite , for given @xmath7 , phase space volume . we should assume therefore that @xmath7 is high enough . the multiple production is the process of colliding particles where kinetic energy is dissipated into the mass of produced particles . then one may validate that the entropy @xmath8 accedes its maximum in the domain ( ) since the multiplicity @xmath4 characterizes the rate of stochastization , i.e. the level of incident energy dissipation over existing ( free ) degrees of freedom . there is also another quantitative definition of our reactions . let @xmath9 be the energy of the fastest particle in the given frame and let @xmath10 be the total incident energy in the same frame . then the difference ( @xmath11 ) is the energy spent on production of the less energetic particles . it is useful to consider the inelasticity coefficient = = 1 - 1 . 0.2ait defines the portion of spent energy . therefore , we wish to consider processes with 1 - 1 . 0.2cfollowing ( ) we will assume that 1->>. 0.2dtherefore , the kinetic energy of produced particles in our processes can not be arbitrarily small . using thermodynamical terminology , we wish investigate the production and properties of comparatively ` cold ' multi - hadron ( mostly of @xmath12-mesons ) state . we would like to note from the very beginning that we have only a qualitative scenario of such states which may be produced and the review of the corresponding why s may be considered as the main purpose of this paper . although at the end of this paper ( appendix k ) we will describe principal features of possible field - theoretical solution of our problem . it should be noted the absence of any authentic experimental information concerning discussed processes . moreover , actually the hadron inelastic interactions with a set peculiar to them of unsolved theoretical problems will be considered . nevertheless we suggest to work in this field in spite of these difficulties because the system with extremal properties may be more transparent since the asymptotics always simplify a picture . we would demonstrate this idea and will try to put it in the basis of developed theoretical methods . absence of experimental information about so high inelastic hadron processes is the consequence of the smallness of corresponding cross sections . besides this it was unclear for what purpose the experimental efforts should be done . we would like to convince the reader that the discussed problem is interesting and important . for instance , we will discuss a possibility that asymptotics over @xmath4 may replace in definite sense the asymptotics over @xmath5 . a short address to experimentalists will be given in concluding sec.4.1.2 . we hope that the paper would be useful both for theorists and experimentalists . by this reason the main text of this paper will contain only the qualitative discussion of the problems and results . the quantitative proof , formulation of pure theoretical methods etc . are added in the appendices . considered extremal problem is a good theoretical laboratory and is described in the appendices and theoretical methods may be applied for other physical problems . we would like to point out that the special technique was built for the problem discussed , see sec.2 . _ having the very high multiplicity ( vhm ) state it is natural to use the thermodynamical methods . _ we will offer for this purpose the real - time @xmath0-matrix interpretation of thermodynamics . it can be shown in what quantitative conditions it will coincide with simpler canonical imaginary - time matsubara formalism . we will give also the generalization of the real - time finite - temperature perturbation theory in the case of local temperature @xmath13 distribution , when @xmath14 . this will allow to use the thermodynamical description if the system is far from equilibrium . _ the particle spectrum in the vhm region is soft . it is just a situation when the collective phenomena should be important . _ to describe this phenomena , the decomposition on correlators will be adopted . the origin of this decomposition lie in the mayer s ` group decomposition ' . in multiple production physics this decomposition is known also as the ` multi - component description ' . it based on the idea that the multiple production process may include various mechanisms . in sec.3 we will investigate model predictions for vhm region . we would like to note two main conclusions : existing multiperipheral - type models are unable to describe the vhm region . the infrared region of the pqcd becomes important even if the constraint ( ) is taken into account . in sec.2.1 we will try to formulate the phenomenology of our problem , i.e. the way the vhm processes may be described and what type of phenomena one may expect . the importance of thermodynamical methods will became evident and we will offer in sec.2.2 a general description of the corresponding formalism . it is important to note that we may classify the possible asymptotics over @xmath4 . we will find that there exist only three classes of asymptotics . this will simplify consideration definitely restricting the possibilities . in sec.2.3 we will use the thermodynamical language to give a physical interpretation of these classes of asymptotics . we will see in result that in our choice of the vhm final state this should lead to reorganization of multiple production dynamics : we will get out of the habitual multiperipheral picture in the vhm domain . moreover , one may assume that the semiclassical approximation becomes exact in the vhm domain . this naturally leads to the idea to search such a scheme of calculation which depends on the choice of final state . quantitative description of this idea may be realized as is described in appendix k. the vhm production phenomena includes two sub - problems . first of all it is the dynamical problem of incident energy degradation into the secondary particle energies and the second one is the description of the final state . we will start discussion in sec.2.1.1 from the second part of the problem to explain that the statistical methods are essential for us . in sec.2.1.2 we will try to outline at least qualitatively the main mechanisms of hadron production . the peculiarity of hadron production phenomena consists in the presence of hidden constraints , the consequence of local non - abelian gauge symmetry . the constraints may prevent thermalization and the incident energy dissipation is confused in this case . just the ` confusing ' effect is dominant in the hadron multiple production processes if @xmath15 . the expected change of dominant mechanism of hadron production is discussed sec.2.1.3 . it is important that , in spite of hidden constraints , the system may freely evolve to define the vhm state . such a vhm state should be in equilibrium . formal definition of the ` equilibrium ' notion will be given in sec.2.2.2 . the problem described contains small parameters @xmath16 and @xmath17 . to have the possibility of estimation of contributions in accordance with these parameters one should include them into the formalism . this becomes possible if and only if the integral quantities are calculated . so , the multiplicity @xmath4 is an index only if the multiple production amplitudes @xmath18 are considered . but the cross section @xmath19 is a nontrivial @xmath20 of @xmath4 . we will calculated by this logic mostly integrals of @xmath21 , living out the consideration the amplitudes , see also sec.2.2 , where the first realization of this idea is offered ( the @xmath22 attempt to realize this idea one may find in ) . this is a general methodological feature of our consideration . the multiple production cross section @xmath19 falls down rapidly in the discussed very high multiplicity ( vhm ) domain ( ) and for this reason the multiplicities @xmath23 are not accessible experimentally . at the lhc energy @xmath24 is valid and we will assume that @xmath25 is just the discussed vhm region ( @xmath26 at the lhc energy ) . we will explain later why n~^2 0.1is chosen for the definition of vhm region . generally speaking , having the state of a large number of particles , it is reasonable to depart from an exact definition of the number @xmath4 of created particles , their individual energies @xmath27 , momenta @xmath28 , etc . since they can not be defined exactly by experiment . indeed , for instance , full reconstruction of kinematics is a practically impossible task because of neutral particles , neutrinos , the more so as @xmath29 is considered . we suppose that nothing will happen if @xmath4 is measured with @xmath30 accuracy since @xmath31 is easily attainable in the vhm region . besides it is practically impossible to deal with theory which operates by the @xmath32 ( @xmath33 ! ) variables . artificial reduction of the set of the necessary variables may lead to a temporary success only . indeed , the last thirty years of multiple production physics development was based on the inclusive approach , when the measured quantities ( cross sections ) depend on a few dynamical parameters only . but later on the experiment and its fractal analyses shows that the situation is not so simple , also as , for instance , for the classical turbulence . so , the event - by - event experimental data shows that the particle density fluctuation is unexpectedly large and the fractal dimension @xmath34 is not equal to zero . we know that if the fractal dimension is non - trivial , then the system is extremely ` non - regular ' . so , @xmath350.3 for the perimeter of great britain and @xmath350.5 for norway . the discrepancy marks the fact that the shore of norway is much more broken then of great britain . it is noticeable that the fractal dimension @xmath34 crucially depends on the type of reaction , incident energy and so on . it is evident that one may choose from @xmath32 an arbitrary finite set of variables to characterize the multiple production process . but the fractal analyses shows that such approach would lead to the same effect as if one may hear , for example , only the first violin of mahler s music . so , it is important to understand when the restricted set of dynamical variables will admit to describe the process ( state ) @xmath36 . the same problem was solved in statistical physics , where the ` rough ' description by a restricted number of ( thermodynamical ) parameters is a basis of its success , see the discussion of rough variables description e.g. in the review of uhlenbeck . we will search the same solution desiring to build a complete theory of the vhm hadron reactions . we want to note that just vhm process may be in this sense ` simple ' : at all evidence , the system becomes ` quiet ' in the vhm region and by this reason its ` rough ' thermodynamical description is available . it seems natural , therefore , to start investigation of multiple production phenomena from the ( extremely rare ) vhm processes . the dominant inelastic hadron processes at @xmath15 are saturated by production of low transverse momentum hadrons . one of the approaches explains this phenomena by the nonperturbative effect of quarks created from vacuum . corresponding dynamics looks as follows . at the expanse of transverse kinetic motion color charges may separated at large distances . nevertheless the transverse motion is suppressed since separation leads to increasing polarization of vacuum , because of confinement phenomenon . then , as in qed , the vacuum becomes unstable in regard to the tunneling creation of real fermions . just on their creation , the transverse kinetic energy is spent and , as a result , particles can not have , with exponential accuracy , high transverse energies . this picture is attractive being simple and transparent , but despite numerous efforts , there is no quantitative description of this phenomenon till now . briefly , the problem is connected with the unknown mechanism of strong coloured electric fields formation among distant colored charges , see also . one may use other terms . the soft channel of multiple production means the long - range correlation among hadron coloured constituents . under this special correlation the non - abelian gauge field theory conservation laws constraint were implied . they are important in dynamics since each conservation law decreases the number of independent degrees of freedom at least on one unity ( this may explain why hadrons @xmath37 ) , i.e. it has the nonperturbative effect . moreover , in so - called integrable systems each independent integral of motion ( in involution ) reduces the number of degrees of freedom on two units . as a result there is no thermalization in such systems and the corresponding mean multiplicity @xmath38 should be equal to zero , see appendix k. existence of multiple production , @xmath39 , testifies the statement that the thermalization phenomena exist in hadron processes , i.e. the system of yang - mills fields is not completely integrable . but the most probable process with @xmath15 did not lead to the final state with maximal entropy since @xmath40 , i.e. the definite restrictions on the dissipation dynamics should be taken into account . such problems , being intermediate , are mostly complicated ones . the quantitative theory of this phenomena may lead to deep revision of the main notions of existing quantum field theory , see appendix k. so , the dynamical display of hidden conservation laws of the hadron system are probably unstable since we expect that the system is not completely integrable , solitary field configurations @xmath41 . then the quantum theory should be able to describe quantum excitations of these fields , i.e. to count the fluctuations of ` curved ' manifolds . the canonical perturbation theory methods , formulated in terms of creation and annihilation of particles in the external field @xmath41 , are too complicated , see and references cited therein . for this reason , existing calculations usually do not exceed the semiclassical approximation . we hope that , as described in appendix k , the quantization scheme would be able to solve this problem ( see also the example described there ) . another approach assumes that the special ` @xmath42-channel ' ladder type feynman diagrams are able to describe the @xmath15 region . this approach did not take into account the confinement nonperturbative effects introducing the hardly controllable supposition that the free quarks and gluons may form a complete set of states . formally this is right , but at all evidence , the decomposition on this fock basis is realized in the non - abelian gauge field theories on zero measure ( see appendix k ) . nevertheless one may reject this argument assuming that the process is happening on a sufficiently small distances . corresponding contribution came from the so called ` hard pomeron ' . but the intrinsic problems of the accuracy of chosen logarithmic approximations , the understanding of the so called non - logarithmic corrections , of the fate of the infrared divergences stay unsolved till now . so , the multiple production of soft hadron phenomena seems unsolvable on the day - to - day level of understanding and we may with pure conscience move it away . this is why the hadron inelastic reactions lost some popularity , migrating the last two decades to the class of ` non - interesting ' problems . yet , in a number of modern fundamental experiments , multiple production plays , at least , the role of background to the investigated phenomena and for this reason we should be ready for the quantitative estimation of it . our hope to describe such a complicated problem as the multiple production phenomenon in the vhm conditions is based on the following idea . at the very beginning of this century , a couple p. and t.euhrenfest , had offered a model to visualize boltzmann s interpretation of the irreversibility phenomena in statistics . the model is extremely simple and fruitful . it considers the two boxes with @xmath43 numerated balls . choosing the label of the balls @xmath44 one must take the ball with the corresponding label from one box and put it into another one . one may repeat this action an arbitrary number of times @xmath42 . starting from the highly ` nonequilibrium ' state with all balls in one box , @xmath45 , it is seen stationary with @xmath42 tendency to equalization of the number of balls in the boxes fig.1 . the stationarity means that the number of balls in the other box rises @xmath46 at least on an early stage of the process . this signifies presence of an @xmath47 flow ( of balls ) toward the preferable ( equilibrium ) state . one can hope that this model reflects a physical reality of nonequilibrium processes with initial state very far from equilibrium . a theory of such processes with ( irreversible ) flow toward a state with maximal entropy should be sufficiently simple being close to the stationary markovian . the vhm production process may be , at least on an early stage , stationary markovian . if this is so then one may neglect long - range effects , non - perturbative as well , since they are not markovian as follows from the experience described in sec.2.1.2 . this is possible if the vhm process is happening so fast ( being the short - range phenomenon ) that the confinement forces became ` frozen ' . it can be shown that the quantitative reason for this phenomena is a fast ( exponential ) reduction with @xmath4 of the soft channel contribution into the hadron production process . so , we expect a change of the multiple production dynamics in the vhm region . thus , the main input idea consists from two general propositions . the first of them is following : ( * i * ) _ the hadron vhm production processes should be close to the stationary markovian . _ ` freezing ' the confinement constraints , the entropy @xmath8 may exceed for given energy @xmath5 its available maximum in the vhm domain . then one can assume that the vhm final state is in ` equilibrium ' , or is close to it . so , ( * ii * ) _ the vhm final state should be close to equilibrium _ + is our second basic proposition . we would select and appreciate particle physics models in accordance with these propositions . the field - theoretical description of statistical systems at a finite temperature is based usually on the formal analogy between imaginary time and inverse temperature @xmath48 . this analogy is formulated by schwinger as the ` euclidean postulate ' and it assumes that ( i ) the system is in equilibrium , i.e. it should allow the arbitrary rearrangement of states of temporal sequence in the described process , and ( ii ) there is not special space - time long - range correlations among states of the process , i.e , for instance , the symmetry constraints should not play a crucial role . we do not know _ ad hoc _ whether or not to apply the ` euclidean postulate ' for given @xmath4 and @xmath7 , even if ( ) is satisfied . for this reason we are forced to formulate the theory in the natural real - time terms . the first important quantitative attempt to build the real - time finite - temperature field theory discovered the formal problem of the so called ` pinch - singularities ' . further investigation of the theory has allowed to demonstrate the cancellation mechanism of these unphysical singularities . this is attained by doubling the degrees of freedom : the green functions of the theory represent @xmath49 @xmath50 @xmath49 matrix . it surely makes the theory more complicated , but the operator formalism of the thermo - field dynamics shows the unavoidable character of this complication . notice that the canonical real - time finite temperature field - theoretical description of the statistical systems based on the kubo - martin - schwinger ( kms ) boundary condition for a field is : ( t)=(t - i ) . 22.1 it , without fail , leads to the @xmath51 fluctuation - dissipation conditions ( see also ) . by this reason it can not be applied in our case , where the dissipation problem is solved . the origin of this boundary condition is shown in appendix a. we will use a more natural microcanonical formalism for particle physics . the thermodynamical ` rough ' variables are introduced in these approach as the lagrange multipliers of corresponding conservation laws . the physical meaning of these ` rough ' variables are defined by the corresponding equations of state . we shall use the @xmath0-matrix approach which is natural for the description of the time evolution . ( the @xmath0-matrix description was used also in . ) for this purpose the amplitudes = a_nm(p , q ) 22.2 of the @xmath52- into @xmath4-particles transition will be introduced . the in- and out - states must be composed from mass - shell particles . using these amplitudes we will calculate & r_nm(p , q)=|a_nm(p , q)|^2= + & = < p_1,p_2, ... ,p_n|q_1,q_2, ... ,q_m > < q_1,q_2, ... ,p_n>. 22.3this will lead to the doubling of the degrees of freedom . the temperature description will be introduced ( see also ) noting that , for instance , & d_n=|a_nm(p , q)|^2d_n(p ) , + & d_n(p)=_1^n , ( p)= ( p^2 + m^2)^1/2 , 22.4 is the differential measure of the final state . it is a first example where the usefulness of the probability - like quantity @xmath53 is seen . the measure ( ) is defined on the energy - momentum shell : _ i=1^np_i = p . 22.4ait should be underlined that @xmath54 are the translationally invariant amplitudes and four equalities : _ i=1^np_i=_i=1^mq_i 22.4bare obeyed identically . so , ( ) are the constraints and to take them into account one may multiply @xmath55 on @xmath56 where @xmath57 is the time - like 4-vector . it is evident that integration over @xmath57 with factor @xmath58 gives the constraints ( ) . one may simplify the calculation assuming that all calculations are performed , for example , in the cm frame @xmath59 . then one may ignore the space components considering @xmath60 . this is the equivalent of the assumption that only the energy conservation law is important . the last step is the substitution @xmath61 , where @xmath62 is our lagrange multiplier . to define its physical meaning one should solve the equation of state : e =- d_n_k=1^ne^-(p_k)-d_n ( ) . 3such a definition of temperature as the lagrange multiplier of the energy conservation law is obvious for microcanonical description . the initial - state temperature will be introduced in the same way , taking into account ( ) . so , we will construct the two - temperature theory . it is impossible to use the kms boundary condition in such a two temperatures description ( the equation of state can be applied at the very end of the calculations ) . it should be notice that the ` density matrix ' @xmath63 , defined in ( [ 22.3 ] ) , describes the ` closed - path motion ' in the functional space . so , if = < n , out|e^is(_+)|m , in > o1and ^*= < q_1,q_2, ... ,p_n>= + = < m , in|e^-is(_-)|n , out > o2then , by definition , _+(_)=_-(_)= ( _ ) , o3with some ` turning - point ' fields @xmath64 , where @xmath65 is the remote hypersurface . the value of @xmath66 specifies the environment of the system . we will show that ( ) coincides with the kms boundary condition in some special cases . here consequences of the vacuum boundary condition : (_)=0 03aare analyzed . one should admit also that below boundary conditions are not unique : one can consider arbitrary organization of the environment of the considered system . the @xmath0-matrix interpretation is able to show the way as an arbitrary boundary condition may be adopted . this should extend the potentialities of the real - time finite - temperature field - theoretical methods . it seems useful to illustrate the above microcanonical approach by the simplest example , see also . by definition , the @xmath4 particles production cross section _ n(s)=d_n(p)(q_1+q_2-^n_i=1p_i)|a_n(p , q)|^2 , 18where @xmath67 is the ordinary @xmath4 particle production amplitude in accelerator experiments . considering the fourier transform of energy - momentum conservation @xmath68-function one can introduce the generating function @xmath69 , see and references cited therein . we may find in the result that @xmath70 is defined by the equality : _ n(e)=^+i_-ie^e_n ( ) , e=(q_1)+(q_2 ) , 3awhere _ n()= \{^n_i=1 } |a_n|^2=d_n ( ) . 19most probable value of @xmath62 in ( ) is defined by the equation of state ( ) . inserting ( ) into ( ) we find the expression ( ) if the momentum conservation shell is neglected . the last one is possible since the cross sections are always measured in the definite frame . let us consider the simplest example of noninteracting particles : @xmath71 where @xmath72 is the bessel function . inserting this expression into ( ) we can find that in the nonrelativistic case ( @xmath73 ) @xmath74 i.e. , we find the well known equality : e_kin = t , o4where @xmath75 is the mean kinetic energy and @xmath76 is the temperature ( boltzmann constant was taken equal to one ) . it is important to note that the equation ( ) has a unique real solution @xmath77 rising with @xmath4 and decreasing with @xmath7 . the expansion of the integral ( ) near @xmath77 unavoidably gives an asymptotic series with zero convergence radii since @xmath78 is the essentially nonlinear function of @xmath62 , see also sec.2.2.2 . this means that , generally speaking , fluctuations in the vicinity of @xmath77 may be arbitrarily high and in this case @xmath77 has no physical sense . but if the fluctuations are gaussian , then @xmath78 coincides with the partition function of the @xmath4 particle state and @xmath77 may be interpreted as the inverse temperature . we will put the observation of this important fact in the basis of our thermodynamical description of the vhm region . the notion of ` equilibrium ' over some parameter @xmath79 in our understanding is a requirement that the fluctuations in the vicinity of its mean value , @xmath80 , have a gaussian character . notice , in this case , one can use this variable for a ` rough ' description of the system . we would like to show now that the corresponding equilibrium condition would have the meaning of the correlations relaxation condition of bogolyubov , see also . let us define the conditions when the fluctuations in the vicinity of @xmath81 are gaussian . firstly , to estimate the integral ( ) in the vicinity of the extremum , @xmath81 , we should expand @xmath82 over @xmath62 : _ n(+_c)=_n(_c)-+^2 _ n(_c)-^3_n(_c)+ ... 21cand , secondly , expand the exponent in the integral ( ) over , for instance , @xmath83 etc . in the result , if higher terms in ( ) are neglected , @xmath84-th term of the perturbation series _ n , k~\{}^k ( ) . 22ctherefore , because of euler s @xmath85 function , the perturbation theory near @xmath81 leads to the asymptotic series . the supposition to define this series formally , for instance , in the borel sense is not interesting from physical point of view . indeed , such a formal solution assumes that the fluctuations near @xmath81 may be arbitrarily high . then , by this reason , the value of @xmath81 loses its significance : arbitrary values of @xmath86 are important in this case . nevertheless it is important to know that our asymptotic series exists in some definite sense , i.e. we can calculate the integral over @xmath62 by expanding it over @xmath86 . therefore , if the considered series is asymptotic , we may estimate it by first term if /^3_c < < ( ^2_n(_c)/^2_c)^3/2 . 23cone of the possible solutions of this condition is /^3_c0 . 24cif this condition is satisfied , then the fluctuations are gaussian with dispersion @xmath87 see ( ) . let us consider now ( ) carefully . we will find by computing derivatives that this condition means the following approximate equality : -3 + 2 0 , 25cwhere @xmath88 means the @xmath84-th derivative . for identical particles , ^(k)_n(_c)=n^k(-1)^kd_n(_c)^k_i=1(q_i ) , 26ctherefore , the left hand side of ( ) is the 3-point correlator @xmath89 since @xmath90 is a density of states for given @xmath62 : & k_3d_3(q)(<^3_i=1(q_i)>__c- 3<^2_i=1(q_i)>__c < ( q_3)>__c+ . + & + .2 ^ 3_i=1<(q_i)>__c ) , 27cwhere the index @xmath81 means that averaging is performed with the boltzmann factor @xmath91 . notice , in distinction with bogolyubov , @xmath89 is the @xmath92 correlation function . so , in our interpretation , one can introduce the notion of temperature @xmath93 if and only if the @xmath94 energy flows , measured by the corresponding correlation functions , are to die out . as a result , to have all the fluctuations in the vicinity of @xmath81 gaussian , we should have @xmath95 , @xmath96 . notice , as follows from ( ) , the set of minimal conditions actually looks as follows : k_m<<k_2 , m3 . 27dif the experiment confirms this conditions then , independently from the number of produced particles , the final state may be described with high enough accuracy by one parameter @xmath81 and the energy spectrum of particles is gaussian . in this conditions one may return to the statistical and the hydrodynamical models . considering @xmath81 as a physical ( measurable ) quantity , we are forced to assume that both the total energy of the system , @xmath97 , and the conjugate to it , variable @xmath81 , may be measured simultaneously with high accuracy . we would like to show now that the ordinary big partition function of the statistical system coincide with _ n , m_(_1,z_1;_2,z_2 ) r_nm(p , q)=(,z ) , densmwhere @xmath98 is defined by ( ) . the summation and integration are performed with constraints that the mean energy of particles in the initial(final ) state is @xmath99 . on may interpret @xmath100 in the first approximation as the temperature and @xmath101 as the activity for initial(final ) state . direct calculation , see appendix b , gives the following expression for generating functional : ( , z)=e^-i*n*(_i^*_j)r_0 ( ) , 22.22awhere the particle number operator ( @xmath102 ) & * n*(_i^*_j)=- dxdx(_+(x)d_+-(x - x,_2,z_2)_-(x)- + & -_-(x)d_-+(x - x,_1,z_1)_+(x ) ) p22and r_0()=z(_+)z^ * (-_-) , p23where @xmath103 is defined in ( ): @xmath104 and , for the vacuum boundary condition @xmath105 , d_+-(x - x,,z)=-id_1(q)e^iq(x - x)e^-(q)z(q ) 22.23a + d_-+(x - x,,z)=id_1(q)e^-iq(x - x)e^-(q)z(q ) 22.24aare respectively the positive and negative frequency correlation functions at @xmath106 . it is evident , r_nm(p , q)=. _ k=1^n\{e^_1(p_k ) } _ k=1^m\{e^_2(q_k ) } ( , z)|_z_i=0 . p2notice , defining @xmath98 through the generating functional we extract the boltzmann factors @xmath107 since the energy - momentum conservation @xmath68-functions were extracted from amplitudes @xmath54 . we suppose that @xmath103 may be computed perturbatively . as a result , ( @xmath108 is the variational derivative ) r(,z)= e^-iv(-i_+)+iv(-i_- ) e^ dx dxj_i ( x)d_ik(x - x;,z)j_k ( x ) , 22.29awhere @xmath109 is the matrix green function . these green functions are defined on the mills time contours @xmath110 in the complex time plane ( @xmath111 ) , see fig.2 . this definition of the time contours coincides with the keldysh time contour . the generating functional ( ) has the same structure as the generating functional of niemi and semenoff . the difference is only in the definition of green functions @xmath112 . this choice is a consequence of the boundary condition ( ) . so , if ( ) is used , then the green function is defined by eq . notice also that if @xmath113 then a new green function obey kms boundary condition , see ( ) . following niemi and semenoff one can write ( ) in the form : ( ) = d_nse^is_ns( ) , nisewhere the functional measure @xmath114 and the action @xmath115 are defined on the closed complex time contour @xmath116 , see fig.3 . the choice of initial time @xmath117 and @xmath118 is arbitrary . then one can perform shifts : @xmath119 and @xmath120 . in result , ( * i * ) if @xmath113 , ( * ii * ) if contributions from imaginary parts @xmath121 and @xmath122 of the contour @xmath116 have disappeared in this limit , ( * iii * ) if the integral ( ) may be calculated perturbatively then this integral is a compact form of the representation ( ) . notice that the requirements ( * i * ) - ( * iii * ) are the equivalent of the euclidean postulate of schwinger . in this frame one can consider another limit @xmath123 . then the @xmath116 contour reduces to the matsubara imaginary time contour , fig.4 . later on we will use this @xmath0-matrix interpretation of thermodynamics . but one should take in mind that corresponding results will hide assumptions ( * i * ) - ( * iii * ) . we would like to mention the ambivalent role of external particles in our @xmath0-matrix interpretation of thermodynamics . in the ordinary matsubara formalism the temperature is measured assuming that the system under consideration is in equilibrium with the thermostat , i.e. temperature is the energy characteristics of @xmath124 particles . in our definition the temperature is the mean energy of produced , i.e. non - interacting , particles . it can be shown that both definitions lead to the same result . explanation of this coincidence is the following . let us consider the point of particle production as the coordinate of fictitious ` particle ' . this ` particle ' interacts since the connected contributions into the amplitudes @xmath125 only are considered , and has the equal to produced particles momentum and so on . the set of these ` particles ' form a system . interaction among these ` particles ' may be described by corresponding correlation functions , see sec.2.3.3 . let us consider now the limit @xmath123 . in this limit ( ) reduces to ( ) = d_me^-s_m( ) , matswhere the imaginary time measure @xmath126 and action @xmath127 are defined on the matsubara time contour . the periodic boundary condition ( ) should be used calculating integral ( ) . the rules and corresponding problems in the integral ( ) can be calculated as described in many textbooks , see also . in the limit considered the time was eliminated in the formalism and the integral in ( ) performed over all states of the ` particles ' system with the weight @xmath128 . notice , the doubling of degrees of freedom has disappeared and our fictitious ` particles ' became real ones . on the other hand , the produced particles may be considered as the probes through which we measure the interacting fields . as was mentioned above , their mean energy defines the temperature , if the energy correlations are relaxed . if even one of the conditions ( * i * ) - ( * iii * ) is not satisfied then one can not reduce our @xmath0-matrix formalism to the imaginary time matsubara theory . then one can ask : is there any possibility , staying in the frame of @xmath0-matrix formalism , to conserve the statistics formalism . this question is discussed in the appendix c , where the wigner functions approach is applied . it may be shown that the formalism may be generalized to describe the kinetic phase of the nonequilibrium process , where the temperature should have the local meaning . the comparison with the ` local equilibrium hypothesis ' is discussed at the end of appendix c. our further consideration will based on the model independent ( formal ) classification of asymptotics . we will consider the generating function : t(s , z)=_n=1^n_max z^n_n(s ) , s=(p_1+p_2)^2>>m^2 , n_max=/m . 21.3this step is natural since the number of particles is not conserved in our problem . so , the total cross section and the averaged multiplicity will be : _ tot(s)=t(s,1)=_n_n(s ) , = _ n n ( _ n(s)/_tot)= .t(s , z)|_z=1 . 21.4 at the same time , the inverse mellin transform gives _ n=.t(s , z)|_z=0= t(s , z)= e^(-nz+t(s , z ) ) . 21.5the essential values of @xmath129 in this integral are defined by the equation ( of state ) : n = zt(z , s ) . 21.6taking into account the definition of the mean multiplicity @xmath3 , given in ( ) , we can conclude that the solution of ( ) @xmath130 is equal to one at @xmath131 . therefore , @xmath132 are essential in the vhm domain . the asymptotics over n ( @xmath133 is assumed ) are governed by the smallest solution @xmath130 of ( ) because of the asymptotic estimation of the integral ( ): _ n(s)e^-nz_c(n , s ) . 21.7 let us assume that in the vhm region and at high energies , @xmath134 , there exist such a value of @xmath135 that we can neglect in ( ) the dependence on the upper boundary @xmath136 . this formal trick with the thermodynamical limit allows to consider @xmath137 as the nontrivial function of @xmath129 for finite @xmath7 . then , it follows from ( ) that z_c(n , s)z_s at n , 21.8where @xmath138 is the leftmost singularity of @xmath137 in the right half plane of complex @xmath129 . one can say that the singularity of @xmath137 attracts @xmath135 if @xmath139vhm . we will put this observation in the basis of vhm processes phenomenology . we would like to underline once more that actually @xmath137 is regular for arbitrary finite @xmath129 if @xmath7 is finite . but @xmath135 behaves in the vhm domain as if it is attracted by the ( imaginary ) singularity @xmath138 . and just this @xmath135 defines @xmath70 in the vhm domain . we want to note that actually the energy @xmath5 should be high enough to use such an estimation . one can notice from the estimation ( ) that @xmath70 weakly depends on the character of the singularity . therefore it is enough to classify only the possible positions of @xmath138 . we may distinguish following possibilities : & ( * a * ) z_s= : _ n < o(e^-n ) + & ( * b * ) z_s=1 : _ n > o(e^-n ) + & ( * c * ) 1<z_s < : _ n = o(e^-n ) , 21.9i.e . , following this classification , the cross section may decrease faster ( * a * ) , slower ( * b * ) , or as ( * c * ) an arbitrary power of @xmath140 . it is evident , if all these possibilities may be realized in nature , then we should expect the asymptotics ( * b * ) . as was explained in sec.2.2.1 , @xmath70 has the meaning of the @xmath4 particle partition function in the energy representation . then @xmath137 should be the ` big partition function ' . taking this interpretation into account , as follows from lee - yang theorem , @xmath137 can not be singular at @xmath141 . at the same time , the direct calculations based on the physically acceptable interaction potentials give the following restriction from above : ( * d * ) _ n < o(1/n ) 21.2this means that @xmath70 should decrease faster than any power of @xmath142 . it should be noted that our classification predicts rough ( asymptotic ) behavior only and did not exclude local increase of the cross section @xmath70 . one may notice that -=z_c(n , s)+o(1/n ) . 21.10using thermodynamical terminology , the asymptotics of @xmath70 is governed by the physical value of the activity @xmath135 . one can introduce also the chemical potential @xmath143 . it defines the work needed for one particle creation , @xmath144 , where @xmath145 is the produced particles mean energy . so , one may introduce the chemical potential if and only if @xmath146 and @xmath135 may be used as the ` rough ' variables . then the above formulated classification has a natural explanation . so , ( * a * ) means that the system is stable with reference to particle production and the activity @xmath135 is the increasing function of @xmath4 , the asymptotics ( * b * ) may realized if and only if the system is unstable . in this case @xmath135 is the decreasing function . the asymptotics ( * c * ) is not realized in equilibrium thermodynamics . we will show that the asymptotics ( * a * ) reflects the multiperipheral processes kinematics : created particles form jets moving in the cm frame with different velocities along the incoming particles directions , i.e. with restricted transverse momentum , see sec.3.1.1 . the asymptotics ( * b * ) assumes the condensation - like phenomena , see sec.3.3 . the third type asymptotics ( * c * ) is predicted by stationary markovian processes with the pqcd jets kinematics , see sec.3.2.2 . the dis kinematics may be considered as the intermediate , see sec.3.2.1 . this interpretation of classes ( ) allows to conclude that we should expect reorganization of production dynamics in the vhm region : the soft channel ( * a * ) of particle production should yield a place to the hard dynamics ( * c * ) , if the ground state of the investigated system is stable with reference to the particle production . otherwise we will have asymptotics ( * b * ) . let us consider the system with several correlation scales . for example , in statistics one should distinguish correlation length among particles ( molecules ) and correlation length among droplets if the two - phase region is considered . in particle physics , one should distinguish in this sense correlation among particles produced in result of resonance decay and correlations among resonances . in pqcd one may distinguish correlations of particles in jet and correlation among jets . there exist many model description of this physical picture . in statistics mayer s group decomposition is well known . in particle physics one should note also the many - component formalism . we will consider the generating functions ( functionals ) formalism considering mostly jet correlations . in many respects it overlaps the above mentioned approaches . the generating function @xmath137 may be written in the form : t(z , s)=_k=1^c_k(s)= _ l=1^z^lb_l , g1where the coefficients @xmath147 are the moments of the multiplicity distribution p_n(s)=_n(s)/_tot(s ) . g2so , c_1(s)=_n np_n(s)= , c_2(s)=_n n(n-1)p_n(s)-^2 g3and so on . using the connection with the inclusive distribution functions @xmath148 : t(z , s)=_k=1^d_k(q ) f_k(q_1,q_2, ... ,q_k;s ) , g4it is easy to find that & c_1(s)=d_1(q ) f_1(q;s)=|f_1(s ) , + & c_2(s)=d_2(q ) \{f_2(q_1,q_2;s)-f_1(q_1;s)f_1(q_2;s)}= _ \{k}_l=0^(_i=1^lk_i - k ) _ i=1^l\ { } , g6where @xmath149 and @xmath150 one may invert formulae ( ): |f_k(s)= _ \{n_k}=0^(_k=1^kn_k - l ) _ k=1^()^n_k g7 the mayer s group coefficients @xmath151 in ( ) have the following connection with @xmath147 : b_l(s)=_k=0^c_k+l(s ) . g23 it seems useful to illustrate the effectiveness of the generating function method by the following example . we will consider the transformation ( multiplicity @xmath152 activity @xmath129 ) to show the origin of the koba - nielsen - olesen scaling ( kno - scaling ) . if @xmath153 , @xmath154 , then @xmath70 is described by the poisson formulae : _ k8it corresponds to the case of absence of correlations . let us consider more weak assumption : c_m(s)=_m(c_1(s))^m , k9where @xmath155 is the energy independent constant , see also , where a generalization of kno scaling on the semi - inclusive processes was offered . then t(z , s)=_m=1\{(z-1)|n(s)}^m . k10to find the consequences of this assumption , let us find the most probable values of @xmath129 . the equation of state @xmath156 has solution @xmath157 increasing with @xmath4 since @xmath137 is an increasing function of @xmath129 , if and only if , @xmath137 is nonsingular at finite @xmath129 . as was mentioned above , the last condition has deep physical meaning and practically assumes the absence of the first order phase transition . let us introduce a new variable : = ( z-1)|n(s ) . k12the corresponding equation of state looks as follows : = ( 1+)t ( ) . k13so , with @xmath158 accuracy , one can assume that _ c(n/|n(s ) ) . k14are essential . it follows from this estimation that such scaling dependence is rightful at least in the neighborhood of @xmath106 , i.e. in vicinity of main contributions into @xmath159 . this gives : |n(s)_n(s)=_tot(s)(n/|n(s ) ) , k15where ( n/|n(s))t(_c(n/|n(s)))\{n/|n(s)_c(n/|n(s))}o(e^-n ) k16is the unknown function . the asymptotic estimation follows from the fact that @xmath160 should be a nondecreasing function of @xmath4 , as follows from nonsingularity of @xmath137 . the estimation ( ) is right at least at @xmath161 . the range validity of @xmath4 , where solution of ( ) is acceptable , depends from exact form of @xmath137 . indeed , if @xmath162 , @xmath163 , then ( ) is right at all values of @xmath4 and it is enough to have the condition @xmath161 . but if @xmath164 , @xmath163 , then ( ) is acceptable if and only if @xmath165 . representation ( ) shows that just @xmath38 is the natural scale of multiplicity @xmath4 . this representation was offered first as a reaction on the so called feynman scaling for inclusive cross section : f_k(q_1,q_2, ... ,q_k)~^k_i=1 . k17 as follows from estimation ( ) , the limiting kno prediction assumes that @xmath166 . in this regime @xmath137 should be singular at @xmath167 . the normalization condition @xmath168 gives : @xmath169 , where @xmath170 is the constant . notice , such behavior of the big partition function @xmath137 is natural for stationary markovian processes described by logistic equations . in the field theory such equation describes the qcd jets . we wish to generalize expansion ( ) to take into account the possibility of many - component structure of the multiple production processes . let us consider particle production through the generation , for instance , of jets . in this case decay of a particle of high virtuality @xmath171 forms a jet of lower virtuality particles . it is evident that one should distinguish correlation among particles in the jet , and correlation among jets . let @xmath172 be the @xmath173 that @xmath174-th jet of mass @xmath175 includes @xmath176 particles , @xmath177 , where _ i=1^n_jn_i = n g8the jets are the result of particles decay . then let us assume that @xmath178 defines the mean number of jets of mass @xmath175 and momentum @xmath179 : & ^(1)_n()=_n_j_\{n}_n_j ( _ i=1^n_jn_i - n ) + & _ i=1^n_j \ { e^-(p_i)|n_1(m_i , p_i)_n_i(m_i ) } , g9where @xmath180 . notice the boltzmann factor @xmath107 , where @xmath181 is the jets energy , play the same role as the corresponding factor in ( ) and introduced to take into account the energy conservation law . we consider the vhm domain and for this reason @xmath182 is assumed . it is useful to avoid the particles number conservation law ( ) . for this purpose we will introduce ^(1)(,z)=_n^(1)_n()=\ { e^-(p)|n_1(m , p)(t(z , m)-1 ) } , g10where t(z , m)=_n z^n_n(m ) . g11comparing ( ) with ( ) we may conclude that @xmath183 plays the role of activity of jets . then the generalization is evident : & ( , z)=_k^(k)(,z)= + & = \{_k_i=1^k\{dm_id_1(p_i ) e^-(p_i)(t(z , m_i)-1)}. + & .|n_k(m_1,p_1, ... ,m_k , p_k ) } , g12where @xmath184 has the same meaning as @xmath147 , i.e. @xmath184 is the correlation function of @xmath84 jets . let us consider the following ` bootstrap ' regime when @xmath185 is defined by the equation : ( , z ) e^-(p)|n_1(m , p)t(z , m ) . g13inserting here the strict expression ( ) we find a nonlinear equation for @xmath183 . the solution of ( ) assumes that e^-(p)|n_1 t > > \{e^-(p ) |n_1t}^2 g14and e^-|n_1 t > > ( e^-t ) ( e^-t ) |n_2 g15 to solve eq . ( ) in the vhm region , where the leftmost singularity over z is important , let us consider the @xmath186 : t(z , m)= , _ 0>0 , g16where @xmath187 is the polynomial function of @xmath129 , @xmath188 . using the normalization condition : |n_j=.t(z , m)|_z=1 g17we can find : a(m)_0=|n_j-(1,m ) , (1,m).(z , m)|_z=1 . g18 the partition function of the jet @xmath183 defined by @xmath186 ( ) is singular at z_s(m)=1 + . g19this singularity would be significant in the vhm region if @xmath189 is decreasing function of @xmath52 . this means an assumption that @xmath190 so , in first approximation we will choose a(m)=|n_j(m ) . g20this choice may be confirmed by concrete model calculations . taking into account the energy conservation law , conditions ( ) and ( ) are satisfied if \{-n } < g21at @xmath139vhm . therefore , ( ) obey eq . ( ) with exponential accuracy in the vhm region , i.e. if n > > = . g22we assume here that one can find so large @xmath4 and @xmath7 that with exponential accuracy the factors @xmath191 did not play an important role . but at low energies the condition ( ) is important and the factors @xmath191 should be taken into account . notice now important consequence of our ` bootstrap ' solution : it means that we can leave production of the heavy jets only , if @xmath192 . on the other hand , let us choose @xmath193 , where @xmath194 is the function of @xmath7 , and consider @xmath195 . then the condition ( ) defines @xmath196 : if z_0 > > g22then we are able to obey the inequalities ( ) and ( ) . the jet mean multiplicity , see sec.3.2.2 , |n_j(s)~. g26then = \{e^/-1}^-1~<<|n_j(s ) . g50at @xmath195 . therefore , ( ) may be satisfied outside the vhm domain . let us compare now the solutions of the equation of state . inserting ( ) into ( ) we can find for a jet of mass @xmath5 that z^1_c = z^1_s-=1 + - . g24the two - jet contribution of the masses @xmath197 gives : z^2_c = z^2_s-=1 + - . g25at arbitrary finite energies @xmath198 and , as follows from ( ) , they decrease @xmath199 with energy . noting the normalization condition , @xmath200 , and assuming that the vacuum is stable , i.e. @xmath201 , we can conclude that if @xmath202 then @xmath138 attracts @xmath130 , i.e. @xmath203 , and if @xmath204 then this contributions should be significant in @xmath159 ; if @xmath195 , then @xmath204 , and if @xmath4 satisfy the inequality ( ) , or if @xmath196 satisfy the inequality ( ) , then considered contributions are significant in @xmath159 . + it is the ( energy - multiplicity ) asymptotics equivalence principle . one of the simplest consequences of this principle is the prediction that the mean transverse momentum of created particles should increase with multiplicity at sufficiently high energies . this principle is the consequence of independence of contributions in the vhm domain on the type of singularity in the complex @xmath129 plane and of the energy conservation law . just the last one shifts the two - jet singularity to the right and @xmath205 . we would like to notice also that this effect , when the mostly ` energetic population ' survives was described mathematically by v.volterra . intuitively evident is that one may find the ` energetic population ' searching the vhm one , or , it is the same , giving it a rich supply , i.e. to give the population enough energy . this is our ( energy - multiplicity ) equivalence ( ( @xmath206)-equivalence ) principle . notice , if the amount of supply is too high then few populations may grow . this is the case when the difference @xmath198 tends to zero at high energies . we would like to note that singular at finite @xmath129 partition functions was predicted in the @xmath207-theory , in qcd jets , in the generalized bose - einstein distribution model . in all of this models decay of the essentially nonequilibrium initial state ( highly virtual parton , heavy resonance , etc ) was described . one may distinguish the phases of the media by a characteristic correlation length . then the phase transition may be considered as the process of changing correlation length . our ` bootstrap ' solution predicts just such phenomena : at low multiplicities the long - range correlations among light jets is dominant . the ` bootstrap ' solution predicts that for vhm processes just the short - range correlations among particles of the heavy jet become dominant . the ( @xmath206)-equivalence means that this transition is a pure dynamical effect . multiple production phenomena was first observed more than seventy years ago . during this time a vast experimental information was accumulated concerning hadron inelastic interactions , see the review papers . now we know that at high energy @xmath208 : + ( i ) the total cross section @xmath209 of hadron interactions is enhanced almost completely by the inelastic channels ; + ( ii ) the mean multiplicity of produced hadrons @xmath3 slowly ( logarithmically ) grows with @xmath208 ; + ( iii ) the interaction radii of hadrons @xmath210 slowly ( logarithmically ) increase with energy . ; + ( iv ) the multiplicity distribution @xmath19 is wider than the poisson distribution . ; + ( v ) the mean value of transverse momentum @xmath84 of produced hadrons is restricted and is independent of the incident energy @xmath5 and produced particle multiplicity @xmath4 ; + ( vi ) the one - particle energy spectrum @xmath211 . first of all , ( i ) means that the high energy hadron interaction may be considered as the ordinary dissipation process . in this process the kinetic energy of incident particles is spent in produced particle mass formation . the vhm process takes place in the vacuum and then it was assumed on the early stages that the multiple production phenomena reflects a natural tendency of the excited hadron system to get to equilibrium with the environment . in this way one can introduce as a first approximation the model that the excited hadron system evolves without any restrictions . in this model we should have @xmath212 . the dissipation is maximal in this case and the entropy @xmath8 exceeds its maximum . this simple model has definite popularity up to 70-th . but the experimental data ( ii ) and ( iii ) prohibit this model and it was forgotten . choosing the model we would like to hope that the considered model takes into account experimental conditions ( i ) - ( vi ) in the @xmath15 domain ; has natural asymptotics over multiplicity to the vhm region . + it is necessary to remember also that new channels of hadron production may arise in the vhm region . + it is impossible to understand all possibilities without those offered in sec.2.3.2 in the classification of asymptotics . thus , we will observe predictions of _ multiperipheral models _ , + distinguishing the soft pomeron models , see . _ the dual - resonance model _ + predictions for the vhm region are described also . it can be shown that this models predict asymptotics ( a ) if @xmath213 . just this result explains why vhm domain is defined by the condition ( ) . we are forced since it allows to include pqcd , forbidden by the multiperipheral models , considering _ hard pomeron model _ + production of mini - jets . but we will find using monte carlo simulations that the pqcd pomeron is unable to adopt the hard channels of hadron production . then _ the deep inelastic processes _ + for vhm region will be considered to generalize the dglap kinematics in the case of heavy qcd jets production . the analysis shows that transition to the vhm leads to the necessity to include low-@xmath214 sub - processes . as a result we get out of the range of pqcd validity . _ multiple reduction of jets . _ + we will see , that at very high energies in the vhm region the heavy jets creation should be a dominant process if the vacuum is stable with reference to the particle production . we will consider also decay of the ` false vacuum ' to describe the consequence of _ phase transition _ + in the vhm domain . this channel is hardly seen for @xmath15 since the confinement constraints may prevent cooling of the system up to phase transitions condition . later on multiple production physics was developed on the basis of experimental observation ( iv ) . the regge pole model naturally explains this experimental data and , at the same time , absorbs all experimental information , ( i)-(vi ) . at the very beginning , adopting the regge poles notion without its microscopical explanation , this description was self - consistent . the efforts to extend the regge pole model to the relativistic hadron reactions was ended by the reggeon diagram technique , see and references cited therein , and it was used later to construct the perturbation theory for @xmath70 . it was shown that the multiplicity distribution is wider then the poissonian because of the multi - pomeron exchanges . the leading energy asymptotics pomeron contribution reflects the created particle kinematics described in appendix d , where the available kinematical scenario in the frame of pqcd are described . so , the longitudinal momentum of produced particles is large and is strictly ordered . at the same time , particles transverse momentum is restricted . let us consider the inelasticity coefficient introduced in ( ) @xmath215 , where @xmath9 is the energy of the fastest particle in the laboratory frame . then the strict ordering of particles in the pomeron kinematics means that @xmath216 is independent of the index of the particle . so , if the fastest particle has the energy @xmath217 , then the following particle should have the energy @xmath218 , and so on . following to this law , @xmath219-th particle would have the energy @xmath220 . in the laboratory frame the energy should degrade to @xmath221 . inserting here the above formulated estimation of @xmath222 we can find that if the number of produced particles is n_0 , n_0=-(1-)^2>0 , = ( s / m^2 ) , 1.1athen we may expect the total degradation of energy . this degradation is the necessary condition noting that the total cross section of slowly moving particles may depend only slightly on energy and is seems necessary for natural explanation of the weak dependence of the hadron cross sections on the energy . this consideration would be lorentz - covariant if one can find the slowly moving particle in an arbitrary frame . resulting estimation of mean multiplicity have good qualitative experimental confirmation . notice that it was assumed deriving ( ) that the energy degrades step by step . in other words , if we introduce a time of degradation , then the time @xmath223 is needed for complete degradation of energy . assuming the random walk in the normal to incident particle plane , we can conclude that the points of particle production are located on a disk ( in the moving frame ) of radii @xmath224 . this means that the interaction radii should grow with the energy of the colliding particles . if @xmath225 is the cross section to observe particle @xmath226 inclusively in the @xmath227 and @xmath228 particles collision , then it was found experimentally that the ratio = = 1.2is universal . this may be interpreted as the direct evidence of fact that the hadron interactions have a large - distance character , i.e. that the interaction radii should be large . this picture assumes that the probability to have total degradation of energy is ~e^-b^2/4 , 1.3where @xmath229 is some dimensional constant ( the slope of regge trajectory ) and @xmath230 is the 2-dimensional impact parameter . this formulae has also the explanation connected to the vacuum instability with reference to the real particle production in the strong colour electric field . the above picture has natural restrictions . we can assume that each of the produced particles may be the source of above described @xmath42-channel cascade of the energy degradation . this means that in the frame of the pomeron phenomenology , we are able to describe the production of n < ^2 1.4particles only . if @xmath231 then the density of particles in the diffraction disk becomes large and ( a ) one should introduce short - distance interactions , or ( b ) rise interaction radii . it will be shown that just ( a ) is preferable . we will build the perturbation theory in the phenomenological frames ( i ) - ( vi ) . considering the system with variable number of particles the generating function t(z , s)=_n z^n _ n(s ) i1would be useful . one can use also the decomposition : t(z , s)=_tot(s)\{(z-1)c_1(s)+(z-1)^2c_2(s)+ ... } , i2where , by definition , c_1(s)==.t(z , s)|_z=1 i3is the mean multiplicity , c_2(s)=.t(z , s)|_z=1 i4is the second binomial momentum , and so on . our idea is to assume that all @xmath232 , @xmath154 may be calculated perturbatively choosing p(0,s)=e^(z-1 ) i5as the born approximation ` superpropagator ' . it is evident that ( ) leads to poisson distribution . then , having in mind ( ii ) , ( iii ) and ( v ) we will use following @xmath186 : p(q , s)=e^(o)-q^2se^(z-1 ) , i6where the transverse momentum @xmath233 is conjugate to the impact parameter @xmath230 . so , the born term ( ) is a fourier transform of the simple product of ( ) and ( ) . it contains only one free parameter , the pomeron intercept @xmath234 . on the phenomenological level it is not important to know the dynamical ( microscopical ) origin of ( ) . for our purpose the laplace transform of @xmath235 would be useful . if @xmath236 , then ( , q^2)=_0^de^-p(q^2,s)= . i8it is the propagator of two - dimensional field theory with mass squared @xmath237 knowing the gribov s reggeon calculus completed by the abramovski - gribov - kancheli ( agk ) cutting rules one can investigate the consequences of this approach . the lla approximation of the pqcd gives = ( 0)-1=_s0.55 , _ s=0.2 . i10but radiative corrections give @xmath238 . we will call this solution as the bfkl model . the quantitative origin of the restriction ( ) is following . the contribution of the diagram with @xmath239 pomeron exchange gives , since the diffraction radii increase with @xmath7 , see ( ) , mean value of the impact parameter decreasing with @xmath239 : @xmath240 where @xmath241 . on the other hand , the number of necessary pomeron exchanges @xmath242 since one pomeron gives maximal contribution ( with factorial accuracy ) at @xmath243 . in result , |^2~a. i11therefore , if the transverse momentum of created particles is a restricted quantity , i.e. @xmath244 , where @xmath245 is a constant , then the mechanism of particle production is valid up to n~^2 . o8 following our general idea , it will be enough for us find the position of singularity over @xmath129 . analysis shows that ( ) predict the singularity at infinity . in appendix e the gribov s reggeon diagram technique with cut pomerons is described and ( ) is derived . it can be shown using this technique that the model with the critical pomeron , @xmath246 , is inconsistent from the physical point of view . as was mentioned above , the model with @xmath247 is natural for the pqcd . the concrete value of @xmath248 will not be important for us . we will assume only that 0<0@xmath249>0@xmath250|^2~b^24(+(n/))@xmath251z@xmath252n~n_max~s@xmath253e^-n@xmath254~s^-1/2@xmath255(m)@xmath256m@xmath257m@xmath258^r(m)@xmath259^r=1/2@xmath260^r_n(m)@xmath261m@xmath262n@xmath263|n^r(m)@xmath264^r_n(m)@xmath265|n^r(m)@xmath266m / m_0@xmath267n@xmath268n|n^2(s)= |n_0^r ( /m_0)@xmath269|n(s)<<n<|n^2(s)@xmath270q|q@xmath271n@xmath272d_ab(x , q^2;n)@xmath273d_ab(x , q^2)@xmath274b@xmath275q^2<0@xmath276a@xmath277~@xmath278>>@xmath279_s()<<1@xmath280q^2@xmath279x@xmath281(1/x)>>1@xmath282d_ab(x , q^2)@xmath283(1/x)@xmath284|q^2|@xmath285~(1/x)/|q^2|@xmath286@xmath287@xmath288 ^ 2<<k_i^2<<-q^2@xmath289k_i^2>0@xmath290~(1/)@xmath291k_i^2>>^2@xmath292|n_j@xmath293|k|@xmath294|n_j @xmath295@xmath296||k_i|@xmath297|k_i|@xmath298k_i@xmath299x@xmath300x0@xmath301t_ab(x , q^2;z)@xmath302n@xmath303z_c@xmath304n@xmath305z_c@xmath306z_c(x , q^2;n)@xmath307n@xmath308d_ab(x , q^2;n)@xmath309t_ab(x , q^2;z)@xmath310z=1@xmath311d_ab(x , q^2;n)@xmath312c_ab(x , q^2;n)@xmath313n@xmath314c_ab(x , q^2;n)~n@xmath315c_ab(x , q^2;n)=const.@xmath316n@xmath317nn_p@xmath318n@xmath319c_ab(x , q^2;n)@xmath320|_ab(x , q^2;n)@xmath321|n_j()@xmath322@xmath323@xmath324n>>|n_g@xmath325(,z)@xmath326x0@xmath327_s(-q^2/^2)~1@xmath279_s(1/x)~1@xmath328|q^2|la^2@xmath329@xmath330_s(^2)<<1@xmath331(^3)_6@xmath332|q^2|la^2@xmath333z_s-1~(1/|n)@xmath334a^(1j)(n , m)@xmath335n@xmath336|n_j(m)@xmath337m@xmath338c_j@xmath339n/|n@xmath340m@xmath341m_1@xmath279m_2@xmath342m_1=m_2=m/2@xmath343m_2<<m_1m@xmath344a^(1j)@xmath345n_1+n_2=n@xmath346@xmath347@xmath348|n_j(m)@xmath349m@xmath350a^1j@xmath336a^2j@xmath351z@xmath352s@xmath353t(z , s)@xmath354z=1@xmath355z_c@xmath356s@xmath279n@xmath357t@xmath358s@xmath359s@xmath360n@xmath361n@xmath362_n(s)@xmath363_n(s)@xmath364n@xmath365w_n(x_1,x_2, ... ,x_n)@xmath366x_k=(u , q)_k@xmath367s@xmath368s@xmath369(|n(s)/n)<<1@xmath370|n(s)@xmath371n@xmath372s@xmath373_max@xmath374(_max / s)0@xmath375s@xmath376n@xmath377n_max=/m@xmath378@xmath336t(z , s)@xmath379z@xmath380z>1@xmath381t(z , s)@xmath382n_max@xmath383t(z , s)@xmath384n@xmath385z@xmath386t(z)@xmath387z@xmath388t(z)@xmath389z=1@xmath390z@xmath391n@xmath392z@xmath393n@xmath394n=@xmath395z=1@xmath396t(z)@xmath397z=1@xmath398z=@xmath399z@xmath400_n@xmath401_n\{-cn^}@xmath402t(z)@xmath403z=1@xmath4041 , 1.3awhere @xmath130 is smallest solution of the equation of state n = zt(z ) . 1.2a therefore , to have the singularity at @xmath106 , we should consider @xmath405 as a decreasing function of @xmath4 . on the other hand , at constant temperature , @xmath406 is the chemical potential , i.e. is the work necessary for creation of one particle . so , the singularity at @xmath106 means that the system is unstable : the less work is necessary for creation of one more particle if @xmath407 is a decreasing function of @xmath4 . the physical explanation of this phenomena is the following , see also . the generating function @xmath408 has following expansion : t(z)=\{_lz^lb_l } , 1.4awhere @xmath151 are known as the mayer s group coefficients . they can be expressed through the inclusive correlation functions and may be used to describe formation of droplets of correlated particles , see sec.2.3.3 . so , if droplet consist of @xmath409 particles , then b_l~e^-l^(d-1)/d 1.5is the mean number of such droplets . here @xmath410 is the surface energy of @xmath411-dimensional droplet . inserting this estimation into ( ) , t(z)~_l e^(l - l^(d-1)/d ) , = z. 1.6the first term in the exponent @xmath412 is the volume energy of the droplet and being positive it tries to enlarge the droplet . the second surface term @xmath413 tries to shrink it . therefore , the singularity at @xmath106 is the consequence of instability : at @xmath132 the volume energy abundance leads to unlimited growth of the droplet . in conclusion we wish to formulate once more the main assumptions . it was assumed first of all that the system under consideration is in equilibrium . this condition may be naturally reached in the statistics , where one can wait the arbitrary time till the system becomes in equilibrium . note , in the critical domain , the time of relaxation @xmath414 , @xmath415 , @xmath416 , @xmath417 is the critical temperature . we can not give the guarantee that in the high energy hadron collisions the final state system is in equilibrium . the reason of this uncertainty is the finite time the inelastic processes and presence of hidden ( confinement ) constraints on the dynamics . but if the confinement forces are frozen in the vhm domain , i.e. the production process is ` fast ' , then the equilibrium may be reached . we may formulate the quantitative conditions , when the equilibrium is satisfied . one should have the gaussian energy spectra of created particles . if this condition is hardly investigated in the experiment , then one should consider the relaxation of ` long - range ' correlations . this excludes the usage of relaxation condition for the ` short - range ' ( i.e. resonance ) correlation ( ii ) . the second condition consists in the requirement that the system should be in the critical domain , where the ( equilibrium ) fluctuations of the system become high . having no theory of hadron interaction at high energies we can not define where lies the ` critical domain ' and even whether it exists or not . but , having the vhm ` cold ' final state , we can hope that the critical domain is achieved . the quantitative realization of this picture is given in appendix j. it is important to note that used there the semiclassical approximation is rightfully in the vhm domain . it seems useful to start the discussion of models outlined the main problems , from authors point of view . * a. * _ soft colour parton problems _ the infrared region of soft colour parton interactions is a very important problem of high energy hadron dynamics . such fundamental questions as the infrared divergences of pqcd , collective phenomena in the coloured particles system and symmetry breaking are the phenomena of the infrared domain . the standard ( most popular ) hadron theory considers pqcd at small distances ( in the scale of @xmath418gev ) as the exact theory . this statement is confirmed by a number of experiments , namely deep - inelastic scattering data , hard jets observation . but the pqcd predictions have a finite range of validity since the non - perturbative effects should be taken into account at distances larger then @xmath419 . it is natural to assume , building the complete theory , that at large distances the non - perturbative effects @xmath420 the perturbative ones . as a result pqcd loses its predictability @xmath421 by the non - perturbative effects . notice , the pqcd running coupling constant @xmath422 becomes infinite at @xmath423 and we do not know what happens with pqcd if @xmath424 . there are few possibilities . for instance , there is a suspicion that at @xmath425 the properties of theory changed so drastically ( being defined on a new vacuum ) that even the notions of pqcd @xmath426 . this means that pqcd should be truncated from below on the ` fundamental ' scale @xmath427 . it seems natural that this infrared cut - off would influence the soft hadrons emission . the new possibility is described in appendix k. this strict formalism allows to conclude that pure pqcd contributions are realized on zero measure , i.e. it is the phenomenological theory only . the successive approach shows that the yang - mills theory should be described in terms of ( action , angle)-like variables . the last one means that the self - consistent description excludes such notions as the ` gluon ' . as a result of this substitution new perturbation theory would be free from infrared divergences , i.e. there is not necessity to introduce the infrared cutoff parameter @xmath427 . ( moreover , in the sector of vector fields ( without quarks ) the theory is ultraviolet stable . ) it seems important for this reason to investigate experimentally just vhm events , where the soft colour partons production is dominant . it is important to try to raise the role of pqcd in the ` forbidden ' area of large distances . the vhm processes are at highly unusual condition , where the non - perturbative effects must be negligible . * b*. _ dissipation problems _ the highly nonequilibrium states decay ( thermalization ) which means in the pqcd terms that the process of vhm formation should be enhanced , at least in asymptotics over multiplicity and energy , by jets . it is the general conclusion of nonequilibrium thermodynamics and it means that the very nonequilibrium initial state tends to equilibrium ( thermalized ) as fast as possible . the entropy @xmath428 of a system is proportional to the number of created particles and , therefore , @xmath428 should tend to its maximum in the vhm region . but the maximum of entropy testifies also the equilibrium of the system . * c*. _ collective phenomena _ we should underline that the collective phenomena may take place if and only if the particles interaction energy is comparable to the kinetic one . the vhm system considered may be ` cold ' and ` equilibrium ' . for this reason the vhm state is mostly adopted for investigation of collective phenomena . one of possible states in which the collective effects , see a.g . , may be important is the ` could coloured plasma ' . the fundamental interest presents the problem of vacuum structure of yang - mills theory . for instance , if the process of cooling is ` fast ' , since the dissipation process of vhm final state formation should be as fast as possible , then one may consider formation of vacuum domains with various properties . then decay of these domains may lead to large fluctuations , for instance , of the isotopic spin . another important question is the collective phenomena in the vhm final state . the last one may be created ` perturbatively ' , for instance , by the formation of heavy jets . then the colour charges should be confined . there are various predictions about this process . one of them predicts that there should be a first order phase transition . now we can ask : what can models say concerning the above problems ? * a*. _ soft coloured partons production _ the multiperipheral models predict fast decreasing of topological cross section in the vhm domain @xmath429 , @xmath430 . at the same time the mean transverse momentum should @xmath431 in this domain since the interaction radii should ` increase ' with @xmath4 . the bfkl pomeron predicts the same asymptotics , @xmath430 , but the pqcd jet predicts @xmath166 . the @xmath22 attempt to insert into the bfkl pomeron the production of particles via ( mini)jets seems impossible . this ` insertion ' can be done into the dis ladder but investigation of the lla kinematics in the vhm domain allows the conclusion that the ` low - x ' contributions should be taken into account . all this experience allows the assumption that in the vhm domain no ` @xmath42-channel ladder ' diagrams play sufficient role . this means the existence of a transition to the processes with jet dominance . the pqcd is unable to predict the transition mechanism . * b*. _ transition into equilibrium _ if the @xmath42-channel ladders are ` destroyed ' in the vhm region , then jets , despite the small factor @xmath432 in the cross sections , are the only mechanism of particle production in the vhm domain . dominance of heavy jets in the vhm domain may naturally explain the tendency towards equilibrium . but the description of thermalization in terms of jets of massless gluons production destroy this hope : the jet contribution @xmath166 assumes ` bremsstrahlung ' of @xmath433 gluons . this prevents the equilibrium since ordering without fail introduce the non - relaxing correlations . * c*. _ collective phenomena _ considering the collective phenomena , we proposer to distinguish ( a ) the collective phenomena connected with the vacuum and ( b ) the collective phenomena produced in the vhm system . following the experience of sec.3.3 we can conclude that the signal of vacuum instability is inequality : @xmath434 . the case ( b ) will not effect the cross section @xmath70 . but if the system became equilibrium in the vhm domain then the collective phenomena may be investigated using ordinary thermodynamical methods . for instance , noting that @xmath435 is the free energy one can measure the thermal capacity ( _ c , z)=c(_c , z ) . o23then , comparing capacities of hadron and @xmath436-quanta systems we can say whether or not the phase transition happened . the connection of the equilibrium and relaxation of correlations is well known . continuing this idea , if the vhm system is in equilibrium one may assume that the colour charges in the pre - confinement phase of vhm event form the plasma . one should note here that the expected plasma is ` cold ' and ` dense ' . for this reason no long - range confinement forces would act among colour charges . then , being ` cold ' , in such a system various , collective phenomena may be important . the experimental possibilities in the vhm domain are not clear up to now . nevertheless first steps toward formulation of trigger system was done , see . by this reason we would like to restrict ourselves by following two general questions . it seems that these questions are mostly important being in the very beginning of vhm theory . 0.3 cm * i*. _ for what values of multiplicity at a given energy the vhm processes become hard _ ? + the answer to this question depends on the value of the incident energy . if we know the answer then it will appear possible to estimate the role of multiperipheral contributions , the jet production rate , the role of vacuum instabilities . it seems that the experimental answer to this question is absent since produced particles are soft , theirs mean transverse and longitudinal momenta have the same value . in our understanding this question means : the total transverse energy may be extremely high . it is interesting also to search the heavy jets , i.e. to observe the fluctuations of particle density in the event - by - event experiment , but this program seems vague since , for all evidence , fractal dimensions tend to zero with increasing multiplicities . 0.3 cm * ii*. _ for what values of multiplicity does the vhm final state become equilibrium _ ? + we hope that having answer on this question we would be able to investigate the status of pqcd , to observe the phase transition phenomena directly , to estimate the role of confinement constraints . the equilibrium means that the energy correlation functions mean values are small . it is interesting also , for example , the charge equilibrium , when the mean value of charge correlation functions are small . notice , the effect of the phase space boundary may lead to ` equilibrium ' . indeed , if @xmath437 and @xmath438 then one may neglect the momentum dependence of the amplitudes @xmath439 . in this case the momentum dependence is defined by the boltzmann exponent @xmath440 , @xmath441 , only and we get naturally to the gaussian law for momentum distribution . correlators should be small in this case since there is no interactions among particles ( @xmath439 are constants ) . but our question assumes that we investigate the possibility of equilibrium when @xmath133 and @xmath442 . there are various approaches to build the real - time finite - temperature field theories of schwinger - keldysh type ( e.g. ) . all of them use various tricks for analytical continuation of imaginary - time matsubara formalism to real time . the basis of the approaches is the introduction of the matsubara field operator _m ( * x * , ) = e^h_s ( * x * ) e^-h , 32where @xmath443 is the interaction - picture operator introduced instead of the habitual heisenberg operator @xmath444 eq . ( ) introduces the averaging over the gibbs ensemble instead of averaging over zero - temperature vacuum states . if the interaction switched on at the moment @xmath117 adiabatically and switched off at @xmath118 then there is the unitary transformation : (x , t)=u(t_i , t_f ) u(t_i , t)_s ( x)u(t , t_i ) . a32introducing the complex mills time contours to connect @xmath117 to @xmath42 , @xmath42 to @xmath118 and @xmath118 to @xmath117 we form a ` closed - time ' contour @xmath445 ( the end - points of the contours @xmath446 together ) this allows to write the last equality ( ) in the compact form : @xmath447 where @xmath448 is the time - ordering on the contour @xmath445 operator . the generating functional @xmath449 of correlation ( green ) functions has the form : @xmath450 where @xmath451 means averaging over the initial state . if the initial correlations have a little effect , we can perform averaging over the gibbs ensemble . this is the main assumption of the formalism : the generating functional of the green functions @xmath449 has the form in this case : @xmath452 with @xmath453 . in accordance with ( ) we have : @xmath454 and , as a result , z(j)=d e^i_c _ d^4 x \ { l(x ) + j(x)(x ) } 33where the path integration is performed with kms periodic boundary condition : @xmath455 in ( ) the contour @xmath456 connects @xmath117 to @xmath118 , @xmath118 to @xmath117 and @xmath117 to @xmath457 . therefore it contains an imaginary - time matsubara part @xmath117 to @xmath457 . a more symmetrical formulation uses the following realization : @xmath117 to @xmath118 , @xmath118 to @xmath458 , @xmath458 to @xmath459 and @xmath460 to @xmath457 ( e.g. ) . this case also contains the imaginary - time parts of the time contour . therefore , eq . ( ) presents the analytical continuation of the matsubara generating functional to real times . one can note that if this analytical continuation is possible for @xmath449 then representation ( ) gives good recipe of regularization of frequency integrals in the matsubara perturbation theory , e.g. . but it gives nothing new for our problem since the matsubara formalism is a formalism for equilibrium states only . the starting point of our calculations is the @xmath4- into @xmath52-particles transition amplitude @xmath125 , the derivation of which is well known procedure in the lehmann - symanzik - zimmermann ( lsz ) reduction formalism framework , see also . the @xmath461-point green function @xmath462 are introduced for this purpose through the generating functional @xmath463 : g_nm(x , y)=(-i)^n+m_k=1^n(x_k)_k=1^m ( y_k)z_j , 22.5where ( x)= , * and z_j = de^is_j ( ) . 22.6 the action s_j ( ) = s_0()-v()+dxj(x)(x ) , 22.7 where @xmath464 is the free part and @xmath465 describes the interactions . at the end of the calculations one can put @xmath466 . to provide the convergence of the integral ( [ 22.6 ] ) over the scalar field @xmath467 the action @xmath468 must contain a positive imaginary part . usually for this purpose feynman s @xmath469-prescription is used . but it is better for us to use the integral on the mills complex time contour @xmath470 . for example , c _ : tt+i , + 0 , -t+o5and after all the calculations return the time contour on the real axis putting @xmath471 . in eq . ( [ 22.6 ] ) the integration is performed over all field configurations with standard vacuum boundary condition : d^4 x _ ( ^)= _ _ d_^=0 , 22.8 which leads to zero contribution from the surface term . let us introduce now field @xmath472 through the equation : - = j(x ) 22.9 and perform the shift @xmath473 in integral ( [ 22.6 ] ) , conserving boundary condition ( [ 22.8 ] ) . considering @xmath472 as the probe field created by the source : ( x)=dy g_0 ( x - y)j(y ) , + ( ^2 + m^2)_x g_0 ( x - y)=(x - y ) , 22.10 only the connected green function @xmath474 will be interesting for us . therefore , g_nm^c(x , y)=(-i)^n+m_k=1^n(x_k ) _ k=1^m(y_k)z ( ) , 22.11 where z()=de^is()-iv(+ ) 22.12 is the new generating functional . to calculate the nontrivial elements of the @xmath0-matrix we must put the external particles on the mass shell . formally this procedure means amputation of the external legs of @xmath474 and further multiplication on the free particle wave functions . as a result the amplitude of @xmath4- into @xmath52-particles transition @xmath125 in the momentum representation has the form : a_nm(q , p)=(-i)^n+m_k=1^n(q_k ) _ k=1^m^ * ( p_k ) z ( ) . 22.13 here we introduce the particle distraction operator ( q)=dxe^-iqx(x ) , ( x)=. 22.14 supposing that the momentum of particles are insufficient for us the probability of @xmath4- into @xmath52-particles transition is defined by the integral : r_nm = d_n ( q ) d_m ( p ) ^(4)(_k=1^nq_k - _ k=1^mp_k ) |a_nm|^2 , 22.15 where d_n(q)=_k=1^nd(q_k)= _ k=1^n , 22.16is the lorentz - invariant phase space element . we assume that the energy - momentum conservation @xmath68-function was extracted from the amplitude . note that @xmath475 is the divergent quantity . to avoid this problem with trivial divergence , connected integration over reference frame , let us divide the energy - momentum fixing @xmath68-function into two parts : ^(4)(q_k -p_k)=d^4 p ^(4)(p - q_k)^(4)(p - p_k ) 22.17 and consider a new quantity : r(p)=_n , md_n ( q ) d_m ( p ) ^(4)(p-_k=1^n q_k ) ^(4)(p-_k=1^n p_k ) |a_nm|^2 22.18defined on the energy momentum shell ( ) . here we suppose that the numbers of particles are not fixed . it is not too hard to see that , up to a phase space volume , r = d^4p r(p ) 22.19 is the imaginary part of the amplitude @xmath476 . therefore , computing @xmath477 the standard renormalization procedure can be applied and the new divergences will not arise in our formalism . the fourier transformation of @xmath68-functions in ( [ 22.18 ] ) allows one to write @xmath478 in the form : r(p)= e^ip(_1+_2 ) ( ) , 22.20where ( ) = _ n , m _ k=1^m\{d(q_k)e^-i_1 q_k } _ k=1^n\{d(p_k)e^-i_2 p_k } |a_nm|^2 . 22.21 introducing the fourier - transformed probability @xmath479 we assume that the phase - space volume is not fixed exactly , i.e. it is proposed that the 4-vector @xmath235 is fixed with some accuracy if @xmath480 are fixed . the energy and momentum in our approach are still locally conserved quantities since an amplitude @xmath125 is translationally invariant . so , we can perform the transformation : _ 1 q_k = ( _ 1 -_1 ) q_k + _ 1 q_k ( _ 1 -_1 ) q_k + _ 1 p since 4-momenta are conserved . the choice of @xmath481 fixes the reference frame . this degree of freedom of the theory was considered in . inserting ( [ 22.13 ] ) into ( [ 22.21 ] ) we find that & ( ) = \{-idx dx ( _ + ( x)d_+-(x - x,_2)_-(x)- + & -_-(x)d_-+(x - x,_1)_+(x ) ) } z(_+)z^ * (-_-) , 22.22where @xmath482 and @xmath483 are the positive and negative frequency correlation functions correspondingly : d_+-(x - x,)=-id_1(q)e^iq(x - x- ) 22.23 describes the process of particles creation at the time @xmath484 and its absorption at @xmath485 , @xmath486 , and @xmath57 is the cm 4-coordinate . function d_-+(x - x,)=id_1(q)e^-iq(x - x+ ) 22.24 describes the opposite process , @xmath487 . these functions obey the homogeneous equations : ( ^2 + m^2)_x g_+-=(^2 + m^2)_x g_-+=0 22.25since the propagation of on mass - shell particles is described . we suppose that @xmath103 may be computed perturbatively . for this purpose the following transformations will be used ( @xmath488 at @xmath489 ) : & e^-iv()= e^-idx ( x)(x ) e^idx j(x)(x ) e^-iv()= + & = e^dx ( x)(x ) e^-iv()= + & = e^-iv(-i ) e^idx j(x)(x ) , 22.26 where @xmath490 was defined in ( [ * ] ) and @xmath491 in ( [ 22.14 ] ) . at the end of the calculations , the auxiliary variables @xmath492 , @xmath493 can be taken equal to zero . using the first equality in ( [ 22.26 ] ) we find that z()= e^-idx ( x)(x ) e^-iv(+ ) e^-dx dx j(x)d_++(x - x)j(x ) , 22.27where @xmath494 is the causal green function : ( ^2 + m^2)_x g_++ ( x - y)=(x - y ) . 22.28 inserting ( [ 22.27 ] ) into ( [ 22.22 ] ) after simple manipulations with differential operators , see ( [ 22.26 ] ) , we may find the expression : & ( ) = e^-iv(-i_+)+iv(-i_-)\ { dx dx + & ( j_+ ( x)d_+-(x - x,_1)j_- ( x)- j_- ( x)d_-+(x - x,_2)j_+ ( x)- + & - j_+ ( x)d_++(x - x)j_+ ( x)+ j_- ( x)d_(x - x)j_- ( x ) ) } , 22.29where d_=(d_++)^ * 22.30 is the anticausal green function . considering the system with a large number of particles , we can simplify the calculations choosing the cm frame @xmath495 . it is useful also to rotate the contours of integration over @xmath496 for the result , omitting the unnecessary constant , we will consider @xmath497 . external particles play a double role in the @xmath0-matrix approach : their interactions create and annihilate the system under consideration and , on the other hand , they are probes through which the measurement of a system is performed . since @xmath498 are the conjugate to the particles energies quantities we will interpret them as the inverse temperatures in the initial ( @xmath499 ) and final ( @xmath500 ) states of interacting fields . they are the ` good ' parameters if and only if the energy correlations are relaxed . 0.2 cm * kubo - martin - schwinger boundary condition * 0.2 cm the simplest ( minimal ) choice of @xmath501 assumes that the system under consideration is surrounded by black - body radiation . this interpretation restores niemi - semenoff s formulation of the real - time finite temperature field theory . indeed , as follows from ( ) , the generating functional @xmath479 is defined by corresponding generating functional & _ 0 ( _ ) = z(_+)z^ * (-_-) = d_+ d_- e^is_0(_+)-is_0(_- ) + & e^-iv(_+ + _ + ) + iv(__- ) , 22.35 see ( [ 22.22 ] ) . the fields @xmath502 and @xmath503 were defined on the time contours @xmath470 and @xmath504 . as was mentioned above , see ( ) , the path integral ( [ 22.35 ] ) describes the closed path motion in the space of fields @xmath467 . we want to use this fact and introduce a more general boundary condition which also guarantees the cancelation of the surface terms in the perturbation framework . we will introduce the equality : _ _ d__+^_+ = _ _ d__- ^_- . 22.36 the solution of eq.([22.36 ] ) requires that the fields @xmath505 and @xmath506 ( and their first derivatives @xmath507 ) coincide on the boundary hypersurface @xmath65 : _(_)=(_)0 , 22.37 where , by definition , @xmath508 is the arbitrary turning - point " field . in the absence of the surface terms , the existence of a nontrivial field @xmath509 has the influence only on the structure of the green functions g_++=<t_+_+ > , g_+-=<_+_- > , + g_-+=<_-_+ > , g_=<_-_- > , 22.38 where @xmath510 is the antitemporal time ordering operator . these green functions must obey the equations : & ( ^2 + m^2)_x g_+- ( x - y)= ( ^2 + m^2)_x g_-+ ( x - y)=0 , + & ( ^2 + m^2)_x g_++ ( x - y)= ( ^2 + m^2)_x^ * g_ ( x - y)=(x - y ) , 22.39 and the general solution of these equations : g_ii = d_ii+g_ii , + g_ij = g_ij , ij 22.40 contain the arbitrary terms @xmath511 which are the solutions of homogenous equations : ( ^2 + m^2)_x g_ij ( x - y)=0 , i , j=+,- . 22.41 the general solution of these equations ( they are distinguished by the choice of the time contours @xmath110 ) g_ij(x - x)=d_1(q ) e^iq(x - x ) n_ij ( q ) 22.42 are defined through the functions @xmath512 which are the functionals of the ` turning - point ' field @xmath508 : if @xmath513 we must have @xmath514 . our aim is to define @xmath512 . we can suppose that @xmath515 the simplest supposition gives : n_ij~<_i_j>~<^2(_)>. 22.43 we will find the exact definition of @xmath512 starting from the @xmath0-matrix interpretation of the theory . it was noted previously that the turning - point field @xmath509 may be arbitrary . we will suppose that on the remote @xmath65 there are only free , on the mass - shell , particles . formally it follows from ( [ 22.40 ] - [ 22.42 ] ) . this assumption is natural also in the @xmath0-matrix framework . in other respects the choice of boundary condition is arbitrary . therefore , we wish to describe the evolution of the system in a background field of mass - shell particles . the restrictions connected with energy - momentum conservation laws will be taken into account and in other respects background particles are free . then our derivation is the same as in . here we restrict ourselves mentioning only the main quantitative points . calculating the product @xmath516 we describe a time ordered processes of particle creation and absorption described by @xmath482 and @xmath483 . in presence of the background particles , this time - ordered picture is slurred over because of the possibility to absorb particles before their creation appears . the processes of creation and absorption are described in vacuum by the product of operators @xmath517 and @xmath518 . we can derive ( see also ) the generalizations of ( [ 22.22 ] ) . the presence of the background particles will lead to the following generating functional : r_cp = e^-i*n*(_i^*_j)r_0 ( _ ) , 22.44 where @xmath519 is the generating functional for the vacuum case , see ( [ 22.35 ] ) . the operator @xmath520 describes the external particles environment . the operator @xmath521 can be considered as the creation and @xmath522 as the annihilation operator and the product @xmath523 acts as the activity operator . so , in the expansion of @xmath524 we can leave only the first nontrivial term : ( _ i^*_j)= d(q ) ^*_i ( q ) n_ij _ j ( q ) , 22.45since no special correlation among background particles should be expected . if the external ( nondynamical ) correlations are present then the higher powers of @xmath525 will appear in expansion ( [ 22.45 ] ) . following the interpretation of @xmath525 , we conclude that @xmath512 is the mean multiplicity of background particles . computing @xmath526 we must conserve the translation invariance of amplitudes in the background field . then , to take into account the energy - momentum conservation laws one should adjust to each vertex of in - going @xmath125 particles the factor @xmath527 and for each out - going particle we have correspondingly @xmath528 . so , the product @xmath529 can be interpreted as the probability factor of the one - particle ( creation+annihilation ) process . the @xmath4-particles @xmath530 process probability is the simple product of these factors if there is no special correlations among background particles . this interpretation is evident in the cm frame @xmath531 . after these preliminaries , it is not too hard to find that in the cm frame we have : & n_++(q_0)=n_(q_0)= = + & = = ( |q_0| ) . 22.47 computing @xmath512 for @xmath532 we must take into account the presence of one more particle : n_+-(q_0)= ( q_0 ) + ( -q_0 ) = + = ( q_0)(1+(q_0 _ 1))+ ( -q_0)(-q_0 _ 1 ) 22.48 and n_-+(q_0)= ( q_0)(q_0 _ 2)+ ( -q_0)(1 + ( -q_0 _ 2 ) ) . 22.49 using ( [ 22.47 ] ) , ( [ 22.48 ] ) and ( [ 22.49 ] ) , and the definition ( [ 22.40 ] ) we find the green functions : g_i , j(x - x,())= e^iq(x - x ) _ ij ( q , ( ) ) 22.50 where i_ij ( q , ( ) ) = ( ) + + + + 2(q^2 -m^2 ) ( ) 22.51 and a_()=-e^(|q_0|q_0 ) . 22.52 the corresponding generating functional has the standard form : r_p(j_)=e^-iv(-i_+)+iv(-i_- ) e^dx dx j_i ( x)g_ij(x - x,())j_j(x ) 22.53where the summation over repeated indexes is assumed . inserting ( [ 22.53 ] ) in the equation of state ( [ 3 ] ) we can find that @xmath533 . if @xmath534 is a ` good ' parameter then @xmath535 coincides with the green functions of the real - time finite - temperature field theory and the kms boundary condition : g_+-(t - t)=g_-+(t - t-i ) , g_-+(t - t)=g_+-(t - t+i ) , 22.54 is restored . the eq.([22.54 ] ) can be deduced from ( [ 22.51 ] ) by direct calculations . we start this consideration from the assumption that the temperature fluctuations are large scale . we can assume that the temperature is a ` good ' parameter in a cell the dimension of which is much smaller then the fluctuation scale of temperature . ( the ` good ' parameter means that the corresponding fluctuations are gaussian . ) let us divide the remote hypersurface @xmath536 on a @xmath537 and let us propose that we can measure the energy and momentum of groups of in- and out - going particles in each cell . the 4-dimension of cells can not be arbitrary small because of the quantum uncertainty principle . to describe this situation we decompose the @xmath68-function of the initial state constraint ( ) on the product of @xmath538 @xmath68-functions : @xmath539 where @xmath540 are the momentum of @xmath84-th in - going particle in the @xmath239-th cell and @xmath541 is the total 4-momenta of @xmath542 in - going particles in this cell , @xmath543 . therefore , @xmath544 the same decomposition will be used for the second @xmath68-function of outgoing particle constraints . we must take into account the multinomial character of particle decomposition on @xmath545 groups . this will give the coefficient : @xmath546 where @xmath547 is the kronecker s symbol . the summation over @xmath548 is assumed . in result , the quantity & r_n_c(p , q)= _ \{n , m}_n_c |a_nm|^2 + & ^n_c_=1\ { ^(4)(q_-^m__k=1q_k , ) ^(4)(p_-^n__k=1p_k , ) } 34defines the probability to find in the @xmath239-th cell the fluxes of in - going particles with total 4-momentum @xmath541 and of out - going particles with the total 4-momentum @xmath549 . the sequence of these two measurements is not fixed . the fourier transformation of @xmath68-functions in ( [ 34 ] ) gives : @xmath550 where @xmath551 has the form : _ n_c()=_=1^n_c\{^m__k=1 e^-i_1,q_k , ^n__k=1 e^-i_2,p_k , } |a_nm|^2 . 35inserting @xmath552 into ( [ 35 ] ) we find : & _ n_c()= \ { i_=1^n_c dx dx [ _ + ( x)d_+-(x - x;_2,)_-(x)- . + & .-_-(x)d_-+(x - x;_1,)_+(x ) ] } _ 0 ( ) , 36 where @xmath553 , and @xmath554 are the positive and negative frequency correlation functions . we must integrate over sets @xmath555 and @xmath556 if the distribution of momenta over cells is not fixed . in the result , r(p)=d^4_1(p ) d^4_2 ( p)_n_c ( ) , 37where the differential measure @xmath557 takes into account the energy - momentum conservation laws : @xmath558 the explicit integration gives that @xmath559 where @xmath560 is 3-vector of the cm frame . choosing cm frame , @xmath561 , @xmath562 in this frame @xmath563 where @xmath564 and @xmath565 was defined in ( [ 36 ] ) with @xmath566 . we will calculate integrals over @xmath498 using the stationary phase method . the equations for the most probable values of @xmath498 : -k(e,\{}_n_c)= _ n_c ( ) , k=1,2 , 38always have unique positive solutions @xmath567 . we propose that the fluctuations of @xmath568 near @xmath569 are small , i.e. are gaussian . this is the basis of the local - equilibrium hypothesis . in this case @xmath570 is the temperature in the initial state in the measurement cell @xmath239 and @xmath571 is the temperature of the final state in the @xmath239-th measurement cell . the last formulation ( [ 37 ] ) implies that the 4-momenta @xmath555 and @xmath556 can not be measured . it is possible to consider another formulation also . for instance , we can suppose that the initial set @xmath555 is fixed ( measured ) but @xmath556 is not . in this case we will have a mixed experiment : @xmath572 is defined by the equation : @xmath573 and @xmath574 is defined by the second equation in ( [ 38 ] ) . considering the continuum limit , @xmath575 , the dimension of the cells tends to zero . in this case we are forced by quantum uncertainty principle to assume that the 4-momenta sets @xmath576 and @xmath577 are not fixed . this formulation becomes pure thermodynamical : we must assume that just @xmath578 and @xmath579 are measurable quantities . for instance , we can fix @xmath578 and try to find @xmath579 as a function of the total energy @xmath10 and the functional of @xmath578 . in this case eqs.([38 ] ) become the functional equations . in the considered microcanonical description , the finiteness of temperature does not touch the quantization mechanism . indeed , one can see from ( [ 36 ] ) that all thermodynamical information is confined in the operator exponent @xmath580 the expansion of which describes the environment , and the ` mechanical ' perturbations are described by the functional @xmath581 . this factorization was achieved by the introduction of the auxiliary field @xmath472 and is independent from the choice of boundary conditions , i.e. from the choice of the systems environment . 0.2 cm * wigner functions * 0.2 cm we will adopt the wigner functions formalism in the carruthers - zachariazen formulation . for the sake of generality , the @xmath52 into @xmath4 particles transition will be considered . this will allow the inclusion of the heavy ion - ion collisions . in the previous section , the generating functional @xmath565 was calculated by means of dividing the ` measuring device ' on the remote hypersurface @xmath536 into @xmath537 cells _ n_c()=e^-i*n*(;,z)_0 ( ) , 39where & * n*(;,z)= \{_=1^n_c dx dx. + & .(_+(x)d_+-(x - x;_2,,z_2)_-(x).- + & .-_-(x)d_-+(x - x;_1,,z_1)_+(x ) ) } 40is the particle number operator . the frequency correlation functions @xmath482 and @xmath482 are defined by equalities : d_+-(x - x;_2,,z_2)=-id_1(q)e^iq_2,(x - x ) e^-_2,(q_2 , ) z_2(q_2 , ) a3.5 + d_-+(x - x,_1,,z_1)=id_1(q)e^-iq_1,(x - x ) e^-_1,(q_1,)z_1(q_1 , ) b3.5it was assumed that the dimension of the device cells tends to zero ( @xmath575 ) . now we wish to specify the cells coordinates . in the result we will get to the wigner function formalism . let us introduce wigner variables : x - x=r , x+x=2y : x = y+r/2 , x=y - r/2 . 3.6then & * n*(;,z)=-i_=1^n_c d(q)dr + & ( _ + ( y+r/2)_-(y - r/2)z_2(q_2,)e^iq_2,r e^-_2,(q_2,).+ + & + ._-(y+r/2)_+(y - r/2)z_1(q_1 , ) e^-iq_1,re^-_1,(q_1,))dy 3.10the boltzmann factor , @xmath582 , can be interpreted as the probability to find a particle with the energy @xmath583 in the final ( @xmath584 ) or initial ( @xmath585 ) state . the total probability , i.e. the process of creation and further absorption of @xmath4 particles , is defined by multiplication of these factors . besides @xmath586 is the out - going particle momentum measured in the @xmath239-th cell . generally it is impossible to adjust the 4-index of cell @xmath239 with coordinate @xmath587 . for this reason the summation over @xmath239 and the integration over @xmath588 are performed in ( ) independently . but let us assume that the 4-dimension of the cell @xmath589 is higher then the scale of the characteristic quantum fluctuations @xmath590 , l>>l_q . o6one can divide the 4-dimensional @xmath587 space into the @xmath589 dimensional cells . then , because of ( ) , the quantum fluctuations can not take away particles from this cell . then we can adjust the index of the measurements cell with the index of the @xmath587 space cell . as a result , & * n*(;,z)=-idyd_i(q)dr + & ( _ + ( y+r/2)_-(y - r/2)z_2(q_2,y)e^iq_2r e^-_2(y)(q_2)+ . + & .+_-(y+r/2)_+(y - r/2)z_1(q_1,y ) ) e^-iq_1re^-_1(y)(q_1 ) , 3.10where dy=__c()dy , o7and @xmath591 is the dimension @xmath589 of the @xmath587 space cell with index @xmath239 . notice that the momentum @xmath592 did not carry the index @xmath239 ( or the index @xmath587 of the space cell ) . our formalism allows the introduction of more general ` closed - path ' boundary conditions . the presence of external black - body radiation will only reorganize the differential operator @xmath593 and a new generating functional @xmath526 has the same form : @xmath594 the calculation of operator @xmath595 is strictly the same as in appendix b. introducing the cells we will find that @xmath596 where the occupation number @xmath597 carries the cell index @xmath587 : @xmath598 and ( @xmath599 ) @xmath600 @xmath601 @xmath602 for simplicity the cm system was used . other calculations are the same as the constant temperature case . first of all , two light - like 4-momenta @xmath603 are introduced . here @xmath604 are momenta of colliding particles . the final state particles momenta have the following representation : & p_1=_1p_2+_1p_1+p_1 , p_2=_2p_2+_2p_1+p_2 , + & k_i=_ip_2+_ip_1+k_i . m1sudakov s parameters , @xmath605 , are not independent . the mass shell conditions and the energy - momentum conservation laws give : & s_1_1=m^2+(p_1)^2=e_1 ^ 2 , s_2_2=e_1 ^ 2 , s_i_i = e_i^2 , + & _1+_2+_i=1 , _1+_2+_i=1 , m2where @xmath606 is the transverse energy . we have for the multiperipheral kinematics : 1_1>>_1>> ... >>_n>>_2~ , + < < _1<<_1<< ... <<_n<<_2~1 m3and the transverse momenta are restricted : considered cm frame : _ i= , |_i|>>|_i| , m5if the particle moves along @xmath607 , and a similar expression exists for particles moving in the opposite direction , where @xmath608 . in the ` central region ' of the cm frame @xmath609 the angles of produced particles are large and energies are small . it should be underlined that all this excludes the ( mini)jets formation . the final - state particles phase space volume element is & d_22+n = c^n_v + & ( _ 2-_1 ) ^n+1_i=1()^2(q_i^2)= dz_n , where @xmath610 and the 4-momentum of produced particle @xmath611 the square of pairs invariant mass : @xmath612 the energy conservation law takes the form : @xmath613 the trajectory of reggeized gluon is @xmath614 where @xmath615 is the gluon ` mass ' . if this virtuality is large , @xmath616 , then the gluon decay creating a pqcd jet . but the constraint on the multiperipheral kinematics prevent this possibility . 0.2 cm * deep inelastic reactions * 0.2 cm for the pure deep inelastic case , when one of the initial hadrons is scattered at the angle @xmath617 have the energy @xmath618 in the cms of beams whereas the another is scattered at small angle and the large transfer momentum @xmath619 is distributed to the some number of the emitted particles due to evolution mechanism we have ( @xmath617 is small ) : & d_n^dis= dd_n ded , + & dd_n=()^n_m^2^q^2 _ m^2^k_n^2_m^2^k_2_x^1 d_n^(1)_n__n^1 d_n-1^(1)_n-1 ... + & _ _ 2 ^ 1d _ 1^(1)_1 p() ... p(_1 ) , p(z)=2 , where @xmath620 and the emission angle @xmath621 . 0.2 cm * large angle production * 0.2 cm for the large - angle particle production process the differential cross section ( as well as the total one ) fall down with cm energy @xmath208 . let us consider for definiteness the process of annihilation of electron - positron pairs to @xmath4 photons : & d_n^dl = df_n , + & df_n=()^n ^n_i=1d x_idy_i(x_i - y_i)(y_i - y_i-1 ) ( y_i)(x_i ) + & ( -x_n)(-y_n),y_i=,= , the similar formulae takes place for subprocess of quark - antiquark annihilation into @xmath4 large - angle moving gluons . at the end , one can consider the following possibilities : + a ) pomeron regime ( p ) ; + b ) evolution regime ( dis ) ; + c ) double logarithmic regime ( dl ) ; + d ) dis+p regime ; + e ) p+dl+p regime . + the description of every regime may be performed in terms of effective ladder - type feynman diagrams . this can be done using the blocks @xmath622 , @xmath623 and @xmath624 . we will consider , see ( ) ( q , ;z)=_0^de^-p(q,;z)= , = ( s / m^2 ) . s9as the ` propagator of the cut pomeron ' . it will be assumed also that @xmath625 so , the resonance short range correlations will be ignored in this definition or propagator . it was assumed also that the ` bare ' slope @xmath229 is @xmath129 independent . it should be underlined that the ` propagator ' ( ) is written phenomenologically . it absorbs the assumptions that ( i ) the diffraction cone shrinks with energy and ( ii ) the inclusive cross sections are universal , see ( ) the set of principal rules concerning multiperipheral kinematics of feynman diagrams is given in appendix d. the reggeon calculus supposes that the virtuality of each line of the feynman diagram is restricted . this ignores ` hard jets ' , later known as the pqcd jets . then the @xmath239 pomeron exchange eikonal diagram has only @xmath626 ways of being cut . if the cut line goes through @xmath627 pomerons , then the corresponding contributions are : ^_( , q)= d_(m^_(q_1, ... ,q_))^2 y^__l=1^(q_i , _i;z)_i=+1^(q_i , _i;z=1 ) , s1where @xmath628 is the ` vertex function ' , the combinatorial coefficient is @xmath629 and the phase space element is d_=_i=1^ ( -_l=1^_l)^2(q-_i=1^q_l ) . s2 as usual , the contribution ( ) leads to the following mean multiplicity of produced particles : ^_=.d_(s / m^2)^^_( , q=0)|_z=1 ~. s3therefore , ~s4is essential in the vhm region , where @xmath630 is assumed . the impact parameter representation : ^_(s , q)=d()^ e^i^_( , q ) s5would be useful also . the contribution ( ) describes interactions with impact parameter 4(s / m^2)/. s6notice @xmath631 is the number of cut pomerons independent of @xmath627 . but , remembering that @xmath632 and that the regge model is only able to describe large distance interactions , @xmath633 , one can conclude that the regge pole description is valid only for n^2 . s7thus is why the vhm region is defined by @xmath634 . then the cut pomeron propagator in the impact parameter representation ( , ; z)=g(,)e^(z-1 ) , l3where g(,)=e^e^-^2/4 l4is the uncut pomeron profile function . using this definition one can find that the contribution of the eikonal diagrams gives a contribution : _ 0 ( , ; z)= ( 1-e^-^2g(,))- ( 1-e^2 ^ 2g(,)(e^(z-1)-1 ) ) , l5where @xmath615 is a constant . first bracket is essential for @xmath636 . so , with exponential accuracy , the first term is equal to @xmath637 let us now consider the second bracket . for @xmath638 ^24^2\{1+(1-e^(z-1))}= 4^2(,z ) l6are essential . it is not hard to see that @xmath639 decreases if @xmath640 and @xmath639 is equal to zero for @xmath106 . in this case , by the definition of the generating function , the integral over @xmath230 of @xmath641 defines the contribution to the total cross section . so , the model predicts the production of particles in the ring : 4^2(,z)^24^2 . l7if @xmath638 , i.e. if @xmath642 . notice also that @xmath643 . then , _ 0 ( , ; z=0)=_0 ( , ; z=1 ) . l8so , the elastic part of the total cross section is half of the total cross section . this means , using the optical analogy , that the scattering on the absolutely black disk is well described . the last conclusion means that the interaction radii should increase with @xmath4 in the vhm region . indeed , as follows from ( ) , _ 0 ( , ; z)(e^2 ^ 2g(,)(e^(z-1)-1)-1 ) , l9at @xmath132 and , therefore , 0 ^ 2b^2=4(+(z-1 ) ) l10are essential . our purpose is to investigate the role of the exponential spectrum ( ) in the asymptotic region over multiplicity @xmath4 . in this case one can validate heavy resonance creation and such formulation of the problem have definite advantages . \(i ) if creation of heavy resonances at @xmath644 is expected , then one can neglect the dependence on the resonance momentum @xmath645 . so , the ` low - temperature ' expansion is valid in the vhm region . \(ii ) having the big parameter @xmath4 , one can construct the perturbations expanding over @xmath142 . \(iii ) we will be able to show at the end the range of applicability of these assumptions . + for this purpose , the following formal phenomena will be used . the grand partition function t(z , s)=_n z^n _ n(s ) , t(1,s)=_tot(s ) , n / m_0n_max(s ) , gpfwill be introduced , see ( ) . then the inverse mellin transformation , see ( ) _ n(s)=t(z , s ) . metrwill be performed expanding it in the vicinity of the solution @xmath646 of the equation of state , see ( ): n = zt(z , s ) . eqst1it is assumed , and this should be confirmed at the end , that the fluctuations in the vicinity of @xmath130 are gaussian . it is natural at first glance to consider @xmath647 as an increasing function of @xmath4 . indeed , this immediately follows from the positivity of @xmath19 and the finiteness of @xmath648 at finite @xmath7 . but one can consider the ` thermodynamical limit ' , see sec.2.3.1 , or the limit @xmath649 . theoretically , the last one is right because of the pcac hypotheses and nothing should happen if the pion mass @xmath649 . in this sense @xmath137 may be considered as the whole function of @xmath129 . then , @xmath647 would be an increasing function of @xmath4 if and only if @xmath137 is a regular function at @xmath106 . the proof of this statement is as follows . we should conclude , as follows from eq . ( ) , that z_c(n , s)z_s at n , and at s = const , asyi.e . the singularity point @xmath138 @xmath650 @xmath130 in asymptotics over @xmath4 . if @xmath651 , then @xmath652 , when @xmath4 tends to infinity . the concrete realization of this possibility is shown in sec.3.3 . but if @xmath653 , then @xmath654 in vhm region , see secs.3.1 , 3.5 . on may use the estimation , see also ( ): -= z_c(n , s)+o(1/n ) , est0where @xmath130 is the @xmath655 solution of ( ) . it should be underlined that this estimation is @xmath656 of the character of singularity , i.e. the position @xmath138 is only important with @xmath657 accuracy . 0.2 cm * partition function * 0.2 cm introducing the ` grand partition function ' ( ) the ` two - level ' description means that & = _ k^k_i=1\{d_k(q)dm_i(q_i , z)e^-_i } + & n_k(q_1,q_2, ... ,q_k;)-(z , s ) , vdwhere @xmath658 . this is our group decomposition . the quantity @xmath659 may be considered as the local activity . so , .|_=1~_totb_1(q ) r1cfif the resonance decay forms a group of particles with total 4-momentum @xmath592 , then @xmath660 is the mean number of such groups . the second derivative gives : .|_=1~_tot \{b_2(q_1,q_2)-b_1(q_1)b_1(q_2)}_totk_2(q_1,q_2 ) r2cfwhere @xmath661 is the two groups correlation function , and so on . one can consider @xmath662 as mayer s group coefficients , see sec.2.3.3 . the lagrange multiplier @xmath62 was introduced in ( ) to each resonance : the boltzmann exponent @xmath663 takes into account the energy conservation law @xmath664 where @xmath10 is the total energy of colliding particles , @xmath665 in the cm frame . this conservation law means that @xmath62 is defined by equation : = t(z , ) . eqst2so , to define the state one should solve two equations of state ( ) and ( ) . the solution @xmath81 of the eq . ( ) has the meaning of inverse temperature of the gas of resonances if and only if the fluctuations in vicinity of @xmath81 are gaussian , see sec.2.2.2 . on the second level , we should describe the resonance decay into hadrons . using ( ) we can write in the vicinity of @xmath106 : ( q , z)=_n z^n ^r_n(q)=g^r()e^(z-1)|n(m ) . rgpfthe assumptions * b * and * d * , see ( ) , were used here so , -(z , s)=_k^k_i=1\{dm_i(m_i , z)}_k(m ; ) , vd1where @xmath666 @xmath667 was defined in ( ) and _ k(m;)=_i=1^k\ { d_k(q)e^-_i(q_i ) } b_k(m;q ) . 1assuming now that @xmath668 are essential , _ k(m;)b_k(m ) _ i=1^k\{e^-m_i } est following the duality assumption , one may write : b_k(m)=|b_k(m ) _ i=1^k \{m_i^e^_0 m_i } stringand @xmath669 is a slowly varying function of @xmath666 : @xmath670 in the result , the low - temperature expansion looks as follows : -(z , s)=_k \{_m_0^dm m^+3/2 e^(z-1)|n^r(m)-(-_0)m}^k . fe1we should assume that @xmath671 . in this sense one may consider @xmath672 as the limiting temperature and the above mentioned constraint means that the resonance energies should be high enough . 0.2 cm * thermodynamical parameters * 0.2 cm remembering that the position of the singularity over @xmath129 is essential , let us assume that the resonance interactions can not renormalize it , i.e. that the sum ( ) is convergent . then , leaving the first term in the sum ( ) , -(z , s)= _ m_0^dm ( m / m_0)^+3/2 e^(z-1)|n^r(m)-(-_0)m . fe2we expect that this assumption is satisfied if & _ m_0^dm m^+3/2 e^(z-1)|n^r(m)-(-_0)m > > + & \{_m_0^dm m^+3/2 e^(z-1)|n^r(m)-(-_0)m}^2 res1for n , s , < < 1 restr so , we would solve our equations of state with the following free energy : -(z , s)= _ m_0^d ( ) ( ) ^-1 e^-(m / m_0 ) , fe3where , using ( ) , =+2(z-1)|n^r_0 + 5/2=2(z-1)|n^r_0 , = m_0(-_0)0 , = const . nw we have in terms of these new variables the following equation for @xmath129 , n = z . eq1cthe equation for @xmath62 takes the form : n_max= , eq2cwhere @xmath673 and @xmath674 is the incomplete @xmath675-function : @xmath676 0.2 cm * asymptotic solutions * 0.2 cm following physical intuition , one should expect the cooling of the system when @xmath644 , for fixed @xmath208 , and heating when @xmath677 , for fixed @xmath4 . but , as was mentioned above , since the solution of eq . ( ) @xmath81 is defined by the value of the total energy , one should expect that @xmath81 decreases in both cases . so , the solution _ c0 , < 0 at n , < 0 at ssol2is natural for our consideration . the physical meaning of @xmath129 is activity . it defines at @xmath678 the work needed for one particle creation . then , if the system is stable and @xmath137 may be singular at @xmath132 only , > 0 at n , < 0 at s. sol3 one should assume solving equations ( ) and ( ) that z_c^_c+1 < < ( _c+1,_c ) . ineq1this condition contains the physical requirement that @xmath133 . in the opposite case , the finiteness of the phase space for @xmath679 should be taken into account . as was mentioned above , the singularity @xmath138 attracts @xmath130 at @xmath644 . for this reason one may consider the following solutions . a. @xmath680 : @xmath681 , @xmath682 . in this case ^-(,)~e^(/ ) . a1this estimation gives the following equations : n = c_1(/)e^(/ ) , = c_2 ()<<1 , a2where @xmath683 are the unimportant constants . the inequality is a consequence of ( ) . these equations have the following solutions : _ c<<1 , _c~n>>1 . a3using them one can see from ( ) that it gives _ n < o(e^-n ) . a4 b. @xmath684 : @xmath685 , @xmath686 . one should estimate @xmath687 near the singularity at @xmath106 and in the vicinity of @xmath246 to consider the consequence of this solution . expanding @xmath687 over @xmath248 at @xmath688 , ( ,)=()-^e^-+o(^+1 ) + o(1 ) . b1this gives the following equations for @xmath689 : n = c_1e^(1/ ) . b2the equation for @xmath248 has the form : n_max = c_2e^(+1)(1/ ) . b3where @xmath690 are unimportant constants . at 0<(1/)-1<<1 , i.e. at ( 1/)<<n<<^2(1/ ) , b4we find : _c~. b5inserting this solution into ( ): _ c~. b6it is remarkable that @xmath691 in the leading approximation is @xmath4 independent . by this reason @xmath692 becomes @xmath4 independent also : _c~ : z_c=1 + . b7this means that _ n = o(e^-n ) b8and obeys the kno scaling with mean multiplicity @xmath693 . considering particle creation in the dis processes , one should distinguish correlation of particles in the ( mini)jets and the correlations between ( mini)jets . we will start from the description of the jet correlations . one should introduce the inclusive cross section for the @xmath239 jets creation @xmath694 where @xmath695 , @xmath696 are the jets 4-momentum in the dis kinetics , @xmath697 . having @xmath698 we can find the correlation functions @xmath699 where @xmath700 and @xmath701 defines the sort of created colour particle . it is useful to introduce the generating functional f^ab(q^2,x;w)=_nd_n(k)_i=1^n w^r_i(k_i ) the amplitude , @xmath702 is the phase space volume and @xmath703 are the arbitrary functions . it is evident that f^ab(q^2,x;w).|_w=1=d^ab(q^2,x ) . 11athe inclusive cross sections _^(r)(k_1,k_2, .... ,k_;q^2,x)=_i=1^f^ab(q^2,x;w).|_w=1 . 12the correlation function n_^(r)(k;q^2,x ) = _ 1=1^f^ab(q^2,x;w).|_w=1 . 13awe can find the partial structure functions @xmath704 , where @xmath4 is the number of produced ( time - like ) gluons , using there definitions . it will be useful to introduce the laplace transform over the variable @xmath705 : f^ab(q^2,x;w)=_j<0()^j f^ab(q^2,j;w ) 14a the expansion parameter of our problem @xmath706 . for this reason one should take into account all possible cuts of the ladder diagrams . so , calculating @xmath707 in the lla all possible cuts of the skeleton ladder diagrams are defined by the factor : \{^ab_rg_r_r^ab } , 15ai.e . the cut line may not only get through the exact green function @xmath708 but through the exact vertex functions @xmath709 also ( @xmath710 are negative ) . we have in the lla ( see appendix d ) @xmath711 and @xmath712 following our approximation , see previous section , we would not distinguish the way in which the cut line goes through the born amplitude @xmath713 we will simply associate @xmath714 to each rung of the ladder . considering the asymptotics over @xmath4 , the time - like partons virtuality @xmath715 should be maximal . here @xmath716 is the fraction of the longitudinal momentum of the jet . then , slightly limiting the jets phase space , k_i^2=|q_i+1|^2(1+o((1/x)/|q_i+1|^2 ) ) . 16 as a result , introducing @xmath717 , where @xmath718 , @xmath719 in the lla variable , we can find the following set of equations : f_ab(q^2,j;w)=_c , r^r_ac(j)w^r ( ) f_ab(q^2,x;w ) , 17awhere ^r_ac(j)=_ac(j)=^1_0x^jp_ac(x ) 18aand @xmath720 is the regular kernel of the bethe - salpeter equation . at @xmath721 this equation is the ordinary one for @xmath707 . we will search the correlation functions from eq . ( ) in terms of the laplace transform @xmath722 let us write : f_ab(q^2,j;w)=d_ab(q^2,j)\ { _ _ i=1^ ( ( w^r_i(_i)-1))n_ab^(r)_(k;q^2,j ) } 19ainserting ( ) into ( ) and expanding over @xmath723 we find the sequence of coupled equations . omitting the cumbersome calculations , we write in the lla that & _ ab^(r)_(_1,_2, .... ,_;q^2,j)= + & = d_ac_1(j,_1)^r_1_c_1c_2(j)d_c_2c_3(j,_2)^r__c_c_+1(j)d_c_+1b(j,_+1 ) . 20one should take into account the conservation laws : _ 1_2_+1= , _ 1<_2< <_+1<. 21computing the laplace transform of this expression we find @xmath724 the kernel @xmath725 was introduced in ( ) . let us write it in the form : d_ab(j,)=_=^_(j ) , 22where d^_qq=_s-_gg , d^_qg=_s-_qq , d^_qg=_gq , d^_gq=_qg 23and _ = \{_qq+_gg+^1/2}. 24if @xmath726 , then @xmath727 are essential . in this case , _ gg~_gq>>_qg~_qq = o(1 ) . 25this means the gluon jets dominance and n^g_gg=_gg+o(1 ) . 26one can find the following estimation of the two - jet correlation function : n_ab^r_1r_2(_1,_2;,)= o(max\{(_1/)^_gg,(_2/)^_gg , ( _ 1/_2)^_gg}}. 27this correlation function is small since in the lla @xmath728 . this means that the jet correlation becomes high if and only if the mass of the correlated jets are comparable . but this condition shrinks the range of integration over @xmath729 and for this reason one may neglect the ` short - range ' correlations among jets . therefore , as follows from ( ) , f_ab(q^2,j;w)=d_gg(,j)\ { _ gg^__0 w^g( ) } 28we will use this expression to find the multiplicity distribution in the dis domain . 0.2 cm * generating function * 0.2 cm to describe particle production , one should replace : @xmath730 where @xmath731 is the @xmath173 of @xmath4 particle production , _ nw^r_n=1 . 29having @xmath239 jets , one should take into account the conservation condition @xmath732 . for this reason , the generating functions formalism is useful . in the result , one can find that if we take ( ) w^g = w^g(,z ) , w^g(,z).|_z=1=1 , 30then @xmath733 defined by ( ) is the generating functional of the multiplicity distribution in the ` @xmath492 representation ' . in this expression @xmath734 is the generating function of the multiplicity distribution in the jet of mass @xmath735 . in the result , see ( ) , f^ab(q^2,x;w)_j<0(1/x)^j e^_gg(,z ) 31where (,z)=^__0w^g(,z ) . 32noting the normalization condition ( ) , (,z=1)=. 33 the integral ( ) may be calculated by the steepest descent method . it is not hard to see that jj_c=1+\{4n(,z)/(1/x)}^1/2 34is essential . notice that @xmath736 should be essential but we find , instead of the constraint ( ) , that (,z)<<(1/x ) . 35 in the frame of this constraint , f^ab(q^2,x;w)\{4}. 36 generally speaking , there exist such values of @xmath129 that @xmath737 . this is possible if @xmath738 is a regular function of @xmath129 at @xmath106 . then @xmath130 should be an increasing function of @xmath4 and consequently @xmath739 would be an increasing function of @xmath4 . therefore , one may expect that in the vhm domain @xmath737 . then @xmath740 would be essential in the integral ( ) . this leads to the following estimation : @xmath741 but this is impossible since @xmath742 should be an increasing function of @xmath129 . this shows that the estimation ( ) has a finite range of validity . solution of this problem with unitarity is evident . one should take into account correlations among jets considering the expansion ( ) . indeed , smallness of @xmath743 may be compensated by large values of @xmath744 in the vhm domain . one may neglect quark jets in the vhm region since the gluons mean multiplicity @xmath745 - quarks multiplicity and in the vhm region the leftmost singularities are important . then we can write : t_j(,z)=t_j(,z ) ^__0d(t_j(,z)-1 ) , j1where @xmath746 and @xmath747 is the generating function of the distribution over the number of gluons @xmath748 : t_j(,z)=_n z^n w_n ( ) , t_j(,z=1)=1 . j2we search a solution in the vhm region , where n>>n_j\ { } , a=12/11 . j3let us consider the following solution : w_n=()^e^-n/|n_j . j4it is useful to introduce _ k()=_n=1^n^k-1w_n ( ) j5for this solution . inserting ( ) in this expression , _ k()=|n^k()_k , j6where @xmath498 ( i ) should be positive and ( ii ) @xmath729 independent . these conditions are satisfied for the following values of @xmath84 . indeed , at @xmath749 , _ k=_n=1^ ( ) ^k+e^-n/|n_j^-(k+ ) . j7 the generating function @xmath750 has the following form in terms of @xmath751 : t_j(,z)=_n=1^(z)^k_k ( ) . j8inserting ( ) into ( ) and assuming that @xmath498 is a @xmath729 independent quantity , we find the following recurrent equation for @xmath498 : _ k=^k_k_1=1_k_1_k - k_1- 2_k . j9therefore , if k > > j10then we can neglect last term in the right hand side of ( ) and in this case @xmath498 are positive and @xmath729 independent . noting that in ( ) @xmath752 are essential the inequality ( ) means that the solution ( ) is correct if n>>|n_j|n_j , j11i.e . , only for this value of @xmath4 @xmath753 has the form ( ) and the corresponding generating functional has singularity at z_s=1 + . j12 it is important for the vhm experiment to have an upper restriction on the asymptotics . we wish to show that @xmath70 decreases faster than any power of @xmath142 : _ n < o(1/n ) . 2.1to prove this estimation , one should know the type of singularity at @xmath106 . one can imagine that the points , where the external particles are created , form the system . here we assume that this system is in equilibrium , i.e. there is not in this system macroscopical flows of energy , particles , charges and so on . the lattice gas approximation is used to describe such a system . this description is quite general and does not depend on details . motion of the gas particles leads to the necessity to sum over all distributions of the particles on cells . for simplicity we will assume that only one particle can occupy the cell . so , we will introduce the occupation number @xmath754 in the @xmath174-th cell : @xmath755 means that we have no particle in the cell and @xmath756 means that a particle exists in a cell . assuming that the system is in equilibrium , we may use the ergodic hypothesis and sum over all ` spin ' configurations of @xmath757 , with the restriction : @xmath758 . it is evident that this restriction introduces the interactions . the corresponding partition function in temperature representation ( , h)=de^-s ( ) 2.2awhere integration is performed over @xmath759 and , considering the continuum limit , @xmath760 . the action s_()=dx\{()^2 -^2+g^4- } 2.3where ~(1- ) , g~ , ~()^1/2h . 2.4and @xmath761 is the critical temperature . 0.2 cm * unstable vacuum * 0.2 cm we start this consideration from the case @xmath762 , i.e. assuming that @xmath763 . in this case the ground state is degenerate if @xmath764 . the extra term @xmath765 in ( ) can be interpreted as the interaction with external magnetic field @xmath766 . this term regulates the number of ` down ' spins with @xmath767 and is related to the activity : z^1/2=e^h , 2.5i.e . @xmath766 coincides with the chemical potential . the potential v()=-^2+g^4 , >0 , 2.6has two minima at @xmath768 if the dimension @xmath769 , no tunnelling phenomena exist . but choosing @xmath770 the system in the correct minimum ( it corresponds to the state without particles ) becomes unstable . the system tunneling into the state with an absolute minimum of energy . the partition function @xmath185 becomes singular at @xmath764 because of this instability . the square root branch point gives ( b , z)=e^-a_2()/h^2 , a_i>0 . 2.7note , @xmath771 at @xmath764 . deforming the contour in the mellin integral over @xmath129 on the branch line , _ n()=^_1 e^-4a_2 ^ 2/^2z . 2.8 in this integral z_c\{}^1/3 2.9is essential . this leads to the following estimation : _ ne^-3(a_2 ^ 2)^1/3n^2/3<o(1/n ) . 2.10 it is useful to note at the end of this section that + ( i ) the value of @xmath69 is defined by @xmath772 and the metastable states , the decay of which gives a contribution into @xmath773 , are not important . + ( ii ) it follows from ( ) that in the vhm domain h~h_c~z_c~(1/n)^1/30 . 2.11so , the calculations are performed for the ` weak ' external field case , when the degeneracy is weakly broken . it is evident that the life time of the unstable ( without particles ) state is large in this case and for this reason the semiclassical approximation is correct . this is an important consequence of ( ) . 0.2 cm * stable vacuum * 0.2 cm let us consider now @xmath774 , i.e. @xmath775 . the potential ( ) has only one minimum at @xmath776 in this case . the inclusion of an external field shifts the minimum to the point @xmath777 . in this case the expansion in the vicinity of @xmath778 should be useful . as a result , ( , z)=\{dx _ c - w(_c ) } , 2.12where @xmath779 can be expanded over @xmath778 : w(_c)=_l_k\{dx_k_c(x_k;h)}_l ( x_1, ... ,x_l ) . 2.13 in this expression , @xmath780 is the one - particle irreducible green function , i.e. @xmath781 is the virial coefficient . then @xmath778 can be considered as the effective activity of the correlated @xmath409-particle group . the sum in ( ) should be convergent and , therefore , @xmath782 if @xmath783 . but in this case the virial decomposition is equivalent to the expansion over the inverse density of particles . in the vhm region it is high and the mean field approximation becomes correct . in result is _ c-()^1/3 : |s_c| if to calculate @xmath69 . in this case z_ce^4gn^3 at n , 2.17is essential and in the vhm domain _ ne^-4gn^4<o(e^-n ) . 2.18this result is an evident consequence of vacuum stability . it should be noted once more that the conditions ( ) considerably simplify calculations . * @xmath0-matrix unitarity constraints * to explain our idea , let us consider the spectral representation of the one - particle amplitude : a_1(x_1 , x_2 ; e)= , + 0 , 3ait describes the transition of a particle with energy @xmath10 from point @xmath784 to @xmath785 . according to our general idea , see introduction to sec.2.1 , we will calculate r_1(e)=dx_1 dx_2 a_1(x_1 , x_2 ; e)a^*_1 ( x_1 , x_2 ; e ) . 4 the integration over the end points @xmath784 and @xmath785 is performed only for the sake of simplicity . inserting ( ) into ( ) and using ortho - normalizability of the wave functions @xmath786 we find that & r_1(e)=_n||^2 = _ n\{-}= + & = _ n=_n(e - e_n ) . 5on other hand , the closed - path amplitude , offered for calculation in , & c_1(e)=_ndx= _ n= + & = _ n\{p+ i(e - e_n)}=_np+ir_1(e ) . 4so , we wish to calculate only the imaginary part of the closed - path contribution : @xmath787 notice the extra factor @xmath788 in the left hand side . the reason for this choice is evident : the real part of @xmath789 is equal to zero at @xmath790 , i.e. did not contribute to the measurable . to calculate the bound states energy spectrum , it is enough to know the only imaginary part of the closed - path amplitude . this property is not accidental . it is known as the optical theorem and is the consequence of the total probability conservation principles . formal realization of it is the unitarity condition for the @xmath791-matrix : @xmath792 . in terms of the amplitudes @xmath793 , @xmath794 , the unitarity condition presents an infinite set of nonlinear operator equalities : i*a**a*^ * = * a * - * a*^*. 2notice that expressing the amplitude by the path integral one can see that the left hand side of this equality offers the double integral and , at the same time , the right hand side is the linear combination of integrals . thus , the continuum contributions into the amplitudes should be canceled to provide the conservation of total probability . in this sense it is a necessary condition . indeed , to see the integral form of our approach , let us use the proper - time representation : a_1(x_1 , x_2 ; e)=_n _ n ( x_1)^*_n ( x_2)i ^_0dte^i(e - e_n+i)t 6 and insert it into ( [ 4 ] ) : r_1(e)=_n ^_0 dt_+dt_- e^-(t_++t_- ) e^i(e - e_n)(t_+-t_- ) . 7 we will introduce new time variables instead of @xmath795 : t_=t , 8 where , as follows from the jacobian of transformation , @xmath796 . but we can put @xmath797 since @xmath798 is essential in the integral over @xmath799 . as a result , _ 1(e)=2_n^_0 dt e^-2 t ^+_- e^2i(e - e_n ) . 9 in the last integral , the continuum of contributions with @xmath800 are canceled . note that the product of amplitudes @xmath801 was ` linearized ' after the introduction of ` virtual ' time @xmath802 . we wish to calculate the density matrix @xmath185 including the consequence of the unitarity condition cancelation of unnecessary contributions . here we demonstrate the result and the intermediate steps we will formulate , without proof , as the statements offered in , where the formalities are described . the statement , see and references cited therein , 0.1 cm _ s1 . the unitarity condition unambiguously determines contributions in the path integrals for @xmath803 _ 0.1 cm looks like a tautology since @xmath804 , where @xmath805 is the action , is the unitary operator which shifts a system along the trajectory . so , it seems evident that the unitarity condition is already included in the path integrals . the rule as the path integrals should be calculated is weel known , see e.g. . nevertheless the general path - integral solution contains unnecessary degrees of freedom ( unobservable states with @xmath806 in the above example ) . we would define the path integrals in such a way that the condition of absence of unnecessary contributions in the final ( measurable ) result be loaded from the very beginning . just in this sense , the unitarity looks like the necessary and sufficient condition unambiguously determining the complete set of contributions . _ s2 . the @xmath52- into @xmath4-particles transition ( unnormalized ) @xmath173 @xmath807 would have on the dirac measure the following symmetrical form _ : & r_nm(p_1, ... ,p_n , q_1, ... ,q_m)= < ^m_k=1|(q_k;u)|^2 ^n_k=1|(p_k;u)|^2>_u= + & = e^-i(j , e)dm(u)e^is_o(u)-iu(u , e ) ^m_k=1|(q_k;u)|^2 ^n_k=1|(p_k;u)|^2 + & ( u)^m_k=1|(q_k;u)|^2 ^n_k=1|(p_k;u)|^2 . 1.1ehere @xmath808 are the in(out)-going particle momenta . it should be underlined that this representation is strict and is valid for arbitrary lagrange theory of arbitrary dimensions . the eikonal approximation for inelastic amplitudes was considered in the functionals @xmath813 and @xmath814 are defined by the equalities : s_o(u)=(s_0(u+e)-s_0(u - e ) ) + 2_c_+dx dte(x , t)(^2+m^2)u(x , t ) , u(u , e)= v(u+e)-v(u - e)-2_c_+ dx dte(x , t)v(u ) , 2.3awhere @xmath815 is the free part of the lagrangian and @xmath816 describes interactions . the quantity @xmath814 is not equal to zero if @xmath817 have nontrivial topological charge ( see also ) . according to s1 , considering motion in the phase space @xmath818 the measure @xmath819 has the dirac form : dm(u , p)= _ x , t du(x , t ) dp(x , t ) ( - ) ( + ) 2.4awith the total hamiltonian h_j(u , p)=dx \ { p^2 + ( u)^2 + v(u)-ju } . 2.5athis last one includes the energy @xmath820 of quantum fluctuations . * a. * only @xmath821 solutions of equations -=0 , + = 0 equwith @xmath466 should be taken into account . this ` rigidness ' of the formalism means the absence of pseudo - solutions ( similar to multi - instanton , or multi - kink ) contribution . the parameter @xmath826 is connected directly with @xmath827 particle energy , momentum , spin , polarization , charge , etc . , and is sensitive to the symmetry properties of the interacting fields system.@xmath828 , if @xmath829 is the magnetics order parameter , and the phase transition means that @xmath830 . @xmath831 has just the same meaning as @xmath832 . ] for the sake of simplicity , @xmath833 is the real scalar field . the generalization would be evident . as a consequence of ( ) , @xmath826 is the function of the external particle momentum @xmath592 and is a @xmath834 functional of @xmath833 : ( q;u)=-dx e^iqx = dx e^iqx(^2 + m^2)u(x ) , q^2=m^2 , 1.2efor the mass @xmath52 field . this parameter presents the momentum distribution of the interacting field @xmath833 on the remote hypersurface @xmath536 if @xmath833 is the regular function . notice , the operator @xmath835 cancels the mass - shell states of @xmath833 . the construction ( ) means , because of the klein - gordon operator and the external states being mass - shell by definition , the solution @xmath836 is possible for a particular topology ( compactness and analytic properties ) of @xmath837 field @xmath833 . so , @xmath826 carries the following remarkable properties : notice , the space - time topology of @xmath838 becomes important in calculating integral ( ) by parts . this procedure is available if and only if @xmath838 is the regular function . but the @xmath837 fields are always singular . therefore , the solution @xmath839 is valid if and only if the quasiclassical approximation is exact . just this situation is realized in the soliton sector of sin - gordon model . despite evident ambiguity @xmath826 carries the definite properties of the order parameter since the opposite solution @xmath836 can only be the dynamical display of an @xmath840 symmetry - matrix was introduced ` phenomenologically ' , see also the example considered in , postulating the lsz reduction formulae , see eq . so , the formal constraints , e.g. the haag theorem , would not be taken into account on the chosen level of accuracy . ] , i.e. of the nontrivial topology of interacting fields , as the consequence of unbroken symmetry . if ( ) have nontrivial solution @xmath41 , then this ` extended objects ' quantization problem arises . we solve it by introducing convenient dynamical variables . the main formal difficulty , see e.g. , of this program consists of transformation of the path - integral measure which was solved in . then + _ s3 . the measure ( ) admits the transformation : u_c : ( u , p)(,)w = g / g_c . o11and the transformed measure has the form : dm(u , p)= _ x , t__c _ _ d(t ) d(t ) ( - ) ( + ) , o12where @xmath841 is the transformed hamiltonian . _ 0.1 cm it is evident that @xmath842 are parameters of integration of eqs . ( ) and they form the factor space @xmath843 . for instance , if one particle dynamics is considered , then one may choose @xmath844 and @xmath845 . one may consider also the following possibility : @xmath846 and @xmath847 in this terms @xmath848 and new hamilton equations have the form : = 1-j , = j. o13so , we have at @xmath466 : @xmath849 and @xmath850 . by this reason the ( action , angle)-type variables are mostly useful_. according to ( ) there exists transformation of the perturbation generating operator : + _ s5 . the operator @xmath851 has following transformed form : 2=dt \{__+ _ _ } , o14 in the factor space , where @xmath852 , @xmath853 , are new auxiliary variables . _ as a result of mapping of the perturbation generating operator @xmath851 on the manifold @xmath854 the equations of motion became linearized : dm=_t(--j _ ) ( -j _ ) . o21then + _ s6 . if the feynman s @xmath469-prescription is adopted , then the green function of eq . ( ) _ g(t - t)=(t - t ) o22 later on we will consider the soliton sector of sin - gordon model . in this case @xmath855 is the coordinate and @xmath856 is the momentum of @xmath174-th soliton and @xmath545 is the number of solitons . expansion of @xmath857 gives the ` strong coupling ' perturbation series . its analysis shows that + _ s7 . action of the integro - differential operator @xmath809 leads to the following representation _ : & r_nm(p , q)=_w\ { d(0)r^_nm(p , q).+ + & . d(0)r^_nm(p , q)}. o20this means that the contributions into @xmath98 are accumulated strictly on the boundary ` bifurcation manifold ' @xmath858 , i.e. depends directly on the topology of @xmath854 . performing the shifts in ( ): _ i ( t ) _ i ( t ) + dt g(t - t ) j_,i(t ) _ i ( t ) + _i ( t ) , + _ i ( t ) _ i ( t ) + dt g(t - t ) j_,i(t ) _ i ( t ) + _i ( t ) , 4.16we can get the green function @xmath859 into the operator exponent : ( ej ) = dt dt ( t - t)\{(t)_(t ) + (t)_(t)}. note the lorentz noncovariantness of our perturbation theory with green function ( ) . using the definition : @xmath860 the functional integrals on the measure ( ) are reduced to the ordinary integrals over initial data @xmath861 . these integrals define zero modes volume . notice that the zero - modes measure was defined without the faddeev - popov @xmath186 . this strategy is necessary since it seems to be important to show the role of quantum corrections noting that for all physically acceptable field theories @xmath862 in the quasiclassical approximation . the @xmath545-soliton solution @xmath863 depends upon @xmath864 parameters . half of them , @xmath545 , can be considered as the position of solitons and the other @xmath545 as the solitons momentum . generally at @xmath865 the @xmath863 solution decomposed on the single solitons @xmath866 and on the double soliton bound states @xmath867 : @xmath868 note that this asymptotic is achieved if @xmath869 or / and @xmath870 . this last one defines the bifurcation line of our model . so , the one soliton @xmath866 and two - soliton bound state @xmath867 would be the main elements of our formalism . its @xmath842 parametrizations have the form : u_s(x;,)=-\{(mx- ) } , = so and u_b(x;,)= -\ { } . bo in the quasiclassical approximation @xmath871 we have : u_n = u_n ( x;_0 , _ 0 + (_0)t ) . 5.7notice that the surface term dx^_(e^iqxu_n)=0 . 5.8then d^2x e^iqx(^2 + m^2)u_n ( x , t ) = -(q^2-m^2)d^2x e^iqxu_n ( x , t ) = 0 5.9since @xmath872 belongs to the mass shell by definition . the condition ( ) is satisfied for all @xmath873 since @xmath863 belongs to schwarz space ( the periodic boundary condition for @xmath838 do not alter this conclusion ) . therefore , in the quasiclassical approximation @xmath874 . expanding the operator exponent in ( [ 5.1 ] ) we find that action of operators @xmath875 , @xmath876 creates terms ~d^2x e^iqx ( t - t ) ( ^2 + m^2)u_n ( x , t ) 0 . 5.12so , generally , if the quantum corrections are included , @xmath877 . now we will show that the quasiclassical approximation is exact in the soliton sector of sin - gordon model . the structure of the perturbation theory is readily seen in the ` normal - product ' form : r_2(q)=_n ^n_i=1 \{d_0 d_0}_i : e^-iu(u_n;/2i)e^is_0(u_n)|(q;u_n)|^2 : , 6.1where = _ - _ = _x 6.2and _ x = dt ( t - t)(t ) 6.3with the @xmath864-dimensional vector @xmath878 . in eq . ( ) @xmath879 is the ordinary symplectic matrix . the colons in ( ) mean that the operator @xmath880 should stay to the left of all functions in the perturbation theory expansion over it . the structure ( ) shows that each order over @xmath881 is proportional at least to the first order derivative of @xmath863 over the variable conjugate to @xmath882 . the expansion of ( ) over @xmath883 can be written in the form ( omitting the quasiclassical approximation ) : r_2(q)=_n ^n_i=1 \{d_0 d_0}_i \{^2n_i=1p_x_i(u_n ) } , 6.4where @xmath884 is the infinite sum of ` time - ordered ' polynomials ( see ) over @xmath863 and its derivatives . the explicit form of @xmath884 is complicated since the interaction potential is nonpolynomial . but it is enough to know , see ( ) , that p_x_i(u_n)~_ij . 6.5 authors would like to take the opportunity to thank j. allaby , a. m. baldin , j. a. budagov , r. cashmore , g. a. chelkov , m. della negra , d. denegri , a. t. filippov , d. froidevaux , f. gianotti , i. a. golutvin , p. jenni , v. g. kadyshevski , o. v. kancheli , z. krumstein , e. a. kuraev , e. m. levin , l. n. lipatov , l. maiani , v. a. matveev , v. a. nikitin , a. nikitenko , a. g. olchevski , g. s. pogosyan , n. a. russakovich , v. i. savrin , m. v. saveliev , ju . schukraft , n. b. skachkov , a. n. tavkhelidze , v. m. ter - antonyan , h. t. torosyan , n.e . tyurin , d. v. shirkov , e. sarkisian , s. tapprogge , o. i. zavialov , g. zinovyev for valuable discussions and important comments . the scientific communications with them on the various stages were important . we acknowledged also to members of the seminar ` symmetries and integrable systems ' of the n.n.bogolyubov laboratory of theoretical physics ( jinr ) , to the scientific seminars of the v. p. dzelepov laboratory of nuclear problems , d. v. scobeltsyn inp of the moscow state university , of the atlas experiment community for a number of important discussions . we would like also to thank m. gostkin and n. shubitidze for the help in the formulation of the monte carlo simulation programs and for the discussions . one of us ( j.m . ) was supported in part by georgian academy of sciences . a.n.sissakian and l.a.slepchenko , preprint jinr , p2 - 10651 , 1977 ; + a.n.sissakian and l.a.slepchenko , fizika , 10 ( 1978 ) 21 ; + s. ch . mavrodiev , v. k. mitryushkin , a. n. sissakian and h. t. torosyan , sov . phys . , 30 ( 1979 ) 245 r.feynman , phys . lett . , 23 ( 1969 ) 1415 ; + a.a.logunov and m.a.mestvirishvili , cern th/1659 ; + a.a.logunov , m.a.mestvirishvili , nguen van hieu , phys . , 25b ( 1967 ) 611 ; + a.a.logunov and m.a.mestvirishvili , cern , th-1707 , 1973 g.giacomelli , nucl . phys . 25b ( 1992 ) 30 ; + g.giacomelli , boll . fis . , nuova ser . 6 ( 1990 ) 20 , int a5 ( 1990 ) 223 ; + h.d.wahl , acta phys . , xxy ( 1983 ) 399 ; + a.b.kaidalov , survey in high energy physics , 7 ( 1994 ) 143 ; + p.g.ratcliff , proc . 8th int . symp . in high - en . spin phys . , minneapolis , vol.1 ( 1988 ) 145 ; + d.grosh et al . , rev . , d49 ( 1994 ) 3113 ; + proc . int . symp . in hadron interaction , eds . j.fischer , p.kolar , v.kundrat , czech , acad . ( 1988 ) 431 e.kuraev , l.lipatov and v.fadin , sov . jetp , 44 ( 1976 ) 443 ; zh . fiz . , 71 ( 1976 ) 840 ; + l.lipatov , sov.j . 20 ( 1975 ) 94 ; + v.gribov and l.lipatov , sov.j . phys . , 15 ( 1972 ) 438 , 675 ; + g.altarelli and g.parisi , nucl . phys . , b126 ( 1977 ) 298 ; + i.v.andreev , chromodynamics and hard processes at high energies ( nauka , moscow , 1981 ) i.p.kadanoff and g.baym , qunantum statistical mechanics ( benjamin , london , 1962 ) ; + g.niklasson and a.sjolander , ann.phys . , 49 ( 1968 ) 249 ; + p.danielevich , ann.phys . , 152 ( 1984 ) 239 + p.kadanoff and p.c.martin , ann.phys.(ny ) , 24 ( 1963 ) 419 e.wigner , phys . , 40 , ( 1932 ) 749 ; + k.hisimi , proc . jap . , 23 ( 1940 ) 264 ; + r.j.glauber , phys . 10 ( 1963 ) 84 ; + e.c.g.sudarshan , phys . rev . 10 ( 1963 ) 177 ; + r.e.cahill and r.g.glauber , phys . , 177 ( 1969 ) 1882 ; + s.mancini , v.i.manko and p.tombesi , quant . , 7 ( 1995 ) 615 ; + v.i.manko , l.rose and p.vitale , hep - th/9806164

neurons are known to exhibit a plethora of dynamical behaviors , represented by the generation of action potential patterns . one of such patterns is bursting , defined by the repeated firing of action potentials followed by quiescent periods . hence the dynamics of bursting neurons has two timescales : a fast scale related to spiking and a slow scale of bursting itself . these timescales are related to different biophysical mechanisms occuring at the level of neuron membrane : there are fast ionic currents ( chiefly @xmath0 and @xmath1 ) responsible for spiking activity and slower @xmath2 currents that modulate this activity . most neurons exhibit bursting behavior if conveniently stimulated . for example , in the neocortical layer @xmath3 pyramidal neurons , when stimulated with dc current pulses , fire an initial burst of spikes followed by shorter bursts @xcite . in layers @xmath4 , @xmath5 , and @xmath6 chattering neurons fire high - frequency bursts of @xmath7 spikes with a short interburst period @xcite . cortical interneurons have been found to exhibit bursting as a response to dc pulses @xcite . pyramidal neurons in the @xmath8 region of hippocampus produce high - frequency bursts after current injection @xcite . thalamocortical neurons and reticular thalamic nucleus inhibitory neurons exhibit bursting as well @xcite . purkinje cells in cerebellum can burst when their synaptic input is blocked @xcite . bursting is also an important feature of sensory systems , because bursts can increase the reliability of synaptic transmission @xcite . in some systems , bursts improve the signal - to - noise ratio of sensory responses and might be involved in the detection of specific stimulus features @xcite . due to both synaptic coupling and common inputs among neurons there are many types of synchronization , which can be generally regarded as the presence of a consistent temporal relationship between their activity patterns @xcite . a strong form of the latter relationship is complete synchronization , where neurons spikes at the same time , i.e. a precise temporal coincidence of events . a weaker relationship is bursting synchronization , in which only the beginning of bursting is required to occur at the same time , even though the repeated spiking may not occur synchronously . there has been observed bursting synchronization in cell cultures of cortical neurons , where uncorrelated firing appeared withing the first three days and transformed progressively into synchronized bursting within a week @xcite . large - scale bursting synchonization in the @xmath9 range has been found in the thalamus during slow - wave sleep , partially originated in the thalamus and gated by modulatory input from the brainstem @xcite . various areas of the basal ganglia have been found to exhibit bursting synchronization related to parkinson s disease and resting tremor @xcite . there exists sound neurophysiological evidence that hypokinetic motor symptoms of parkinson s disease such as slowness and rigidity of voluntary movements are closely related to synchronized bursting in the @xmath10 range @xcite . the connection between bursting synchronization and pathological conditions like parkinson s disease , essential tremor and epilepsy has led to the proposal of many control strategies aiming to suppress or mitigate bursting synchronization @xcite . one of such strategies is deep - brain stimulation ( dbs ) , which consists of the application of an external high - frequency ( @xmath11 ) electrical signal by depth electrodes implanted in target areas of the brain like the thalamic ventralis intermedius nucleus or the subthalamic nucleus @xcite . the effect of dbs would be similar to that produced by tissue lesioning and has proved to be effective in suppression of the activity of the pacemaker - like cluster of synchronously firing neurons , and achieving a suppression of the peripheral tremor @xcite . there is strong clinical evidence that dbs is a highly effective technique for treatment of patients with parkinson s disease @xcite . in spite of these results , dbs is yet far from being completely understood . many results in this field have been obtained from empirical observations made during stereotaxic neurosurgery , but further progress can be obtained with proper mathematical modelling of dbs @xcite . the effects of dbs in networks of bursting neurons have been investigated when dbs is implemented through an harmonic external current @xcite and a delayed feedback signal @xcite . on modelling the response of a neuronal network to an external perturbation like dbs it is of paramount importance to keep the model simple enough such that large - scale simulations ( using a large number of neurons ) can be performed in a reasonable computer time . in such reductionist point of view a minimal model could be one in which we can assign a geometrical phase to the bursting activity . the bursting neuron is thus regarded as a phase oscillator undergoing spontaneous oscillations with a given frequency . thus bursting synchronization becomes a special case of phase synchronization , a phenomenon well understood for coupled oscillators with and without external excitation @xcite . a simple model for the dynamics of nonlinearly coupled phase oscillators is the kuramoto model , which in its original version considers a global ( all - to - all ) coupling @xcite . it can be generalized by considering an arbitrary coupling architecture ( generalized kuramoto model ) @xcite . the particular interest in such models is that many analytical and numerical results are known for them , specially the global case for which a mean - field theory exists for the transition between a non - synchronized to a ( phase-)synchronized behavior @xcite . for generalized kuramoto models it is possible to derive analytical expressions for the critical value of the coupling strength for which the abovementioned transition occurs @xcite . hence such a body of knowledge can be applied to networks of bursting neurons , helping to design strategies of synchronization control and/or suppression like dbs . the main goal of this paper is to show , using analytical and numerical arguments , that a system of coupled bursting neurons described by rulkov s model can be reduced to a generalized kuramoto model . this reduction is valid as long as phase synchronization is concerned , since for frequency synchronization the behaviors can be quite different , though . we consider , in particular , some widely used connection topologies , like random ( erds - renyi ) , small - world , and scale - free networks . we show that the analytical results for the critical coupling strength to synchronized behavior , originally derived for the generalized kuramoto model , can be used to describe the synchronization transition also for networks of bursting neurons . as a matter of fact , since bursting activity presents two timescales it can be also approached from the point of view of a relaxation oscillator @xcite . in our work , however , we describe bursting using a single phase . this simplification is justified since phase synchronization of bursting is chiefly related to the slow timescale . in other words , the fast spikes can be nonsynchronized even though the slow dynamics is synchronized . this paper is organized as follows : in section ii we describe the model we used to describe bursting neurons . section iii considers networks of coupled bursting neurons using different connection topologies . section iv reviews some results on the generalized kuramoto model , and section v includes the comparisons we made between kuramoto model and the network of bursting neurons . our conclusions are left to the final section . the choose of a suitable model describing the dynamics of biological neurons is dictated by some requirements . first the model must take into account the kind of dynamics one wishes to describe @xcite . for example , if all one needs is to describe a spiking neuron , for the sake of neural coding simulations for example , a simple leaky integrate - and - fire ( lif ) model would be enough @xcite . however if one needs to describe the interplay between different ionic currents flowing through the neuron membrane , the hodgkin - huxley ( hh ) model would be a natural choice @xcite . on the other hand , the hh model would require far more computational power than the lif since the former involves four complicated first - order differential equations whereas the latter just one simple equation . with bursting neurons this criterion also holds . given that bursting results from the interplay between fast and slow ionic currents , hodking - huxley - type models would need at least one more equation to describe slow @xmath12 modulation @xcite . a model of thermally sensitive neurons exhibiting bursting has been proposed by huber and braun @xcite , which describes spike train patterns experimentally observed in facial cold receptors and hypothalamic neurons of the rat @xcite , electro - receptors organs of freshwater catfish @xcite , and caudal photo - receptor of the crayfish @xcite . however , the huber - braun model has @xmath3 differential equations for each neuron , and computational limitations impose restrictions to its use for large networks @xcite . if numerical simulations do not need to take into account the effect of system parameters and only the phenomenological aspects of bursting are relevant , then a good choice is the two - dimensional mapping equations proposed by rulkov @xcite @xmath13}^2 } + y(n ) , \\ \label{rulkovy } y(n+1 ) & = & y(n ) - \sigma x(n ) - \beta,\end{aligned}\ ] ] where @xmath14 is the fast and @xmath15 is the slow dynamical variable , whose values are taken at discrete time @xmath16 , with @xmath17 and @xmath18 . time evolution of the ( a ) fast and ( b ) slow variables in the rulkov map ( [ rulkovx])-([rulkovy ] ) for @xmath19 , @xmath20.,scaledwidth=60.0% ] the parameter @xmath21 affects directly the spiking timescale , its values being chosen in the interval @xmath22 $ ] so as to produce chaotic behavior for the evolution of the fast variable @xmath23 , characterized by an irregular sequence of spikes . the parameters @xmath24 and @xmath25 , on their hand , describe the slow timescale represented by the bursts , and take on small values ( namely @xmath26 ) so as to model the action of an external dc bias current and the synaptic inputs on a given isolated neuron @xcite . the rulkov model was derived using dynamical rather than biophysical hypotheses . we choose the parameter @xmath21 so as to yield chaotic behavior for the characteristic spiking of the fast variable @xmath27 [ fig . [ rulkovfig](a ) ] . the bursting timescale , on the other hand , comes about the influence of the slow variable @xmath28 , which provide a modulation of the spiking activity due to a saddle - node bifurcation [ fig . [ rulkovfig](b ) ] . on comparing the dynamics of the rulkov map with similar results of the huber - braun model one is led to an approximate correspondence between variables of the rulkov map and variables with biophysical significance : the discrete time @xmath29 in the map corresponds to @xmath30 of the continuous time ; and @xmath14 stands for the membrane potential , in such a way that each unit of @xmath14 in fig . [ rulkovfig](a ) corresponds to _ circa _ @xmath31 . the rulkov model has been used in several numerical investigation of coupled neuron models @xcite due to the chaoticity of the @xmath14-dynamics of the rulkov map , the duration of each burst suffers a slight variability . the beginning of each burst can be chosen rather arbitrarily , but it turns out that a useful choice is to consider the local maxima of the variable @xmath15 : let @xmath32 be the time at which the @xmath33th burst begins [ see fig . [ rulkovfig](b ) ] . hence the duration of this burst is @xmath34 . we can define a geometric bursting phase by considering a variable @xmath35 that increases of @xmath36 after a bursting event . a linear interpolation gives @xcite [ phase ] ( n ) = 2 + 2 , ( n_n n_+1 ) . we can also define a bursting ( angular ) frequency , which gives the time rate of the phase evolution : @xmath37 in the rulkov model ( [ rulkovx])-([rulkovy ] ) the variable @xmath27 plays the role of the membrane potential at discrete time @xmath16 , where @xmath17 . hence the difference @xmath38 can be interpreted as the time derivative of the potential . if the membrane capacitance is scaled to the unity , it amounts to the transmembrane current . hence the effect of coupling is to inject a synaptic current in the equation for the @xmath14-variable ( [ rulkovx ] ) . let us denote by @xmath39 the membrane potential of the @xmath40th neuron ( @xmath41 ) . the equations for a network of coupled neurons are @xmath42}^2 } + y_i(n ) + \varepsilon \sum_{j=1}^{n } a_{ij } x_j(n ) , \\ \label{y } y_i(n+1 ) & = & y_i(n ) - \sigma x_i(n ) - \beta,\end{aligned}\ ] ] in which @xmath43 is the adjacency matrix , which defines the type of synapse which connects the @xmath40th and @xmath44th neurons ( respectively post - synaptic and pre - synaptic ) . if the spatial distance between them is taken into account , we can model an electrical synapse by a band - diagonal adjacency matrix which includes only near - neighbors of a given neuron . chemical synapses are thus represented here by off - band - diagonal elements , since they include the effect of distant neurons . this is a simplified model , though , since it does not take into account the number of open channels in the post - synaptic neuron . for studies of synaptic - dependent phenomena , like plasticity , memory storage , learning , pattern recognition , etc . this simplified form of the coupling term may not be sufficient for numerical simulations . moreover , we consider all interactions as bidirectional , in such a way that the adjacency matrix is symmetric ( @xmath45 ) , what is more likely to be the case in electrical than chemical synapses . however these simplifications are acceptable in a model as long as we are interested only in the effect of the network topology on the neuron dynamics . an example is the investigation of the dbs effects related to tremor associated with bursting synchronization in the thalamus @xcite . the only parameter in our model that is to characterize the intensity of the synaptic connections is the coupling strength @xmath46 . it takes on positive values for excitatory synapses and negative values for inhibitory ones . since @xmath46 is the same for all neurons in our model we can not consider here the case in which part of the synapses are excitatory ( @xmath47 ) and part inhibitory . a modification such as this , although simple to implement in principle , would make the model more difficult to compare with minimal models of phase dynamics . the only restriction on the values of @xmath46 is that the coupling term itself can not be too large so as to drive the neuron dynamics off a bursting state . this parameter must be adjusted by trial - and - error . since in neuronal assemblies the neurons are likely to exhibit some diversity , we choose randomly the values of @xmath21 within the interval @xmath22 $ ] according to a specified probability distribution function ( pdf ) @xmath48 @xcite . in spite of this we keep the other parameters with the same values ( namely @xmath49 ) since this do not change our results in an appreciable way . we can compute the bursting phase @xmath50 and the corresponding angular frequencies @xmath51 using ( [ phase ] ) and ( [ frequency ] ) , respectively , for the coupled neurons in the same way as we did for isolated ones . when the @xmath21 parameter is chosen in the interval @xmath52 $ ] there is a linear relation between @xmath21 and the mean burst frequency . for this reason when we choose the @xmath21 parameter for each neuron according with a given pdf @xmath48 we generate a pdf for the bursting frequency of uncoupled neurons denoted by @xmath53 with the same shape as @xmath54 . in this work we consider two types of distributions : ( i ) a waterbag ( or uniform ) distribution : [ uniform ] g_w ( ) = \ { ccc & & ab + 0 & . , where @xmath55 and @xmath56 ; and ( ii ) a truncated cauchy distribution : [ cauchy ] g_t ( ) = ^-1 1 + ( ) ^2 ^-1 , where @xmath55 , @xmath56 , @xmath57 is the position of the distribution peak , and @xmath58 is the half - width at half - maximum of the pdf @xcite . time evolution of ( a ) mean field and ( c , e ) fast variables of two selected neurons in a network of uncoupled rulkov neurons . ( b , d , f ) stand for coupled neurons.,scaledwidth=70.0% ] we shall defer the discussion of the nature and properties of the adjacency matrix to the following section . for the moment let us assume that a convenient form of @xmath43 has been given to the system of coupled rulkov neurons ( [ x])-([y ] ) . one distinctive effect of synaptic coupling is bursting synchronization : two or more neurons begin their bursting activity at the same times ( up to a given tolerance ) , regardless of whether or not the ensuing sequence of spikes coincides . a example is provided by fig . [ bs ] , where we consider two randomly selected neurons in a network : in the uncoupled case the neurons burst at different and uncorrelated times [ fig . [ bs](c ) and ( e ) ] , what makes the mean field [ meanfield ] x(n ) = _ i=1^n x_i(n ) , to have small - amplitude fluctuations [ fig . [ bs](a ) ] . for large enough values of @xmath46 the coupled system of neurons present bursting synchronization as illustrated by the two selected neurons [ fig . [ bs](d ) and ( f ) ] . the mean field in this case undergoes large - amplitude oscillations with the same frequency as of the bursting neurons [ fig . [ bs](b ) ] . another quantitative characterization is provided by the order parameter magnitude and its time average given , respectively , by @xcite [ order ] r(n ) = _ i=1^n e^i_i(n ) , = _ n=1^n r(n ) , where @xmath59 is chosen so as to yield stationary values of @xmath60 . if the neurons burst exactly at the same time their phases coincide and hence the normalized sum in ( [ order ] ) gives @xmath61 . on the other hand , if the neuron bursting is so uncorrelated that the times at which they burst are practically random the summation gives nearly zero and @xmath62 for such an extreme case . in finite networks there is likely to be chance correlations , hence we consider @xmath63 as a threshold for the transition from a non - synchronized to a partially synchronized state . intermediate cases ( @xmath64 ) thus represent partially synchronized states . neurons are connected by axons and dendrites , so that we can regard those neurons as embedded in a three - dimensional lattice . however , due to the high connectivity of the neurons it is necessary to use complex networks to describe the properties of neuronal assemblies . these complex networks can be viewed in two basic levels : a microscopic , neuroanatomic level , and a macroscopic , functional level . studies in the former level are limited to those few examples in which there is available data on the neuronal connectivity , as the worm _ c. elegans _ @xcite . in the second ( macroscopic ) level of description of neural networks , the use of non - invasive techniques as electroencephalography , magnetoencephalography and functional magnetic resonance imaging provides anatomical and functional connectivity patterns between different brain areas @xcite . this information provides a way to study the brain cortex , considering the latter as being divided into anatomic and functional areas , linked by axonal fibers . scannell and coworkers have investigated the anatomical connectivity matrix of the visual cortex for the macaque monkey and the cat @xcite . the basic elements of a complex network are nodes and links . a link connects two or more nodes , and these connections can be unidirectional or bidirectional . in the language of neuronal networks a node can be either a single neuron or a cortical area , depending on the level of characterization . if we consider a microscopic description the links are synapses ( electrical or chemical ) , whereas they are axonal fibers in networks of cortical areas or simply they stand for the functional relationship between the corresponding regions in the cortex . a complex network can be characterized by various topological and metrical properties @xcite . for our purposes we will need only two of them : the average path length @xmath65 and the average clustering coefficient @xmath66 . the path length between two nodes @xmath67 and @xmath68 is the minimum number of links that must be traversed through the network to travel from @xmath67 to @xmath68 . the average path length @xmath65 is the path length averaged over all pairs of nodes in the network . the average clustering coefficient @xmath66 of a network is defined as the average fractions of pairs of neighbors of a given node which happen also to be neighbors of each other @xcite . for example , if a node @xmath67 is connected to nodes @xmath68 and @xmath69 , and if @xmath68 and @xmath69 are themselves connected , then @xmath67 , @xmath68 and @xmath69 form a triangular cluster . the value of @xmath66 turns to be the fraction of triangular clusters that exist in the network with respect to the total possible number of such clusters . from the graph - theoretical perspective a complex network can be described by an adjacency matrix @xmath43 , whose elements are equal to the unity if the nodes are connected and zero otherwise . in studies of cortical area networks , both in the anatomic and functional levels , each link is sometimes given a specific weight , which can be even time - varying in investigations of plasticity . since we discard self - interactions , the diagonal elements of the adjacency matrix are zero . in a globally coupled network all nodes are connected with all other nodes . since in this case the contribution of the coupling would increase with the number of nodes , the coupling term is usually divided by the number of nodes . in such a network each node can be regarded as being coupled to the mean field of other nodes . the adjacency matrix of global networks is constant : @xmath70 for all @xmath71 . there have been numerical investigations of such networks with respect to strategies of dbs @xcite . the coupling term in ( [ x ] ) will be divided by the number of neurons , hence we shall redefine it as [ xi ] = ( color online ) variation of the order parameter magnitude with the coupling strength for a globally coupled ( a ) , and a randomly coupled ( b ) network of @xmath72 rulkov neurons and probability @xmath73 . black and red lines stand , respectively , for a waterbag and a truncated cauchy distribution of values of the @xmath21 parameter in the @xmath74 $ ] . the green curves represent a polynomial fit given by eq . ( [ rfit ] ) . the insets zoom the behavior near the transition to bursting synchronization.,scaledwidth=100.0% ] the variation of the order parameter magnitude of the coupled network with the coupling strength above is depicted in fig . [ g](a ) , where we plot @xmath75 as a function of @xmath76 when the values taken on by the parameter @xmath77 are drawn from a waterbag pdf ( black line ) and a truncated cauchy distribution ( red line ) . since we used randomly chosen initial conditions for the neurons @xmath78 we considered averages over @xmath79 different realizations of the initial pattern . in both cases the behavior is qualitatively similar : for small coupling values there is no bursting synchronization and , after a critical coupling value @xmath80 we begin to observe partial synchronization and , further on , complete phase synchronization . the approach to the latter differs with respect to the pdf @xmath54 , being slightly faster for a truncated cauchy distribution than for a waterbag pdf [ see the inset of fig . [ g](a ) ] . the `` physical '' distance between two neurons does not play any role in globally coupled networks , since the contribution is counted in the same way for all nodes . this deficiency has been circumvented by considering a network in which the coupling strength decreases with the lattice distance as a power - law , with exponent @xmath81 @xcite . if @xmath82 the lattice distance does not matter and we recover the globally coupled case . as the exponent @xmath81 tends to infinity we approach a locally coupled network , where only the nearest - neighbor nodes have to be taken into account . unlike global networks , random network present a typically small density of links . random ( erds - renyi ) networks are obtained by from @xmath83 initially uncoupled nodes and building @xmath84 links between randomly chosen pairs with a uniform probability @xmath85 ( avoiding self - interactions ) @xcite . since the total possible number of links is @xmath86 one has [ prober ] p = . the average path length of er networks is typically very small since it scales as the logarithm of the network size ( @xmath87 ) . moreover the clustering coefficient of er networks is likewise small since @xmath88 . since the contribution of the coupling term is typically small we do not need here to rescale it as we did before . we show , in fig . [ g](b ) , the variation of the order parameter magnitude with the coupling strength @xmath46 , showing a qualitatively similar picture but with a considerably lower critical coupling strength @xmath89 for the transition from a non - synchronized to a partially synchronized bursting . if the probability @xmath85 is too small there may not occur bursting synchronization at all , and we require a minimum number of links to observe such phenomenon , as shown in fig . [ g](b ) , where @xmath73 and we have @xmath90 links . random networks of the erds - renyi type represent one limit case of a spectrum of networks of which the other end comprises regular networks , which are lattices of @xmath83 nodes for which each node has links to its @xmath91 nearest and next - to - the - nearest neighbors , hence there are local connections only . the average path length of one - dimensional regular lattices is @xmath92 and the clustering coefficient is given by @xmath93/[4(z-1)]$ ] @xcite . in between those limiting cases we have small - world networks , for which we typically have @xmath94 and @xmath95 . it is possible to obtain them from regular lattices essentially by two procedures : ( i ) watts - strogatz ( ws ) and ( ii ) newman - watts ( nw ) ones . ws networks are obtained from regular lattices by going through each of its links and , with some specified probability @xmath85 , randomly rewiring that link , moving one of its ends to a new node randomly chosen from the rest of the lattice @xcite . the coordination number of the network is still @xmath91 on average , but the number of links per node may be greater or smaller than @xmath91 due to the possible existence of nonlocal shortcuts , in addition to the local links . one possible disadvantage of this construction is that , if @xmath83 is small enough , the rewiring process may create clusters disconnected to the rest of the network , for which nodes the path length would be obviously infinite . the latter problem can be circumvented in nw networks , that are constructed similarly to the ws ones , but the nonlocal shortcuts are added to the lattice with probability @xmath85 , instead of being rewired @xcite . a small - world network of either ws or nw types is essentially described by the probability of nonlocal shortcuts @xmath85 . by computing both @xmath65 and @xmath66 we can get a range of @xmath85 for which the small - world property is fulfilled by the network . the small - world property requires the network to exhibit a relatively small path length while retaining an appreciable degree of clustering , i.e. the following conditions @xmath94 and @xmath95 may hold for many different types of networks of the ws or nw types @xcite . the adjacency matrix for such networks is symmetric and band diagonal ( with zero diagonal elements ) and the non - diagonal parts are sparse , most of their elements being equal to zero . ( color online ) ( a ) dependence of the ratios @xmath96 ( black line ) and @xmath97 ( red line ) with the probability of shortcuts in a small - world network of the newman - watts type with @xmath72 and @xmath98 ; ( b ) dependence of the merit figure @xmath24 with @xmath85 ; ( c ) variation of the order parameter magnitude with the coupling strength for a small - world network of rulkov neurons with probability @xmath99 . black and red lines stand , respectively , for a waterbag and a truncated cauchy distribution of values of the @xmath21 parameter in the @xmath22 $ ] . the green curve represents a polynomial fit given by eq . ( [ rfit ] ) . the insets zoom the behavior near the transition to bursting synchronization.,scaledwidth=90.0% ] if the probability of shortcuts is zero we have the average path length and cluster coefficient of a regular networks , denoted as @xmath100 and @xmath101 , respectively , for a network of @xmath102 neurons with @xmath103 local connections . in fig . [ s](a ) we plot the ratios @xmath96 and @xmath97 as a function of the probability of randomly chosen shortcuts @xmath85 in the nw procedure . the small - world property holds as long as the ratio @xmath96 is large and @xmath97 is small , what yields an interval @xmath104 $ ] . a further confirmation of this interval is provided by the merit figure @xmath24 , defined by [ sigma ] = , where [ ratios ] = , = , in such a way that the larger is @xmath24 the better the small - world property holds for the network . we plot the merit figure as a function of @xmath85 in fig . [ s](b ) , which indeed assumes larger values in the interval @xmath104 $ ] . the variation of the order parameter magnitude for a small - world network with @xmath99 is depicted in fig . [ s](c ) as a function of @xmath46 , showing a transition to partial bursting synchronization for @xmath105 . in scale - free networks the connectivity @xmath106 ( the number of links per node @xmath40 ) satisfies a power - law pdf [ sf ] p(k ) ~k^- , ( > 1 ) , in such a way that highly connected neurons are connected , on the average , with highly connected ones , a property also found in many social and computer networks @xcite . functional magnetic resonance imaging experiments have suggested that some brain activities can be assigned to scale - free networks , with a scaling exponent @xmath107 between @xmath108 and @xmath109 , with a mean connectivity @xmath110 @xcite . in fact , this scale - free property is consistent with the fact that the brain network increases its size by the addition of new neurons , and the latter attach preferentially to already well - connected neurons . in this paper we use the barabsi - albert ( ba ) coupling prescription through a sequence of steps , starting from an initial random ( erds - rnyi ) network of size @xmath111 nodes and @xmath112 random connections @xcite . every step we add a new node to the network which makes two connections : the first is determined by a uniform probability of connection among the vertices in the network and the second link is chosen such that the probability of connection decays according with the degree of connectivity of each node . hence this second link is more likely to be attached to the most connected node in the network than to the less connected one . when the network reaches the desired size @xmath83 we stop adding new nodes . ( color online ) ( a ) probability distribution function for the number of connections per site in a network obtained from ba procedure and @xmath72 nodes ; ( b ) variation of the order parameter magnitude with the coupling strength for a small - world network of rulkov neurons with probability @xmath99 . black and red lines stand , respectively , for a waterbag and a truncated cauchy distribution of values of the @xmath21 parameter in the @xmath74 $ ] . the green curve represents a polynomial fit given by eq . ( [ rfit ] ) . the insets zoom the behavior near the transition to bursting synchronization.,scaledwidth=100.0% ] we applied the ba procedure until the final network has @xmath72 nodes . we then obtained a numerical approximation for the pdf of connections per node @xmath113 , depicted in fig . [ law](a ) . the solid red line is a least squares fit giving a power - law dependence of the form ( [ sf ] ) with an exponent @xmath114 , confirming that the network is scale - free indeed . the variation of the order parameter magnitude with @xmath46 is plotted in fig . [ law](b ) , illustrating the transition to bursting synchronization with @xmath115 . the behavior of the average order parameter magnitude with @xmath46 after the transition to bursting synchronization at @xmath116 can be fitted , when the frequency distribution is a truncated cauchy pdf , by the following expression [ rfit ] ^s , where the exponents are different according to the type of network [ table [ fitab ] ] . .[fitab ] critical coupling strength and fitting parameters for some types of networks with @xmath102 nodes and a truncated cauchy distribution . [ cols="^,^,^,^,^",options="header " , ] after having chosen adequate values of the parameters appearing in the pdf of both rulkov and kuramoto models , we can compare the results for them with respect to the transition to bursting synchronization . in figs . [ compfig](a ) to ( d ) we compute the order parameter magnitude for the rulkov ( black line : waterbag , red line : cauchy ) and kuramoto ( green line : waterbag , blue line : cauchy ) as a function of the coupling strength for the network topologies considered in this paper . the evolution of @xmath60 is nearly the same for both models , with small differences . for example the value of @xmath60 in the gkm reaches its maximum slightly before ( i.e. with smaller coupling strengths ) the rulkov network does . this precedence of gkm in comparison with rulkov is observed for all connection topologies we have considered in this work . ( color online ) order parameter magnitude vs. coupling strength for a network of rulkov neurons ( black : waterbag , red : cauchy ) and kuramoto oscillators ( green : waterbag , blue : cauchy ) with @xmath72 nodes , for ( a ) global coupling , ( b ) erds - rnyi ; ( c ) small - world , and ( d ) scale - free.,scaledwidth=100.0% ] in this work we propose minimal models for the description of bursting neurons using some of the most used connection topologies for neuronal networks which can be used in numerical experiments . one of the simplest models that describe bursting is a two - dimensional discrete - time mapping proposed by rulkov , and we considered networks of coupled rulkov neurons using global , random ( erds - renyi ) , small - world ( newman - watts ) , and scale - free ( barabasi - albert ) topologies . we can define a geometrical phase for coupled rulkov neurons , such that bursting synchronization is equivalent to some form of phase synchronization . using the time rate of the bursting phase we can define a bursting frequency and thus investigate frequency synchronization in such networks . using a complex order parameter we quantify the state of phase synchronization of the system . as the coupling strength is increased , there is a transition from a non - synchronized to a partially phase - synchronized state , which evolves to a completely phase - synchronized one for larger values of the coupling strength . we computed the corresponding critical values of the coupling strength for the connection topologies considered in this work . the existence of a phase reduction for the coupled rulkov neuron system suggests the usefulness of a generalized kuramoto model which , in the present context , can be considered a minimal model for bursting neurons . we presented in this paper some quantitative results that justify this claim . in order to compare both models , however , it is necessary to adjust the corresponding probability distribution functions for the natural frequencies . we did so by using theoretical estimates for the critical coupling strength for the generalized kuramoto model in different connection topologies . after choosing the pdf parameters in a suitable way the behavior of the order parameter is very similar for both the rulkov and kuramoto models . in this appendix we show that , in the scenario near a phase synchronized regime , one can perform a phase reduction from a network of coupled oscillators to a generalized kuramoto model , what explains why the properties of a coupled rulkov neuronal network are similar to the kuramoto model . we stress that this similarity does not necessarily apply to the situation near a frequency synchronized regime , since the dynamical requirements are different . our presentation follows closely that in ref . @xcite and the general treatment given in ref . @xcite . let a general @xmath117-dimensional flow be given by [ flow ] = * f*(*x * ) , where @xmath118 is a @xmath117-dimensional vector in the system phase space and @xmath119 a vector field . we assume that there is a stable period-@xmath120 orbit [ orbit ] * x*_0(t ) = * x*_0(t+t ) . the `` slow '' dynamics along this periodic orbit can be described by a phase @xmath121 which time evolution is given by [ phasedef ] ( * x * ) = _ * x * ( * x * ) = 1 . now let us consider a network of coupled oscillators with a slight mismatch in their parameters described by [ coupled ] _ i = * f*(*x*_i ) + * f*_i(*x*_i ) + _ j a_ij * v*(*x*_i,*x*_j ) , ( i=1,2, n ) , where @xmath122 is different for each oscillator and stands for the vector field part containing slightly mismatched parameters , @xmath123 is the coupling strength , @xmath124 the adjacency matrix , and @xmath125 is a coupling function . from ( [ phasedef ] ) , the slow phase @xmath126 of the coupled oscillators is implicitely defined by the function [ z ] * z*(_i ) = _ * x * ( * x*_0(_i ) ) , where @xmath127 is a period-@xmath120 stable orbit . the time evolution of the phase is then governed by = 1 + _ j a_ij * z*(_i ) * v*(*x*_i,*x*_j ) + * z*(_i)*f*_i(*x*_i ) , for @xmath128 . on introducing an auxiliary phase @xmath129 and using a time average over a period @xmath120 , by keeping @xmath130 fixed we can write , in a first - order approximation , an equation governing the time evolution of @xmath130 : [ psi ] _ i = w_i + _ j a_ij ( _ i,_j ) , where [ w ] w_i & = & _ 0^t * z*(t+_i)*f*_i(*x*_i ) dt , + [ gama ] ( _ i,_j ) & = & _ 0^t * z*(t+_i)*v*(*x*_0(t+_i),*x*_0(t+_j ) ) dt , play the roles of frequencies and coupling functions , respectively , for the auxiliary phases @xmath130 . if we are close to a global phase synchronized state , for which @xmath131 for any pairs of oscillators @xmath132 , we have an approximate form for the coupling function @xmath133 . introducing the time - dependent variable @xmath134 and supposing that @xmath135 there results [ gama2 ] ( _ i,_j ) = _ 0^t+_j * z*(+_i-_j ) * v*(*x*_0(+_i-_j),*x*_0(+_j ) ) dt . by expanding the coupling function @xmath136 in a power series and assuming that @xmath137 is nearly constant over the periodic orbit @xmath127 , we get [ gama3 ] ( _ i,_j ) a + b ( _ i-_j ) , where [ a ] a & = & _ 0^t * z * * v*(*x*_0(),*x*_0 ( ) ) d , + [ b ] b & = & _ 0^t * z * ( _ * x**v*(*x*,*y * ) ) _ * x*=*y*=*x*_0 ( ) d , notice that this approximation holds whenever [ cond ] _ * x**v*(*x*,*y * ) 0 for any @xmath118 and @xmath138 belonging to the periodic orbit @xmath127 . substituting ( [ gama3 ] ) into ( [ psi ] ) yields [ psi1 ] _ i _ i + s_i a + b _ j a_ij ( _ i-_j ) , where @xmath139 is the intensity of the @xmath40th node of the network . as long as we deal with networks for which the intensities @xmath140 present only a small variation over the network , we can take @xmath140 as practically constant , i.e. , independent of @xmath40 . moreover , since we are by hypothesis near a global phase synchronized state , the phase difference @xmath141 is small enough to justify the replacement @xmath142 , in such a way that the equation governing the time evolution of the auxiliary phases ( near a phase - synchronized situation ) is a generalized kuramoto model [ kuramoto1 ] _ i w_i + _ j a_ij ( _ i-_j ) , where @xmath143 and @xmath144 . this work was made possible through financial support from the brazilian research agencies , capes and cnpq . fabiano ferrari thanks to ruedi stoop and the institute of neuroinformatics ( uzh , ethz ) for having received him as a visiting student during 2014 through the program `` science without boundaries '' . we thank dr . j. c. p. coninck for fruitful discussions and helpful suggestions . 00 b. w. connors and m. j. gutnick , _ intrinsic firing patterns of diverse neocortical neurons _ , trends in neuroscience * 13 * , 99 - 104 ( 1990 ) d. a. blank and r. stoop , _ collective bursting in populations of intrinsically nonbursting neurons _ , z. naturforsch . a , * 54 * , 617 - 627 ( 1999 ) . c. m. gray and d. a. mccormick , _ chattering cells : superficial pyramidal neurons contributing to the generation of synchronous oscillations in the visual cortex _ , science * 274 * , 109 - 113 ( 1996 ) . r. stoop , d. blank , a. kern , j.j . vyver , m. christen , s. lecchini , c. wagner , _ collective bursting in layer iv synchronization by small thalamic inputs and recurrent connections _ , cognitive brain research , * 13 * , 293 - 304 ( 2002 ) . h. markram , m. toledo - rodriguez , y. wang , a. gupta , g. silberberg , and c. wu , _ interneurons of the neocortical inhibitory system _ , nature reviews neuroscience * 5 * , 793 - 807 ( 2004 ) . h. su , g. alroy , e. d. kirson , and y. yaari , _ extracellular calcium modulates persistent sodium current - dependent burst - firing in hippocampal pyramidal neurons _ , journal of neuroscience , * 21 * , 4173 - 4182 ( 2001 ) . e. j. ramcharan , j. w. gnadt , and s. murray sherman , _ burst and tonic firing in thalamic cells of unanesthetized , behaving monkeys _ , visual neuroscience * 17 * , 55 - 62 ( 2000 ) . m. d. womack and k. khodakhah , _ active contribution of dendrites to the tonic and trimodal patterns of activity in cerebellar purkinje neurons _ , j. neurosci . * 22 * , 10603 - 10612 ( 2002 ) . r. krahe and f. gabbiani , _ burst firing in sensory systems _ , nature reviews neuroscience * 5 * , 13 - 23 ( 2004 ) . w. metzner , c. koch , r. wessel , and f. gabbiani , _ feature extraction by burst - like spike patterns in multiple sensory maps _ , j. neurosci . * 18 * , 22832300 ( 1998 ) . r. c. elson , a. i. selverston , r. huerta , n. f. rulkov , m. i. rabinovich and h. d. i. abarbanel . _ synchronous behavior of two coupled biological neurons . * 81 * , 56925695 ( 1998 ) . p. varona , j. j. torres , h. d. i. abarbanel , m. i. rabinovich , and r. c. elson . _ dynamics of two electrically coupled chaotic neurons : experimental observations and model analysis . _ biological cybernetics * 84 * , 91101 ( 2001 ) . v. makarenko and r. llins . _ experimentally determined chaotic phase synchronization in a neuronal system . sci . * 95 * , 1574715752 ( 1998 ) . m. i. rabinovich , p. varona , j. j. torres , r. huerta and h. d. i. abarbanel . _ slow dynamics and regularization phenomena in ensembles of chaotic neurons_. physica a , * 263 * , 405414 ( 1999 ) . h. kamioka , e. maeda , y. jimbo , h. p. c. robinson , and a. kawana , _ spontaneous periodic synchronized bursting during formation of mature patterns of connections in cortical cultures _ , neuroscience letters , * 206 * , 109112 ( 1996 ) . m. steriade , d. a. mccormick , and t. j. sejnowski , _ thalamocortical oscillations in the sleeping and aroused brain _ , science , * 262 * , 679685 ( 1993 ) . m. bevan , p. magill , d. terman , j. bolam , and c. wilson , _ move to the rhythm : oscillations in the subthalamic nucleus - external globus pallidus network _ , trends neurosci . * 25 * , 525 - 531 ( 2002 ) . p. j. uhlhaas and w. singer , _ neural synchrony in brain disorders : relevance for cognitive dysfunctions and pathophysiology _ , neuron * 52 * , 155 - 168 ( 2006 ) . p. brown , _ abnormal oscillatory synchronisation in the motor system leads to impaired movement _ , curr opin neurobiol , * 17 * , 656 - 664 ( 2007 ) . w. d. hutchison , j.o . dostrovsky , j.r . walters and _ et al . _ , _ neuronal oscillations in the basal ganglia and movement disorders : evidence from whole animal and human recordings _ , j. neurosci . * 24 * , 9240 - 9243 ( 2004 ) . c. park , r. m. worth , and l. l. rubchinsky , _ fine temporal structure of beta oscillations synchronization in subthalamic nucleus in parkinson s disease _ , j. neurophysiol . * 103 * , 2707 - 2716 ( 2010 ) . c. hammond , c. , h. bergman , and p. brown , _ pathological synchronization in parkinson s disease : networks , models and treatments _ , trends neurosci . * 30 * , 357 - 364 ( 2007 ) . a. l. benabid , p. pollak , c. gervason , d. hoffmann , d. m. gao , m. hommel , j. e. perret , and j. de rougemont , _ long - term suppression of tremor by chronic stimulation of the ventral intermediate thalamic nucleus _ , lancet * 337 * , 403 - 406 ( 1991 ) . s. blond , d. caparros - lefevre , f. parker , r. assaker , h. petit , j .- d . guieu , and j .- l . christiaens , _ control of tremor and involuntary movement disorders by chronic stereotaxic stimulation of the ventral intermediate thalamic nucleus _ , journal of neurosurgery * 77 * , 62 - 68 ( 1992 ) . m. c. rodriguez - oroz , j. a. obeso , a. e. lang , j .- l . houeto _ _ , _ bilateral deep brain stimulation in parkinson s disease : a multicentre study with @xmath6 years follow - up _ , brain * 128 * , 2240 - 2249 ( 2005 ) . a. albanese and l. romito , _ deep brain stimulation for parkinson s disease : where do we stand ? _ , frontiers in neurology , * 2 * , 33 ( 2011 ) . p. a. tass , _ a model of desynchronizing deep brain stimulation with a demand - controlled coordinated reset of neural subpopulations _ , biol . cybern . * 89 * , 81 - 88 ( 2003 ) j .- p . pfister , p.a . tass , _ stdp in oscillatory recurrent networks : theoretical conditions for desynchronization and applications to deep - brain stimulation _ , frontiers in computational neuroscience , * 4 * , 22 ( 2010 ) . c. hauptmann , o. popovych , p.a . tass , _ effectively desynchronizing deep brain stimulation based on a coordinated delayed feedback stimulation via several sites : a computational study _ , biological cybernetics * 93 * , 463 - 470 ( 2005 ) . c. a. s. batista , a. m. batista , j. c. a. de pontes , r. l. viana and s. r. lopes . _ chaotic phase synchronization in scale - free networks of bursting neurons _ , phys . e * 76 * , 016218 ( 2007 ) . c. a. s. batista , s. r. lopes , r. l. viana , and a. m. batista , _ delayed feedback control of bursting synchronization in a scale - free neuronal network _ , neural networks * 23 * , 114 - 124 ( 2010 ) . a. s. pikovsky , m. g. rosenblum , g. osipov , and j. kurths . _ phase synchronization of chaotic oscillators by external driving_. physica d * 104 * , 219 - 238 ( 1997 ) . y. kuramoto , _ chemical oscillations , waves , and turbulence _ ( dover publications , new york , 1984 ) . j. a. acebrn , l. l. bonilla , c. j. p. vicente , f. ritort , and r. spigler , _ the kuramoto model : a simple paradigm for synchronization phenomena _ , reviews of modern physics * 77 * , 137 - 185 ( 2005 ) s. strogatz , _ from kuramoto to crawford : exploring the onset of synchronization in populations of coupled oscillators _ , physica d * 143 * , 120 ( 2000 ) . g. restrepo , e. ott , b. r. hunt , _ onset of synchronization in large networks of coupled oscillators _ , phys . e , * 71 * , 036151 ( 2005 ) . d. l. wang . _ the time dimension for scene analysis_. ieee transcation on neural networks , * 16 * , 1401 ( 2005 ) . b. ibarz , j. m. casado , and m. a. f. sanjun . _ map - based models in neuronal dynamics_. phys * 501 * , 1 ( 2011 ) . c. koch and i. segev , _ methods in neuronal modeling : from ions to networks _ ( the mit press , cambridge , mass . 1999 ) a. l. hodgkin and a. f. huxley , _ a quantitative description of membrane current and its application to conduction and excitation in nerve _ , j. physiology * 117 * , 500 - 544 ( 1952 ) . r. e. plant and m. kim . _ mathematical descriptions of a bursting pacemaker neuron by a modification of hodgkin - huxley equations . _ biophys . j. * 16 * , 227 - 244 ( 1976 ) . p. r. shorten and d. j. n. wall . _ a hodgkin huxley model exhibiting bursting oscillations _ , bulletin of mathematical biology , * 62 * , 695 - 715 ( 2000 ) . h. a. braun , m. t. huber , m. dewald , k. schfer , and k. voigt._computer simulations of neuronal signal transduction : the role of nonlinear dynamics and noise_. int . j. bifurcat . chaos * 8 * , 881 ( 1998 ) . w. braun , b. eckhardt , h. a. braun , and m. huber . _ phase - space structure of a thermoreceptor_. phys . e * 62 * , 6352 - 6360 ( 2000 ) . h. a. braun , m. t. huber , n. anthes , k. voigt , a. neiman , x. pei , and f. moss . _ noise - induced impulse pattern modifications at different dynamical period - one situations in a computer model of temperature encoding_. biosystems , * 62 * , 99 - 112 ( 2001 ) h. a. braun , m. dewald , k. schfer , k. voigt , x. pei , k. dolan , and f. moss . _ low - dimensional dynamics in sensory biology 2 : facial cold receptors of the rat . _ j. comput . neurosci . * 7 * , 17 - 32 ( 1999 ) . k. schfer , h. a. braun , r. c. peters , and f. bretschneider . _ periodic firing pattern in afferent afferent discharges from electroreceptor organs of catfish_. pflgers arch . j. physiol . * 429 * , 378 - 385 ( 1995 ) . u. feudel , a. neiman , x. pei , w. wojtenek , h. braun , m. huber , and f. moss . _ homoclinic bifurcation in a hodgkin - huxley model of thermally sensitive neurons_. chaos * 10 * , 231 - 239 ( 2000 ) . n. f. rulkov . _ regularization of synchronized chaotic bursts_. phys . * 86 * , 183 - 186 ( 2001 ) . d. s. andres , f. gomez , f. s. ferrari , d. cerquetti , m. merello , r. l. viana , and r. stoop , _ diffusive coupling modulates neuronal coding in the parkinsonian globus pallidus _ , submitted . n. f. rulkov . _ modeling of spiking - bursting neural behavior using two - dimensional map_. phys . e * 65 * , 041922 ( 2002 ) . s. nadarajah and s. kotz , _ a truncated cauchy distribution _ , internat . techn . * 37 * , 605 - 608 ( 2006 ) . n. f. rulkov , i. timofeev , and m. bazhenov . _ oscillations in large - scale cortical networks : map - based model_. j. comp . * 17 * , 203 - 223 ( 2004 ) . c. a. s. batista , a. m. batista , j. c. a. de pontes , s. r. lopes , and r. l. viana . _ bursting synchronization in scale - free networks . _ chaos , solit . & fractals , * 41 * , 22202225 ( 2009 ) . h. yu , j. wang , b. deng , x. wei , _ et al . _ _ chaotic phase synchronization in small - world networks of bursting neurons . _ chaos , * 21 * , 013127 ( 2011 ) . r. l. viana , a. m. batista , c. a. s. batista , j. c. a. de pontes , f. a. dos s. silva , and s. r. lopes . _ bursting synchronization in networks with long - range coupling mediated by a diffusing chemical substance . _ comm . simulat . , * 17 * , 2924 - 2942 ( 2012 ) . m. v. ivanchenko , g. v. osipov , v. d. shalfeev , j. kurths . _ phase synchronization in ensembles of bursting oscillators_. phys lett . * 93 * , 134101 ( 2004 ) . m. g. rosenblum and a. s. pikovsky . _ delayed feedback control of collective synchrony : an approach to suppression of pathological brain rhythms . e * 70 * , 041904 ( 2004 ) . a. arenas , a. fernndez , and s. gmez , _ a complex network approach to the determination of functional groups in the neural system of caenorhabditis elegans . notes comput . * 515 * , 918 ( 2008 ) . c. c. hilgetag and m. kaiser , organization and function of complex cortical networks , _ in _ `` lectures in supercomputational neuroscience ( dynamics in complex brain networks ) '' , eds . p. b. graben , c. zhou , m. thiel , and j. kurths ( springer , berlin - heidelberg - new york , 2008 ) . g. zamora - lpez , c. zhou , and j. kurths . _ exploring brain function from anatomical connectivity_. frontiers in neuroscience * 5 * , 83 ( 2011 ) . j. w. scannell and m. p. young . _ the connectional organization of neural systems in the cat cerebral - cortex_. curr . biol . * 3 * , 191 - 200 ( 1993 ) . j. w. scannell , c. blakemore , and m. p. young . _ analysis of connectivity in the cat cerebral - cortex_. j. neurosci . * 15 * , 1463 ( 1995 ) . r. albert and a .- _ statistical mechanics of complex networks_. rev . phys . * 74 * , 47 ( 2002 ) . d. j. watts , _ small worlds _ ( princeton university press , princeton , 2000 ) . p. erds and a. rnyi . _ on the evolution of random graphs_. bulletin of the international statistical institute , * 38 * , 343 - 347 ( 1959 ) . m. e. j. newman . _ models of the small world_. j. stat . * 101 * , 819 - 841 ( 2000 ) . d. j. watts and s. h. strogatz . _ collective dynamics of small - world networks_. nature , * 393 * , 440 - 442 ( 1998 ) . m. e. j. newman and d. j. watts . _ renormalization group analysis of the small - world network model_. phys . lett . a * 263 * , 341 - 346 ( 1999 ) . j. c. a. de pontes , r. l. viana , s. r. lopes , c. a. s. batista , and a. m. batista , _ bursting synchronization in non - locally coupled maps _ , physica a , * 387 * , 4417 - 4428 ( 2008 ) . d. r. chialvo , _ critical brain networks _ , physica a * 340 * , 756 - 765 ( 2004 ) . o. sporns , d. r. chialvo , m. kaiser , and c. c. hilgetag , _ organization , development and function of complex brain networks _ , trends cogn . * 8 * , 418 - 425 ( 2004 ) . v. m. eguiluz , d. r. chialvo , g. a. cecchi , m. baliki , and a. v. apkarian , _ scale - free brain functional network _ , phys . lett . * 94 * , 018102 ( 2005 ) . a. l. barabasi and r. albert , _ emergence of scaling in random networks . _ science * 286 * , 509512 ( 1998 ) . m. breakspear , s. heitmann , and a. daffertshofer , _ generative models of cortical oscillations : neurobiological implications of the kuramoto model . _ frontiers in human neuroscience , * 4 * , 114 ( 2010 ) . l. timms and l. q. english . _ synchronization in phase - coupled kuramoto oscillator networks with axonal delay and synaptic plasticity _ , physical review e , * 89 * , 032906 ( 2014 ) . a. arenas , a. daz - guilera , j. kurths , y. moreno , and c. zhou , _ synchronization in complex networks_. phys . rep . * 469 * , 93153 ( 2008 ) . m. jalili , _ failure tolerance of spike phase synchronization in coupled neural networks . _ chaos * 21 * , 033126 ( 2011 ) . c. a. s. batista , e. l. lameu , a. m. batista , s. r. lopes , t. pereira , g. zamora - lpez , j. kurths , and r. l. viana . _ phase synchronization of bursting neurons in clustered small - world networks . e * 86 * , 016211 ( 2012 ) . a. nordenfelt , j. used , and miguel a.f . sanjun , _ bursting frequency versus phase synchronization in time - delayed neuron networks . _ e * 87 * , 052903 ( 2013 ) . t. de l. prado , s. r. lopes , c. a. s. batista , j. kurths , and r. l. viana , _ synchronization of bursting hodgkin - huxley - type neurons in clustered networks _ , phys . e * 90 * , 032818 ( 2014 ) .

strongly correlated fermionic systems , responsible for relevant many - body phenomena such as high - temperature superconductivity , the fractional quantum hall effect or metal - insulator transitions , represent one of the most important theoretical challenges in condensed matter physics . among the simplest possible models of interacting fermions in a 2d lattice is the hubbard model,@xcite which is believed to be one of the keys to understanding the theoretical riddle of high - temperature superconductivity,@xcite and which serves as a good example to illustrate the nature and scale of the difficulties encountered . in spite of a titanic effort by the condensed matter community spanning several decades , still today the phase diagram of the 2d hubbard model and its relation to high - temperature superconductors remain highly controversial . in the absence of exactly solvable models , accurate numerical simulations are essential in order to gain further insight into the physics of strongly correlated systems . while quantum monte carlo ( qmc ) techniques are very powerful in simulating bosonic systems , they suffer from the so - called _ negative sign problem _ in the case of fermionic and frustrated models . @xcite on the other hand , generic 1d lattice systems can be accurately addressed with the density matrix renormalization group ( dmrg ) method , @xcite but this approach scales inefficiently with the lattice size in 2d systems . recent progress in the simulation of 2d fermionic models has been made with a variety of methods . @xcite however , results obtained with different methods are often inconsistent , highlighting the need for further improvement and for alternative approaches . a promising new route to studying strongly correlated fermion systems in a 2d lattice , presently under intense investigation , @xcite is based on using a _ tensor network _ as ground state variational ansatz . for bosonic ( e.g. spin ) 2d lattice models , tensor network anstze include _ projected entangled - pair states _ ( peps ) for inhomogeneous@xcite and homogeneous systems , @xcite and the _ _ multi - scale entanglement renormalization ansatz__@xcite ( mera ) . [ homogeneous peps are also known with names such as ( vertex ) _ _ tensor product states__@xcite ] . the interest in these approaches resides in the fact that they manage to retain some of the useful features of dmrg and qmc , while avoiding their main shortcomings . indeed , peps and mera approaches are free from the negative sign problem that prevents the application of qmc to fermionic and frustrated models . at the same time , and unlike dmrg , both peps and mera can efficiently represent ground states of 2d lattice models . in addition , compared to other variational approaches , peps and mera are relatively unbiased towards specific ground states . still at an early stage of development , the major limitation of these methods is that the cost of simulations increases sharply with the amount of entanglement in the ground state . this limits the range of models that can be analyzed accurately at present . nevertheless , several systems of frustrated antiferromagnets beyond the reach of dmrg and qmc have already been addressed.@xcite in recent months , generalizations of tensor network algorithms to fermionic systems have been put forward independently by several groups.@xcite as a result , it is now possible to study interacting fermions in 2d lattices both within the context of the mera @xcite and peps . @xcite the fundamental new step , common in all the proposals , is to incorporate the fermionic character of the ground state wave function directly into the ansatz . this is accomplished by considering a _ network of fermionic operators _ , that is , a set of linear maps , made of anticommuting operators , that are connected according to a network pattern , as first proposed in refs . for the mera and in ref . for the peps . in actual computations , one is still forced to store and manipulate tensors ( i.e. multi - dimensional arrays of coefficients ) corresponding e.g. to the expansion coefficients of the fermionic maps . the process of ` _ _ fermionization _ _ ' of a tensor network algorithm , i.e. its extension to fermionic systems , can in practice be achieved and visualized in a variety of ( ultimately equivalent ) ways , depending on how the underlying network of fermionic operators is translated into a set of tensors and rules for their manipulation . examples include the use of a jordan - wigner transformation , @xcite or the introduction of additional bond indices between tensors.@xcite a particularly simple form of ` _ _ fermionization _ _ ' of tensor networks was introduced in ref . , where it was applied in the context of the mera . we emphasize that ref . is based on reformulating previous work by corboz , evenbly , verstraete and vidal on fermionic mera,@xcite which in turn had its origins in a key observation by verstraete.@xcite the fermionization procedure of ref . , which applies to any tensor network , is remarkedly simple . it does not require the introduction of a jordan - wigner transformation in the bond indices , or to have to explicitly keep track and dynamically modify a global fermionic order ; neither does it require the introduction of additional bond indices in the tensors . instead , the fermionic character of the tensor network is engraved in two simple rules : ( i ) use of parity invariant tensors and ( ii ) replacement of line crossings with so - called fermionic swap gates . the net result of applying these rules is a modified variational ansatz that can be manipulated using standard tensor network operations ( tensor multiplications , etc . ) , thus producing a straightforward fermionic version of existing tensor network algorithms . importantly , the computational cost of bosonic and fermionic algorithms scales in the same way with the amount of entanglement in the ground state . @xcite this remarkable result was also independently derived in ref . . this paper has two main goals . the first is to explain how to obtain peps algorithms for fermionic systems by applying the above ` _ _ fermionization _ _ ' rules to existing bosonic peps algorithms . fermionic peps were originally proposed by kraus , schuch , verstraete and cirac in ref . and have also been discussed by barthel , pineda and eisert in ref . . our formulation of fermionic peps must be , at some level , equivalent to those proposals . however , the present formulation , which is based on previous independent work,@xcite is remarkably straightforward and appears to be comparatively much simpler . in particular , it allowed us to numerically implement a fermionic peps algorithm for infinite systems by only introducing a small number of changes to existing code for bosonic systems . a second main goal of this paper is to demonstrate the usefulness of fermionic peps . in spite of the several existing formulations of fermionic peps , @xcite and with the exception of ref . , where some qualitative results are reported for the @xmath0 model , no evidence has been presented yet showing that fermionic peps are a good variational ansatz for interacting fermion systems . [ notice , however , that ref . shows that gaussian fermionic peps can represent states of non - interacting fermions ] . here we do present such evidence , in the form of ground state computations for several 2d models . specifically , we use a fermionic version of the infinite peps ( ipeps ) algorithm@xcite to address models on an infinite square lattice . first , results for free spinless fermions are compared with the corresponding exact solution , showing that a peps with small bond dimension is capable of reproducing the ground state energy with several digits of accuracy . then a model of interacting spinless fermions is addressed . qualitatively , the simulation reproduces the phase diagram predicted within hartree - fock , with metal and charge - density wave phases separated by a line of first order phase transitions . at a quantitative level , however , we obtain ground state energies that are lower than those obtained with hartree - fock , and this shifts the boundary between phases significantly . finally , for the @xmath0 model , we obtain ground state energies that are close to those of a specialized gutziller - projected ansatz . the rest of the paper is organized as follows : sec . [ sec : peps ] reviews the peps formalism for bosonic systems and the general fermionization procedure of tensor networks introduced in ref . , which is then applied to peps algorithms . [ sec : ctm ] considers in more detail the fermionic version of the ipeps algorithm for infinite 2d lattices , which was employed to obtain the benchmark results presented in this paper [ sec : results ] describes ground state calculations for systems of free and interacting fermions in an infinite 2d lattice . [ sec : conclusions ] contains some conclusions , while appendix [ app : genoperators ] defines generalized fermionic operators and appendix [ app : twosite ] describes in detail one step of the update in the fermionic ipeps algorithm . _ note on terminology. _ for the purposes of this paper , a _ tensor _ is simply a multi - dimensional array of complex coefficients , and a _ tensor network _ is a set of tensors some of whose indices are connected according to a network pattern , where being connected means that there is a sum or trace over that index , in the sense of tensor multiplication . accordingly , in this paper a bosonic / fermionic tensor network is a tensor network used in the context of simulating a bosonic / fermionic system . thus , in the present formulation a fermionic peps is simply a `` tensor network that serves as a variational ansatz for fermionic systems '' . it is different from a bosonic peps in the presence of special gates called fermionic swap gates ( and in that its tensors are necessarily parity preserving ) . in particular , even though the rules used to create a fermionic tensor network , as introduced in ref . and reviewed here , were obtained by studying how to mimic a network of fermionic operators ( that is , of operators that obey anticommuting relations ) , here a fermionic peps is _ not _ a network of fermionic operators . one of the merits of the present formulation is precisely that it replaces the considerable complexity involved in dealing with a network of fermionic operators with a simple set of rules . in particular , it avoids having to explicitly define , keep track and dynamically modify a fermionic order for the bond indices . the equivalence between our formulation of a fermionic tensor network and a network of fermionic operators was already established in ref . for the case of the mera . a general derivation of this equivalence would distract from the purpose of this paper and will be presented elsewhere . the goal of this section is to introduce a fermionic version of bosonic peps algorithms,@xcite so that they can be applied to simulate fermionic systems in a 2d lattice . we start by reviewing some key aspects of the peps formalism for bosonic systems . this allows us to introduce the notation and the diagrammatic representation of tensors used throughout secs . [ sec : peps ] and [ sec : ctm ] . then we describe the fermionization rules of ref . , which we also extensively review . we apply these rules to obtain a fermionic peps ansatz ( see also refs . ) , and provide a discussion of how fermionic peps algorithms can be obtained by modifying existing bosonic peps algorithms . let us consider a quantum many - body system in a lattice @xmath1 made of @xmath2 sites , labelled by an integer @xmath3 . each site @xmath4 is described by a complex vector space @xmath5 of finite dimension @xmath6 , with basis states @xmath7 . the vector space @xmath5 could represent , for instance , the possible states of a quantum spin sitting on that site of @xmath1 . the system is further characterized by a _ local _ ( _ bosonic _ ) hamiltonian . this is a hermitian operator @xmath8 that ( when expressed in terms of bosonic operators , i.e. operators that commute when acting on different sites ) decomposes as a sum of terms each involving only a small number of sites . let @xmath9 be a pure state , @xmath10 where index @xmath11 labels a basis on site @xmath4 . a task of interest is to compute a specific state @xmath12 somehow related to @xmath13 , e.g. its ground state , and to evaluate the expectation value @xmath14 of some local observable @xmath15 . however , representing a vector @xmath16 requires a number of complex coefficients @xmath17 that grows exponentially in @xmath2 . this poses a serious computational challenge . exact diagonalization techniques are only affordable for small systems ( e.g. , at most @xmath18 for @xmath19 ) , and alternative numerical strategies are required to analyze large systems . projected entangled - pair states@xcite ( peps ) were introduced as a means to obtain an efficient description for some states @xmath16 of a 2d lattice @xmath1 . for concreteness , in this work we consider the case of a square lattice , although all the discussions can be extended to other type of lattices . to each site @xmath20 there corresponds a vector of integers @xmath21 , and we also write @xmath22 to denote a site of lattice @xmath1 . a peps is made of a collection of @xmath2 tensors @xmath23}\}$ ] , one for each site @xmath24 , connected through _ bond indices _ that follow the pattern of links of the lattice @xmath1 . upon tracing over all bond indices , a peps yields a tensor @xmath25 with the @xmath26 complex coefficients @xmath17 of a state @xmath27 . with coefficients @xmath28 for a state @xmath29 of a @xmath30 square lattice this tensor is expressed in terms of a peps made of a set of 9 tensors @xmath31}\}$ ] , one for each site @xmath32 . ( b ) bulk tensor @xmath33}$ ] with components @xmath33}_{ulsdr}$ ] . notice that the legs corresponding to indices @xmath34 , @xmath35 , @xmath36 , @xmath6 and @xmath37 emerge from the circle in anti - clockwise order . ( c ) hermitian conjugate @xmath33\dagger}$ ] of the peps tensor @xmath33}$ ] , represented as its mirror image . notice that the legs corresponding to indices @xmath37 , @xmath6 , @xmath36 , @xmath35 and @xmath34 of @xmath38\dagger})_{rdslu}$ ] emerge again in anti - clockwise order . ( d ) hermitian conjugates of @xmath25 and its peps representation.,width=302 ] throughout this paper a diagrammatic representation of tensors and tensor networks is used , see fig . [ fig : bosonpeps ] . each tensor is depicted as a shape ( circle , square , diamond , etc ) and its indices as emerging lines . a line connecting two shapes ( or starting and ending at the same shape ) denotes an index over which a trace is taken . as an example , fig . [ fig : bosonpeps](a ) represents , in the case of a lattice @xmath1 made of @xmath30 sites , a tensor @xmath25 with 9 indices , corresponding to the coefficients @xmath39 of @xmath40 , followed by a peps made of 9 tensors @xmath31}\}$ ] . the number of indices in a tensor @xmath33}$ ] depends on the number of nearest neighbors of site @xmath22 , with tensors in the bulk having more indices than tensors at a boundary . specifically , a bulk tensor has components @xmath33}_{ulsdr}$ ] , with one _ physical _ index @xmath36 and four _ bond _ indices @xmath41 . the physical index @xmath36 labels the basis of the vector space @xmath5 for site @xmath24 and therefore takes @xmath6 different values , whereas each bond index connects the tensor with a tensor in a nearest neighbor site and ranges from @xmath42 to @xmath43 , where @xmath43 is the so - called _ bond dimension _ of the peps . correspondingly , a bulk peps tensor is represented by a circle with five legs , see fig . [ fig : bosonpeps](b ) . as in the rest of the paper , here we follow the prescription that the indices of a tensor are drawn in anti - clockwise order . notice also that the open indices of the peps in fig . [ fig : bosonpeps](a ) reach the exterior of the tensor network in exactly the same ( anti - clockwise ) order that they appear in @xmath44 . these notational and diagrammatical details are somewhat superfluous in the bosonic case ( since one can change the order of indices in a tensor by simply permuting its components ) but will become important in the extension to fermions . a peps on a @xmath45 lattice , where @xmath46 , contains @xmath47 bulk tensors , each depending on @xmath48 complex coefficients . therefore the peps is characterized by @xmath49 parameters . if @xmath43 has a fixed value independent of @xmath2 , then the peps is indeed an efficient encoding of some states @xmath16 , and it can be used e.g. as a variational ansatz for the ground state of @xmath13 . however , for the peps to be a useful ansatz , we also need to provide an efficient strategy to optimize its tensors and manipulate them in order to extract physically relevant information , as discussed next . two major tasks to be accomplished with a peps algorithm@xcite are : ( i ) optimization of its @xmath49 parameters so as to obtain a good approximation to e.g. the ground state of the local hamiltonian @xmath13 ; ( ii ) given a peps for a state @xmath12 , computation of expectation values @xmath14 of local observables @xmath15 , as given by @xmath50 these two tasks happen to be closely related . they involve taking the trace over all the bond indices of a composite tensor network , an operation referred to as _ contracting _ the tensor network . written in terms of tensors @xmath25 and @xmath51 . ( b ) the same scalar product , but written in terms of a peps together with its hermitian conjugate . ( c)-(d ) using the jump move of fig . [ fig : jump ] , this tensor network can be modified so that each tensor @xmath33}$ ] is drawn next to its hermitian conjugate @xmath33}$ ] . ( e ) tensor network @xmath52 in terms of reduced tensors @xmath53}$ ] . ( f ) @xmath53}$ ] defined in terms of @xmath33}$ ] and @xmath33\dagger}$ ] , eq . .,width=302 ] an emblematic example of tensor network contraction required in a peps algorithm concerns the computation of the scalar product @xmath54 . we need to introduce a few more definitions and notation . let @xmath55 denote the hermitian conjugate of a tensor @xmath56 , obtained by reversing the order of the indices of @xmath56 and taking the complex conjugate of each of its coefficients ( and diagrammatically represented as the mirror image of @xmath56 ) . for instance , tensor @xmath51 in fig . [ fig : bosonpeps](d ) , corresponding to @xmath57 , @xmath58 has coefficients @xmath59 fig . [ fig : bosonpeps](c ) represents the hermitian conjugate @xmath33\dagger}$ ] of a bulk tensor @xmath33}$ ] , which has coefficients @xmath60\dagger})_{rdslu } = { a^{[\vec{r}]}_{ulsdr}}^*.\ ] ] we build the hermitian conjugate of the peps for @xmath25 as the set of tensors @xmath31\dagger}\}$ ] connected according to the mirror image of the network of connections in the peps for @xmath25 , see fig . [ fig : bosonpeps](d ) . for each site @xmath22 , let us define the reduced tensor @xmath53}$ ] in terms of tensor @xmath33}$ ] and @xmath33\dagger}$ ] by tracing over the physical index @xmath36 . for instance , in the bulk , the reduced tensor @xmath53}$ ] has components @xmath61}_{u \bar{u } l \bar{l } \bar{d } d \bar{r } r } \equiv \sum_{s=1}^d a^{[\vec{r}]}_{ulsdr } ( a^{[\vec{r}]\dagger})_{\bar{r } \bar{d } s \bar{l } \bar{u } } , \label{eq : a}\ ] ] see fig . [ fig : bosonscalar](f ) . then the scalar product @xmath54 results from contracting a tensor network @xmath52 that consists of all reduced tensors @xmath53}$ ] connected according to the links in @xmath1 , see fig . [ fig : bosonscalar](a)-(e ) , where the ( trivial ) jump move of fig . [ fig : jump ] is used . contracting the tensor network @xmath52 to obtain the scalar product @xmath54 comes with a cost that grows exponentially in the linear size @xmath62 of lattice @xmath1 , and therefore can not be accomplished efficiently . a key ingredient of peps algorithms is precisely a strategy to _ efficiently _ but approximately contract the tensor network @xmath52 , thus producing an approximation to @xmath54 . this can be done in several ways , depending on the size and topology of lattice @xmath1 . in a finite lattice with open boundary conditions , one can use matrix product state ( mps ) techniques . @xcite in the case of a torus , coarse - graining techniques known as _ tensor entanglement renormalization group _ ( terg ) @xcite can be used . finally , in an infinite lattice , both infinite mps@xcite and _ corner transfer matrix _ ( ctm)@xcite techniques have been employed . in order to optimize a peps so as to approximate the ground state of @xmath13 , as well as to evaluate the expectation value of local observables , it is useful to contract certain class of tensor networks called _ environments_. the environment @xmath63}$ ] of a site @xmath22 is the tensor network obtained from @xmath52 by removing tensor @xmath53}$ ] , and can be used to compute the expectation value of a local observable acting on that site . , made of reduced tensors @xmath53}$ ] , corresponding to the scalar product @xmath54 in an @xmath45 lattice @xmath1 with open boundary conditions . ( b ) environment @xmath64}$ ] for two contiguous sites @xmath65 , obtained from @xmath52 by removing tensors @xmath66}$ ] and @xmath67}$ ] . ( c ) approximate environment @xmath68}$ ] consisting of six tensors @xmath69 . ( d ) tensor @xmath70 of coefficients for a two - site local operator @xmath15 , eq . . ( e ) tensor network made of the approximate environment @xmath68}$ ] and tensors @xmath71}$ ] , @xmath71\dagger}$ ] , @xmath72}$ ] , @xmath72\dagger}$ ] and @xmath70 . its contraction produces an approximation to @xmath73.,width=321 ] similarly , a two - site environment @xmath64}$ ] is the tensor network obtained by removing tensors @xmath66}$ ] and @xmath67}$ ] from @xmath52 , and can be used e.g. to compute the expectation value of a two - site observable @xmath74 acting on sites @xmath75 . figure [ fig : bosonenv](b ) shows the environment @xmath64}$ ] corresponding to two nearest neighbor sites @xmath76 . again , the exact contraction of the environment can not be performed efficiently , but efficient schemes , analogous to those employed to contract @xmath52 , can be used in order to approximately contract @xmath77}$ ] . the whole environment is in this way approximated by a smaller tensor network @xmath68}$ ] made of 6 tensors @xmath78 , see fig . [ fig : bosonenv](c ) . these tensors can then be connected to tensors @xmath71}$ ] , @xmath71\dagger}$ ] , @xmath72}$ ] , @xmath72\dagger}$ ] and @xmath70 to produce an approximation to @xmath73 , see fig . [ fig : bosonenv](e ) . in a system with a hamiltonian @xmath13 made of two - site interactions @xmath79}$ ] between nearest neighbors , the approximate environment @xmath68}$ ] is particularly useful . on the one hand , it is employed in the optimization of tensors @xmath71}$ ] and @xmath72}$ ] after a gate has been applied on the two sites , as part of simulating an _ imaginary - time evolution _ according to @xmath13 , @xmath80 which offers one way of obtaining a peps approximation to the ground state @xmath81 of @xmath13 . on the other hand , @xmath82}$ ] can also be used to compute the expectation value of the energy on that link , @xmath83 } { \mbox{$| \psi \rangle$}}$ ] , as part of an algorithm to optimize the peps by minimizing its energy , @xmath84}\ } } \frac { { \mbox{$\langle \psi_{\{a^{[\vec{r}]}\ } } |$}}~ \hat{h}~ { \mbox{$| \psi_{\{a^{[\vec{r}]}\ } } \rangle$ } } } { { \mbox{$\langle \psi_{\{a^{[\vec{r}]}\ } } | \psi_{\{a^{[\vec{r}]}\ } } \rangle$ } } } \label{eq : mine}\ ] ] which is another way of obtaining a peps approximation to the ground state @xmath81 of @xmath13 . we refer to refs . for more details . other peps algorithms@xcite bypass the computation of environments . this concludes our short review of peps algorithms for bosonic 2d lattice models . let us now consider a fermionic lattice system . for the sake of simplicity , we first assume that each site @xmath20 is described by a complex vector space @xmath5 of dimension @xmath19 that is associated to a fermionic annihilation operator @xmath85 , with anticommutation relations @xmath86 [ we will shortly extend the discussion to sites with vector space of finite dimension @xmath87 , see also appendix [ app : genoperators ] ] . a basis of the vector space @xmath88 of the lattice system is given by @xmath89 recall that fermionic operators can be expressed in terms of pauli matrices @xmath90 by means of a jordan - wigner transformation , @xmath91 the fermionic lattice system @xmath1 is further characterized by a local fermionic hamiltonian @xmath13 . this is a hamiltonian that , when expressed in terms of the fermionic operators @xmath92 , decomposes as the sum of terms involving only a small number of sites . as in eq . , let @xmath16 be a pure state of lattice @xmath1 , @xmath93 here we will assume that @xmath12 is somehow related to the fermionic hamiltonian @xmath13 , for instance it is its ground state . once more , we would like to find a variational ansatz depending on @xmath47 parameters to efficiently encode the tensor @xmath25 containing the @xmath26 coefficients @xmath17 of a pure state @xmath12 . one possibility would be to use a peps exactly as in the bosonic case . however , this might not be a good idea . remember that the label @xmath94 provides an order to the set of sites of @xmath1 , whose position in the lattice is given by @xmath95 . two nearest neighbor sites @xmath96 and @xmath97 on the square lattice might correspond to values @xmath98 and @xmath99 that are far apart . then , when expressed in terms of pauli matrices , the local fermionic hamiltonian @xmath13 will no longer look local . for instance , a nearest neighbor hopping term @xmath100 develops a string of @xmath101 s . this might be harmful in two ways . on the one hand , the presence of strings of @xmath101 s would require important modifications in the algorithms to approximate the ground state of @xmath13 with a peps , either through imaginary - time evolution or energy minimization . on the other hand , it is unclear that the peps itself , which was originally designed as an ansatz for ground states of local bosonic hamiltonians , will be as good an ansatz also for ground states of fermionic hamiltonians , given that the latter are non - local when expressed in bosonic variables . below we will explain how to modify the peps so that it is suitable to study fermionic systems ( see also refs . ) . before , however , we introduce the notation necessary to deal with local vector spaces @xmath5 of dimension @xmath87 . fermionic systems are governed by hamiltonians that preserve the _ parity of the fermionic particle number _ , to which we refer simply as ` _ _ parity _ _ ' . that is , fermions can only be created or annihilated in pairs , and parity is a constant of motion . as a result , we can assume that the pure state @xmath102 of lattice @xmath1 has a well - defined parity , and observables @xmath15 and reduced density matrices are block diagonal in parity . let us consider again the vector space @xmath5 of a single site , now with finite dimension @xmath87 . it is natural to decompose @xmath5 as the direct sum of an even parity subspace @xmath103 and an odd parity subspace @xmath104 , @xmath105 and to choose a basis of vectors with well - defined parity . accordingly , the physical index @xmath36 describing one such basis is decomposed as @xmath106 , where @xmath107 is the parity and @xmath108 ( denoted @xmath109 and @xmath110 ) enumerates the different basis states with parity @xmath111 . the parity operator @xmath112 then acts on this basis as @xmath113 in the case of spinless fermions , with a local dimension @xmath19 , the two possible states of a site are the local vacuum @xmath114 , signaling the absence of a fermion , and the state @xmath115 , corresponding to one fermion . in our notation these states read @xmath116 in the case of the @xmath0 model there are three possible states per site , @xmath117 , where @xmath118 and @xmath119 denote an electron with spin up and spin down , respectively . in our notation these states read @xmath120 finally , in the case of the hubbard model with local dimension @xmath121 there is an additional state , @xmath122 , corresponding to a doubly occupied site , which in our notation reads @xmath123 in analogy with the physical index @xmath106 , we introduce a parity operator also on the bond indices of a peps . accordingly , the tensor @xmath33}_{ulsdr}$ ] has bond indices @xmath124 , etc . this means that e.g. bond index @xmath34 can take values @xmath125 where the bond dimension @xmath43 is given by @xmath126 . the actual values of @xmath127 and @xmath128 can be chosen at convenience . ) ; ( b ) line crossings are replaced with a special gate , a fermionic swap gate @xmath129.,width=264 ] given a peps for bosonic systems , cf . [ fig : bosonpeps](a ) , in this work we obtain a peps for fermionic systems by applying the two rules used in ref . to fermionize the mera . these rules are applied both to the peps and to all related tensor networks that are involved e.g. in optimizing the ansatz or computing expectation values of local observables . * rule 1 : * each tensor @xmath56 in a tensor network is chosen to be _ parity preserving _ , i.e. @xmath130 where @xmath131 denotes the _ parity _ of the basis state labelled by @xmath132 , see fig . [ fig : rules](a ) . * rule 2 : * each crossing of lines in the tensor network is replaced with a _ fermionic swap gate _ @xmath133 , see fig . [ fig : rules](b ) . this gate implements a fermionic exchange and has the form @xmath134with @xmath135 given by @xmath136 of a @xmath30 lattice @xmath1 , obtained from the bosonic peps in fig . [ fig : bosonpeps](a ) by replacing line crossings with fermionic swap gates . the coefficients @xmath140 are obtained by contracting this fermionic peps using standard tensor multiplication techniques ( see text ) . the presence of fermionic swap gates introduces a complex structure of minus signs in @xmath140 . ( b ) bulk tensor @xmath33}$ ] , which is chosen to be parity preserving , eq . . ( c ) hermitian conjugate of @xmath33}$ ] . ( d ) scalar product @xmath54 written as a tensor network of reduced tensors @xmath53}$ ] defined in ( e ) . notice that , in a @xmath45 lattice , @xmath141 fermionic swap gates pervade the peps and yet , thanks to the jump move , fig . [ fig : jumpmove ] , it is possible to write the scalar product @xmath54 in terms of only @xmath142 fermionic swap gates and in such a way that all of them are near a pair @xmath38},a^{[\vec{r}]\dagger})$ ] and can therefore be absorbed into the definition of the reduced tensor @xmath53}$ ] . [ for a detailed derivation , replace line crossings with fermionic swap gates in fig . [ fig : bosonscalar ] ] . as a result , there are no fermionic swap gates left in @xmath52 , and therefore this tensor network can be contracted using the techniques employed for bosonic peps.,width=264 ] rule 1 is very convenient from a computational perspective : it ensures that the parity of the wave function is exactly preserved during ( otherwise approximate ) calculations , while the block diagonal structure of tensors can be exploited to reduce computational costs . in addition , rule 1 is important in order to account for the antisymmetric character of fermionic wavefunctions ( rule 2 ) in a simple way . however , we emphasize that tensor networks made of parity preserving tensors are also useful to describe bosonic systems ( e.g. , a @xmath143 invariant spin system , such as the quantum ising model ) and therefore rule 1 is not what turns a bosonic tensor network ansatz into a fermionic one . rule 2 accounts for the fermionic character of the tensor network , in the sense that it is employed to mimic the effect of anticommutators in a network of fermionic operators , as justified in ref . in the context of the mera ( see also the note on terminology in the introduction of this paper ) . several additional remarks concerning fermionic tensor networks and their manipulations are in order . we start with a number of comments on fermionic tensor network representations that are relevant to the present formulation of the fermionic peps ansatz . \(i ) _ fermionic order_. the label @xmath144 for the sites of @xmath1 establishes an order on these sites . this order has been used to define a local basis of the hilbert space in eq . , and can also be used to translate the local fermionic lattice model into a non - local bosonic one through the jordan - wigner transformation of eq . ( although this is not the strategy that we follow here ) . according to our prescription to graphically represent tensors and tensor networks , this order is also the order in which the open indices are drawn in fig . [ fig : fermipeps](a ) . notice that the structure of line crossings in the peps depends on the fermionic order and therefore the number and location of fermionic swap gates , see fig . [ fig : orders ] . notice also that , in contrast with refs . , we do not explicitly introduce fermionic operators ( and corresponding order ) on the bond indices of the peps , but use instead a simple graphical notation and two rules that already account for the complex pattern of fermionic - exchange minus signs in tensor @xmath145 . corresponds to a specific choice of fermionic order , which appears in the definition of the local basis of eq . and jordan - wigner transformation of eq . . ( b ) another possible fermionic peps , associated to another fermionic order . the two tensor networks are not equivalent in that they can not be mapped into each other by jump moves alone ( where one is not allowed to drag a line over the end of another open line ) , but the mapping is possible if we allow some additional fermionic exchanges to modify the order of the open lines . importantly , it can be seen that the tensors @xmath31}\}$ ] in both peps appear in the same way in any expectation value . in particular , they would be optimized using exactly the same figure of merits . this shows that the tensors @xmath31}\}$ ] are independent of the choice of fermionic order.,width=264 ] \(ii ) _ local operators_. a local fermionic operator @xmath15 is characterized by a tensor of components that describes the action of @xmath15 on a given basis of states . for instance , in the simple case where each site has a vector space @xmath5 of dimension @xmath19 , we expand a two - site operator @xmath15 as @xmath146 where @xmath147 and @xmath148 and @xmath149 are the annihilation operators acting on the two sites ( see appendix [ app : genoperators ] for the @xmath150 case ) . the coefficients @xmath151 are given by @xmath152 for instance , a hopping term @xmath153 reads @xmath154 since the only non - vanishing coefficient is @xmath155 which is obtained using the anticommutation relations . when the two - site operator @xmath15 acts on two sites @xmath156 , the state @xmath12 of the system is modified into some other state @xmath157 . this is implemented simply by connecting the tensor ( or tensor network ) that represents @xmath158 with the four - index tensor @xmath159 as indicated in fig . [ fig : operator](a ) . in particular , when the two sites are not contiguous in the fermionic order , @xmath160 , a number of line crossings appear . this is reflected in the computation of the expectation value @xmath14 , see fig . [ fig : operator](b)-(d ) . is described by a tensor @xmath161 of coefficients obtained by connecting tensors @xmath25 and @xmath70 . the example corresponds to a @xmath162 lattice and a choice of fermionic order such that the two sites on which @xmath15 acts , which are nearest neighbors in @xmath1 , are not contiguous in the fermionic order ( they occupy positions @xmath163 and @xmath164 ) . the line crossings ( fermionic swap gates ) that appear in connecting tensors @xmath25 and @xmath70 can be interpreted as changes in the fermionic order needed in order to bring the two sites on which @xmath15 acts together , then bringing them back to their original position . ( b ) expectation value @xmath73 . ( c ) the same expectation value in terms of a fermionic peps . ( d ) after some manipulation , an approximation of the expectation value is expressed in terms of an approximate two - site environment @xmath82}$ ] and tensors @xmath71}$ ] , @xmath71\dagger}$ ] , @xmath72}$ ] , @xmath72\dagger}$ ] and @xmath70 , as it was done for a bosonic system in fig . [ fig : bosonenv ] . the only difference here is the presence of a few fermionic swap gates.,width=302 ] \(iii ) _ parity changing tensors_. parity preserving tensors allow us to represent both states @xmath12 with even fermionic particle number ( i.e. parity @xmath165 ) and local operators @xmath15 that are parity preserving . but they also allow us to represent states with an odd fermionic particle number ( parity @xmath166 ) and parity changing operators . this is so because a parity changing tensor @xmath167 , @xmath168 can be represented as a parity preserving tensor @xmath56 , @xmath169 where the additional index @xmath170 only takes one value , @xmath171 . for instance , in order to represent a state @xmath102 with an odd number of particles by means of a peps , an additional index @xmath172 is attached to one of the peps tensors , see fig . [ fig : pchanging](a ) . similarly , a fermionic annihilation operator @xmath173 , which changes parity , can be represented as a tensor with three indices , one of which is fixed to @xmath171 , see fig . [ fig : pchanging](b ) . it is sometimes computationally convenient ( e.g. in the computation of two - point correlators , fig . [ fig : correlator ] ) to represent a parity preserving operator that is the product of two parity changing operators , such as the hopping operator @xmath174 in eq . , by two parity preserving tensors connected by a fixed index @xmath172 , see fig . [ fig : pchanging](b ) . , of a @xmath175 lattice , with odd parity @xmath166 . all peps tensors are parity preserving . one of the tensors has an additional index @xmath176 with fixed parity @xmath166 . ( b ) the tensor @xmath70 of coefficients @xmath177 for a two - site operator such as a hopping term @xmath178 in eq . , that is the product of two parity changing operators , @xmath179 and @xmath149 , can be decomposed as the product of two parity - preserving tensors @xmath180 and @xmath181 connected by an index @xmath176 with fixed parity @xmath166 . this decomposition is useful e.g. in the computation of a correlator involving two distant sites , see fig . [ fig : correlator ] . notice that parity preserving tensor @xmath180 can also be used independently of @xmath181 in order to represent the parity changing operator @xmath148 . , width=302 ] \(iv ) _ simplification_. in the particular case of a pair of crossing lines @xmath137 and @xmath170 where index @xmath170 only takes one possible value @xmath111 of the parity , the fermionic swap gate @xmath133 reduces to a product of two gates . namely , to a product of two identity operators @xmath182 if @xmath183 , and to a product of the parity @xmath112 on line @xmath137 and identity @xmath184 on line @xmath170 if @xmath166 . it follows , for instance , that the jump move applied to a line @xmath137 and a parity preserving tensor @xmath56 with an index @xmath185 ( used e.g. to represent a parity changing tensor @xmath167 in eqs . -[eq : tt ] ) allows us to ignore index @xmath170 in @xmath56 at the price of applying the parity operator @xmath112 to line @xmath137 , see fig . [ fig : jumpmove](b ) . this simplification appears to be useful e.g. in the calculation of the expectation value @xmath186 , see fig . [ fig : correlator ] . a fermionic tensor network can be manipulated simply by performing a sequence of tensor multiplications . it turns out , however , that by considering a special property ( jump move ) of fermionic tensor networks obeying rules 1 and 2 , the fermionic swap gates can be treated in a very special and advantageous way : they can be ignored until they correspond to a crossing of two indices connected to the same tensor , in which case they can be absorbed into the tensor using a low cost operation . as a result , in a fermionic tensor network algorithm we can follow the same sequence of tensor multiplications that we would have employed to manipulate its bosonic counterpart . in particular , we can use the same optimal sequence of contractions , with the same computational cost ( to leading order ) . let us discuss all these aspects in more detail . is represented with a parity preserving tensor @xmath56 with an extra index @xmath172 that remains open . in this case the jump rule introduces an additional fermionic swap gate @xmath129 , involving @xmath170 and the jumping line , that reduces to a parity operator @xmath187 on the jumping line ( see text ) . ( c ) a line can also be dragged over a set of open lines provided that the latter have even overall parity ( in other words , provided they could be connected to a parity preserving tensor),width=302 ] \(i ) _ jump move_. it follows from rules 1 and 2 that in a fermionic tensor network , as in the bosonic case , lines can be dragged over tensors , see fig . [ fig : jumpmove](a ) . this invariance under ` jump ' moves allows us to modify fermionic tensor networks in such a way that the fermionic swap gates do not increase the leading cost of manipulations . for a proof of this property , which exploits the fact that tensors are parity - preserving , we refer again to ref . . notice that the jump move does not include dragging a line over the open end of another line ( this would amount to a change in the local fermionic order ) . the latter transformation is only allowed when it involves a set of open lines with even overall parity , as explained next . \(ii ) _ open lines_. by construction , line ends only appear in the diagrams grouped in such a way that they could be connected to a parity preserving tensor ( that is , their combined parity is always even ) without introducing additional crossings to the network . for instance , the set of all open legs of the peps in figs . [ fig : fermipeps](a ) or [ fig : pchanging](a ) have even overall parity , since the peps as a whole has parity @xmath183 . another example is given by the open legs of an environment , fig . [ fig : corner](c ) . since such groups of open legs with even overall parity could be connected to a parity preserving tensor , the jump move extends naturally to them , as illustrated in fig . [ fig : jumpmove](c ) . \(iii ) _ absorption of a fermionic swap gate_. given a tensor @xmath56 , with coefficients @xmath188 , such that the two contiguous indices @xmath189 are connected to a fermionic swap gate , it is possible to absorb the latter into the former and produce a new tensor @xmath167 with coefficients @xmath190 the cost of the absorption is thus just proportional to the number of coefficients in @xmath56 and comparable to permuting two adjacent indices in a bosonic tensor network . this cost is , in particular , smaller than the tensor multiplication that may have produced @xmath56 or a subsequent tensor multiplication involving @xmath167 . therefore the absorption of fermionic swap gates only makes a sub - leading contribution to the total cost of contracting a fermionic tensor network ( in the same way that the permutation of indices only makes a sub - leading contribution to the total cost of contracting a bosonic tensor network ) . and @xmath191 connected by two lines . ( b ) by means of a jump move , all lines standing between @xmath192 and @xmath191 can be dragged away . ( c ) any fermionic swap gate involving the lines that connect @xmath192 and @xmath191 can be absorbed into e.g. tensor @xmath192 , which is mapped into some @xmath193 . ( d ) finally tensors @xmath194 and @xmath191 are multiplied together , producing tensor @xmath56 . the cost of all these manipulations is dominated by the last step.,width=302 ] \(iv ) _ same leading cost_. in a bosonic system , the contraction of a tensor network is implemented by a sequence of multiplications of tensors , where each multiplication involves two tensors and the sequence is chosen carefully to minimize computational costs . in the fermionic case , it is natural to ask whether the presence of fermionic swap gates will modify the optimal sequence of tensor multiplications and increase the computational cost of the contraction . it turns out that one can always follow the same sequence of multiplications as in the bosonic case . indeed , as illustrated in the example of fig . [ fig : contractjump ] , in order to multiply two tensors , one can use jump moves and fermionic swap gate absorptions to eliminate any fermionic swap gate between the two tensors , which can then be multiplied together exactly as in the bosonic case . the leading cost of these manipulations is given by tensor multiplications , and it is therefore the same for bosonic and fermionic systems . we have explained how to build a fermionic peps for the state @xmath16 of a fermionic lattice system . the next step is to consider algorithms to compute the expectation value @xmath195 of a local operator @xmath15 from a fermionic peps , as well as to optimize the coefficients in its tensors @xmath31}\}$ ] . it turns out that these algorithms can be obtained by just introducing simple modifications to existing bosonic peps algorithms . here we discuss these modifications broadly . let us consider first the computation of the scalar product @xmath54 . as we did in the bosonic case , see fig . [ fig : bosonscalar ] , we build this scalar product by connecting the open indices of a peps for @xmath12 with those for a peps for @xmath57 obtained by hermitian conjugation . the resulting tensor network may contain , in the fermionic case , a large number of fermionic swap gates . however , an important point is that , by using the jump move of fig . [ fig : jumpmove](a ) , the scalar product @xmath54 can again be re - expressed in terms of a tensor network @xmath52 made of reduced tensors @xmath196}\}$ ] , see fig . [ fig : fermipeps](d ) . as shown in fig . [ fig : fermipeps](e ) , the fermionic reduced tensor @xmath53}$ ] differs from its bosonic counterpart by the presence of four fermionic swap gates . importantly , in a finite or infinite lattice with open boundary conditions , _ all _ fermionic swap gates present in the scalar @xmath54 are absorbed into the reduced tensor @xmath53}$ ] . thus , the tensor network @xmath52 contains no fermionic swap gates . as a result , an approximation to the scalar product @xmath54 can be obtained by contracting @xmath52 with _ exactly _ the same approximate contraction techniques ( namely mps techniques for finite peps@xcite and either infinite mps or ctm techniques for infinite peps@xcite ) as in the bosonic case . [ in some implementations , it is possible to lower computational costs by considering the components @xmath33}$ ] and @xmath33\dagger}$ ] of @xmath53}$ ] separately . in this case the approximate contraction techniques have to be slightly modified so as to account for the four fermionic swap gates involved in the definition of @xmath53}$ ] ] . figure [ fig : cylinder ] considers a peps for a fermionic lattice system on a cylinder . its scalar product can again be expressed in terms of a tensor network @xmath52 , made of reduced tensors @xmath196}\}$ ] , that contains no line crossings . therefore its contraction can also be performed exactly as in the bosonic case . instead , in a fermionic lattice system on a torus , the analogous tensor network @xmath52 contains a number of fermionic swap gates , see fig . [ fig : torus ] . however , it can be seen that even in this case @xmath52 can still be coarse - grained using the same terg techniques of ref . for the bosonic case , by properly coarse - graining the swap fermionic gates at each step of the coarse - graining . of fermionic @xmath30 lattice with cylindrical boundary conditions . ( b ) tensor network for the scalar product @xmath54 . ( c ) tensor network @xmath52 of reduced tensors @xmath196}\}$ ] . as in a system with open boundary conditions , @xmath52 has no line crossings and can be contracted as in the bosonic case . , width=321 ] of fermionic @xmath30 lattice with toroidal boundary conditions . ( b ) tensor network for the scalar product @xmath54 . ( c ) tensor network @xmath52 of reduced tensors @xmath196}\}$ ] . on the torus , it is not possible to eliminate all line crossings in @xmath52 , which must contain fermionic swap gates and therefore differs from the bosonic case . however , terg techniques@xcite can still be applied by noting that the coarse - graining of the reduced tensors @xmath196}\}$ ] can take place independently of the fermionic swap gates , which are also coarse - grained in an obvious way . the ( coarse - grained ) fermionic swap gates only need to be absorbed into the rest of the tensor network at the latest coarse - graining step , when the torus has been reduced to only a few lattice sites . , width=321 ] as discussed in sec . [ sec : peps : optim ] , given an environment @xmath197}$ ] , an approximation @xmath198}$ ] can be used both for the computation of the expectation value @xmath199 of a local operator @xmath15 and for the optimization of the tensors @xmath200}\}$ ] defining the peps . the approximate contraction of @xmath197}$ ] leading to @xmath198}$ ] is very similar to the contraction of @xmath52 for the scalar product @xmath54 , and can again be accomplished in the fermionic case using the same techniques than in the bosonic case . from @xmath198}$ ] , obtaining an approximation to the expectation value @xmath73 involves contracting the tensor network of fig . [ fig : operator](d ) , which differs from its bosonic counterpart in the presence of 12 fermionic swap gates . an analogous tensor network is central to the update of the peps tensors @xmath31}\}$ ] during an imaginary - time evolution towards the ground state @xmath81 of a nearest neighbor fermionic hamiltonian @xmath13 , cf . more details of this update will be provided in sec . [ sec : ctm ] in the context of an infinite lattice system , where also the computation of two - point correlators between distant sites will be addressed . to summarize , the differences between bosonic and fermionic peps algorithms are in practice reduced to : ( i ) use of parity preserving tensors @xmath31}\}$ ] to efficiently encode the state @xmath9 ; and ( ii ) presence of fermionic swap gates in some of the tensor networks that need to be contracted , e.g. in order to compute the expected value of a local observable or optimize the tensors @xmath31}\}$ ] of the variational ansatz . however , thanks to the jump move , the optimal sequence of tensor multiplications involved in the contraction of a given fermionic tensor network is the same than in the bosonic case , as is the incurred computational cost . all in all , we see that fermionic peps algorithms can be obtained by introducing a few modifications to bosonic peps algorithms . this is further illustrated in the next section , where we provide more details on the specific peps algorithm used in sec . [ sec : results ] , namely the ipeps algorithm.@xcite in sec . [ sec : results ] the formulation of fermionic peps presented in this paper is tested by computing the ground state of a number of models on an infinite lattice . in this section we provide additional details on the fermionic ipeps ansatz and algorithm used for those computations . we start by reviewing the bosonic ipeps algorithm.@xcite then we describe the modifications required in order to address a fermionic system . } $ ] for two contiguous sites @xmath65 , in terms of the reduced tensors @xmath205 and @xmath206 corresponding to the peps tensors @xmath202 and @xmath203 . ( b ) approximate environment @xmath82}$ ] expressed in terms of 10 tensors , corresponding to the 10 shaded regions in ( a ) , as used in the ctm approach@xcite . some of the tensor are used in fig . [ fig : correlator](b ) . ( c ) approximate environment @xmath82}$ ] written in terms of 6 tensors @xmath69 , as used in the computation of the expectation value of a two - site observable @xmath15 ( fig . [ fig : correlator](a ) ) and the update of tensors @xmath202 and @xmath203 during an imaginary - time evolution ( fig . [ fig : revupdate3]).,width=340 ] let us first consider how to compute the expectation value of a two - site operator @xmath15 , see eq . . this requires computing both @xmath54 and @xmath73 . the scalar product @xmath54 is expressed in terms of an infinite 2d tensor network @xmath52 of reduced tensors @xmath205 and @xmath206 distributed according to a checkerboard pattern . as in a finite system , an environment @xmath207}$ ] for two nearest sites @xmath76 is then built from @xmath52 by removing two reduced tensors , see fig . [ fig : corner](a)-(b ) . contracting @xmath207}$ ] requires an approximation scheme , which produces an approximate environment @xmath82}$ ] consisting of six tensors @xmath69 connected through bond indices that take @xmath208 different values , see fig . [ fig : corner](d ) . here @xmath208 quantifies the degree of approximation in @xmath82}$ ] . one possible approximation scheme consists in using infinite mps techniques , as was discussed in the original ipeps algorithm in ref . . another possible approximation scheme consists in using ctm techniques,@xcite as discussed in the context of the ipeps algorithm in ref . . an approximation @xmath209 to the expectation value @xmath195 of a two - site observable @xmath15 is then obtained by computing @xmath210 where @xmath184 denotes the identity operator on the space @xmath211 of two sites of @xmath1 and we use the same approximate environment @xmath82}$ ] to compute @xmath212 and @xmath213 , where the latter is computed as an expectation value for @xmath214 . the approximate value @xmath215 will in general differ from the exact value @xmath195 . one expects , however , that in the limit of a large @xmath208 one recovers the exact value , @xmath216 in practice , we compute @xmath215 for increasingly large values of @xmath208 , e.g. @xmath217 , until the expectation value @xmath215 does no longer depend substantially in @xmath208 , and assume that this corresponds to @xmath195 . the cost of computing @xmath215 using ctm techniques scales with the peps bond dimension @xmath43 and the environment bond dimension @xmath208 as @xmath218 . notice that for any value of the bond dimension @xmath43 , a peps produces a variational energy @xmath219 , @xmath220 that is @xmath221 , where @xmath222 is the exact ground state energy . therefore , if we could compute @xmath219 exactly , we would obtain an upper bound to @xmath222 , which could e.g. be compared with another upper bound corresponding to another variational ansatz . however , the approximate value @xmath223 is not guaranteed to be an upper bound to @xmath222 . in this paper we assume that @xmath223 is an upper bound to @xmath222 once it has converged for large values of @xmath208 . this assumption is seen to be correct in the case of free fermions in sec . [ sec : results ] . in the ipeps algorithm of refs . an approximation to the ground state is obtained by simulating an evolution in imaginary time , eq . . the optimization of tensors @xmath202 and @xmath203 during the imaginary - time evolution is achieved by breaking the evolution into steps ( suzuki - trotter decomposition ) and then considering updates that involve a single link . there are four types of updates , corresponding to the four inequivalent links given by the bond indices @xmath41 of tensor @xmath202 . for each link , new tensors @xmath224 and @xmath225 are produced as a result of an optimization that aims to account for the action of a two - site gate @xmath226 } \delta t)$ ] acting on that link . we refer to refs . for a detailed explanation on how to use the approximate environment @xmath82}$ ] to obtain updated tensors @xmath224 and @xmath225 . the algorithm proceeds by iteratively updating tensors @xmath202 and @xmath203 until they converge to some pair of tensors that depend on @xmath208 and the time step @xmath227 . a finite time step @xmath227 introduces errors in the imaginary - time evolution . a better approximation to the ground state is obtained by gradually reducing @xmath227 until its value does no longer affect significantly quantities of interest ( e.g. the energy ) . for instance , this occurs for @xmath228 in the simulations of sec . [ sec : results ] . the cost of simulating an imaginary - time evolution is proportional to the number of time steps that are being simulated and scales as @xmath229 , since each update requires reconverging an approximate two - site environment @xmath68}$ ] . a simplified way of simulating an evolution in imaginary time was proposed in ref . . this simplified update does not involve the environment @xmath68}$ ] and has a much lower cost per iteration , namely @xmath230 when applied in a straightforward way and reducible to @xmath231 after more careful considerations . on the one hand , ignoring the environment ( that is , the rest of the wavefunction ) implies that the update may not be optimal , and indeed there are cases , such as the 2d quantum ising model near its critical point , where it produces a less accurate approximation to the ground state,@xcite although we also found many cases where it only performs marginally worse than when using the environment . on the other hand , its much lower cost accelerates the simulations considerably . notice , however , that the computation of expectation values with the resulting peps still requires computing @xmath68}$ ] , which has cost @xmath229 . defined in terms of peps tensors @xmath202 and @xmath232 and four fermionic swap gates . ( b ) reduced tensor @xmath206 defined in terms of peps tensors @xmath203 and @xmath233 and four fermionic swap gates.,width=340 ] to the expectation value @xmath14 of a local observable @xmath15 acting on two contiguous sites . this tensor network only differs from the one in fig . [ fig : bosonenv](e ) for a bosonic system in the presence of 12 fermionic swap gates . ( b ) computation of two - point correlators . for concreteness , we consider the expected value @xmath234 discussed in eq . . the figure shows the tensor network representing @xmath235 , where sites @xmath189 are in the same row of @xmath1 but separated by 4 sites . notice the line connecting the two tensors corresponding to operators @xmath236 and @xmath237 , which crosses a number of other lines introducing a number of fermionic swap gates . since this line corresponds to an index @xmath172 with well - defined parity @xmath166 , the fermionic swap gates simplify into @xmath238.,width=340 ] as in the bosonic case , the fermionic ipeps ansatz exploits translation invariance of a system on an infinite lattice @xmath1 to store the state @xmath12 using only a small number of peps tensors , which are repeated throughout the lattice . whether we are interested in approximating a ground state invariant under translations by one site or with checkerboard order , we consider again just two ( parity preserving ) peps tensors @xmath202 and @xmath203 as in eq . . it may not seem obvious that a fermionic ipeps , characterized by tensors @xmath202 and @xmath203 , represents a state @xmath12 of the infinite lattice @xmath1 that is invariant under diagonal shifts ( checkerboard order ) . indeed , the presence of the ubiquitous fermionic swap gates , which are not homogeneously distributed across the tensor network corresponding to the ipeps ( see fig . [ fig : fermipeps ] for an illustration in the case of a finite peps ) may seem incompatible with the checkerboard order . however , the checkerboard order becomes manifest during the computation of expectation values for local observables . given a fermionic ipeps , the computation of the expectation value of e.g. a two - site observable @xmath15 is very similar to the bosonic case . since there are no fermionic swap gates in the tensor network @xmath52 for the scalar product @xmath54 or the two site environment @xmath64}$ ] , these tensor networks can be contracted exactly in the same way as in a bosonic ipeps . [ if one wants to use the decomposition of the reduced tensors @xmath205 and @xmath206 in terms of @xmath202 and @xmath203 , fig . [ fig : reducedab ] , the presence of four fermionic swap gates needs to be taken into account ] . in the simulations of sec . [ sec : results ] , we have used the directional ctm approach discussed in ref . in order to produce an approximate environment @xmath68}$ ] . then an approximation @xmath209 to the expected value @xmath195 is computed using the tensor network of fig . [ fig : correlator](a ) . figure [ fig : correlator](b ) describes the tensor network that needs to be contracted in order to compute a two - point correlators . the simulation of time evolution proceeds in a very similar way as in the bosonic ipeps algorithm , with the difference that the tensor networks involved contain fermionic swap gates instead of simple line crossings . a detailed description of the two updates used in this paper is presented in appendix . [ app : twosite ] . before we move to presenting the results of ground state computations , we conclude this section with a summary of the fermionic ipeps algorithm : \(i ) _ ansatz . _ the state @xmath12 of the system on an infinite square lattice @xmath1 is encoded in two parity symmetric tensors @xmath202 and @xmath203 . these tensors depend on @xmath239 parameters , where @xmath6 is the dimension of the vector space @xmath5 of one site of @xmath1 and @xmath43 is the bond dimensions of the ipeps . \(ii ) _ computation of expectation values_. given tensors @xmath202 and @xmath203 , the reduced tensors @xmath205 and @xmath206 are computed according to fig . [ fig : reducedab ] . from the reduced tensors @xmath205 and @xmath206 , an approximation @xmath68}$ ] to a two - site environment @xmath64}$ ] is obtained using the ctm algorithm , see fig . [ fig : corner ] . the computational cost scales as @xmath229 , where @xmath208 is the bond dimension of the approximate environment . then expectation values are computed by contracting small tensor networks involving @xmath68}$ ] . for instance , an approximation @xmath215 to the expected value @xmath195 , eq . , for an operator @xmath15 on two nearest neighbor sites is obtained according to fig . [ fig : correlator](a ) , whereas an approximation to a two - point correlator is obtained according to fig . [ fig : correlator](b ) . \(iii ) _ approximation of the ground state_. starting from e.g. random tensors @xmath202 and @xmath203 , an imaginary - time evolution is used to find an approximation to the ground state of a local hamiltonian @xmath13 . the two updates of appendix [ app : twosite ] can be used : the _ standard _ update , which in general produces better ground state approximations and has cost @xmath229 ; and the _ simplified _ update , which has a significantly lower cost and is often only marginally less accurate than the standard update . in this section we test the fermionic ipeps algorithm , as summarized at the end of the last section , for several models of free and interacting fermions in an infinite square lattice @xmath1 . we start by considering an exactly solvable model of free spinless fermions , which allows us to compare the numerical results with the exact solution , and therefore assess the accuracy of the approach . it turns out that , similarly as in the 2d mera , @xcite the accuracy of the numerical results depends on the amount of entanglement in the system . we also compare the use of the standard and simplified updates discussed in appendix [ app : twosite ] . the second model , describing interacting spinless fermions on a square lattice , is no longer exactly solvable . we compare our results for the phase diagram with the hartree - fock ( hf ) solution from ref . , and show that our results for large @xmath240 are an improvement upon the hf result . the final example is the @xmath0 model , where we compare energies obtained with ipeps with previous variational monte carlo results based on gutzwiller - projected ansatz wave functions . in all these examples we present ipeps results for different bond dimensions @xmath43 , and study the convergence of the energy as a function of @xmath208 , the bond dimension of the environment . in all these examples we chose @xmath241 , see eq . . the first model under consideration is an exactly solvable model of free spinless fermions,@xcite given by the hamiltonian @xmath242 - 2\lambda \sum_i { \hat c}_{i}^{\dagger } { \hat c}_i , \label{eq : free}\ ] ] with @xmath243 denoting the sum over nearest - neighbor pairs , @xmath244 the chemical potential , and @xmath245 the pairing potential . figure [ fig : ff_pd ] shows the exact phase diagram of the model.@xcite for @xmath246 the model reduces to the usual tight - binding model of free fermions , with a metal phase for @xmath247 and a band insulator for @xmath248 . also the line @xmath249 , @xmath250 corresponds to a metal phase with a one dimensional fermi surface . for @xmath250 , @xmath251 the system is superconducting , with a critical phase for @xmath252 and a gapped phase for @xmath253 . as a function of chemical potential @xmath244 and pairing potential @xmath245 . for @xmath250 , @xmath251 the system is superconducting.,width=264 ] , for different values of @xmath245 and @xmath43 . the dimension @xmath208 is 20 and 40 for @xmath254 and @xmath255 , respectively . , width=302 ] fig . [ fig : e_ff ] shows the relative error in the ground state energy obtained by simulating an imaginary - time evolution with the standard update ( cf . appendix [ app : twosite ] ) . results for bond dimensions @xmath254 and @xmath255 and at different locations in the phase diagram are shown . in the gapped phase , @xmath253 , accurate results are already obtained for @xmath254 , and for @xmath255 these accuracies are increased by one to three orders of magnitude . [ a special case is the band insulator for @xmath246 , @xmath256 , ( not shown ) which corresponds to an unentangled or product ground state and can be reproduced exactly even with bond dimension @xmath257 . ] the accuracy in the critical phase , @xmath258 , is of the order of @xmath259 for @xmath254 . increasing the bond dimension to @xmath255 , an order of magnitude is gained for @xmath260 , but the gain is smaller for @xmath261 . finally , the free fermion regime with a 1d fermi surface ( @xmath246 , @xmath262 ) is the most challenging case . note that in this case , the entanglement entropy exhibits a logarithmic correction to the area law.@xcite this shows that the accuracy of the results depends on the amount of entanglement in the system , similarly to the findings with the 2d mera . @xcite as a function of distance r ( in x - direction ) , see eq . , for two different values of @xmath263 of the free - fermion model . middle panels : absolute value of the correlation function @xmath264 in semi - logarithmic scale . lower panels : the difference between the simulation result @xmath265 and the exact result @xmath266 for different values of @xmath43 . notice that better accuracies are obtained in the gapped phase , which corresponds to a less entangled ground state.,width=302 ] next we study the accuracy obtained for the two - point correlator @xmath267 where @xmath268 , i.e. site @xmath170 is in the same row as site @xmath170 but separated by @xmath269 columns , see fig . [ fig : correlator](b ) . the ipeps results shown in fig . [ fig : ff_corrs ] are seen to approach the exact values with increasing @xmath43 . in the critical phase ( left panels ) only the correlations at short distances are reproduced accurately . the precision of the correlator at longer distance is rather poor for small @xmath43 , but improves upon increasing @xmath43 . in the gapped phase ( right panels ) the accuracy is clearly better , already for small bond dimension . for @xmath240 accurate results are obtained up to distance @xmath270 where the magnitude of the correlator is of the order @xmath271 . these results are obtained from simulations with the simplified update , and we checked that they are similar to the ones obtained with the standard update ( for @xmath254 and @xmath255 ) . we compare and discuss the two updates further below . , the bond dimension of the environment , for @xmath255 . in these examples the energy decreases monotonously with @xmath208 , i.e. the energy from the simulation is always larger than the true ground state energy.,width=302 ] since the ipeps is a variational ansatz , the ground state energy @xmath219 of a state represented by an ipeps is an upper bound of the exact ground state energy @xmath222 . with increasing bond dimension @xmath43 the true ground state can be represented more accurately , and consequently the energy @xmath219 becomes lower , i.e. a better upper bound of @xmath222 . however , as discussed in the previous section , quantities such as the energy can only be extracted in an approximate way from the ipeps . the error of this approximation depends on the bond dimension of the environment @xmath208 . in fig . [ fig : chidep ] the dependence of the relative error of the energy as a function of @xmath208 is plotted for an ipeps with bond dimension @xmath255 . we observe that , with increasing @xmath208 , the energy @xmath223 converges to some value that is indeed an upper bound for @xmath222 . this is consistent with the assumption that @xmath223 has converged to the true energy @xmath219 of our @xmath255 ipeps . in the other models analyzed in this section we will also assume that once @xmath223 does no longer change with increasing @xmath208 , it has attained @xmath219 and therefore is an upper bound to @xmath222 . [ fig : chidep ] shows that , in the present free model , the approximate energy @xmath223 monotonically decreases with increasing @xmath208 , and therefore any value of @xmath223 is already an upper bound to @xmath222 . however , this behavior is not true in general , as we will see further below . it is therefore important to study the convergence in @xmath208 in each case separately . , and @xmath272 ( @xmath273 for @xmath274 ) . the environment in the gapped phase converges considerably faster than in the critical phase.,width=302 ] let us make a few comments about the convergence of the environment in the ipeps algorithm , for fixed sets of tensors @xmath202 and @xmath203 which approximate the ground state of different phases . figure [ fig : envdep ] illustrates how the energy converges with the number of renormalization steps in the directional ctm algorithm @xcite for different points in the phase diagram . the environment in the gapped phase clearly converges faster than in the critical phase . here , the tensors at the boundary of the ipeps have been chosen randomly at the beginning . depending on the initial conditions of these boundary tensors the number of steps needed to converge can vary significantly . in practice , when using the standard update , we use the environment from the previous imaginary time step as an initial condition , so that only a few renormalization steps are needed to re - converge the environment . note also that for different initial boundary tensors the energy might converge to slightly different values ( especially in the critical phase ) . it is therefore advisable to check results for different initial boundary tensors . for this model , we have also compared the precision of the ground state energy obtained with the standard and simplified updates , see fig . [ fig : ff_compare ] . the accuracies obtained with the standard update are typically slightly better than the ones obtained with simplified updates . however , the simulations with the simplified update are computationally considerably cheaper , because the environment has to be computed only once at the end of the simulation for the evaluation of observables , whereas in the standard update the environment it has to be recomputed at each step of the imaginary - time evolution . the leading cost of the two methods is the same , but the computational cost differs by a rather large factor , which depends on the total number of imaginary - time steps . for the remaining examples of this paper we will consider only simplified updates this allowed us to perform simulations with @xmath240 and @xmath275 on a standard computer in roughly one day . we checked for @xmath254 and @xmath255 that the results obtained with the two updates give similar accuracies . it is conceivable , however , that the difference in accuracy between the two update schemes become larger with increasing @xmath43 , depending on the model under consideration . .,width=302 ] the second model under consideration is a model of interacting spinless fermions , which is not exactly solvable . it is defined by the hamiltonian @xmath276 - \mu \sum_i { \hat c}_{i}^{\dagger } { \hat c}_i + v \sum_{\langle ij \rangle } { \hat c}_i^{\dagger}{\hat c}_i { \hat c}_j^{\dagger}{\hat c}_j , \label{eq : hint}\ ] ] with @xmath277 the nearest - neighbor hopping amplitude , the ( repulsive ) nearest - neighbor interaction strength , and @xmath278 the chemical potential . the hartree - fock ( hf ) phase diagram @xcite as a function of @xmath279 and particle density @xmath280 is given fig . [ fig : if_pd ] ( solid line ) . the hf calculation predicts a gapped charge - density - wave ( cdw ) phase at half filling ( @xmath281 ) and a translational invariant normal state ( metal phase ) far away from half filling . in between these two phases they find a thermodynamically unstable region , which we identify as a phase separation ( ps ) region , i.e. where the system splits into two parts , one in the metal phase and the other in the cdw phase . is the interaction strength , and @xmath280 is the particle density ( filling ) . the charge - density wave phase ( checkerboard pattern ) at exactly half - filling ( @xmath281 ) is separated from the metal phase by a first order phase transition , with an intermediate region corresponding to phase separation ( ps ) . with increasing @xmath43 , the boundary between the metal phase and the ps region moves away from the hartree - fock ( hf ) result from ref . . dotted and dashed lines are a guide to the eye . the uncertainty of the phase boundaries obtained with ipeps is smaller than the symbol size ( in x - direction ) . , width=302 ] for @xmath282 , obtained by ipeps and hartree - fock ( hf).@xcite the first order phase transition between the metal phase and the charge - density wave ( cdw ) phase occurs at a value @xmath283 where the two corresponding energies cross . lower panel : particle density @xmath280 as a function of chemical potential . at the first order phase transition point @xmath283 , @xmath280 jumps from a certain value @xmath284 in the metal phase to @xmath281 in the cdw phase . for densities in between @xmath284 and @xmath281 the system exhibits phase separation . the numbers in brackets indicate the uncertainty in the last digit . , width=302 ] our results for the phase diagram , obtained with the simplified update using an ipeps with bond dimension @xmath255 and @xmath240 , are given by the squares and triangles in fig . [ fig : if_pd ] , respectively . the phase diagram obtained with ipeps qualitatively agrees with the hf solution . however , with increasing @xmath43 the phase boundary to ps region moves away from the hf result . in the following we explain how we computed the phase boundary and discuss the origin of this deviation . we focus only on the left half of the phase diagram , @xmath285 , since it is mirror symmetric with the @xmath281 line . we determined the phase boundary between the metal phase and the ps region for several values of @xmath279 , for @xmath255 and @xmath240 . for each value of @xmath279 , we studied the first order phase transition , which occurs at a certain value @xmath286 . the ipeps algorithm known to be particularly suitable to study first order phase transitions , thanks to displaying some degree of hysteresis . @xcite specifically , if we start a simulation with an ipeps that represents a state in e.g. the metal phase , it will remain in the metal phase upon increasing @xmath278 , even for values ( slightly ) larger than @xmath283 ( where the ground state is no longer metallic ) . this allows us to compute the energy of the metal phase in the region @xmath287 even though the cdw - state has a lower energy , and vice versa . we can therefore compute the energies of the ground states of the two phases individually , and the phase transition occurs where the two energies cross , as shown in the upper panel of fig . [ fig : if_energy ] for @xmath282 . at the transition point @xmath283 , the density @xmath280 jumps from a certain value @xmath288 in the metal phase to the value @xmath281 in the cdw phase , as plotted in the lower panel of fig . [ fig : if_energy ] . for densities with values between @xmath284 and @xmath281 there exists no homogenous ground state , i.e. the system exhibits phase separation ( cf . [ fig : if_pd ] ) . let us now compare the energies obtained with ipeps with the hf solution in fig . [ fig : if_pd ] . while the results in the cdw phase coincide ( up to @xmath289 ) , we obtain lower energies for @xmath240 than the hf result ( of the order of @xmath290 ) . this leads to a shift of the transition point @xmath283 to a larger value of @xmath278 , and thus the phase boundary appears at a larger value of @xmath284 than in the hf case . the error of the energy due to the finite @xmath208 is smaller than the symbol size ( cf . [ fig : if_chidep ] ) . within this error , we regard the results as `` variational '' energies , i.e. an upper bound of the true ground state energy . accordingly , our @xmath240 results in the metal phase are closer to the true ground state energy , and the phase boundary for @xmath240 is an improvement upon the hf solution . note that the convergence of the energy with @xmath208 in fig . [ fig : if_chidep ] is not monotonous as in the previous examples in fig . [ fig : chidep ] . still , for large @xmath208 the energies do not seem to change significantly anymore . for some simulations we observe that the environment does not converge to a fixed point , but rather oscillates slightly . in such a case the energy fluctuates around a certain value . we take this error also into account in our study . the final uncertainty of the phase boundary obtained by ipeps is smaller than the symbol sizes ( in x - direction ) in fig . [ fig : if_pd ] . all simulation results for this model were obtained with the simplified update of appendix [ app : twosite ] . is not monotonous . for large @xmath208 the values do not seem to change significantly anymore . we associate an error bar to the values of the energy , depending on the convergence behavior in @xmath208 . this error is smaller than the symbol sizes in fig . [ fig : if_energy].,width=302 ] as a final example we consider the @xmath0 model , @xmath291 with @xmath292 the spin index , @xmath293 the electron density and @xmath294 the spin @xmath295 operator on site @xmath137 , and @xmath296 . the @xmath0 model is an effective model of the hubbard model in the limit of strong on - site repulsion . the local hilbert space of each site contains three basis states @xmath297 , i.e. two electrons with opposite spins can not occupy the same lattice site as in the hubbard model . the @xmath0 model is an important model in the context of high - t@xmath298 superconductivity and its phase diagram is still controversial . we focus here on the parameter @xmath299 which lies in the relevant parameter range of cuprate superconductors . at half filling , i.e. particle density @xmath300 , the model corresponds to the antiferromagnetic spin 1/2 heisenberg model , where the ground state has long - range antiferromagnetic ( nel ) order . far away from half filling the model is a metal . at small , but finite doping @xmath301 , several studies predict a @xmath302-wave superconducting phase , see e.g. refs . and references therein . in studies with dmrg also a phase with stripe order @xcite was found at low doping . some studies suggest that the formation of stripes is due to phase separation of the undoped antiferromagnet and the superconducting phase . @xcite here we focus on the variational monte carlo ( vmc ) results from ref . , based on gutzwiller - projected ansatz wave functions . this study was done for finite lattices of size @xmath303 , where the finite size corrections of the energy are estimated to be of the order of @xmath304 . the monte carlo sampling error is of the same order of magnitude . at low doping ( @xmath305 ) the best variational energies are obtained by an ansatz wave function including superconductivity and antiferromagnetic order . at larger doping @xmath306 a better variational energy is obtained with an ansatz wave function that is superconducting without antiferromagnetic order . in the following we compare our energies with the best values obtained in this previous study . figure [ fig : tj_energy ] shows the energy per site ( with the chemical potential term subtracted ) as a function of particle density . one can see that the ipeps results approach the vmc results from ref . with increasing @xmath43 . for @xmath240 the ipeps energies are roughly @xmath259 higher than the vmc energies , and for @xmath307 the deviation is of the order of @xmath304 , which is the same order of magnitude as the error bar of the vmc results . note also , that some of the energies for @xmath307 are lower than the vmc results . however , the @xmath307 results are not necessarily `` variational '' energies , as we discuss below . of the @xmath0 model as a function of filling @xmath280 , with @xmath299 . with increasing @xmath43 , the energies obtained by ipeps are approaching the values from the variational monte carlo ( vmc ) study in ref . note that the chemical potential term has been subtracted from the energy . the error bars of the vmc results are of the order @xmath308 . lower panel : relative deviation between the energies obtained by ipeps and vmc . full symbols indicate that ipeps energy is lower than the vmc energy.,width=302 ] these preliminary results are encouraging for the future study of the phase diagram of the @xmath0 model using peps algorithms , because the current energies with the largest dimension @xmath307 are compatible with previous variational studies , and we expect to be able to increase @xmath43 by properly exploiting the global symmetries of the model . we also would like to emphasize that ipeps is a general ansatz , i.e. the same ansatz is used for any model on a square lattice , whereas in other variational studies the ansatz wave function is typically based on the specific physics of the model . it is remarkable that ipeps , which starts from a random initial state , yields comparable energies as the ones obtained by specialized ansatz wave functions . finally , it will be interesting to verify whether the phases obtained by ipeps correspond to the ones predicted by the variational study , or whether a phase with different dominant correlations appears , e.g. stripe - ordered phase , as previously found in dmrg studies . @xcite figure [ fig : tj_chi ] shows that the energies as a function of @xmath208 are sufficiently converged , both for @xmath255 and @xmath240 , i.e. the uncertainty due to a finite @xmath208 is much smaller than the symbol sizes in fig.[fig : tj_energy ] . we can thus view the energies for @xmath309 as `` variational '' in the sense that they are an upper bound of the true ground state energy . however , for at finite doping the energy increases with increasing @xmath208 , and it is at present not clear what value it will reach for larger ( and presently unaffordable ) values @xmath208 . thus , in this case the energy is not believed to necessarily be an upper bound to the exact ground state energy . . the energies for @xmath309 do not seem to change significantly anymore for large @xmath208 , whereas for @xmath307 the energy is still increasing at the largest value of @xmath208 used.,width=302 ] in recent months several theoretical proposals have appeared describing fermionic versions of tensor network algorithms for 2d lattice systems , namely fermionic mera@xcite and fermionic peps@xcite . in this paper we have explained how to obtain fermionic peps algorithms by applying the general fermionization procedure of tensor networks introduced in ref . . a highlight of our formulation of fermionic peps is its simplicity : it replaces the complexity involved in dealing with a network of fermionic operators with a tensor network built by following two rules , namely the use of parity preserving tensors and the substitution of line crossings with fermionic swap gates . we have then used the fermionic ipeps algorithm to compute an approximation to the ground state of several fermionic models on an infinite lattice . by simulating an exactly solvable model of free fermions,@xcite we have been able to see that even a fermionic peps with small bond dimension @xmath43 ( @xmath310 ) is already capable of reproducing up to several digits of the exact ground state energy , as well as two - point correlators . the results also showed that , similarly to what had been observed with the fermionic mera,@xcite the accuracy of the approach depends on the amount of entanglement in the system . generally speaking , gapped systems are less entangled than critical ones and , accordingly , a fermionic ipeps with a given bond dimension @xmath43 produces better accuracies for the former . the simulation of model of interacting spinless fermions has provided us with a first clear indication of the usefulness of the fermionic peps as a variational ansatz for the ground state of systems of interacting fermions . the fermionic peps approach reproduces the hartree - fock phase diagram@xcite , with metal and charge - density wave phases ; but even with bond dimension @xmath255 and @xmath240 , it improves the ground state energies on the metallic phase and this results in a significant shift of the phase boundary . finally , results for the @xmath0 model in the relevant parameter range for cuprate superconductors are particularly encouraging , given that fermionic peps , still at an early stage of development , already produce ground state energies comparable with those of previous variational studies.@xcite the main limitation in present calculations is due to the scaling @xmath229 of simulation costs with the peps bond dimension @xmath43 and the environment bond dimension @xmath208 , which restricts @xmath43 and @xmath208 to relatively small values . there are , however , several ways in which larger values of @xmath43 and @xmath208 could become affordable , thereby leading to more accurate results . on the one hand , one can incorporate the internal global symmetries of a model ( e.g. particle number conservation ) into the tensors and exploit their block structure to reduce computational costs.@xcite this strategy is expected to be decisive for the characterization of the phase diagram of the @xmath0 and hubbard models . on the other hand , monte carlo sampling techniques could be used to reduce the formal dependence of the simulation costs in @xmath43 and @xmath208 . @xcite finally , the use parallel computing on a large cluster would also lead to improved fermionic peps simulations . _ acknowledgements.- _ the authors thank j. de woul , e. langmann , and d. ivanov for providing us their data from refs . and . we also acknowledge inspiring discussions with j. jordan , r. pfeifer , l. tagliacozzo and h .- q . support from the university of queensland ( ecr2007002059 ) and the australian research council ( ff0668731 , dp0878830 ) is acknowledged . in the case of a lattice where each site is described by a generic vector space @xmath5 of finite dimension @xmath6 , we decompose @xmath5 into even and odd parity sectors , @xmath311 and use an index @xmath106 to label a basis with well defined parity , @xmath312 see sec . [ sec : peps : parity ] . we can also introduce a set @xmath313 of generalized fermionic operators on each site , where @xmath314 is defined as @xmath315 then a local operator @xmath15 acting on just one site can be expanded as @xmath316 notice that @xmath317 is parity preserving if @xmath318 and parity changing if @xmath319 . fermionic operators @xmath320}_{s_1}$ ] and @xmath321}_{s_2}$ ] acting on two different sites @xmath322 fulfill @xmath323}_{s_1 } { \hat f}^{[\vec{r}_2]}_{s_2 } = s(s_1,s_2 ) { \hat f}^{[\vec{r}_2]}_{s_2 } { \hat f}^{[\vec{r}_1]}_{s_1},\ ] ] where @xmath324 a two - site operator @xmath15 acting on sites @xmath325 can then be written as @xmath326 where @xmath327\dagger } { \hat f}_{s'_2}^{[\vec{r}_2]\dagger } { \mbox{$| 0_10_2 \rangle$ } } { \mbox{$\langle 0_10_2 |$ } } { \hat f}_{s_2}^{[\vec{r}_2 ] } { \hat f}_{s_1}^{[\vec{r}_1]}.\ ] ] and @xmath328.,width=302 ] is obtained from tensor @xmath202 by crossing the legs @xmath36 and @xmath6 . ( b ) tensor @xmath329 is obtained from tensor @xmath328 by crossing the legs @xmath330 and @xmath331 . ( c ) tensors @xmath332 and @xmath333 are obtained by joining the @xmath334 indices and the @xmath335 indices of @xmath336 , and doing a singular value decomposition of the resultant matrix . the singular values are included in the definition of tensor @xmath333 . ( d ) tensors @xmath337 and @xmath338 are obtained by joining indices @xmath339 and indices @xmath340 of @xmath203 together , then performing a singular value decomposition of the resulting matrix , then splitting the composed indices back apart . the singular values are included in tensor @xmath337 . ( e ) tensors @xmath341 and @xmath342 are obtained by joining indices @xmath343 and indices @xmath344 of @xmath345 , then performing a singular value decomposition of the resulting matrix , then splitting the composed indices back apart . the singular values are included in tensor @xmath342 . ( f ) tensors @xmath346 and @xmath347 are obtained by joining indices @xmath348 and indices @xmath349 of @xmath329 , then performing a singular value decomposition of the resulting matrix , then splitting the composed indices back apart . the singular values are included in tensor @xmath346.,width=340 ] is obtained by contracting a tensor network that contains the approximate environment @xmath82 } = \{e_1 , \cdots , e_6\}$ ] , tensors @xmath332 , @xmath341 , @xmath350 and @xmath347 , and a few fermionic swap gates . ( b ) tensor @xmath351 is obtained from the contraction of @xmath352 and the gate @xmath353.,width=321 ] this appendix describes how to update the tensors @xmath202 and @xmath203 that define a fermionic ipeps . we consider the two strategies employed to obtain the results of sec . [ sec : results ] , namely ( i ) the _ standard _ update , used in refs . for bosonic systems , which requires the approximate environment @xmath82 } = \{e_1,\cdots , e_6 \}$ ] of fig . [ fig : corner](d ) ; and ( ii ) a _ simplified _ update , used in ref . , which does not involve an environment . in these two schemes , a gate @xmath354}\delta t)$ ] is applied to the two sites @xmath355 of a given link , with tensors @xmath202 and @xmath203 , and new tensors @xmath224 and @xmath225 are chosen in order to best account for the action of the gate . here we simply list the steps required in order to obtain the updated tensors @xmath224 and @xmath225 . for a justification of the schemes we refer to refs . . we emphasize that the only difference between these updates and the ones used in a bosonic system is the presence of fermionic swap gates . in particular , if the fermionic swap gates are eliminated ( by setting @xmath356 in eqs . -[eq : s ] ) we obtain update algorithms for bosonic peps . in some models , such as the quantum ising model near criticality , @xcite the standard update produces significantly more accurate results than the simplified update , but this did not seem to be the case in the gapless phases studied in this paper . in all models away from criticality that we could test , the simplified update produces only marginally worse accuracies . on the other hand , the much lower computational cost of the simplified update allows to consider larger bond dimension @xmath43 than with the standard update . for concreteness , in the following we assume that the gate @xmath353 is applied on a horizontal link where tensors @xmath202 and @xmath203 are at the left and right , respectively . similar derivations apply to the other three types of links . and @xmath357 are obtained by iteratively solving two linearized equations . specifically , we iterate the following two steps until convergence : ( a ) given @xmath337 and @xmath346 , we obtain @xmath358 by solving the linear equation @xmath359 , where @xmath360 and @xmath361 are obtained as indicated in the diagram . then , we set @xmath362 , and compute @xmath363 . ( b ) given @xmath333 and @xmath342 , we obtain @xmath357 by solving the linear equation @xmath364 , where @xmath360 and @xmath361 are again obtained as indicated in the diagram . then , we set @xmath365 , and compute @xmath366.,width=245 ] given tensors @xmath202 and @xmath203 , an approximate environment @xmath82 } = \{e_1,\cdots , e_6 \}$ ] is obtained as described in sec . [ sec : ctm ] : first build the reduced tensors @xmath205 and @xmath206 , fig . [ fig : reducedab ] , which are the building blocks of the exact environment @xmath207}$ ] ; then use e.g. ctm techniques@xcite to produce an approximation @xmath82}$ ] to @xmath207}$ ] . on the other hand , it is convenient to introduce a number of additional tensors . first compute tensors @xmath345 and @xmath328 according to fig . [ fig : revupdate1 ] , as well as tensors @xmath336 and @xmath329 according to fig . [ fig : revupdate2](a , b ) . then perform a singular value decomposition of tensors @xmath336 , @xmath203 , @xmath345 and @xmath329 as explained in the caption of fig . [ fig : revupdate2](c - f ) . here we use the notation @xmath367 for the singular value decomposition of a matrix @xmath56 , where @xmath368 and @xmath369 are isometries and @xmath370 is the diagonal matrix of singular values . notice that the singular value decompositions are performed by regarding a tensor as a matrix after grouping its indices into two sets . is obtained from the contraction of @xmath332 and @xmath358 . ( b ) the updated tensor @xmath371 is computed by crossing the @xmath36 and @xmath6 indices of @xmath372 . ( c ) the updated tensor @xmath373 is obtained from the contraction of @xmath357 and @xmath350.,width=340 ] tensor @xmath374 is then obtained by contracting the tensor network of fig . [ fig : revupdate3](a ) , which in turn is used to produce tensor @xmath351 in fig . [ fig : revupdate3](b ) . from tensors @xmath374 and @xmath351 , updated tensors @xmath358 and @xmath357 are obtained by iterating until convergence the process explained in the caption of fig . [ fig : revupdate4 ] . for the simplified update from ref . , the structure of the peps tensor network is slightly different to the one considered so far in this paper . here the infinite peps is specified by two tensors @xmath384 and @xmath375 , and four diagonal matrices @xmath376 with non - negative diagonal entries that assign weights to the indices of @xmath384 and @xmath375 , see fig . [ fig : revupdate6](a ) . notice that the usual expression in terms of tensors @xmath202 and @xmath203 can be recovered e.g. by multiplying the square root of the weight matrices @xmath376 with the tensors @xmath384 and @xmath375 , see fig . [ fig : revupdate6](b)-(c ) . the simplified update consists of the following steps : first , a tensor @xmath385 is computed in analogous way as tensor @xmath336 in fig . [ fig : revupdate2](a ) . then tensor @xmath380 is obtained by contracting the network in fig . [ fig : revupdate7](a ) , and subsequently decomposed through a singular value decomposition as shown in fig . [ fig : revupdate7](b ) . this results in tensors @xmath377 and @xmath378 and the matrix of weights @xmath379 , which corresponds to the singular values of @xmath380 . at this stage @xmath379 is truncated into @xmath386 , which keeps only the @xmath43 largest diagonal entries of @xmath379 . then tensors @xmath377 and @xmath378 are also truncated accordingly . next , tensors @xmath387 and @xmath383 are obtained from ( the truncated version of ) tensors @xmath377 and @xmath378 as shown in fig . [ fig : revupdate8](a)-(b ) . finally , tensor @xmath388 is obtained from @xmath387 again in an analogous way as @xmath224 in fig . [ fig : revupdate5](b ) . h. c. jiang , z. y. weng , and t. xiang , phys . lett . 101 , 090603 ( 2008 ) . z. y. xie , h. c. jiang , q. n. chen , z. y. weng , and t. xiang , phys . 103 , 160601 ( 2009 ) . chen , c .- y . lai , and m .- f . yang , j. stat . mech . : theory exp . ( 2009 ) p10001 . g. vidal , phys . 99 , 220405 ( 2007 ) . g. vidal , phys . . lett . 101 , 110501 ( 2008 ) . l. cincio , j. dziarmaga , and m. m. rams phys . 100 , 240603 ( 2008 ) . g. evenbly and g. vidal , phys . 102 , 180406 ( 2009 ) . g. evenbly and g. vidal , phys . rev . b 79 , 144108 ( 2009 ) . v. murg , f. verstraete , and j. i. cirac , phys . rev . b 79 , 195119 ( 2009 ) . g. evenbly and g. vidal , arxiv:0904.3383 . b. bauer , g. vidal , and m. troyer , j. stat . mech . : theory exp . ( 2009 ) p09006 . p. chen , c .- y . lai , and m .- f . yang , arxiv:0910.5081 . h. yokoyama and m. ogata . 65 , 3615 ( 1996 ) . d. a. ivanov . phys . rev . b 70 , 104503 ( 2004 ) . m. lugas , l. spanu , f. becca , and s. sorella . phys . rev . b 74 , 165122 ( 2006 ) . l. spanu , m. lugas , f. becca , and s. sorella . phys . rev . b 77 , 024510 ( 2008 ) . s. r. white and d. j. scalapino . 80 , 1272 ( 1998 ) . s. r. white and d. j. scalapino . 81 , 3227 ( 1998 ) . s. r. white and d. j. scalapino . phys . rev . b 61 , 6320 ( 2000 ) . v. j. emery , s. a. kivelson , and h. q. lin . 64 , 475 ( 1990 ) . c. s. hellberg and e. manousakis , phys . 78 , 4609 ( 1997 ) .

the spectra of snrs in radio continuum are usually represented by a power - law , reflecting the pure synchrotron radiation from the snr shell:@xmath3 where @xmath4 is the spatially - integrated flux density and @xmath5 is the radio spectral index . test - particle diffusive shock acceleration ( dsa ) theory predicts that for strong shocks , radio spectral index @xmath5 is approximately 0.5 ( bell 1978ab ) . indeed , mean value of the observed radio spectral indices , for galactic snrs , is around 0.5 and correspondent distribution is roughly gaussian @xcite . on the other hand , the dispersion in that distribution is non - negligible ( see fig . 1 in reynolds et al . 2012 ) . in the case of shocks with sufficiently low mach number ( less than around ten ) , steeper spectra ( @xmath6 ) are generally expected . in test - particle dsa , for the synchrotron spectral index @xmath5 , the following relations hold : @xmath7 where @xmath2 is the post - shock thermal gas adiabatic index , @xmath8 is the compression ratio for parallel shock , @xmath9 is the ( upstream adiabatic ) mach number that represents the ratio of shock velocity to the local sound speed and @xmath10 is the energy spectral index . on the other hand , only few old snrs would be expected to have such weak shocks @xcite . in fact , snrs with steep spectrum ( @xmath6 ) are usually young objects @xcite . steeper radio spectra may be explained by a cumulative effect of non - linear and oblique - shock steepening in the case of young snrs @xcite . an important prediction of the so called non - linear dsa ( ndsa ) , where the reaction effects of cosmic - ray particle pressure is taken into account , is that the spectrum of young snrs should flatten at higher energies so that a `` concave up '' spectrum is formed @xcite . there is also a considerable number of galactic snrs with @xmath0 @xcite . contamination with flat spectrum thermal emission can not be completely ruled out in all of the cases so it may be partially responsible for these lower ( flat ) spectral index values . [ cols="^,^,^,^,^,^,^,^,^,^",options="header " , ] \(1 ) abdo at al . ( 2010a ) , ( 2 ) dubner et al . ( 2000 ) , ( 3 ) gabici et al . ( 2010 ) , ( 4 ) giuliani et al . ( 2010 ) , ( 5 ) sawada & koyama ( 2012 ) , ( 6 ) tian et al . ( 2007 ) , ( 7 ) paron et al . ( 2012 ) , ( 8) dubner et al . ( 1999 ) , ( 9 ) brogan et al . ( 2005 ) , ( 10 ) chen et al . ( 2004 ) , ( 11 ) castro & slane ( 2010 ) , ( 12 ) green ( 2009 ) , ( 13 ) sun et al . ( 2011 ) , ( 14 ) abdo et al . ( 2010b ) , ( 15 ) castelletti et al . ( 2007 ) , ( 16 ) uchida et al . ( 2012 ) , ( 17 ) su et al . ( 2011 ) , ( 18 ) patnaik et al . ( 1990 ) , ( 19 ) anderson & rudnick ( 1993 ) , ( 20 ) abdo at al . ( 2010c ) , ( 21 ) moffett & reynolds ( 1994 ) , ( 22 ) zhou et al . ( 2011 ) , ( 23 ) gao et al . ( 2011 ) , ( 24 ) reichardt et al . ( 2012 ) , ( 25 ) pannuti et al . ( 2010 ) , ( 26 ) foster ( 2005 ) , ( 27 ) jeong et al . ( 2012 ) , ( 28 ) troja et al . ( 2006 ) , ( 29 ) troja et al . ( 2008 ) , ( 30 ) castelletti et al . ( 2011 ) , ( 31 ) abdo et al . ( 2010d ) @xmath11 @xcite noted that spectral index of this snr is varying . + @xmath12 or 0.54 at frequencies higher and 0.02 at frequencies lower than 1 ghz ( sun et al . 2011 ) . + the most of the ( shell , composite and mixed - morphology ) snrs with @xmath0 are in fact ( evolutionary ) old objects expanding in a high density environment . because the shock velocities are much lower than in young snrs , the oblique - shock effects considered by bell et al . ( 2011 ) are unlikely to contribute to the flattening of the radio spectral indices of these snrs . also , in the case of older snrs , it is not yet clear whether the mechanism for producing relativistic electrons is local acceleration at a shock front or compression of the existing population of galactic background relativistic electrons ( leahy 2006 and references therein ) . it must be noted that the majority of members of the so called mixed - morphology class of snrs have flat spectral indices ( see table 4 in vink 2012 ) . the mixed - morphology ( thermal composite ) snrs are characterized with bright ( thermal ) interiors in x - rays , and bright ( non - thermal ) shell - like radio morphology @xcite . they are mainly evolutionary old snrs expanding in a high density environment ( many of them interacting with molecular clouds ) and many of them are detected in @xmath2-rays @xcite . in greens s catalog of galactic snrs @xcite these remnants are classified either as shell or composite type ( they are not separated from the rest of snrs ) . it should be pointed out that , from the list of the snrs with flat spectrum , plerions ( pulsar wind or synchrotron nebulae ) are excluded since they are not of interest for this paper . in fact , as the radio and x - ray emission from these objects are powered by the pulsar wind , not by the supernova explosion , they should not be referred as snrs ( at least in a classical context ) . for pulsar wind nebulae , radio spectral indices are usually less than around @xmath13 @xcite . finally , for a completeness of this summary , it should be noted that among the galactic and some large magellanic cloud ( lmc ) snrs , existence of significant spectral break and steepening at higher radio frequencies were also observed ( crawford et al . 2008 ; xiao et al . 2008 ; bozzetto et al . 2010 ; de horta et al . 2012 ) . the best example is the radio continuum spectrum of galactic snr g180.0 - 1.7 ( s147 ) for which the spectral break is around 1.5 ghz , with low - frequency spectral index around 0.3 and high - frequency spectral index around 1.20 ( xiao et al . the compression of the local magnetic field and shift of the turn - over in the galactic radio spectrum to higher frequencies is one of the possible explanation of this kind of spectrum . on the other hand , synchrotron losses during the early phase of the snr would cause a bend at a rather high frequency , but subsequent expansion of the snr would shift the frequency toward about 1.5 ghz ( xiao et al . some of the models that try to explain `` concave down '' spectra of some galactic and extragalactic objects ( including synchrotron losses as well as finite emission region ) are presented in the following papers : bregman ( 1985 ) , biermann & strittmatter ( 1987 ) , heavens & meiseheimer ( 1987 ) . it is very difficult to derive accurate integrated flux densities . the small number of data points ( with associated errors ) and the dispersion of flux densities at the same frequencies pose a practical problem in determining the ( integrated ) radio spectral index and , generally , the shape of radio continuum . in that sense , it should be noted that present knowledge of radio spectral indices as well as of the overall shape of the galactic snr radio continuum spectra is far from being precise ( at least for the majority of galactic snrs ) . new observations usually lead to non - negligible changes in the measured spectral indices . for example , although for some mixed - morphology snrs greens s catalog @xcite gives rather steep spectral index , new observations given in @xcite and @xcite give slightly flatter , but still steep ( @xmath6 ) spectra ( e.g. for g53.6 - 2.2 , @xmath14 ; g93.7 - 0.2 , @xmath15 ; g116.9 + 0.2 , @xmath16 ; g160.9 + 2.6 , @xmath17 ) . in this paper are considered only those snrs for which the flux densities are on the scale of @xcite and have known errors @xmath18 . snrs with only four or less data points are not discussed in this paper . table 1 shows the list of several galactic snrs , with well defined radio spectra in the sense of previously mentioned criteria , and their basic properties : type , integrated radio spectral index @xmath5 , presence of low - frequency turn - over in overall spectrum ( t.o . ) , distance @xmath19 , mean diameter @xmath20 , approximate age @xmath21 , association / interaction with molecular cloud ( mc ) , and detection in @xmath2-rays ( @xmath2 ) . the question of nature of the flat spectral indices in some galactic snrs is , in fact , still open . in sections 2 4 a brief revision of proposed explanations ( with the strongest arguments in their favor ) is presented . section 4 also deals with high energy properties of the snrs with flat radio spectra focusing on their detection in @xmath2-rays with the large area telescope onboard fermi @xmath2-ray space telescope ( _ fermi _ lat ) . in section 5 , model which includes significant intrinsic thermal bremsstrahlung radiation is discussed . finally , section 6 concludes the paper . although fermi i ( dsa ) is generally believed to be a primary mechanism for particle acceleration in snrs , fermi ii ( stochastic acceleration , see liu et al . 2008 for details ) process in the shock turbulent vicinity could also be significant in some cases . @xcite used results of drge et al . ( 1987 ) who solved the steady - state transport equation for the volume - integrated phase space distribution function of relativistic electrons in the vicinity of a plane - parallel shock wave allowing simultaneously for energy gain by multiple shock crossing and by second - order fermi momentum diffusion upstream and downstream of the shock . they concluded that the observed dispersion in spectral index values below @xmath22 is attributed to a distribution of low upstream plasma @xmath23 values for different snrs . plasma @xmath23 is defined as follows : @xmath24 where @xmath25 is the gas pressure , @xmath26 the magnetic pressure , @xmath27 the magnetic induction , @xmath28 the number density , @xmath29 the boltzmann constant , @xmath30 and @xmath31 the electron and ion temperatures ( if different , or just temperature @xmath32 of the medium ) , @xmath33 the adiabatic sound speed , @xmath34 the alfv@xmath35n speed and @xmath2 the adiabatic index . for example , temperatures of around @xmath36 k ( @xmath37 1 ev ) and number densities of around @xmath38 lead to the plasma @xmath23 of 0.16 for the average value of galactic magnetic field ( magnetic induction ) of around @xmath39 . higher densities ( as well as higher temperatures ) would rise the plasma @xmath23 . also , higher magnetic fields will result in lower values of plasma @xmath23 for the same set of temperatures and densities . @xcite also pointed out that values of @xmath40 are sufficient to yield @xmath41 for shock compression ratios higher than ( or equal ) 2.5 ( assuming , also , constant spatial diffusion coefficient ; see equations 6 and 7 in schlickeiser & frst 1989 ) . on the other hand , in their work , they did not account for the possibility of compression ratios higher than 4 that could be expected in some ( evolutionary ) older snrs ( spectral index becomes 0.5 for compression ratio of 4 independently of plasma @xmath23 ) . drge et al . ( 1987 ) emphasized that the flat spectra could be obtained for compression ratios that tend to the value of 4 if the spatial diffusion coefficient increases with momentum and @xmath42 . however , in the case of @xmath42 plasma and constant spatial diffusion coefficient , flat spectral indices are obtained for compression ratios higher than 4 ( which is not discussed in drge et al . adiabatic index is assumed to be @xmath43 in all the relevant equations . @xcite discussed the particle acceleration process at a ( parallel ) shock wave , in the presence of the second - order fermi acceleration in the turbulent medium near the shock , as an alternative explanation for the observed flat radio spectra of some galactic snrs in molecular clouds . he found that even very weak shocks may produce very flat cosmic - ray particle spectra in the presence of momentum diffusion ( using power - law form for the spatial diffusion coefficient @xmath44 , assuming shock compression ratio of 4 and applying an effective jump condition for the alfv@xmath35n velocity @xmath45 , where indices 1 and 2 refer to the upstream and downstream of the shock , respectively ) . ostrowski ( 1999 ) pointed out that as a consequence , the dependence of the particle spectral index of the shock compression ratio can be weaker than that predicted for the case of pure first - order acceleration . his analysis was based on the model of ostrowski & schlickeiser ( 1993 ) that improved the work of drge et al . ( 1987 ) and @xcite deriving a simplified kinetic equation in the momentum space valid for the case of small momentum diffusion and also allowing the description of stationary spectra at the shock . @xcite emphasized that shock waves with alfv@xmath35n speed non - negligible in comparison to the shock velocity are responsible for generating the flat particle distribution . he noted that for the ratio of alfv@xmath35n speed and shock velocity ( inverse of the so called alfv@xmath35n mach number , see equation 4 ) around 0.1 the second - order fermi process can substantially modify the energy spectrum of the particles accelerated at the shock and that the model holds for values of that ratio less than 0.2 . previously mentioned alfv@xmath35n mach number of a shock is determined by @xcite : @xmath46 where @xmath47 is the shock velocity , @xmath34 is the alfv@xmath35n velocity and @xmath48 is the shock velocity in @xmath49 , @xmath50 is the ionized ambient gas number density in @xmath51 and @xmath52 is the local magnetic induction , just before the shock , measured in @xmath53 . alfv@xmath35n mach number is generally an important parameter as the ratio of acceleration rates of the fermi i ( dsa ) and fermi ii ( second - order ) processes scales as @xmath54 @xcite . as the bulk post - shock flow is expected to be super - alfv@xmath35nic it is logical to conclude that the dsa is more rapid process . the corresponding analysis was successfully applied to the snrs w44 and ic443 . although @xcite could not reproduce the steep spectral index ( 0.55 ) in the case of snr 3c391 , using the spectral index value of 0.49 @xcite , this model could be justified ( for shock velocity of around 300 km / s and inter - clump density of around @xmath55 , as in ostrowski 1999 , this model could account for observable spectral index of 0.49 if magnetic induction is @xmath56 ; see equations 2.6 - 2.9 from ostrowski 1999 ) . it is worth mentioning that compression ratios higher than 4 ( not discussed in ostrowski 1999 ) would give , in this model , lower magnetic induction estimates . it is worth noting that snrs expanding in molecular clouds mostly evolve in the clumpy medium . the dominant part of the cloud mass is contained in the compact dense clumps filling only around 10% or less of the cloud volume and the rest of the cloud is filled with inter - clump gas of average number density of @xmath55 ( chevalier 1999 ; ostrowski 1999 ) . however , one must bear in mind that snr precursors could contribute to modification of physical conditions in the immediate pre - shock ism ( arthur 2012 ) . the fact that many snrs with flat spectral indices are also connected with molecular cloud environment could be a sign that models like @xcite are more appropriate in these cases . as the radiative shock has been revealed by the _ spitzer _ irs slit observation of snr 3c396 , that expands through the clumpy environment , model of ostrowski ( 1999 ) could possibly account for the observed flat radio spectral index ( hewitt et al . 2009 ; su et al . 2011 ) . for inter - clump density of around @xmath57 and for shock velocities of 120 km / s and 350 km / s ( hewitt et al . 2009 ) , magnetic induction of @xmath58 is needed , respectively , to account for observable radio spectral index of 0.34 ( see equations 2.6 - 2.9 from ostrowski 1999 ) . for a blast wave velocity of 870 km / s ( su et al . 2011 ) , needed value of magnetic induction is higher around @xmath59 . in the case of snr w28 , for shock velocity of around 80 km / s and inter - clump density of around @xmath60 ( rho & borkowski 2002 ) , model of ostrowski ( 1999 ) could account for observable spectral index of 0.35 if magnetic induction is @xmath61 . for higher inter - clump density of @xmath55 , needed magnetic induction would be around @xmath62 . finally , hb21 is known to be interacting with clumpy molecular clouds ( pannuti et al . 2010 and references therein ) . if we assume , approximately , 130 km / s for the velocity of radiative shell and the mean ambient density of @xmath63 ( koo et al . 2001 and references therein ) , model of ostrowski ( 1999 ) could account for observable radio spectral index of 0.36 if magnetic induction is @xmath64 . in all of these cases compression ratio of 4 was assumed . finally , it must be emphasized that there are several snrs , with flat spectral indices , that are generally associated with relatively lower densities such as g82.2 + 5.3 ( w63 ) , with @xmath65 ( higgs et al . 1991 ; mavromatakis 2004 ; gao et al . 2011 ) , and g166.0 + 4.3 ( vro 42.05.01 ) , with @xmath66 ( guo & burrows 1997 ; leahy & tian 2005 ; bocchino et al . 2009 ; gao et al . both snrs are classified as mixed - morphology ones . w63 is probably expanding in the wind blown bubble and vro 42.05.01 has probably broken out of the cloud within which it formed , expanded across an interstellar tunnel or cavity , and is now interacting with the material that forms the opposite tunnel wall ( pineault et al . 1987 ; burrows & guo 1994 ; guo & burrows 1997 ; mavromatakis 2004 ) . if the model of @xcite is acceptable for these snrs , then the observed spectral indices lead to the plasma @xmath23 around 0.17 and 0.05 for w63 and g166.0 + 4.3 , respectively , in the case of compression ratio of 3.9 and constant spatial diffusion coefficient . slightly higher values of plasma @xmath23 are obtained for compression ratio of 2.5 . if power - law index of the spatial diffusion coefficient dependence of momentum equals 0.67 , then observed spectral indices could not be obtained by the model of drge et al . ( 1987 ) . finally , if plasma @xmath42 than compression ratios of around 6.95 and 12.3 correspond to the observed spectral indices of w63 and g166.0 + 4.3 , respectively ( for constant spatial diffusion coefficient ) . these estimates will , on the other hand , lead to ( unrealistically ) low isothermal mach number , fully radiative snr shocks . the flatter spectra may also be due to compression ratios greater than four at fully radiative ( isothermal ) shocks ( in the post sedov - taylor phases of snr evolution ) . in the case of parallel isothermal shock , compression ratio @xmath8 is equal to a square of isothermal mach number ( @xmath67 ) so that test particle dsa gives : @xmath68 in figure 1 it is shown that this explanation , for observable flat spectral index range , @xmath69 , holds only for very low isothermal mach number fully radiative shocks ( isothermal mach numbers around 23 ) in the test - particle dsa . of course , one must bear in mind that the onset of the so called momentum - conserving snowplow phase occur only near the end of a snr lifetime , or not occur at all and even a pressure - driven snowplow model is just the asymptotic case @xcite . of course , local regions of the snr blast wave may leave the sedov - taylor phase earlier at those places where the ambient density is much higher than average , though the bulk evolution still corresponds to energy conserving phase @xcite . this means that different parts of an snr may be in different phases so that different shock velocities and compression ratios within the snr are possible , which could have repercussions on the integrated radio spectrum . it must be noted that the values of adiabatic index slightly less than @xmath43 ( due to high temperatures behind the shock front and/or because the molecules in the cloud , with which snr interacts , will cool more efficiently than the ambient medium or the ejecta as discussed in davis 2007 ) would reproduce some of the observable flat spectra in the case of parallel adiabatic shock wave and test - particle dsa ( see equation 2 ) . on the other hand , such values are generally not expected in the post - shock medium . finally , anand ( 2012 ) derived and discussed the jump relations across a shock in non - ideal gas flow . it is interesting to note that even for very small non - idealness of gas , using equation ( 5 ) from anand ( 2012 ) , for the compression ratio , flat spectral indices are obtained for the large range of shock velocities in the framework of dsa . of course , applicability of the non - ideal gas equation of state in the case of the particular ism plasmas is questionable . since there is a significant connection between the majority of galactic snrs with flat integrated radio spectrum and their detection in @xmath2-rays , the high energy properties of these snrs are now discussed ( see table 1 ) . apart from historical ( e.g. tycho and cassiopeia a abdo et al . 2010e ; giordano et al . 2012 ) and young tev - bright snrs ( e.g. rx j1713.7 - 3946 and vela jr . abdo et al . 2011 ; tanaka et al . 2011 ) , among the snrs recently detected with _ fermi _ lat , there is a huge class of snrs that interact with molecular clouds as well as of the evolved snrs without molecular cloud interaction ( e.g. cygnus loop , s147 katagiri et al . 2011 ; katsuta et al . 2012 ) . @xmath2-ray emission from snrs can be produced by few different , non - exclusive , mechanisms . electrons and positrons in the snr shell can interact by inverse compton scattering with ambient photon fields ( such as cosmic microwave background and infrared radiation ) to produce @xmath2-rays ( helder et al . 2012 ) . also , electron - atom or electron - molecule interactions in a dense medium result in bremsstrahlung radiation ( reichardt et al . mentioned mechanisms involve electrons accelerated up to the sufficient energies to produce the observed @xmath2-rays ( leptonic scenario ) . @xmath2-rays can also be produced by neutral pion decay , where neutral pions result from the collision of accelerated protons ( or heavier nuclei ) with nucleons of the ambient gas ( hadronic scenario ) . in the case of snrs that interact with molecular clouds , hadronic scenario is probable and one of the spectral signatures is steepening in the gev band . generally , absence of non - thermal x - ray emission favors hadronic origin of the observed @xmath2-ray emission . the rapid steepening of the spectrum above few gev is the kind of signature expected from the re - acceleration of pre - existing cosmic - rays , where high energy cosmic - rays escape from the snr confinement region ( reichardt et al . 2012 and references therein ) . uchiyama et al . ( 2012 ) emphasized that it has not been clear yet whether the @xmath2-ray emission is produced by escaping or by trapped cosmic - rays , since shocked molecular clouds inside an snr could also be the sites of efficient @xmath2-ray production . although the majority of snrs that interact with molecular clouds and are detected in @xmath2-rays are those with flat radio spectra ( e.g. w44 , ic443 , w28 , w49b , 3c391 , w51c , ctb37a , hb21 abdo et al . 2009 ; abdo et al . 2010abcd ; castro & slane 2010 ; reichardt et al . 2012 ) , it must be mentioned that several snrs that interact with molecular clouds and have spectral indices of 0.5 are also detected in @xmath2-rays with _ fermi _ lat ( e.g. g8.7 - 0.1 , g109.1 - 1.0 , g349.7 + 0.2 castro & slane 2010 ; ajello et al . 2012 ; castro et al . 2012 ) . reichardt et al . ( 2012 ) emphasized that snr hb21 belongs to the group of low - luminosity , gev - emitting snrs ( such as cygnus loop or s147 ) , which are clearly less luminous than the first gev - emitting snrs that were discovered ( such as w51c , ic443 , w49b ) . the spectral break is found at lower energy in hb21 than in the case of the luminous snrs . reichardt et al . ( 2012 ) suggested that the @xmath2-ray emission from hb21 can be understood as a combination of emission from shocked / illuminated molecular clouds , one of them coincident with the snr shell itself . snr g260.4 - 3.4 ( puppis a ) , with radio spectral index of 0.5 ( green 2009 ) , was also observed at @xmath2-rays ( hewitt et al . 2012 ) and it represents an interesting transitional case between young snrs still evolving into a circumstellar medium ( e.g. cas a ) , and older snrs which are interacting with large , dense molecular clouds ( e.g. ic 443 ) . in the case of kes 67 , not yet detected by _ lat , the bright pulsar wind nebula hess j1825 - 137 is extended near the position of the snr ( grondin et al . 2011 ) thus making the analysis highly complicated . similarly , near the position of snr 3c396 , a radio - faint @xmath2-ray pulsar ( psr j1907 + 0602 ) powering a bright tev pulsar wind nebula ( mgro j1908 + 06 ) is located , substantially complicating the analysis ( abdo et al . . possible interaction with molecular cloud in the case of snr 3c434.1 is indicated by jeong et al . ( 2012 ) , which makes this remnant an interesting target for analysis of possible @xmath2-radiation in its direction . on the other hand , there are no point sources included in the lat 2-year point source catalog ( nolan et al . 2012 ) closer than 2 degrees to the position of snr 3c434.1 . also , there are no extended sources closer than 10 degrees to the position of this remnant . a detailed study of the @xmath2-emission in the direction of snr 3c434.1 is planned for a work in the near future . @xcite analyzed @xmath2-ray emission from snrs interacting with molecular clouds . they showed that the simple re - acceleration of pre - existing cosmic - rays by the process of diffusive shock acceleration at a cloud shock is generally sufficient to power the observed @xmath2-ray emission through the decays of neutral pions . neutral pions are produced in hadronic interactions between high - energy protons ( nuclei ) and gas in the compressed - cloud layer . @xcite proposed that the radio emission may be additionally enhanced by the presence of secondary electrons / positrons , i.e. the products left over from the decay of charged pions , created due to cosmic - ray nuclei colliding with the background plasma . they concluded that presence of secondary electrons / positrons may also explain the flat spectral radio indices of some mixed - morphology snrs . @xcite applied their analysis to snrs w51c , w44 and ic443 . of course , this kind of explanation holds only for those snrs that interact with molecular clouds assuming further that hadron scenario for @xmath2-radiation is significant . another restriction of this model is the fact that @xcite used relations that are applicable in the case of the sedov - taylor phase which are not suitable for mixed - morphology snrs . on the other hand , the simple re - acceleration of pre - existing cosmic - rays and subsequent compression alone would not fully explain the @xmath2-rays associated with cloud - interacting snrs . it must be noted that gev and tev @xmath2-rays outside the southern boundary of snr w28 may be explained by molecular clouds illuminated by runaway cosmic - rays ( uchiyama et al . 2010 and references therein ) . @xcite also assumed pre - existing cosmic - rays in the cloud to have the same spectra as the galactic cosmic - rays in the vicinity of the solar system although the ambient cosmic - rays in the pre - shock cloud may deviate from the galactic pool due to the runaway cosmic - rays that have escaped from snr shocks at earlier epochs . another model by @xcite was also proposed assuming the possibility of electron acceleration by magnetohydrodynamical turbulence behind high density shocks ( vink 2012 ) . in their model , the calculated radio spectrum fits that observed from the snr ic443 shell because of the large shock compression ratio and to a lesser extent because of the effect of the second - order fermi acceleration . it is worth mentioning that this model involves a lepton scenario of @xmath2-ray emission from snrs interacting with molecular clouds . recently , it was shown that the observations over a very broad range of radio frequencies reveal a curvature in the spectra of some evolutionary older galactic snrs expanding in high density environment ( tian & leahy 2005 ; uroevi ' c & pannuti 2005 ; leahy & tian 2006 ; uroevi ' c , pannuti & leahy 2007 ; oni & uroevi 2008 ) . the ndsa effects could not be responsible for a characteristic shape of the integrated radio spectra in these cases . the presence of intrinsic thermal bremsstrahlung radiation was proposed as an explanation of the `` concave up '' radio spectrum ( uroevi ' c 2000 ; uroevi ' c et al . 2003ab ; oni et al . 2012 and references therein ) . presence of the cooled thermal x - ray electrons in post sedov - taylor snrs , higher ambient densities , interaction with or expansion in molecular cloud are main arguments used by the model that includes significant intrinsic thermal bremsstrahlung emission ( thermal model ) . detections of the low - frequency turn - overs , in the integrated radio spectrum , related with thermal absorption linked to snrs ( brogan et al . 2005 ; castelletti et al . 2011 ) are important in justification of this model as one can then predict the significance of thermal emission at higher frequencies . generally , linear polarization measurements ( that give the upper limits for the thermal component of the radio emission ) , possible detections in @xmath70 as well as the observations of radio recombination lines associated with snrs should accompany this kind of analysis . it must be emphasized that most of the potential targets for testing the thermal model hypothesis fall in the mixed - morphology class of snrs . because of the fact that these snrs appear to be in the radiative phases of their evolution , with shock velocities less than 200 km / s ( vink 2012 ) , the ensemble of cooled thermal x - ray electrons could exist , supporting the thermal model @xcite . in addition , recent discovery of strong radiative recombination continua ( rrc ) , seen in the x - rays of several snrs of this type , represents definite evidence that their plasma is recombining ( yamaguchi et al . rrcs in their x - ray spectra are indicative of higher densities in interiors of these remnants ( vink 2012 ) . thus , mixed - morphology snrs are characterized by a more / less uniform high interior density with medium hot temperatures , rather than very high temperatures and very low interior densities , as in the case of typical shell remnants . also , among the snrs with detected rrc , the most of them ( if not all ) have flat radio spectral index ( sawada & koyama 2012 ; uchida et al . 2012 ; vink 2012 ) . finally , as it was emphasized earlier , the morphology of these snrs is difficult to explain with standard snr evolutionary phases @xcite and several models were proposed in the current literature ( white & long 1991 ; cox et al . an x - ray emitting overionized plasma is the result of an early heating followed by a rapid cooling , but the mechanism responsible for such processes in mixed - morphology snrs is still not completely understood ( miceli 2011 ) . @xcite have recently analyzed the integrated radio spectra of 3 snrs ( ic443 , 3c391 , 3c396 ) for which the thermal model could be a natural explanation . all of the 3 snrs are characterized by flat spectral indices ( see table 1 ) based on a simple power - law ( synchrotron ) fit and all of them interact with molecular clouds . for ic443 and 3c391 the low - frequency turn - over is due to thermal absorption linked to the snrs and for snr 3c396 the nature of the observed low - frequency turn - over is not yet clear . it must be emphasized that most of the snrs that are the best targets for testing thermal model are also characterized by the flat spectral indices . the observed flat spectral indices could be just the apparent manifestation of the presence of thermal component . it must be noted that there are no predictions of curved `` concave up '' integrated radio spectra in the cases of the explanations presented in sections 24 for the flat radio spectra seen in some snrs . the shock wave dissociates molecules , or raises the gas temperature so that previously inaccessible degrees of freedom become accessible . the compression ratio for the parallel shock wave , including the possibility of jump in adiabatic index , is as follows : @xmath71 where @xmath72 and @xmath73 are adiabatic indices in upstream and downstream region , respectively . if @xmath74 ( diatomic molecules like @xmath75 ) and @xmath76 ( fully ionized gas ) it is easily seen that ( upstream ) adiabatic mach number @xmath77 leads to @xmath78 so that for weaker shocks the steeper spectra will be preferable . as the ambient gas is usually pre - ionized by radiative precursor , the similar conclusion ( for slightly smaller adiabatic mach numbers ) could be drawn from the simple relation for constant adiabatic index of @xmath43 ( described in section 1 using equation 2 , see figure 2 ) . this means that we expect that evolutionary old snrs , with velocities smaller than around @xmath79 , have steeper synchrotron spectral indices in the framework of test - particle dsa . this can partially account ( except high parameter uncertainties ) for steeper synchrotron spectral indices determined from the thermal plus non - thermal model fits @xcite . radio spectral index variations , both with frequency and location within several galactic snrs were detected ( frst & reich 1988 ; anderson & rudnic 1993 ; zhang et al . 1997 ; leahy & roger 1998 ; leahy et al . 1998 ; tian & leahy 2005 ; leahy 2006 ; ladouceur & pineault 2008 ) . in order to have spectral changes with position , physical conditions that alter the radio spectrum must change with position . of course , spatial radio spectral index variations are naturally expected in the case of composite snrs due to the presence of pulsar wind nebula . on the other hand , variations in @xmath5 are detected in both shell and mixed - morphology snrs . if the spectral index is changing with frequency , generally , there is a possibility that two different radiation mechanisms are producing the radio continuum emission . integrated radio spectrum with `` concave up '' curvature in the case of older snrs , where efficient ndsa process is not expected , can be produced by adding two different emission spectra along the same line of sight or within the same emitting volume ( leahy & roger 1998 ) . these kind of spectra could be due to different electron populations or because of the significant intrinsic thermal bremsstrahlung radiation @xcite . linear polarization measurements are of the great importance as spectral flattening associated with high density regions of an snr which do not show high linear polarization could be naturally explained by the thermal model . it is worth mentioning that there are some other processes that could shape the integrated radio continuum at higher frequencies ( between 10 - 100 ghz ) in several galactic snrs , such as the spinning dust emission ( draine & lazarian 1998 ; scaife at al . the spinning dust model will be responsible for a creation of a `` hump '' in the high - frequency radio spectrum ( see figure 5 from scaife at al . of course , as it was emphasized it the section 1 , present knowledge on integrated radio spectrum is generally unsatisfactory for firm quantitative analysis of the proposed models so as to the contribution of dust emission . more data at radio frequencies higher than 1 ghz , e.g. those from atca ( the australia telescope compact array ) , are expected to shed a light on the existence of the so called `` radio thermally active '' snrs as well as on the issue of the flat spectral indices . it must be mentioned that there is a general problem in observing at the higher radio continuum frequencies ( @xmath8030 ghz ) due to atmospheric absorption issues . ground - based radio astronomy is limited to high altitude sites such as in the case of alma ( atacama large millimeter array ) . except the supernova remnants g31.9 + 0.0 ( 3c391 ) , g39.2 - 0.3 ( 3c396 ) , g189.1 + 3.0 ( ic443 ) , analyzed in @xcite , that represent the best targets for justification of the thermal model , in this paper , the integrated radio spectra of snrs g6.4 - 0.1 ( w28 ) , g18.8 + 0.3 ( kes 67 ) , g34.7 - 0.4 ( w44 ) , g43.3 - 0.2 ( w49b ) , g89.0 + 4.7 ( hb21 ) , g94.0 + 1.0 ( 3c434.1 ) were also discussed . all of these snrs are associated with dense environment ( see table 1 ) . integrated flux densities are collected from the literature keeping in mind the criteria defined in section 1 . in the case of presence of spectral turn - over at low frequencies only data which are not influenced by the possible thermal absorption ( or maybe by synchrotron self - absorption ) were used in thermal model fits @xcite . simple sum of non - thermal and thermal components , represented by corresponding power - laws , ( see equation 18 and corresponding discussion in oni et al . 2012 ) is assumed . for frequencies in @xmath81 , the relation for the integrated flux density can be written as follows : @xmath82,\ ] ] where : @xmath83 and @xmath84 are flux densities corresponding to thermal and non - thermal components , respectively . weighted least squares fit ( influenced by the instrumental errors ) were applied . in the case of all of these six snrs thermal model fit is either not significant ( contribution of the thermal component at 1 ghz is near zero ) or the fit quality is not satisfactory ( and associated parameter errors exceed values of the corespondent parameters ) . pure non - thermal fit parameters for these snrs correspond to ones given in the present literature ( see table 1 ) . in the case of snr w28 , low - frequency spectral turn - over in the integrated spectrum is possible as flux density at 30.9 mhz has a rather low value in the sense of scatter from the power - law fit ( see figure 5 in dubner et al . the associated error for that flux density is 20% ( kassim 1989 ) . on the other hand , there is a great spread in flux density values , especially at 2.7 ghz , which prevents firm conclusions . the great spread in the integrated flux densities at the same frequencies is very good seen in the case of snrs w49b and w44 ( morsi & reich 1987 ; kassim 1989 ; taylor et al . 1992 ; kovalenko et al . 1994 ; lacey et al . 2001 ; castelletti et al . 2007 ; sun et al . 2011 ) . for snr w44 there is no apparent low - frequency spectral turn - over in the integrated radio spectrum , only the localized one , towards the southeast border of the snr , likely due to the free - free absorption from ionized gas in the post - shock region at the snr / molecular cloud interface ( castelletti et al . 2007 ) . detected turn - over at lower frequencies for w49b is likely due to the extrinsic free - free absorption by an intervening cloud of thermal electrons ( lacey et al . there is no apparent low - frequency spectral turn - over in integrated spectra of snrs kes 67 , hb21 and 3c434.1 ( reich et al . 1983 ; goss et al . 1984 ; landecker et al . 1985 ; kassim 1989 ; milne et al . 1989 ; tatematsu et al . 1990 ; kovalenko et al . 1994 ; dubner et al . 1996 ; zhang et al . 2001 ; reich et al . 2003 ; foster 2005 ; kothes et al . 2006 ; gao et al . 2011 ; sun et al . 2011 ) . in the case of other snrs that are listed in @xcite as those with flat spectral indices ( not plerions ) , expanding in high density environment , there are not enough data ( under criteria presented in section 1 ) for a detailed discussion . generally , there is a significant number of snrs listed in @xcite for which the spectral indices are determined only by two or three data points . some of the snrs that could be interesting targets for analysis , in future , when more reliable data at different frequencies are obtained , are g49.2 - 0.7 ( w51c ) and g290.1 - 0.8 ( msh 11 - 61a ) . a few words of caution regarding the justification of thermal model . snr s expansion in complex environment and possible association with regions ( and so possible thermal contamination ) makes the study more complicated . the best example is the case of steep radio spectrum snr g132.7 + 1.3 ( hb3 ) for which the intrinsic thermal radiation was dismissed due to the actual overlap between the snr and corespondent regions ( shi et al . 2008 ) . of course , the question on the presence of significant thermal radio emission from snr hb3 , in fact , still remains open as the possible region radiation contamination can not rule out the possibility of thermal bremsstrahlung emission from the snr itself , it can just mask it . final conclusion is that our present knowledge on integrated flux densities prevents us from possibility of strong justification of the thermal model and limits the discussion to just a few galactic snrs analyzed in @xcite . a considerable fraction of galactic snrs are characterized by the flat spectral indices ( @xmath0 ) . in this paper , known models are summarized and discussed : 1 . there are several explanations of the observable flat radio spectra of some snrs . most of the models involve a significant contribution of second - order fermi mechanism , but some of them discuss higher shock compressions , contribution of secondary electrons / positrons left over from the decay of charged pions , as well as the possibility of thermal contamination . 2 . in the case of expansion in high density environment , thermal bremsstrahlung could theoretically shape the spectrum of snrs . lack of more high quality data constraints the discussion only to several snrs . on the other hand , this model can naturally account to observable curved `` concave up '' radio spectra of some galactic snrs . new observations are expected to create a clear picture about the high frequency range ( 1 - 100 ghz ) of the radio continuum , as well as about the question of the significant contribution of intrinsic thermal bremsstrahlung radiation . with instruments like the jvla and atca , it should be far easier to measure the spectral indices in individual observations . 3 . the fundamental question about the origin of flat spectral indices still remains open . the more realistic picture would involve the action of more than one of the mentioned processes , in which some of them could be more or less prominent depending on the particular setup . since there is a significant connection between the majority of galactic snrs with flat integrated radio spectra and their detection in @xmath2-rays , as well as detection of rrc in their x - ray spectra , the analysis of high energy properties of these snrs will be potentially very important . abdo , a. a. , ackermann , m. , ajello , m. , baldini , l. , ballet , j. , barbiellini , g. , baring , m. g. , bastieri , d. , baughman , b. m. , bechtol , k. , and 167 coauthors : * 706 * , 1 ( 2009 ) abdo , a. a. , ackermann , m. , ajello , m. , allafort , a. , baldini , l. , ballet , j. , barbiellini , g. , bastieri , d. , bechtol , k. , bellazzini , r. , and 144 coauthors : * 718 * , 348 ( 2010a ) abdo , a. a. , ackermann , m. , ajello , m. , baldini , l. , ballet , j. , barbiellini , g. , baring , m. g. , bastieri , d. , baughman , b. m. , bechtol , k. , and 171 coauthors : sci * 327 * , 1103 ( 2010b ) abdo , a. a. , ackermann , m. , ajello , m. , baldini , l. , ballet , j. , barbiellini , g. , bastieri , d. , bechtol , k. , bellazzini , r. , bloom , e. d. , and 126 coauthors : * 722 * , 1303 ( 2010c ) abdo , a. a. , ackermann , m. , ajello , m. , baldini , l. , ballet , j. , barbiellini , g. , bastieri , d. , baughman , b. m. , bechtol , k. , bellazzini , r. , and 162 coauthors : * 712 * , 459 ( 2010d ) abdo , a. a. , ackermann , m. , ajello , m. , allafort , a. , baldini , l. , ballet , j. , barbiellini , g. , baring , m. g. , bastieri , d. , baughman , b. m. , and 162 coauthors : * 710 * , 92 ( 2010e ) abdo , a. a. , ackermann , m. , ajello , m. , baldini , l. , ballet , j. , barbiellini , g. , bastieri , d. , baughman , b. m. , bechtol , k. , bellazzini , r. , and 170 coauthors : * 711 * , 64 ( 2010f ) abdo , a. a. , ackermann , m. , ajello , m. , allafort , a. , baldini , l. , ballet , j. , barbiellini , g. , baring , m. g. , bastieri , d. , bellazzini , r. , and 154 coauthors : * 734 * , 28 ( 2011 ) ajello , m. , allafort , a. , baldini , l. , ballet , j. , barbiellini , g. , bastieri , d. , bechtol , k. , bellazzini , r. , berenji , b. , blandford , r. d. , and 139 coauthors : * 744 * , 80 ( 2012 ) anand , r. k. : * 342 * , 377 ( 2012 ) anderson , m. c. , rudnick , l. : * 408 * , 514 ( 1993 ) arthur , s. j. : * 281 * , 267a ( 2002 ) baars , j. w. m. , genzel , r. , paulinu - toth , i. i. k. , witzel , a. : * 61 * , 99 ( 1977 ) bandiera , r. , petruk , o. : * 419 * , 419 ( 2004 ) bell , a. r. : * 182 * , 147 ( 1978a ) bell , a. r. : * 182 * , 443 ( 1978b ) bell , a. r. , schure , k. m. , reville , b. : * 418 * , 1208 ( 2011 ) biermann , p. l. , strittmatter , p. a. : * 322 * , 643 ( 1987 ) bocchino , f. , miceli , m. , troja , e. : * 498 * , 139 ( 2009 ) bozzetto , l. m. , filipovi , m. d. , crawford , e. j. , bojii , i. s. , payne , j. l. , medik , a. , wardlaw , b. , de horta , a. y. : serb . j. * 181 * , 43 ( 2010 ) bregman , j. n. : * 288 * , 32 ( 1985 ) brogan , c. l. , lazio , t. j. , kassim , n. e. , dyer , k. k. : * 130 * , 148 ( 2005 ) burrows , d. n. , guo , z. : * 421 * , l19 ( 1994 ) bykov , a. m. : adv . space res . * 33 * , 366 ( 2004 ) bykov , a. m. , chevalier , r. a. , ellison , d. c. , uvarov , y. , a. : * 538 * , 203 ( 2000 ) castelletti , g. , dubner , g. , brogan , c. , kassim , n. e. : * 471 * , 537 ( 2007 ) castelletti , g. , dubner , g. , clarke , t. , kassim , n. e. : * 534 * , 21 ( 2011 ) castro , d. , slane , p. : * 717 * , 372 ( 2010 ) castro , d. , slane , p. , ellison , d. c. , patnaude , d. j. : * 756 * , 88 ( 2012 ) chen , y. , su , y. , slane , p. o. , wang , q. d. : * 616 * , 885 ( 2004 ) chevalier , r. a. : * 511 * , 798 ( 1999 ) cox , d. p. , shelton , r. l. , maciejewski , w. , smith , r. k. , plewa , t. , pawl , a. , rozicka , m. : * 524 * , 179 ( 1999 ) crawford , e. j. , filipovi , m. d. , de horta , a. y. , stootman , f. h. , payne , j. l. : serb . j. * 177 * , 61 ( 2008 ) davis , k. w. : ph.d . thesis , clemson university ( 2007 ) de horta , a. y. , filipovi , m. d. , bozzetto , l. m. , maggi , p. , haberl , f. , crawford , e. j. , sasaki , m. , uroevi ' c , d. , pietsch , w. , gruendl , r. , and 5 coauthors : * 540 * , 25 ( 2012 ) draine , b. t. , lazarian , a. : * 494 * , l19 ( 1998 ) drge , w. , lerche , i. , schlickeiser , r. : * 178 * , 252 ( 1987 ) dubner , g. m. , giacani , e. b. , goss , w. m. , moffett , d. a. , holdaway , m. : * 111 * , 1304 ( 1996 ) dubner , g. , giacani , e. , reynoso , e. , goss , w. m. , roth , m. , green , a. : * 118 * , 930 ( 1999 ) dubner , g. m. , velazquez , p. f. , goss , w. m. , holdaway , m. a. : * 120 * , 1933 ( 2000 ) foster , d. : * 1043 * , 1054 ( 2005 ) frst , e. , reich , w. : lnp * 316 * , 33 ( 1988 ) gabici , s. , casanova , s. , aharonian , f. a. , rowell , g. : proceedings of the annual meeting of the french society of astronomy and astrophysics * 313 g * ( 2010 ) gao , x. y. , han , j. l. , reich , w. , reich , p. , sun , x. h. , xiao , l. : * 529 * , 159 ( 2011 ) giordano , f. , naumann - godo , m. , ballet , j. , bechtol , k. , funk , s. , lande , j. , mazziotta , m. n. , rain , s. , tanaka , t. , tibolla , o. , uchiyama , y. : * 744 * , 2 ( 2012 ) giuliani et al . : * 516 * , l11 ( 2010 ) goss , w. m. , mantovani , f. , salter , c. j. , tomasi , p. , velusamy , t. : * 138 * , 469 ( 1984 ) green , d. a. : a catalogue of galactic supernova remnants ( 2009 march version ) , astrophysics group , cavendish laboratory , cambridge , united kingdom ( available at http://www.mrao.cam.ac.uk/surveys/snrs/ ) ( 2009 ) grondin , m .- h . , funk , s. , lemoine - goumard , m. , van etten , a. , hinton , j. a. , camilo , f. , cognard , i. , espinoza , c. m. , freire , p. c. c. , grove , j. e. , and 10 coauthors : * 738 * , 42 ( 2011 ) guo , z. , burrows , d. n. : * 480 * , l51 , ( 1997 ) heavens , a. f. , meisenheimer , k. : * 225 * , 335 ( 1987 ) helder , e. a. , vink , j. , bykov , a. m. , ohira , y. , raymond , j. c. , terrier , r. : * 173 * , 369 ( 2012 ) hewitt , j. w. , rho , j. , andersen , m. , reach , w. t. : * 694 * , 1266 ( 2009 ) hewitt , j. w. , grondin , m .- h . , lemoine - goumard , m. , reposeur , t. , ballet , j. , tanaka , t. : * 759 * , 89 ( 2012 ) higgs , l. a. , landecker , t. l. , israel , f. p. , bally , j. : j. roy . canada * 85 * , 24 ( 1991 ) jeong , i - g , byun , d - y , koo , b - c , lee , j - j , lee , j - w , kang , h. : * 342 * , 389 ( 2012 ) kassim , n. e. : * 71 * , 799 ( 1989 ) katagiri , h. , tibaldo , l. , ballet , j. , giordano , f. , grenier , i. a. , porter , t. a. , roth , m. , tibolla , o. , uchiyama , y. , yamazaki , r. : * 741 * , 44 ( 2011 ) koo , b - c , rho , j. , reach , w. t. , jung , j. , mangum , j. g. : * 552 * , 175 ( 2001 ) kothes , r. , fedotov , k. , foster , t. j. , uyanker , b. : * 457 * , 1081 ( 2006 ) katsuta , j. , uchiyama , y. , tanaka , t. , tajima , h. , bechtol , k. , funk , s. , lande , j. , ballet , j. , hanabata , y. , lemoine - goumard , m. , takahashi , t. : * 752 * , 135 ( 2012 ) kovalenko , a. v. , pynzar , a. v. , udaltsov , v. a. : astron . rep . * 38 * , 95 ( 1994 ) lacey , c. k. , lazio , t. j. w. , kassim , n. e. , duric , n. , briggs , d. s. , dyer , k. k. : * 559 * , 954 ( 2001 ) ladouceur , y. , pineault , s. : * 490 * , 197 ( 2008 ) landecker , t. l. , higgs , l. a. , roger , r. s. : * 90 * , 1082 ( 1985 ) leahy , d. a. , roger , r. s. : * 505 * , 784 ( 1998 ) leahy , d. a. , xizhen , z. , xinji , w. , jiale , l. : * 339 * , 601 ( 1998 ) leahy , d. , tian , w. : * 440 * , 929 ( 2005 ) leahy , d. a. : * 647 * , 1125 ( 2006 ) leahy , d. a. , tian , w. w. : * 451 * , 251 ( 2006 ) liu , s. , fan , z - h . , fryer , c. l. , wang , j - m . , li , h. : * 683 * , 163 ( 2008 ) mavromatakis , f. , aschenbach , b. , boumis , p. , papamastorakis , j. : * 415 * , 1051 ( 2004 ) milne , d. k. , caswell , j. l. , kesteven , m. j. , haynes , r. f. , roger , r. s. : publ . * 8 * , 187 ( 1989 ) miceli , m. : mem * 82 * , 709 ( 2011 ) moffett , d. a. , reynolds , s. p. : * 425 * , 668 ( 1994 ) morsi , h. w. , reich , w. : * 71 * , 189 ( 1987 ) nolan , p. l. , abdo , a. a. , ackermann , m. , ajello , m. , allafort , a. , antolini , e. , atwood , w. b. , axelsson , m. , baldini , l. , ballet , j. , and 227 coauthors : * 199 * , 31 ( 2012 ) oni , d. , uroevi ' c , d. : serb . j. * 177 * , 67 ( 2008 ) oni , d. uroevi ' c , d. , arbutina , b. , leahy , d. : * 756 * , 61 ( 2012 ) ostrowski , m. : * 345 * , 256 ( 1999 ) ostrowski , m. , schlickeiser , r. : * 268 * , 812 ( 1993 ) pannuti , t. g. , rho , j. , borkowski , k. j. , cameron , p. b. : * 140 * , 1787 ( 2010 ) paron , s. , ortega , m. e. , petriella , a. , rubio , m. , dubner , g. , giacani , e. : * 547 * , 60 ( 2012 ) patnaik , a. r. , hunt , g. c. , salter , c. j. , shaver , p. a. , velusamy , t. : * 232 * , 467 ( 1990 ) pineault , s. , landecker , t. l. , routledge , d. : * 315 * , 580 ( 1987 ) reich , w. , fuerst , e. , sieber , w. : iau symp . * 101 * , 377r ( 1983 ) reich , w. , zhang , x. , frst , e. : * 408 * , 961 ( 2003 ) reichardt , i. , de oa - wilhelmi , e. , rico , j. , yang , r. : * 546 * , 21 ( 2012 ) reynolds , s. p. : * 46 * , 89 ( 2008 ) reynolds , s. p. : * 336 * , 257 ( 2011 ) reynolds , s. p. , gaensler , b. m. , bocchino f. : * 166 * , 231 ( 2012 ) rho , j. , borkowski , j. : * 575 * , 201 ( 2002 ) rho , j. , petre , r. : * 503 * , l167 ( 1998 ) sawada , m. , koyama , k. : * 64 * , 81 ( 2012 ) scaife , a. , green , d. a. , battye , r. a. , davies , r. d. , davis , r. j. , dickinson , c. , franzen , t. , g@xmath35nova - santos , r. , grainge , k. , hafez , y. a. , and 11 coauthors : * 377 * , l69 ( 2007 ) schlickeiser , r. , frst , e. : * 219 * , 192 ( 1989 ) shi , w. b. , han , j. l. , gao , x. y. , sun , x. h. , xiao , l. , reich , p. , reich , w. : * 487 * , 601 ( 2008 ) su , y. , chen , y. , koo , bon - chul , zhou , x. , lu , deng - rong , jeong , il - gyo , delaney , t. : * 727 * , 43 ( 2011 ) sun , x. h. , reich , p. , reich , w. , xiao , l. , gao , x. y. , han , j. l. : * 536a * , 83s ( 2011 ) tanaka , t. , allafort , a. , ballet , j. , funk , s. , giordano , f. , hewitt , j. , lemoine - goumard , m. , tajima , h. , tibolla , o. , uchiyama , y. : * 740 * , 51 ( 2011 ) tatematsu , k. , fukui , y. , landecker , t. l. , roger , r. s. : * 237 * , 189 ( 1990 ) taylor , a. r. , wallace , b. j. , goss , w. m. : * 103 * , 931 ( 1992 ) tian , w. w. , leahy , d. : * 436 * , 187 ( 2005 ) tian , w. w. , leahy , d. a. , wang , q. d. : * 474 * , 541 ( 2007 ) troja , e. , bocchino , f. , reale , f. : * 649 * , 258 ( 2006 ) troja , e. , bocchino , f. , miceli , m. , reale , f. : * 485 * , 777 ( 2008 ) uchida , h. , koyama , k. , yamaguchi , h. , sawada , m. , ohnishi , t. , tsuru , t. g , tanaka , t. , yoshiike , s. , fukui , y. : * 64 * , 0141 ( 2012 ) uchiyama , y. , roger d. blandford , r. d. , funk , s. , tajima , h. , tanaka , t. : * 723 * , l122 ( 2010 ) uchiyama , y. , funk , s. , katagiri , h. , katsuta , j. , lemoine - goumard , m. , tajima , h. , tanaka , t. , torres , d. f. : * 749 * , l35 ( 2012 ) uroevi ' c , d. : ph.d . thesis , university of belgrade ( 2000 ) uroevi ' c , d. , duric , n. , pannuti , t. : serb . j. * 166 * , 61 ( 2003a ) uroevi ' c , d. , duric , n. , pannuti , t. : serb . j. * 166 * , 67 ( 2003b ) uroevi ' c , d. , pannuti , t. g. : astropart . * 23 * , 577 ( 2005 ) uroevi ' c , d. , pannuti , t. g. , leahy , d. : * 655 * , l41 ( 2007 ) vink , j. : * 20 * , 49 ( 2012 ) white , r. l. , long , k. s. : * 373 * , 543 ( 1991 ) xiao , l. , frst , e. , reich , w. , han , j. l. : * 482 * , 783 ( 2008 ) yamaguchi , h. , ozawa , m. , ohnishi , t. : adv . space res . * 49 * , 451 ( 2012 ) zhang , x - z , higgs , l. a. , landecker , t. l. , wu , x - j , qian , s - j . : * 1 * , 443 ( 2001 ) zhang , x. , zheng , y. , landecker , t. l. , higgs , l. a. : * 324 * , 641 ( 1997 ) zhou , x. , miceli , m. , bocchino , f. , orlando , s. , chen , y. : * 415 * , 244 ( 2011 )

brown dwarfs are self - gravitating objects with physical and atmospheric properties intermediate between stars and planets . with masses below @xmath10.075 m@xmath2 depending on age and metallicity . see @xcite for an extensive discussion of evolutionary models . ] @xcite , these objects can not sustain hydrogen fusion , and hence cool and dim as they age , radiating primarily at infrared wavelengths . the evolution of their spectra spans the spectral classes m , l , t , and y , with transitions demarcated by the appearance and disappearance of absorption lines and bands as molecules form and condense out of their atmospheres at different temperatures and pressures ( and references therein ) . despite having a basic understanding of their evolution , brown dwarf formation remains an open question . standard jeans collapse of molecular clouds requires high densities so that gravity can overcome thermal pressure . once the collapse has begun , halting the accretion becomes problematic . several mechanisms have been proposed to resolve this issue , including turbulent fragmentation of protostellar clouds @xcite , fragmentation of pre stellar disks @xcite , ejection by dynamical interactions with other protostars @xcite , and photoerosion of prestellar cores . in principle , these formation mechanisms should leave traces on the statistical properties of brown dwarfs , including the occurrence of multiple systems and distributions of their separation , relative masses and eccentricity . observationally , it has been shown that multiplicity increases with primary mass , even at the lower mass end of the main sequence , with the g dwarf binary fraction being 57@xmath3 higher than that for m dwarfs @xcite . current estimates of the binary fraction of very low mass ( vlm ) late - m to t dwarfs ( vlm @xmath4 ) are @xmath5 , with a peak in separation at @xmath6 4 au and a mass ratio distribution peaking at nearly equal masses @xcite . however , these multiplicity statistics have been largely determined from resolved imaging programs , sampling separations greater than 3 au . @xcite pointed out that the current peak in the binary angular separation distribution is coincident with the resolution limit of hst and ground - based adaptive optics ( ao ) imaging , indicating that tight ( @xmath7au ) vlm binaries could be undercounted . likewise , @xcite and @xcite report a higher unresolved binary fraction ( @xmath8 ) based on overluminous binary candidates in color - magnitude plots . conversely , spectroscopic radial velocity ( rv ) studies find binary fractions of 2.5@xmath3 in systems separated by @xmath7 au @xcite and @xmath9 up to 3 au . for the 0 - 6 au range , @xcite estimate a binary fraction of 26@xmath10 . however , the difficulty of obtaining high resolution spectra of faint vlm dwarfs results in small sample size . since total binary fractions for vlm stars and brown dwarfs could range between @xmath11 , it is imperative to constrain this statistic to make conclusions about brown dwarf formation . an alternative method for detecting tight unresolved binaries , developed by @xcite , involves identifying blended light pairs , or spectral binaries . we will refer as _ spectral _ binaries to those objects whose combined - light spectrum shows distinct peculiarities that come from the highly structured spectra of single m , l and t dwarfs when blended together , as opposed to _ spectroscopic _ binaries which are binaries that show rv variations . the first brown dwarf spectral binary , 2mass j05185995@xmath122828372 , was serendipitously identified by @xcite based on its hybrid characteristics containing features of both l and t dwarfs . the superposition of l plus t dwarf spectra proved to be the simplest model of its peculiar spectrum and it was later resolved as a binary using the hubble space telescope @xcite . the unusually blue l dwarf sdss j080531.84 + 481233.0 was next identified as a spectral binary with l4.5 and t5 components by @xcite , based on a peculiar methane absorption band starting at @xmath13 m , and was later confirmed as an astrometric variable by @xcite . a third system , 2mass j03202839@xmath120446358 , was concurrently identified as an unresolved m9+t5 spectral binary @xcite and a rv variable with an orbital period of eight months @xcite . these examples serve to illustrate how spectral binaries can encompass a broad range of system architectures . to date , 34 vlm spectral binaries and candidates have been reported ( see table [ tab : sb ] ) , and ten have been confirmed by direct imaging , rv or astrometric variability (; gelino et al . in prep ) . detecting binaries using the spectral binary method is particularly useful for multiplicity statistics , as the method is independent of separation within 0@xmath145 , which translates to @xmath1510 - 20 au for field brown dwarfs at distances of 20 - 40 pc . the closest separation pairs can be followed - up to measure orbits and component masses , as well as infer ages by comparison to evolutionary models @xcite . systems with independent age constraints can also be used to test the evolutionary models directly @xcite . additionally , unresolved binaries are strong contaminants in luminosity functions that later lead to uncertainties in mass functions and studies of formation history through stellar populations @xcite , so their identification is extremely important . finally , spectral binaries with late - m / early - l primaries and t dwarf secondaries can straddle the hydrogen burning limit , thus giving additional insight into brown dwarf evolution . in this paper we adapt the technique of @xcite to search for spectral binaries composed of late - m or early - l dwarf primaries with t dwarf secondaries . m dwarfs are the most common stars in the galaxy @xcite , and are the brightest vlm objects , enabling better statistics through larger magnitude - limited search volumes and sample sizes . m - dwarf spectra are also intrinsically distinct from t - dwarf spectra , but differ in brightness by several magnitudes , rendering peculiar features extremely subtle . in [ sec : obs ] we describe our spectral sample used to find spectral binaries , drawn from the spex prism libraries and new observations . in [ sec : mtbin ] we explain our two methods to identify spectral binary candidates : by visual examination ( [ sec : visual ] ) and through spectral indices ( [ sec : specind ] ) . in [ sec : specfit ] , we perform single and binary template fitting to identify fourteen binary candidates . in [ sec : candidates ] , we describe the properties of the candidates . in [ sec : blue ] and [ sec : separation ] , we discuss our major contaminant , blue l dwarfs , and show preliminary evidence that the separations of spectral binaries are tighter than the resolved population . our results are summarized in [ sec : summary ] . the spex prism library is composed of low resolution ( @xmath16 ) spectra acquired with the spex @xmath17 m spectrograph , mounted on the 3.0 m nasa infrared telescope facility ( irtf ) , located in mauna kea , hawaii @xcite . all spectra were obtained using the prism - dispersed spex mode , which continuously samples wavelengths between @xmath18 m at a dispersion of @xmath19 pixel@xmath20 . the library includes close to 2,000 sources , both previously published data and 530 new spectra acquired between november 2000 and december 2013 ( table [ tab : newobs ] ) . the new observations were obtained with the 0@xmath145 or 0@xmath147 slit , generally aligned with the parallactic angle . total integration times ranged between 360 s and 1200 s , depending on source brightness and atmospheric conditions , and were obtained in an abba dither patter along the slit . spectra of nearby a0 v stars were used to flux calibrate the raw spectra and correct for telluric absorption . internal flat fields and argon arc lamps were observed with each flux standard for pixel response and wavelength calibration . all data were reduced with the spextool package @xcite using standard settings . a detailed description of our reduction procedures is given in @xcite . the sources observed have optical and/or near - infrared spectral classifications reported in the literature . to obtain a self - consistent set of spectral types , we computed spex spectral types based on spectral indices , following the method described in @xcite . from these , we selected two samples : the `` candidate '' sample which has been purged of spectral types outside the m7-l7 range , optical subdwarfs , giants , and poor quality spectra , but keeping binaries , objects suspected of being binaries from previous studies , young objects and unusually red and blue dwarfs , and the `` template '' sample which has been purged of binaries , candidate binaries , giants and poor quality spectra ( as determined by visual inspection ) only . the `` candidate '' sample contains 815 spectra of 738 objects with spex spectral types between m7-l7 , as those would be the potential primaries for late - m / early - l plus t binaries . the `` template '' sample comprise 1110 spectra of 992 single sources whose spectral types range between m7-l7 for primaries and t1-t8 for secondaries used in spectral fitting . the distribution of spectral types for both samples is shown in figure [ fig : sample ] . in both samples , the number of spectra decreases toward later spectral types due to declining space densities for l dwarfs @xcite and sensitivity limits for late l and t dwarfs . since there are significantly more sources with late - m spectral types in our samples , it is more likely to find binaries with a late - m primary . the sources included were observed as part of several different programs , including our ongoing program to compile a magnitude - limited sample of l dwarfs ( burgasser et al . in prep . ) . as such , we do not claim the sample to be complete or unbiased . the spectral morphology of unresolved late - m / early - l plus t dwarf binary systems gives rise to a distinctive feature in blended - light spectra : a small `` dip '' centered at 1.63 @xmath21 m , which is the combination of ch@xmath22 absorption from the secondary and feh from the primary @xcite . methane does not exist in the spectra of late - m / early - l dwarfs , so its presence indicates a t dwarf companion . however , this feature is very weak in blended - light spectra since a t dwarf is significantly fainter than the m / l primary ( e.g. @xmath23 3.5 mag between an m8 and a t5 , which is the case for 2mass j03202839@xmath120446358 ) . moreover , variations in the spectral slope for a blue or red l dwarf , can make this feature ambiguous , as can poor correction of hydrogen lines in the a0v calibrators . alternative indicators such as a relatively higher flux around the 1.25 @xmath21 m peak and * an * inflated bump short ward of 2.2 @xmath21 m , may also reveal the presence of a t dwarf companion , or that the spectrum of the source is unusually blue . to facilitate our visual inspection , we fit the candidate sample to templates of single objects , following the same chi - squared minimization routine as in section [ sec : specfit ] , and then subtracted the median combination of the ten best fitting single sources from each spectrum . the objects with residuals consistent with a t dwarf spectrum were selected as visual candidates . to validate this procedure , we also performed the same template subtraction on four confirmed spectral binaries : sdss j000649.16@xmath12085246.3 , 2mass j03202839@xmath120446358 , sdss j080531.89 + 481233.0 , and 2mass j13153094@xmath122649513 ( see table [ tab : sb ] ) . the residuals from these subtractions clearly exhibited t dwarf - like morphologies . twelve sources were selected as visual candidates . in addition to visual inspection , we also used spectral indices to identify additional candidate binaries due to the subtlety of t dwarf features in combined - light spectra @xcite . we initially examined standard classification indices from @xcite , as well as the `` _ _ h__-dip '' index from @xcite , and further defined five new indices . the new indices were designed by comparing the residuals of the four known binary spectra after subtracting their best single template fits . as a control sample , we also examined single templates subtracted from each other , which showed no evidence for a t dwarf companion . the new spectral indices specifically designed in this paper are : * _ h_-bump : measures the peak in the continuum from the t dwarf in the _ h _ band relative to the dip centered around 1.63 @xmath21 m seen in m+t binaries , making this index complimentary to _ h_-dip . a higher value of _ h_-bump implies a larger flux at 1.55 @xmath21 m , possibly caused by the presence of a t dwarf . * _ j_-curve : designed to detect the flux coming from both the 1.05 @xmath21 m and 1.27 @xmath21 m peaks of a t dwarf , as compared to the deep methane absorption at 1.12 @xmath21 m . * _ j_-slope and _ k@xmath24_-slope : measure the slope of the peaks in the _ j _ and _ k@xmath24 _ bands at 1.27 @xmath21 m and 2.10 @xmath21 m . in both cases , the peaks in a single late - m / early - l should look somewhat flat , giving values close to one , whereas in a late - m / early - l plus t dwarf binary the slope of the _ j _ and _ k@xmath24 _ band peaks are slightly negative and positive , respectively . * h@xmath25o-_y _ : measures the prominence of the _ y_-band peak of the t dwarf at @xmath26 m compared to the water and methane absorption around @xmath27 m . m and l dwarfs do not present peaks in the _ y_-band . the thirteen indices examined are described in table [ tab : index ] . we also used _ j - k@xmath24 _ , _ j - h _ and _ h - k@xmath24 _ colors synthesized from the spectra themselves , and the source spectral type , for a total of seventeen parameters . comparing all seventeen parameters against each other yielded 136 pairings . after visual examination to determine which pairings best segregated the four known m / l+t binaries , twelve combinations were selected ( figure [ fig : paramspace ] ) . we then used two techniques to define regions of interest in each combination for candidate selection . if a trend among all sources was clear , we fit the points to a second order polynomial and defined a region demarcated in the y - axis by the + 1@xmath28 or -1@xmath28 curves from the fit function , and in the x - axis by the horizontal spread of the binary benchmarks . conversely , if the points did not indicate any trends , then the region was demarcated such that it included the four binary benchmarks . the limits to these regions are described in table [ tab : selectreg ] . objects falling in eight or more selection regions were considered strong index candidates , those falling in four to eight regions were considered weak index candidates ( figure [ fig : histogram ] ) . the number of selected sources rises steadily below four or five combinations , suggesting that sources selected fewer than four times could be spurious . three of our benchmarks were selected by all twelve combinations , while sdss j0006@xmath120852 missed only the spt / ch@xmath22-h cut , since it falls within one standard deviation from the fitting curve . in total , eight strong and twenty - two weak candidates were selected , including the previously identified spectral binaries 2mass j20261584@xmath122943124 @xcite and 2mass j13114227 + 3629235 @xcite . seven visual candidates overlapped with the index candidates : five as strong and two as weak . to statistically test the binary hypothesis for our visual and index - selected candidates , we compared each spectrum to templates of both single sources and binary systems , using the method described in @xcite . the candidates determined by visual inspection or spectral index selection were first rejected from the template pool . then , all spectra were interpolated onto a common wavelength scale from @xmath29 to @xmath30 m and normalized to the peak flux between @xmath31 m . each candidate spectrum @xmath32 $ ] was directly compared to all single templates @xmath33 $ ] and ranked by a weighted chi - squared statistic . @xmath34\left[\frac{c[\lambda]-\alpha t [ \lambda]}{\sigma_c [ \lambda]}\right]^2\ ] ] where @xmath35 $ ] is a vector of weights proportional to the wave band size of each pixel ( see @xcite ) , @xmath36 is a scaling factor minimizing @xmath37 and @xmath38 $ ] is the noise spectrum for each candidate . the statistic was computed over the wavelength range : @xmath39 m , @xmath40 m and @xmath41 m , avoiding regions of strong telluric absorption . binary templates were constructed by first scaling each template spectrum to absolute fluxes using the 2mass m@xmath42 versus spectral type relation of @xcite , and then combining all pairs of single templates , such that the spectral type of the primary was earlier than that of the secondary resulting in a total of @xmath43 binary templates . more specifically , the primary spectral type was fixed to lie between m7-l7 while the secondary spectral type ranged between t1-t8 , since types earlier than t1 do not evidence strong methane features yet . the best binary fits were ranked using a chi - squared minimization routine . we determined the true significance that a binary template is superior to a single template by comparing the @xmath37 distributions of the binary and single fits using the one - sided f - test statistic @xmath44 : @xmath45 here , dof is the degrees of freedom for each fit ( equation 2 in @xcite ) . candidates with an f - statistic falling under the @xmath46 confidence level were rejected , including five visual candidates . in particular , 2mass j14493784 + 2355378 and 2mass j14232186 + 6154005 ( also a weak index - selected candidate ) , two previously identified spectral binary candidates from @xcite and @xcite were rejected due to their low confidence level that the binary fit was statistically better than the single fit . since our template sample includes a wide range of objects such as young and unusually blue and red dwarfs , the peculiarities of these candidates may be better explained by factors other than unresolved binarity . one exception to the index selection was 2massi j1711457 + 223204 whose spex spectral type was too late to be included in the candidate sample , yet it was a visual candidate and passed the binary fit f - test . figures [ fig : strongfit ] and [ fig : weakfit ] show the best single ( left ) and binary ( right ) template fits to our strong and weak candidates . table [ tab : cand ] summarizes the results of these fits . upon further visual examination , some binary fits still proved unsatisfactory . this was the case for the blue l dwarfs 2mass j11181292@xmath120856106 , 2mass j14162409 + 1348267 , 2mass j15150083 + 4847416 , 2mass j17114558 + 40285779 , and the subdwarfs 2mass j03303847@xmath122348463 , 2mass j03301720 + 3505001 , 2mass j04024315 + 1730136 , 2mass j15412408 + 5425598 and 2mass j23311807 + 4607310 . section [ sec : blue ] discusses these issues in more detail . as a result , fourteen candidates have been recognized , of which eleven are newly identified . in an effort to balance the tradeoff between fidelity of binary candidates and completeness , we are leaning towards the former . our binary selection criteria are conservative and it is likely that other spectral binaries may be identified with slightly looser constraints . in summary , from the @xmath6800 sources compiled in the candidate sample , twelve were selected by visual inspection and thirty were selected by spectral indices . seven sources overlapped the results of these selection methods . after fitting all thirty - five candidates , seventeen were rejected due to their confidence level being lower than @xmath46 , and four more due to their unusually blue colors ( see section [ sec : blue ] ) , leading to a final count of fourteen : six strong , seven weak and one visual candidate not overlapping with the index - selected . labels of strong and weak candidates come from index selection . originally discovered by @xcite , 2mass j0236 + 0048 was classified as an l6 in the optical and l6.5 in the infrared by @xcite . in their study , @xcite comment that this object may belong to the pleiades moving group , given its proper motion of @xmath47 = [ 161.33\pm10.10 , 176.33\pm19.16]$ ] mas yr@xmath20 and agreement between photometric and moving group distances at d = 26 pc . however , @xcite reclassified this object as an l9 , reducing its spectroscopic distance to 18 pc while its strong feh band at 0.99@xmath21 m argues against low surface gravity @xcite . nevertheless , the spectrum of this source does not show any signs of youth @xcite . 2mass j0236 + 0048 is selected by eleven out of twelve spectral index combinations , and its binary fit is significantly better than its single fit , making this a strong binary candidate with l5.0@xmath480.6 and t1.9@xmath481.1 components . @xcite discovered sdss j0931 + 2802 in the sdss catalog and classified it as an l3 at a mean distance of 29@xmath489 pc . its spectrum shows excess flux in the _ j_-band at @xmath49 m and a noticeable dip in the _ h_-band in the vicinity of @xmath50 m , as expected for a t dwarf component . this source was selected as a visual candidate , and by eleven out of twelve spectral index combinations , and our spectral fitting predicts component types of l1.4@xmath480.1 and t2.6@xmath480.9 . identified as a brown dwarf candidate by , 2mass j1311 + 3629 is a peculiar l5 . while also classified as unusually blue in wavelengths longer than _ j _ band @xcite , it lacks evidence of low metallicity or h@xmath25 collision - induced absorption ( cia ) in _ h _ and _ k _ bands . @xcite identified the methane feature in the _ h _ band centered around 1.63 @xmath21 m suggesting unresolved binarity . in this study , it was selected by eleven spectral index combinations and also as a visual candidate due to its methane absorption band starting at 1.60 @xmath21 m . template fitting gives spectral types of l4.8@xmath480.6 and t4.1@xmath482.7 . 2mass j1341@xmath123052 was discovered by @xcite and classified as an l3 in the optical by @xcite , who also measured its parallax and distance ( @xmath51 pc ) . 2mass j1341@xmath123052 was selected by eight spectral indices , and its spectral fitting suggests component spectral types of l1.2@xmath480.3 and t6.3@xmath481.0 . 2mass j1453 + 1420 was classified as an l1 in both the infrared @xcite and the optical @xcite , where it clearly shows excess flux in the _ j _ band and a dip in the _ h _ band . it is selected by eleven out of twelve spectral index combinations , and it is slightly blue with a _ j - k@xmath24 _ color of 1.18@xmath480.05 as compared to the median for * the * l1 spectral type 1.34@xmath480.19 @xcite . it is best fit by l1.1@xmath480.0 and t6.0@xmath481.1 components . 2mass j2026@xmath122943 had already been identified as a spectral binary candidate by @xcite , but it failed to be resolved by keck ao , thus setting an upper limit in separation of 0@xmath1425 or a projected separation of 9 au at a distance of 36@xmath485 pc @xcite . this source clearly shows a dip in its spectrum centered at 1.63 @xmath21 m , and it is best fit by a combination of l1.0@xmath480.5 and t5.8@xmath481.0 components . 2mass j0206 + 2235 was discovered and classified as an l5.5 by @xcite , and characterized as a blue l dwarf by @xcite . it was selected by seven spectral index combinations and fit to l5.1@xmath480.5 and t3.2@xmath482.3 components . 2mass j0427@xmath121127 was discovered and classified as an m7 by . it was selected by five spectral index combinations and best fit by m7.4@xmath480.2 and t5.1@xmath481.5 components . 2mass j1036@xmath123441 was classified as an l6 @xcite at a distance of 21@xmath483 pc . it almost made the cut for a strong candidate , since it was selected by seven spectral index combinations . this source was best fit by components with l5.2@xmath480.4 and t1.4@xmath480.4 spectral types . despite not having a pronounced methane absorption feature centered at 1.63 @xmath21 m , the binary fit is significantly better than the single fit , especially at the _ j _ band peak . 2mass j1059@xmath122113 is an l1 @xcite at a distance of 32.1 @xmath482.2 pc . this source was selected by four spectral index combinations and its best binary fit yields components with l0.6@xmath480.4 and t3.4@xmath481.3 spectral types . its spectrum shows a strong absorption feature centered at 1.63 @xmath21 m , as well as a flux excess at 1.23@xmath21 m and 2.20@xmath21 m . sdss j1422 + 2215 was identified and classified as an l6 in the nir by @xcite and also as an unusually blue l dwarf , showing strong h@xmath25o and feh absorption bands , which may be due to subsolar metallicity and/or thinner condensate cloud decks . it was selected by four out of twelve spectral index combinations with most likely component spectral types of l4.2@xmath480.6 and t4.1@xmath482.3 . wise j1623@xmath120508 was classified as an l1 in the nir @xcite . this source was selected by four spectral index combinations and best fit by l0.6@xmath480.3 and t6.0@xmath480.3 components . 2mass j1707@xmath120138 was discovered and classified as an l2 by . selected by five spectral index combinations , its spectrum is best fit by components with l0.7@xmath480.5 and t4.3@xmath482.0 spectral types . its spectrum shows a strong absorption feature centered at 1.63@xmath21 m . 2mass j1711 + 2232 was first identified and classified as an l6.5 in the optical by @xcite . due to its feh and ch@xmath22 absorption features in the _ h _ band , @xcite suggested it could be a spectral binary with l5.0 and t5.5 components . we find slightly different component spectral types of l1.5@xmath480.6 and t2.5@xmath481.0 , yet this source was not selected by spectral indices because of its late spex spectral type of l8.8 . despite having been imaged with hst / wfpc , it remains unresolved @xcite . four of the candidates selected by spectral indices were rejected after spectral fitting due to their poor binary fits . when we investigated these sources in detail , we found they were classified as blue objects in the literature and/or showed an unusually blue spectrum . 2mass j11181292@xmath120856106 was classified as a metal - poor subdwarf by @xcite . sdss j141624.09 + 134826.7 is part of a resolved binary system with a t7.5 companion that is itself a blue outlier . @xcite rejected unresolved binarity for the primary based on a qualitative comparison to the unusually blue l dwarf 2mass j11263991@xmath125003550 . the l6 2mass j15150083 + 4847416 shows a stable rv of -29.97@xmath480.14 km s@xmath52 @xcite and no signs of binarity from its spectrum . finally , 2mass j17114558 + 40285779 was discovered by @xcite as an unusually blue wide companion to the k star g203 - 50 . they discuss the possibility that the object may be unusual due to unresolved binarity , but argue in favor of low metallicity . for all of these sources , the lack of single templates akin to blue objects resulted in statistically better binary fits , yet the match is still relatively poor around the 1.63 @xmath21 m methane absorption feature . a few more previously unidentified nir subdwarfs were also selected as weak candidates and subsequently rejected due to their poor binary fits . the best binary fits for 2mass j03303847@xmath122348463 , 2mass j03301720 + 3505001 , 2mass j04024315 + 1730136 , 2mass j15412408 + 5425598 , and 2mass j23311807 + 4607310 use another subdwarf as a primary , which again indicates that they are part of a rare blue population that has a short supply of examples in this sample . m+t binaries have slightly bluer spectra caused by the extra flux in the _ j _ band corresponding to the peak in the t dwarfs . particularly , some sources originally classified as unusually blue have been later identified as spectral binaries ( e.g. sdss j0805 + 4812 @xcite ) . in contrast , intrinsically blue l dwarfs have low metallicity , thin cloud coverage , large - grain clouds or a combination of these , causing a blue tilt to the nir spectrum @xcite . @xcite have defined red and blue photometric outliers as the objects whose _ j - k@xmath24 _ color placed them 2@xmath28 or 0.4 mag away from the average for their spectral type , while pointing out the difficulty to distinguish outliers beyond a spectral type of l9 due to the small sample of objects . figure [ fig : j - k ] shows the _ j - k _ colors for our sample as compared to their spectral types , including the median and @xmath53 lines as calculated from the sample ( solid lines ) and reported in the literature ( dashed lines ) by @xcite and @xcite for samples of m and l dwarfs , respectively . figure [ fig : j - k ] suggests that blue l dwarfs are a major contaminant in our sample since a significant fraction of both known binaries and candidates have similar colors and thus lie in the same region as blue sources . we conclude that the blue l dwarf contaminants can be recognized if rejected due to their poor fits to binary template spectra . true confirmation of our candidates requires observational follow - up to either resolve the systems or measure rv or astrometric variability . as noted in the introduction , spectral binaries can be used to devise an unbiased method to measure the vlm binary separation distribution . therefore , it is worth examining the separation distribution of vlm and brown dwarf spectral binaries confirmed to date , to see if there are any differences compared to the resolved population . figure [ fig : separation ] shows the distribution of projected separations from 122 confirmed vlm binaries . among the observational methods for detecting binaries , such as direct imaging , radial velocity variations , astrometric variations , and microlensing , direct imaging has proven to be the most successful so far ( 73@xmath3 of confirmed vlm binaries ) , but its biggest drawback is its limit in resolution . at minimum angular scales of 0@xmath141 - 0@xmath142 for ao and hst programs , and typical distances of field brown dwarfs of 20 - 30 pc , telescope sensitivity reaches its limit at separations of around 2 - 6 au . at 2.90 au , the mean projected separation of eight independently - confirmed spectral binaries plotted in figure [ fig : separation ] falls at the lower end of this sensitivity limit , at less than the mean of known vlm binaries excluding the spectral binaries ( 3.75 au ) , raising the possibility that there may be significantly more tightly bound systems . to assess whether this is a significant difference , we performed a two - sample kolmogorov - smirnov test comparing the projected separation distributions of all binary systems to the confirmed spectral binaries . specifically , the distributions were constrained in angular separation to 50 - 500 mas , where the lower limit corresponds to the smallest possible imaging resolution in good seeing , while the upper limit restricts the maximum size of the slit . in addition , the distance was constrained to less than 30 pc , since objects that are farther away would be more difficult to confirm as binaries . in this way , we intend to fairly compare the spectral binary method to the other available methods for binary detection . these constraints reduced the number of spectral binaries to six . the result was a d statistic of 0.41 and a probability of @xmath54 . while the low probability is indicative of a difference between the samples , the small sample size makes this statistic inconclusive . many more of the existing spectral binaries need to be characterized before a significant difference can be confirmed or ruled out . we have identified fourteen brown dwarf binary candidates with late - m / early - l plus t dwarf components based on visual inspection of low resolution data , and analysis with spectral indices and template fitting . we combined five new spectral indices , with previously defined ones , spectral type , and _ j - h _ , _ h - k@xmath24 _ , and _ j - k@xmath24 _ colors to define pairings that effectively select spectral binary candidates , and confirmed them by comparison to both single and binary template spectra from the spex prism library . unusually blue l dwarfs were the main contaminant of this analysis , with four candidates classified as unusually blue but nonetheless being poorly matched to binary spectra . exploring the separation distribution of binary systems we find suggestive evidence that spectral binaries are more closely separated than other binaries , but the confirmed sample is too small to be conclusive . we are now undertaking follow - up ao imaging and rv monitoring of these candidates to confirm them and measure orbital properties . the authors thank telescope operators for their assistance during our observations . dbg would like to thank the friends of the international center at ucsd for their generous scholarship as well as davy kirkpatrick and fellow graduate students alex mendez and david vidmar for their helpful discussion and coding tips . this publication makes use of data from the spex prism spectral libraries , maintained by adam burgasser at http://www.browndwarfs.org/spexprism ; the dwarf archives compendium , maintained by chris gelino at http://dwarfarchives.org ; and the vlm binaries archive maintained by nick siegler at http://vlmbinaries.org . the authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of mauna kea has always had within the indigenous hawaiian community . we are most fortunate to have the opportunity to conduct observations from this mountain . + _ facilities : _ irtf ( spex ) . lccccccccccc sdss j000649.16 - 085246.3 & m9 & & m8.5@xmath480.5 & t5@xmath481 & 14.14@xmath480.04 & 1.01@xmath480.05 & 3.15@xmath480.31 & 0.29@xmath480.01 & rv & 2 ; 2 + ulas j004757.41 + 154641.4 & & t2.0@xmath482.0 & l8.0 & t7.0 & 17.83@xmath480.05 & 1.41@xmath480.07 & & @xmath55 & & 22 + 2mass j00521232 + 0012172 & l5 & & l4 & t3 & 16.36@xmath480.11 & 0.90@xmath480.19 & & @xmath56 & & 24 + sdss j011912.22 + 240331.6 & & t2 & t0@xmath480.7 & t4@xmath480.4 & 17.02@xmath480.18 & @xmath57 & -0.42@xmath480.19 & @xmath58 & & 21 + ulas j020529.62 + 142114.0 & & t1.0@xmath480.5 & t1.0 & t3.0 & 17.99@xmath480.04 & 1.06@xmath480.07 & & @xmath59 & & 22 + 2mass j02060879 + 22355930 & & l5.5 & l5.1@xmath480.5 & t3.2@xmath482.3 & 16.56@xmath480.11 & 1.39@xmath480.17 & 1.61@xmath480.89 & & & 26 + 2mass j02361794 + 0048548 & l6.5 & & l5.0@xmath480.6 & t1.9@xmath481.1 & 16.10@xmath480.08 & 1.43@xmath480.12 & 1.05@xmath480.48 & @xmath60 & & 26 + sdss j024749.90 - 163112.6 & & t2.0@xmath481.5 & t0@xmath480.2 & t7@xmath480.3 & 17.19@xmath480.18 & 1.57@xmath480.27 & 0.68@xmath480.10 & @xmath61 & & 21 + 2mass j03202839 - 0446358 & m8 : & l1 & m8.5@xmath480.3 & t5@xmath480.9 & 12.13@xmath480.03 & 1.13@xmath480.04 & 3.5@xmath480.2 & @xmath62 & rv & 5 ; 1 + sdss j035104.37 + 481046.8 & & t1.0@xmath481.5 & l6.5@xmath480.7 & t5@xmath480.7 & 16.47@xmath480.13 & 1.47@xmath480.18 & 0.31@xmath480.31 & @xmath63 & & 21 + denis - p j04272708 - 1127143 & & dm7 & m7.4@xmath480.2 & t5.1@xmath481.5 & 13.74 & 0.99@xmath480.14 & 4.13@xmath480.62 & & & 26 + 2mass j05185995 - 2828372 & & & l6 & t4 & 15.98@xmath480.10 & 1.82@xmath480.12 & 0.13@xmath480.19 & 1.80@xmath480.50 & di & 10 ; 7 + 2mass j07354882 + 2720167 & l1 & & l1 & l4 & 16.94@xmath480.13 & 1.28@xmath480.21 & & @xmath64 & & 23 + sdss j080531.84 + 481233.0 & l4 & l9.5 & l4.5 & t5 & 14.73@xmath480.04 & 1.46@xmath480.05 & 1.50@xmath480.09 & 0.9 - 2.3 & av & 6 ; 11 + sdss j090900.73 + 652527.2 & & t1.5 & t1.5@xmath480.5 & t2.5@xmath480.3 & 16.03@xmath480.09 & 0.86@xmath480.17 & -0.12@xmath480.10 & @xmath65 & & 21 + sdss j092615.38 + 584720.9 & t4.5 & & t3 & t6 & 16.77@xmath480.14 & @xmath66 & 0.4@xmath480.2 & 2.6@xmath480.5 & di & 16 ; 7 , 8 + sdss j093113.23 + 280227.1 & l3 & & l1.4@xmath480.1 & t2.6@xmath480.9 & 14.98@xmath480.04 & 1.25@xmath480.06 & 2.22@xmath480.23 & @xmath67 & & 26 + 2mass j09490860 - 1545485 & & t2 & t1@xmath480.2 & t2@xmath480.2 & 16.15@xmath480.12 & 0.92@xmath480.20 & -0.07@xmath480.05 & @xmath68 & & 21 + 2mass j10365305 - 3441380 & l6 & & l5.2@xmath480.4 & t1.4@xmath480.4 & 15.62@xmath480.05 & 1.82@xmath480.06 & 0.51@xmath480.32 & @xmath69 & & 26 + sdss j103931.35 + 325625.5 & & t1 & l7@xmath480.2 & t4@xmath480.2 & 16.41@xmath480.15 & 1.25@xmath480.22 & 0.26@xmath480.09 & @xmath63 & & 21 + 2mass j10595138 - 2113082 & l1 & & l0.6@xmath480.4 & t3.4@xmath481.3 & 14.56@xmath480.04 & 1.35@xmath480.06 & 2.58@xmath480.32 & @xmath1552.5 & & 26 + 2mass j11061197 + 2754225 & & t2.5 & t0@xmath480.2 & t4.5@xmath480.2 & 14.82@xmath480.04 & 1.02@xmath480.07 & -0.37@xmath480.06 & @xmath152.67 & ol & 4 , 14 ; 15 + sdss j120747.17 + 024424.8 & l8 & t0 & l6.5@xmath480.7 & t2.5@xmath480.5 & 15.58@xmath480.07 & 1.59@xmath480.09 & 0.48@xmath480.28 & @xmath1517 & & 21 + 2mass j12144089 + 6316434 & & t3.5 & t2 & t6 & 16.59@xmath480.12 & 0.71@xmath480.26 & & @xmath1524 & & 24 + 2mass j13114227 + 3629235 & l5pec & & l4.8@xmath480.6 & t4.1@xmath482.7 & 15.54@xmath480.05 & 1.40@xmath480.09 & 2.19@xmath481.02 & @xmath70 & & 27 + 2mass j13153094 - 2649513 & l5 & & l3.5@xmath482.5 & t7@xmath480.6 & 15.07@xmath480.05 & 1.63@xmath480.07 & 3.03@xmath480.03 & 6.60@xmath480.90 & di & 3 ; 3 + 2mass j13243559 + 6358284 & & t2 : & l8@xmath480.2 & t3.5@xmath480.2 & 15.60@xmath480.07 & 1.54@xmath480.09 & -0.05@xmath480.06 & @xmath71 & & 21,2 4 + 2mass j13411160 - 30525049 & l3 & & l1.2@xmath480.3 & t6.3@xmath481.0 & 14.61@xmath480.03 & 1.53@xmath480.04 & 3.28@xmath480.53 & @xmath72 & & 26 + sdss j141530.05 + 572428.7 & & t3@xmath481 & l8@xmath480.5 & t5@xmath480.3 & 16.73@xmath480.16 & @xmath73 & -0.13@xmath480.20 & @xmath1529 & & 21 + sdss j142227.20 + 221557.5 & l6.5 & & l4.2@xmath480.6 & t4.1@xmath482.3 & 16.87@xmath480.03 & 1.23@xmath480.04 & 2.36@xmath480.78 & @xmath58 & & 26 + 2mass j14232186 + 6154005 & l4 & & l2 & l5 & 16.63@xmath480.15 & 1.35@xmath480.2 & & & & & 23 + sdss j143553.25 + 112948.6 & & t2@xmath481 & l7.5@xmath480.4 & t6@xmath480.3 & 17.14@xmath480.23 & @xmath74 & 0.41@xmath480.12 & @xmath72 & & 21 + sdss j143945.86 + 304220.6 & & t2.5 & t1@xmath480.2 & t5@xmath480.6 & 17.22@xmath480.23 & @xmath75 & 0.06@xmath480.24 & @xmath76 & & 21 + 2mass j14532589 + 1420418 & l1 & & l1.1@xmath480.0 & t6.0@xmath481.1 & 15.07@xmath480.04 & 1.18@xmath480.05 & 3.27@xmath480.46 & @xmath77 & & 26 + sdss j151114.66 + 060742.9 & & t0@xmath482 & l5.5@xmath480.8 & t5@xmath480.4 & 16.02@xmath480.08 & 1.47@xmath480.13 & 0.54@xmath480.32 & @xmath1521 & ol & 4 ; 12 + sdss j151603.03 + 025928.9 & & t0 : & l7.5@xmath481.1 & t2.5@xmath482.2 & 17.23@xmath480.20 & 1.80@xmath480.27 & 0.30@xmath480.65 & @xmath1527 & & 21 , 24 + wise j16235970 - 0508114 & & l1 & l0.6@xmath480.3 & t6.0@xmath480.8 & 14.94@xmath480.04 & 1.39@xmath480.06 & 3.39@xmath480.40 & & & 26 + 2mass j17072529 - 0138093 & & & l0.7@xmath480.5 & t4.3@xmath482.0 & 14.29@xmath480.03 & 1.22@xmath480.04 & 2.87@xmath480.75 & @xmath78 & & 26 + 2massi j17114573 + 2232044 & l6.5 & & l1.5@xmath480.6 & t2.5@xmath481.0 & 17.09@xmath480.18 & 2.36@xmath480.20 & 1.20@xmath480.40 & @xmath79 & & 21 + 2mass j17310140 + 5310476 & l6 & & l5 & l8 & 16.37@xmath480.11 & 1.52@xmath480.18 & & @xmath1548 & & 23,24 + 2mass j17373467 + 5953434 & l9 & & l5 & t5 & 16.88@xmath480.16 & 1.16@xmath480.31 & & @xmath80 & & 23 + 2mass j20261584 - 2943124 & l1 : & & l0.1@xmath480.5 & t5.8@xmath481.0 & 14.80@xmath480.03 & 1.44@xmath480.05 & 3.42@xmath480.40 & @xmath159.00 & di & 13 + sdss j205235.31 - 160929.8 & & t1@xmath481 & l7.5@xmath480.6 & t2@xmath480.2 & 16.33@xmath480.12 & 1.21@xmath480.19 & 0.04@xmath480.18 & 3.20@xmath480.50 & di & 4 ; 17 + 2mass j21265916 + 7617440 & l7 & t0 & l7 & t3.5 & 14.34@xmath480.03 & 1.18@xmath480.05 & & & & 28 + 2mass j21392676 + 0220226 & & t1.5 & l8.5@xmath480.7 & t3.5@xmath481.0 & 15.26@xmath480.05 & 1.68@xmath480.07 & @xmath810.21 & @xmath82 & & 21 + ulas j222958.30 + 010217.2 & & t3.0@xmath480.5 & t1.0 & t5.0 & 17.89@xmath480.04 & 0.67@xmath480.10 & & @xmath83 & & 22 + ulas j223348.82 + 002214.0 & & t0.0@xmath480.5 & l5.0 & t4.0 & 18.07@xmath480.05 & 1.43@xmath480.07 & & @xmath84 & & 22 + ulas j235618.01 + 075420.4 & & t0.0@xmath481.0 & l7.0 & t7.0 & 18.09@xmath480.05 & 1.87@xmath480.06 & & @xmath85 & & 22 + ulas j232315.39 + 071931.0 & & t2.0@xmath480.5 & t0.0 & t2.0 & 17.30@xmath480.02 & 1.01@xmath480.04 & & @xmath56 & & 22 + kelu-1a & l3pec@xmath481.5 & l2.0@xmath481.0 & l0.5@xmath480.5 & t7.5@xmath481 & 13.88@xmath480.03 & 1.57@xmath480.04 & & @xmath86 & & 25 + lllllccccl 2mass j0000286 - 124515 & j00002867 - 1245153 & m8.5 & & m9.2 & 13.20 & 1.23 & 2013 aug 14 & 120 & 44 ; 44 + lehpm 1 - 162 & j00054768 - 2157176 & m8.5 & m8 & m8.7 & 13.27 & 1.07 & 2006 sep 03 & 120 & 32 ; 75 + 2massi j0006205 - 172051 & j00062050 - 1720506 & l2.5 & & l3.0 & 15.66 & 1.65 & 2008 sep 08 & 120 & 25 ; 25 + sdss j000632.60 + 140606.4 & j00063260 + 1406064 & l1 & & l0.4 & 15.85 & 0.79 & 2009 jun 30 & 120 & 48 ; 48 + sdss j000646.81 + 151225.8 & j00064681 + 1512258 & l4 : : & & l4.3 & 16.22 & 1.41 & 2013 sep 03 & 120 & 62 ; 62 + 2mass j0007078 - 245804 & j00070787 - 2458042 & m7 & & m7.7 & 13.12 & 1.05 & 2013 oct 23 & 120 & 44 ; 44 + 2mass j00100009 - 2031122 & j00100009 - 2031122 & l0 & & m8.8 : & 14.13 & 1.25 & 2008 jul 14 & 120 & 44 ; 44 + 2massi j0013578 - 223520 & j00135779 - 2235200 & l4 & & l5.7 & 15.78 & 1.74 & 2009 nov 04 & 120 & 25 ; 25 + 2massw j0015447 + 351603 & j00154476 + 3516026 & l2 & & l1.0 & 13.88 & 1.61 & 2008 sep 08 & 120 & 13 ; 13 + sdss j001608.44 - 004302.3 & j00160843 - 00430209 & & l5.5 & l4.3 & 16.33 & 1.78 & 2013 sep 03 & 120 & 34 ; 34 + sdss j001637.62 - 103911.2 & j00163762 - 1039112 & l0 & & m8.9 & 15.46 & 0.92 & 2009 nov 04 & 120 & 48 ; 48 + sdss j001911.65 + 003017.8 & j00191165 + 0030176 & l1 & & l0.4 & 14.92 & 1.35 & 2009 nov 07 & 120 & 20 ; 20 + denis - p j0020.9 - 4414 & j00205982 - 4414340 & m8v & & m9.4 & 14.90 & 1.19 & 2007 nov 10 & 120 & 6 ; 6 + lehpm 1 - 494b & j00210589 - 4244433 & m9.5 & & l0.6 : & 13.52 & 1.22 & 2006 aug 28 & 120 & 6 ; 15 + sdss j00220934 - 0110397 & j00220934 - 0110397 & l0 & & m8.9 & 15.82 & 1.12 & 2012 sep 27 & 120 & 60 ; 60 + bri 0021 - 0214 & j00242514 - 0158166 & m9.5 & & l0.6 & 11.99 & 1.45 & 2007 jul 04 & 120 & 80 ; 44 + & & & & & & & 2008 jul 14 & 120 & 80 ; 44 + lhs 1074 & j00255117 - 0748069 & sdm6 & & m7.9 & 14.68 & 0.81 & 2004 sep 08 & 120 & ; 39 + lehpm 1 - 606 & j00271049 - 1813083 & & m8 & m7.5 & 13.45 & 1.02 & 2006 sep 03 & 120 & 32 ; 75 + 2massi j0028394 + 150141 & j00283943 + 1501418 & l4.5 & l3 & l6.1 : & 16.51 & 1.95 & 2009 dec 07 & 120 & 13 ; 13 , 34 + 2mass j00285545 - 1927165 & j00285545 - 1927165 & l0 : & & l0.7 & 14.19 & 1.35 & 2009 nov 04 & 120 & 49 ; 49 + 2massw j0030438 + 313932 & j00304384 + 3139321 & l2 & & l3.2 & 15.48 & 1.45 & 2008 sep 08 & 120 & 11 ; 11 + sips j0031 - 3840 & j00311925 - 3840356 & l2.5 & l2 & l1.2 & 14.10 & 1.18 & 2006 aug 28 & 120 & 36 ; 59 , 36 + 2mass j00320509 + 0219017 & j00320509 + 0219017 & l1.5 & m9 & l0.4 & 14.32 & 1.52 & 2009 jun 30 & 120 & 49 ; 49 , 30 + & & & & & & & 2009 nov 07 & 120 & 49 ; 49 , 30 + 2massi j0032431 - 223727 & j00324308 - 2237272 & l1 & & m9.7 : & 15.39 & 1.43 & 2009 dec 07 & 120 & 25 ; 25 + eros - mp j0032 - 4405 & j00325584 - 4405058 & l0gamma & & l0.5 & 14.78 & 1.51 & 2008 sep 07 & 120 & 8 ; 56 + 2massi j0034568 - 070601 & j00345684 - 0706013 & l3 & & l4.3 & 15.53 & 1.59 & 2009 nov 07 & 120 & 25 ; 25 + sdss j003843.99 + 134339.5 & j00384397 + 13433950 & l1 & & l0.6 & 15.91 & 1.15 & 2012 sep 27 & 120 & 20 ; 20 + sdss j004154.54 + 134135.5 & j00415453 + 1341351 & l0 & & l0.6 & 14.45 & 1.22 & 2009 jan 24 & 120 & 20 ; 20 + & & & & & & & 2009 nov 08 & 120 & 20 ; 20 + lhs 1135 & j00433134 + 2054316 & sdm6.5 & & m7.0 & 13.95 & 0.81 & 2004 jul 24 & 120 & ; 39 + 2massw j0045214 + 163445 & j00452143 + 1634446 & l2beta & l3.5 & l1.3 : : & 13.06 & 1.69 & 2007 sep 16 & 120 & 30 ; 56 , 30 + 2mass j00464841 + 0715177 & j00464841 + 0715177 & l0 : : & m9 & m9.5 : & 13.89 & 1.34 & 2008 jul 13 & 120 & 30 ; 49 , 30 + sips j0050 - 1538 & j00502444 - 1538184 & l1 : & & l0.5 & 13.78 & 1.13 & 2008 sep 08 & 120 & 36 ; 44 + sdssp j005406.55 - 003101.8 & j00540654 - 00310180 & l1 & & l2.2 & 15.73 & 1.35 & 2012 sep 25 & 120 & 19 ; 20 + 2mass j00550460 - 3052000 & j00550460 - 3052000 & m8 : & & m7.2 & 13.03 & 1.12 & 2012 dec 27 & 120 & ; + lhs 1166 & j00554418 + 2506235 & sdm6.5 & & m7.2 & 14.26 & 0.69 & 2004 sep 07 & 120 & ; 39 + sdss j00570556 - 0846241 & j00570556 - 0846241 & l0 & & m9.3 & 15.70 & 1.34 & 2012 sep 25 & 120 & 60 ; 60 + 2massw j0058425 - 065123 & j00584253 - 0651239 & l0 & & l0.0 & 14.31 & 1.41 & 2008 jul 14 & 120 & 13 ; 13 + lhs 132 & j0102510 - 373743 & m8 & & m8.1 & 11.13 & 1.06 & 2008 sep 07 & 120 & ; + sdss j010311.51 - 004417.0 & j01031151 - 0044170 & m9.5 & & m8.5 & 14.29 & 1.12 & 2013 oct 29 & 120 & 62 ; 62 + 2massi j0103320 + 193536 & j01033203 + 1935361 & l6 & & l5.4 & 16.29 & 2.14 & 2006 sep 02 & 120 & 13 ; 13 + 2massi j0104075 - 005328 & j01040750 - 0053283 & l4.5 & & l5.8 : & 16.53 & 1.20 & 2009 nov 07 & 120 & 29 ; 81 + sdss j010637.33 + 151855.0 & j01063733 + 1518550 & m8.5 & m8 pec & m9.2 & 14.36 & 0.93 & 2013 sep 03 & 120 & 62 ; 62 , 66 + sdss j01071600 - 1517570 & j01071600 - 1517570 & m7 & & m7.2 & 13.34 & 1.06 & 2012 dec 27 & 120 & ; + sdssp j010752.33 + 004156.1 & j01075242 + 0041563 & l8 & l5.5 & l6.6 & 15.82 & 2.12 & 2005 oct 19 & 120 & 21 ; 20 , 34 + sdss j01084048 + 1347392 & j01084048 + 1347392 & l0 & & m9.3 & & & 2012 sep 27 & 120 & 60 ; 60 + 2mass j01165457 - 1357342 & j01165457 - 1357342 & m9 & & m8.8 & 14.21 & 1.24 & 2009 nov 08 & 120 & 49 ; 49 + 2mass j01170586 + 1752568 & j01170586 + 1752568 & m9 & m8 & m8.6 & 14.14 & 1.20 & 2009 nov 04 & 120 & 61 ; 61 + ssspm j0124 - 4240 & j01235905 - 4240073 & m8 & l2.5 & m8.4 & 13.15 & 1.12 & 2008 sep 07 & 120 & 37 ; 49 , 37 + 2massi j0125369 - 343505 & j01253689 - 3435049 & l2 & & l1.2 & 15.52 & 1.62 & 2008 sep 07 & 120 & 25 ; 25 + 2mass j01273917 + 2805536 & j01273917 + 2805536 & m8.5 & & m9.3 & 14.04 & 1.18 & 2013 aug 14 & 120 & ; + sdss j012743.51 + 135420.9 & j01274352 + 13542099 & l5 & l6+/-1 & l5.5 & 16.83 & 1.40 & 2012 sep 25 & 120 & 20 ; 20 , 34 + 2massw j0129122 + 351758 & j01291221 + 3517580 & l4 & & l4.7 & 16.78 & 2.08 & 2009 nov 07 & 120 & 11 ; 11 + 2mass j01335299 + 0033017 & j01335299 + 0033017 & m9 : & & m9.1 & 15.84 & 1.21 & 2009 jan 24 & 120 & 62 ; 62 + nltt 5282 & j0134592 - 090442 & esdm9.5 ? & & m8.8 & 15.25 & 0.49 & 2008 jul 14 & 120 & ; + 2massw j0135358 + 120522 & j01353586 + 1205216 & l1.5 & & l0.6 & 14.41 & 1.49 & 2009 nov 04 & 120 & 13 ; 13 + 2massw j0141032 + 180450 & j01410321 + 1804502 & l1 & l4.5 & l0.8 & 13.88 & 1.38 & 2008 sep 07 & 120 & 30 ; 44 , 30 + 2mass j01460119 - 4545263 & j01460119 - 4545263 & m9 & & m9.5 & 14.40 & 1.36 & 2008 oct 12 & 120 & 49 ; 49 + & & & & & & & 2009 dec 07 & 120 & 49 ; 49 + 2massw j0147334 + 345311 & j01473344 + 3453112 & l0.5 & & l1.0 & 14.95 & 1.37 & 2008 sep 08 & 120 & 11 ; 11 + 2mass j01490895 + 2956131 & j01490895 + 2956131 & m9.5 & & m9.7 & 13.45 & 1.47 & 2008 sep 07 & 120 & ; + sdss j01535423 + 1404528 & j01535423 + 1404528 & l0 & & m9.0 & 15.21 & 1.25 & 2012 sep 27 & 120 & 60 ; 60 + lehpm 2153 & j02042212 - 3632308 & m8 : & & m8.1 & 13.27 & 1.08 & 2013 dec 05 & 120 & 49 ; 49 + sdss j020608.97 + 223559.2 & j02060879 + 22355930 & & l5.5 & l6.9 : & 16.56 & 1.39 & 2013 dec 05 & 120 & 40 ; 40 + denis j02065660 - 0735190 & j02065660 - 0735190 & m8.5 & & m8.5 & 14.34 & 1.35 & 2013 aug 14 & 120 & 59 ; 59 + sdss j020735.60 + 135556.3 & j02073556 + 13555639 & l3 & l3+/-1.5 & l0.6 & 15.46 & 1.65 & 2012 sep 27 & 120 & 20 ; 20 , 34 + 2massw j0208549 + 250048 & j02085498 + 25004880 & l5 & & l5.9 & 16.21 & 1.80 & 2012 sep 25 & 120 & 13 ; 13 + 2massi j0218291 - 313322 & j02182913 - 3133230 & l3 & & l3.8 & 14.73 & 1.57 & 2009 nov 08 & 120 & 27 ; 27 + 2mass j02192196 + 0506306 & j02192196 + 0506306 & l1 : & l1 & l3.0 & 14.97 & 1.49 & 2008 sep 23 & 120 & ; 55 + ssspm j0219 - 1939 & j02192807 - 1938416 & l1 & l2.5 & l0.6 & 14.11 & 1.20 & 2008 sep 08 & 120 & 83 ; 37 + 2massi j0224367 + 253704 & j02243669 + 25370419 & l2 & & l2.3 & 16.58 & 1.89 & 2012 sep 27 & 120 & 13 ; 13 + 2mass j02284243 + 1639329 & j02284243 + 16393299 & l0 : & & m8.7 & 13.17 & 1.35 & 2012 oct 27 & 120 & 49 ; 49 + 2mass j02301551 + 2704061 & j02301551 + 2704061 & l0 : & & m9.4 & 14.29 & 1.31 & 2008 sep 07 & 120 & 44 ; 44 + denis j02304500 - 0953050 & j02304500 - 0953050 & l0 & & l0.9 & 14.68 & 1.69 & 2013 aug 14 & 120 & 59 ; 59 + sdss j023547.56 - 084919.8 & j02354755 - 08491980 & l2 & & l2.0 : & 15.57 & 1.38 & 2012 sep 27 & 120 & 20 ; 20 + sdssp j023617.93 + 004855.0 & j02361793 + 00485479 & l6 & l6.5 & l6.6 : & 16.10 & 1.43 & 2012 sep 25 & 120 & 21 ; 20 , 34 + 2massi j0239424 - 173547 & j02394245 - 1735471 & l0 & & m9.0 & 14.29 & 1.25 & 2009 nov 08 & 120 & 27 ; 27 + 2massi j0241536 - 124106 & j02415367 - 1241069 & l2 : & & l1.9 & 15.61 & 1.67 & 2008 sep 08 & 120 & 27 ; 27 + 2massw j0242435 + 160739 & j02424354 + 16073920 & l1.5 & & l1.7 & 15.78 & 1.43 & 2012 oct 27 & 120 & 11 ; 11 + sdss j02511323 + 00473631 & j02511323 + 00473631 & m8 & & m8.2 & 13.77 & 1.09 & 2013 sep 03 & 120 & 48 ; 48 + 2mass j02522628 + 0056220 & j02522628 + 0056220 & m8 & & m8.3 & 13.13 & 1.16 & 2012 dec 27 & 120 & ; + sdss j02540582 - 1934523 & j02540582 - 1934523 & m9 & m9 & m8.1 & 13.08 & 1.17 & 2009 jan 25 & 120 & 49 ; 85 + & & & & & & & 2012 dec 27 & 120 & 49 ; 85 + 2mass j02594471 + 2254443 & j02594471 + 2254443 & m9 & & m9.5 & 14.11 & 1.25 & 2009 dec 07 & 120 & 49 ; 49 + 2massi j0302012 + 135814 & j03020121 + 13581419 & l3 & & l2.7 : & 16.53 & 1.90 & 2005 oct 19 & 120 & 13 ; 13 + & & & & & & & 2012 sep 27 & 120 & 13 ; 13 + lehpm 1 - 3070 & j03061185 - 3647417 & m8 & l0 & m8.7 & 11.69 & 1.06 & 2006 sep 03 & 120 & 32 ; 37 + 2massw j0306268 + 154514 & j03062684 + 15451370 & l6 : & & l6.0 : & 17.11 & 1.97 & 2012 sep 25 & 120 & 13 ; 13 + sdss j03083243 - 08105138 & j03083243 - 08105138 & m8 & & m7.6 & 14.98 & 0.89 & 2013 sep 03 & 120 & 48 ; 48 + 2massw j0309088 - 194938 & j03090888 - 19493870 & l4.5 & & l4.9 & 15.75 & 1.69 & 2012 sep 25 & 120 & 13 ; 13 + 2mass j03101401 - 2756452 & j03101401 - 2756452 & l5 & & l6.5 & 15.80 & 1.84 & 2009 nov 08 & 120 & 44 ; 44 + 2mass j03140344 + 1603056 & j03140344 + 16030560 & l0 & & m9.4 & 12.53 & 1.29 & 2012 sep 25 & 120 & 49 ; 49 + 2massi j0316451 - 284852 & j03164511 - 28485209 & l0 : & & l1.0 & 14.57 & 1.46 & 2013 dec 05 & 120 & 27 ; 27 + 2mass j03201710 - 1026120 & j03201710 - 1026120 & m8 & & m7.6 & 13.87 & 1.15 & 2013 aug 14 & 120 & 44 ; 44 + 2mass j03250136 + 2253039 & j03250136 + 22530390 & l3 & & l3.3 & 15.43 & 1.65 & 2013 sep 03 & 120 & 44 ; 44 + 2massw j0326137 + 295015 & j03261367 + 2950152 & l3.5 & & l4.6 & 15.48 & 1.65 & 2009 dec 07 & 120 & 11 ; 11 + 2mass j03264453 + 1919309 & j03264453 + 1919309 & m8.5 & & m8.6 & 13.12 & & 2012 sep 25 & 120 & 49 ; 49 + sdssp j032817.38 + 003257.2 & j03281737 + 00325719 & l3 & & l2.5 : & 15.99 & 1.83 & 2012 sep 27 & 120 & 19 ; 20 + lspm j0330 + 3504b & j03301720 + 3505001 & & d / sdm7 & m8.6 & 16.00 & 0.68 & 2005 oct 17 & 120 & ; 75 + & & & & & & & 2005 dec 31 & 120 & ; 75 + sdssp j033035.13 - 002534.5 & j03303511 - 0025346 & l4 & & l6.5 & 15.31 & 1.47 & 2009 nov 04 & 120 & 12 ; 20 + lehpm 1 - 3365 & j03303847 - 2348463 & & esdm7 & m8.8 & 15.80 & 0.86 & 2004 sep 09 & 120 & 32 ; 75 + 2mass j03320043 - 2317496 & j03320043 - 2317496 & m8 & & m8.4 & 13.64 & 1.10 & 2013 oct 29 & 120 & 49 ; 49 + 2mass j03335134 + 0014068 & j03335134 + 0014068 & sdl & & m7.6 & 16.43 & -0.35 & 2013 dec 05 & 120 & 62 ; 62 + lehpm 1 - 3396 & j03340904 - 4953382 & & m8 & m9.4 & 11.38 & 0.98 & 2004 sep 06 & 120 & 32 ; 75 + 2mass j03350236 + 2342407 & j03350236 + 2342407 & m8.5 & & m7.4 & 12.25 & 0.99 & 2013 dec 05 & 120 & 14 ; 67 + 2mass j03354535 + 0658058 & j03354535 + 0658058 & m8 & & m8.0 & 13.41 & 1.15 & 2012 oct 27 & 120 & 49 ; 49 + 2massw j0337036 - 175807 & j03370359 - 1758079 & l4.5 & & l6.0 : & 15.62 & 2.04 & 2009 nov 08 & 120 & 13 ; 13 + 2mass j03395284 + 2457270 & j03395284 + 2457270 & m8 & & m7.7 & 12.84 & 1.10 & 2012 dec 27 & 120 & ; + sdss j03440891 + 0111249 & j03440891 + 0111249 & l1 & & l2.9 : & 14.74 & 1.22 & 2012 oct 27 & 120 & 60 ; 60 + sdss j03510002 - 00524587 & j03510002 - 00524587 & m8 & & m7.3 & 11.30 & 1.07 & 2012 oct 27 & 120 & 48 ; 48 + 2mass j03521086 + 0210479 & j03521086 + 02104797 & m9 & l0+/-1 & m8.6 & 13.08 & 1.12 & 2012 oct 27 & 120 & 47 ; 49 , 47 + sdss j035308.54 + 103056.0 & j03530854 + 1030560 & l1 & & l0.9 & 15.45 & 1.27 & 2013 oct 29 & 120 & 60 ; 60 + 2mass j03540135 + 2316330 & j03540135 + 2316330 & m8 & & m7.9 & 13.12 & 1.13 & 2012 dec 27 & 120 & ; + lehpm 2 - 471 & j03551067 - 1858173 & & m7 & m7.6 & 13.81 & 0.93 & 2006 sep 03 & 120 & 32 ; 75 + 2mass j0355201 + 143929 & j0355201 + 143929 & m8 & & m8.8 & 13.81 & 1.11 & 2013 oct 29 & 120 & 44 ; 44 + 2mass j03552337 + 1133437 & j03552337 + 1133437 & l5gamma & & l4.8 : : & 14.05 & 2.52 & 2009 dec 07 & 120 & 49 ; 56 + 2massw j0355419 + 225702 & j03554190 + 22570159 & l3 & & l2.9 : & 16.11 & 1.83 & 2012 sep 27 & 120 & 11 ; 11 + denis - p j035729.6 - 441730 & j03572695 - 4417305 & l0beta & & l0.6 : & 14.37 & 1.46 & 2007 sep 16 & 120 & 38 ; 56 + 2mass j04012977 - 4050448 & j04012977 - 40504488 & l0 : & & l0.0 & 14.53 & 1.36 & 2012 dec 27 & 120 & 49 ; 49 + lspm j0402 + 1730 & j04024315 + 1730136 & & sdm7 & m7.0 & 15.59 & 0.56 & 2005 oct 17 & 120 & ; 75 + & & & & & & & 2005 dec 31 & 120 & ; 75 + 2mass j04070752 + 1546457 & j04070752 + 1546457 & l3.5 & & l3.9 & 15.48 & 1.92 & 2009 dec 07 & 120 & 49 ; 49 + 2mass j0407089 - 234829 & j0407089 - 234829 & m8 : & & m8.5 & 13.77 & 1.15 & 2013 oct 29 & 120 & 44 ; 44 + 2mass j04081032 + 0742494 & j04081032 + 0742494 & m8 & & m8.1 & 13.59 & 1.17 & 2013 dec 05 & 120 & 49 ; 49 + 2massi j0408290 - 145033 & j04082905 - 1450334 & l2 & l4.5 & l1.3 & 14.22 & 1.40 & 2009 nov 04 & 120 & 30 ; 27 , 30 + 2massi j0409095 + 210439 & j04090950 + 2104393 & l3 & & l3.4 & 15.51 & 1.66 & 2009 dec 07 & 120 & 13 ; 13 + 2mass j0417474 - 212919 & j0417474 - 212919 & m8 & & m8.3 & 13.85 & 1.18 & 2013 oct 29 & 120 & 44 ; 44 + sdss j04270723 + 0859027 & j04270723 + 0859027 & m8 & & m8.8 & 12.92 & 1.19 & 2012 dec 27 & 120 & ; + denis j0427270 - 112713 & j04272708 - 1127143 & m7 & & m7.9 & 13.74 & 1.07 & 2013 dec 05 & 120 & 59 ; 59 + 2massi j0428510 - 225323 & j04285095 - 22532270 & l0.5 & & m9.9 & 13.51 & 1.39 & 2012 sep 25 & 120 & 25 ; 25 + lp 655 - 23b & j04305157 - 0849007 & m8 & & m8.5 & 12.90 & 1.12 & 2013 dec 05 & 120 & 28 ; 46 + lp 775 - 31 & j0435161 - 160657 & & m7 & m8.0 & 10.41 & 1.05 & 2004 sep 07 & 120 & ; + 2mass j04362054 - 4218523 & j04362054 - 42185236 & l0 : & & l0.4 & 14.49 & 1.35 & 2012 dec 27 & 120 & 49 ; 49 + 2massi j0443058 - 320209 & j04430581 - 32020899 & l5 & & l6.4 : & 15.27 & 1.40 & 2012 sep 27 & 120 & 25 ; 25 + sdss j044337.61 + 000205.1 & j04433761 + 0002051 & m9 & & m9.0 & 12.51 & 1.29 & 2013 dec 05 & 120 & 20 ; 44 + 2mass 04441479 + 0543573 & j04441479 + 0543573 & m8 & & m7.9 & 13.67 & 1.15 & 2013 dec 05 & 120 & 49 ; 49 + 2massi j0451009 - 340214 & j04510092 - 34021500 & l0.5 & & l0.6 & 13.54 & 1.25 & 2012 sep 27 & 120 & 27 ; 27 + 2massi j0453264 - 175154 & j04532647 - 1751543 & l3 : & & l3.1 & 15.14 & 1.68 & 2009 nov 04 & 120 & 27 ; 27 + 2mass j05002100 + 0330501 & j05002099 + 03305010 & l4 & & l4.1 & 13.67 & 1.61 & 2012 dec 27 & 120 & 49 ; 49 + 2massi j0502134 + 144236 & j05021345 + 1442367 & l0 & & m8.9 & 14.27 & 1.32 & 2013 dec 05 & 120 & 60 ; 55 + 2mass j05120636 - 2949540 & j05120636 - 29495400 & l4.5 & & l5.7 : : & 15.46 & 2.18 & 2013 oct 29 & 120 & 27 ; 52 + 2mass j05170548 - 4154413 & j05170548 - 4154413 & m9 & & m8.8 & 13.46 & 1.19 & 2012 dec 27 & 120 & ; + lehpm 2 - 183 & j05173729 - 3348593 & m8 & m9 & m9.4 & 12.00 & 1.17 & 2004 sep 06 & 120 & 32 ; 27 , 75 + 2mass j05184616 - 2756457 & j05184616 - 2756457 & l0 : & & l1.0 & 15.26 & 1.65 & 2009 jan 25 & 120 & 44 ; 44 + 2mass j05264348 - 4455455 & j05264348 - 4455455 & m9.5 & & m9.8 & 14.08 & 1.38 & 2013 dec 12 & 120 & 49 ; 49 + 2mass j05301261 + 6253254 & j05301261 + 6253254 & l1 & & l1.1 & 14.05 & 1.34 & 2008 jan 09 & 120 & 49 ; 49 + 2mass j0534584 - 151143 & j0534584 - 151143 & m9 & & m8.9 & 13.15 & 1.15 & 2013 oct 29 & 120 & 44 ; 44 + sdssp j053951.99 - 005902.0 & j05395199 - 00590189 & l5 & l5 & l5.0 & 14.03 & 1.51 & 2009 jan 24 & 120 & 12 ; 12 , 34 + 2mass j05431887 + 6422528 & j05431887 + 6422528 & l1 & & l1.2 & 13.57 & 1.52 & 2008 jan 09 & 120 & 49 ; 49 + 2mass j05441150 - 2433010 & j05441150 - 2433010 & m8 & & m7.6 & 12.53 & 1.07 & 2012 dec 27 & 120 & ; + 2mass j06022216 + 6336391 & j06022216 + 6336391 & l1 : & & l1.5 & 14.27 & 1.58 & 2008 jan 08 & 120 & 49 ; 49 + lsr 0602 + 3910 & j06023045 + 3910592 & l1 & & l0.8 & 12.30 & 1.44 & 2009 nov 04 & 120 & 88 ; 88 + 2mass j06050190 - 2342260 & j06050190 - 2342260 & l0 : & & m9.8 & 14.51 & 1.37 & 2012 dec 27 & 120 & 44 ; 44 + 2mass j06050196 - 2342270 & j06050190 - 2342260 & l0 : & & l1.1 : & 14.51 & 1.37 & 2009 jan 25 & 120 & 44 ; 44 + 2mass j06085283 - 2753583 & j06085283 - 2753583 & m8.5 low g & l0 lg & m9.6 & 3.600 & -8.78 & 2006 sep 02 & 120 & 27 ; 59 , 71 + sips j0614 - 2019 & j06141196 - 2019181 & & l4 & l1.3 & 14.78 & 1.41 & 2006 sep 01 & 120 & ; 72 + denis - p j0615493 - 010041 & j06154934 - 01004158 & l2+/-1 & & l1.0 & 13.75 & 1.21 & 2010 jan 02 & 120 & 53 ; 53 + & & & & & & & 2012 dec 27 & 120 & 53 ; 53 + 2mass j06244595 - 4521548 & j06244595 - 45215479 & l5 : & & l5.1 & 14.48 & 1.88 & 2013 oct 29 & 120 & 49 ; 49 + sdss j062621.22 + 002934.2 & j06262121 + 00293410 & l1 & & l0.3 & 15.93 & 1.07 & 2010 jan 02 & 120 & 20 ; 20 + 2mass j06411840 - 4322329 & j06411839 - 43223290 & l1.5 & & l2.4 : & 13.75 & 1.30 & 2012 dec 27 & 120 & 49 ; 49 + denis - p j0652197 - 253450 & j06521977 - 25345058 & l0 & & m9.2 & 12.76 & 1.24 & 2009 jan 25 & 120 & 53 ; 53 + & & & & & & & 2009 nov 04 & 120 & 53 ; 53 + denis - p j0652197 - 253450 & j06521977 - 2534505 & l0 & & m9.2 & 12.76 & 1.24 & 2006 sep 02 & 120 & 53 ; 53 + lehpm 2 - 461 & j06591011 - 4747002 & m6.5 & m7 & m7.4 & 13.64 & 0.92 & 2006 dec 20 & 120 & 32 ; 87 , 75 + lehpm 2 - 436 & j07075333 - 4900574 & m8.5 & m8 & m8.8 & 13.23 & 1.12 & 2006 dec 20 & 120 & 32 ; 75 + & & & & & & & 2006 dec 21 & 120 & 32 ; 37 , 75 + 2massw j0708213 + 295035 & j07082132 + 29503500 & l5 & & l5.4 : & 16.72 & 1.95 & 2013 dec 12 & 120 & 13 ; 13 + denis - p j0716478 - 063037 & j07164790 - 06303696 & l1+/-1 & & l1.1 & 13.90 & 1.33 & 2010 jan 02 & 120 & 53 ; 53 + & & & & & & & 2011 apr 04 & 120 & 53 ; 53 + 2mass j07200325 - 0846499 & j07200325 - 0846499 & m9 : & & m9.8 & 10.63 & 1.16 & 2013 dec 05 & 120 & 70 ; 70 + 2mass j07231462 + 5727081 & j07231462 + 5727081 & l1 & & l0.2 & 13.97 & 1.36 & 2008 jan 08 & 120 & 49 ; 49 + sdss j07342570 + 30065792 & j07342571 + 3006583 & m8 & & m7.6 & 15.13 & 1.05 & 2009 jan 25 & 120 & 48 ; 48 + sdss j07351959 + 4108503 & j07351959 + 4108503 & l0 & & m8.9 & 15.78 & 0.93 & 2010 jan 26 & 120 & 60 ; 60 + & & & & & & & 2011 mar 09 & 120 & 60 ; 60 + 2massw j0740096 + 321203 & j07400965 + 32120320 & l4.5 & & l4.0 & 16.19 & 1.97 & 2010 jan 26 & 120 & 61 ; 61 + & & & & & & & 2012 dec 27 & 120 & 13 ; 13 + sdss j074756.31 + 394732.9 & j07475631 + 3947329 & l0 & & m8.3 & 15.08 & 1.35 & 2008 jan 11 & 120 & 20 ; 20 + sdss j074838.61 + 174332.9 & j07483861 + 1743329 & l7 & & l6.0 & 16.27 & 1.85 & 2013 oct 29 & 120 & 54 ; 54 + denis - p j0751164 - 253043 & j07511645 - 2530432 & l2.5 & & l1.1 & 13.16 & 1.17 & 2009 nov 04 & 120 & 53 ; 53 + 2massi j0753321 + 291711 & j07533216 + 29171190 & l2 & & l2.0 & 15.52 & 1.67 & 2010 jan 28 & 120 & 13 ; 13 + & & & & & & & 2011 apr 03 & 120 & 13 ; 13 + 2mass j07575274 + 0914103 & j07575274 + 0914103 & l4 : : & & l3.7 : & 15.86 & 1.77 & 2013 apr 26 & 120 & 54 ; 54 + sdss j080048.13 + 465825.5 & j08004815 + 46582560 & l2 & & l1.3 & 15.51 & 1.20 & 2011 mar 09 & 75 & 20 ; 20 + 2mass j08041429 + 0330474 & j08041429 + 0330474 & m8.5 & & m9.7 & 13.69 & 1.25 & 2013 oct 29 & 120 & 49 ; 49 + sdss j080959.01 + 443422.2 & j08095903 + 4434216 & & l6 & l6.3 : & 16.44 & 2.02 & 2008 jan 12 & 120 & 34 ; 40 + denis - p j0812316 - 244442 & j08123170 - 24444239 & l2.5+/-1 & & l0.9 & 13.82 & 1.43 & 2010 jan 03 & 120 & 53 ; 53 + & & & & & & & 2012 dec 27 & 120 & 53 ; 53 + sdss j08175749 + 1824050 & j08175749 + 1824050 & l1 & & l2.0 & 15.09 & 1.27 & 2013 apr 26 & 120 & 60 ; 60 + sdss j08181228 + 3310482 & j08181228 + 3310482 & l0 & & l1.3 & 15.99 & 0.93 & 2011 apr 03 & 120 & 60 ; 60 + sdss j081946.02 + 165853.9 & j08194602 + 1658539 & m9 & & m8.6 & 13.79 & 1.17 & 2013 dec 05 & 120 & 74 ; 74 + denis j0823031 - 491201 & j08230313 - 4912012 & l1.5 & l3 & l2.7 & 13.55 & 1.48 & 2013 nov 24 & 120 & 53 ; 53 , 68 + 2mass j08230838 + 6125208 & j08230837 + 61252079 & l2 : & & l2.5 & 14.82 & 1.62 & 2011 apr 03 & 120 & 49 ; 49 + sdss j08264265 + 19392195 & j08264262 + 1939224 & l0 & & m9.2 & 14.78 & 1.20 & 2010 jan 25 & 120 & 48 ; 48 + 2massw j0829066 + 145622 & j08290664 + 1456225 & l2 & & l1.0 & 14.75 & 1.58 & 2008 jan 08 & 120 & 13 ; 13 + gj 1111 & j08294949 + 2646348 & m6.5 & & m7.2 & 8.230 & 0.97 & 2013 dec 12 & 120 & 2 ; 58 + 2massw j0829570 + 265510 & j08295706 + 26550990 & l6.5 & & l6.4 : : & 17.11 & 2.15 & 2013 dec 12 & 120 & 13 ; 13 + 2mass j08303256 + 0947150 & j08303256 + 0947150 & m8 & & m7.8 & 11.89 & 1.13 & 2012 dec 27 & 120 & ; + 2mass j08315564 + 1025466 & j08315564 + 1025466 & m9 & & m8.6 & 13.62 & 1.17 & 2009 mar 21 & 120 & ; 49 + 2massw j0832045 - 012835 & j08320451 - 0128360 & l1.5 & & l1.2 & 14.13 & 1.42 & 2008 jan 10 & 120 & 13 ; 13 + 2mass j08352366 + 1029318 & j08352366 + 1029318 & m7 & & m7.9 & 13.14 & 1.09 & 2013 dec 05 & 120 & 49 ; 49 + sdss j08354533 + 2224308 & j08354537 + 2224310 & l0 & & m8.7 & 15.74 & 1.29 & 2010 jan 25 & 120 & 60 ; 60 + 2mass j08355829 + 0548308 & j08355829 + 05483080 & l3 & & l1.7 & 14.53 & 1.36 & 2010 jan 24 & 120 & 49 ; 55 + & & & & & & & 2011 apr 04 & 120 & 49 ; 55 + sdss j08362199 + 4949315 & j08362199 + 4949315 & l0 & & m9.3 : & 15.42 & 0.88 & 2011 mar 09 & 120 & 60 ; 60 + sdss j08364634 + 0526426 & j08364635 + 0526426 & l0 & & m9.4 & 14.58 & 1.29 & 2008 sep 23 & 120 & 60 ; 60 + 2mass j08391608 + 1253543 & j08391608 + 1253543 & m9 & & m8.8 & 13.75 & 1.16 & 2013 dec 12 & 120 & 74 ; 74 + lhs 2034 & j08402975 + 1824091 & m6 & & m7.0 & 11.05 & 1.01 & 2013 dec 12 & 120 & 1 ; 57 + sdss j08410685 + 6035063 & j08410685 + 6035063 & l4 & & l1.5 & 15.94 & 1.25 & 2011 apr 03 & 120 & 60 ; 60 + sdss j08430794 + 3141292 & j08430794 + 3141292 & l3 & & l2.5 & 15.99 & 1.34 & 2011 apr 03 & 120 & 60 ; 60 + sdss j08433328 + 1024435 & j08433328 + 1024435 & l1 & & l2.7 : & 14.87 & 1.20 & 2013 dec 12 & 120 & 60 ; 60 + sdss j084403.46 + 043436.19 & j08440346 + 04343619 & m8 & & m8.7 & 13.46 & 1.05 & 2013 dec 12 & 120 & 48 ; 48 + sdss j08475148 + 0138110 & j08475148 + 0138110 & l3 & & l3.2 & 16.23 & 1.82 & 2013 dec 05 & 120 & 60 ; 60 + 2mass j08490052 + 0220155 & j08490052 + 0220155 & m8 & & m7.3 & 12.93 & 1.03 & 2013 dec 05 & 120 & 74 ; 74 + 2mass j08533434 - 0329432 & j08533434 - 0329432 & m9e & & m8.6 & 15.04 & 0.52 & 2009 mar 21 & 120 & ; + sdss j08583697 + 2710508 & j08583697 + 2710508 & l0 & & l0.0 & 15.05 & 1.39 & 2011 mar 09 & 120 & 60 ; 60 + 2massi j0859254 - 194926 & j08592522 - 1949279 & l6 : : & l8 & l6.8 & 15.53 & 1.78 & 2008 jan 12 & 120 & 27 ; 27 , 69 + 2mass j08593854 + 6341355 & j08593854 + 6341355 & l0 & & m8.6 & 13.70 & 1.31 & 2008 jan 11 & 120 & 49 ; 49 + 2mass j08594029 + 1145325 & j08594029 + 1145325 & m8 & & m8.4 & 12.74 & 1.25 & 2012 dec 27 & 120 & ; + lhs 2090 & j09002359 + 2150054 & m6.5 & & m7.3 & 9.440 & 1.00 & 2013 dec 12 & 120 & 17 ; 58 + sdss j09002368 + 2539343 & j09002368 + 2539343 & l7 & & l6.7 & 16.43 & 1.77 & 2013 dec 12 & 120 & 60 ; 60 + sdss j090206.90 + 003319.36 & j09020690 + 00331936 & m7 & & m7.3 & 12.11 & 0.95 & 2013 dec 05 & 120 & 48 ; 48 + sdss j09094813 + 1940439 & j09094813 + 1940439 & l1 & & m9.9 & 14.73 & 1.29 & 2013 apr 26 & 120 & 60 ; 60 + denis - p j0909 - 0658 & j09095749 - 0658186 & l0 & & l0.2 & 13.89 & 1.35 & 2008 jan 10 & 120 & 9 ; 52 + 2mass j09161504 + 2139512 & j09161504 + 2139512 & m9 & & m9.8 & 13.22 & 1.15 & 2013 dec 12 & 120 & 74 ; 74 + 2massw j0918382 + 213406 & j09183814 + 21340580 & l2.5 & & l2.7 & 15.66 & 1.76 & 2011 apr 03 & 120 & 11 ; 11 + sdss j09230870 + 2340137 & j09230870 + 2340137 & l1 & & l2.3 & 13.85 & 1.04 & 2011 mar 09 & 120 & 60 ; 60 + 2massw j0928397 - 160312 & j09283972 - 1603128 & l2 & & l1.7 & 15.32 & 1.71 & 2008 jan 12 & 120 & 13 ; 13 + sdss j09311323 + 2802271 & j09311323 + 2802271 & l3 & & l2.5 : & 14.98 & 1.25 & 2011 mar 09 & 120 & 60 ; 60 + sdss j093128.22 + 052821.93 & j09312822 + 05282193 & m7 & & m8.4 & 12.86 & 1.06 & 2013 dec 05 & 120 & 48 ; 48 + sdss j09323747 + 6725145 & j09323747 + 6725145 & l0 & & m9.4 & 15.91 & 0.92 & 2011 apr 03 & 120 & 60 ; 60 + 2mass j09340617 + 0536234 & j09340617 + 0536234 & m8 pec & & m7.5 & 15.57 & 0.81 & 2004 mar 11 & 120 & 45 ; 45 , 75 + 2mass j09352803 - 2934596 & j09352802 - 29345959 & l0 & & l0.3 & 14.04 & 1.21 & 2012 dec 27 & 120 & 49 ; 49 + sdss j09385888 + 0443439 & j09385888 + 0443439 & l0 & & m7.8 & 15.24 & 1.24 & 2013 apr 26 & 120 & 60 ; 60 + sdss j09404788 + 2946530 & j09404788 + 2946530 & l1 & & l0.4 & 15.29 & 1.37 & 2013 apr 26 & 120 & 60 ; 60 + sdss j09413492 + 1009421 & j09413492 + 1009421 & l0 & & m8.7 & 14.58 & 1.21 & 2013 apr 26 & 120 & 60 ; 60 + 2massw j0944027 + 313132 & j09440279 + 3131328 & l2 & & l1.7 & 15.50 & 1.49 & 2008 jan 09 & 120 & 13 ; 13 + 2mass 09474477 + 0224327 & j09474477 + 0224327 & m8 : & & m8.2 & 13.17 & 1.09 & 2013 dec 05 & 120 & 49 ; 49 + lhs 2195 & j09492223 + 0806450 & m8.5 & & m8.5 & 12.31 & 1.10 & 2013 dec 12 & 120 & 49 ; 49 + 2mass j09524622 + 0620410 & j09524622 + 0620410 & m8 & & m7.3 & 12.45 & 0.99 & 2013 dec 12 & 120 & 74 ; 74 + 2mass j09532126 - 1014205 & j09532126 - 1014205 & l0 & & m9.4 & 13.47 & 1.33 & 2008 jan 12 & 120 & 44 ; 44 + 2mass j09532126 - 1014205 & j09532126 - 10142056 & l0 & & l0.0 & 13.47 & 1.33 & 2006 dec 20 & 120 & 44 ; 44 + nltt 22851b & j09532455 + 0526583 & m9.5 & & m9.8 & 15.67 & 1.28 & 2013 dec 05 & 120 & 62 ; 62 + sdss j100319.17 - 010508.15 & j10031917 - 01050815 & m7 & & m8.0 & 12.33 & 1.09 & 2013 dec 12 & 120 & 48 ; 48 + g 196 - 3b & j10042066 + 5022596 & l3beta & & l5.6 : : & 14.83 & 2.05 & 2006 dec 23 & 120 & 7 ; 56 + lhs 5166b & j10043929 - 33351889 & l4 & & l4.5 : & 14.48 & 1.56 & 2013 apr 26 & 120 & 22 ; 22 + lhs 2243 & j10163470 + 2751497 & m8 v & & m7.4 & 11.99 & 1.03 & 2013 dec 12 & 120 & 5 ; 5 + 2massi j1017075 + 130839 & j10170754 + 1308398 & l2 : & l1 & l1.1 & 14.10 & 1.39 & 2008 jan 12 & 120 & 27 ; 27 , 30 + sdss j10174251 + 4310579 & j10174251 + 4310579 & l1 & & l0.9 & 15.55 & 1.03 & 2011 mar 09 & 120 & 60 ; 60 + 2massw j1018588 - 290953 & j10185879 - 29095349 & l1 & & l0.3 & 14.21 & 1.42 & 2012 dec 27 & 120 & 22 ; 22 + denis j1019245 - 270717 & j10192447 - 2707171 & l0.5 & & m9.1 & 13.53 & 1.06 & 2013 dec 12 & 120 & 59 ; 59 + 2mass 10213232 - 2044069 & j10213232 - 2044069 & m9 & & m8.5 & 13.19 & 1.13 & 2013 dec 12 & 120 & 49 ; 49 + 2mass j10220489 + 0200477 & j10220489 + 0200477 & l0 & & m9.0 & 14.10 & 1.20 & 2008 jan 10 & 120 & 49 ; 49 + 2mass j10224821 + 5825453 & j10224821 + 5825453 & l1beta & & l0.6 & 13.50 & 1.34 & 2008 jan 11 & 120 & 49 ; 49 + sdss j102552.43 + 321234.0 & j10255227 + 3212349 & & l7.5+/-2.5 & l5.8 : & 15.91 & 0.84 & 2008 jan 08 & 120 & 40 ; 40 + 2massi j1029216 + 162652 & j10292165 + 1626526 & l2.5 & & l2.8 & 14.29 & 1.67 & 2008 jan 12 & 120 & 13 ; 13 + sdss j10330910 + 1216259 & j10330910 + 1216259 & l0 & & m8.9 & 15.07 & 1.12 & 2013 apr 26 & 120 & 60 ; 60 + sdss j10340567 + 0350163 & j10340567 + 0350163 & l0 & & m8.7 & 14.70 & 1.07 & 2013 apr 26 & 120 & 60 ; 60 + 2massw j1035245 + 250745 & j10352455 + 2507450 & l1 & & l1.1 & 14.76 & 1.47 & 2008 jan 08 & 120 & 13 ; 13 + denis j104617.0 - 421237 & j10461703 - 4212372 & & m8 & m8.2 & 14.29 & 1.01 & 2008 jan 24 & 120 & ; 84 + lp 213 - 68 & j10471381 + 4026493 & m8 & & m8.2 & 12.44 & 1.18 & 2013 dec 12 & 120 & 16 ; 16 + denis - p j1047 - 1815 & j10473108 - 18155739 & l2.5 & & l0.3 & 14.20 & 1.31 & 2012 dec 27 & 120 & 10 ; 10 + sdss j104922.45 + 012559.2 & j10492244 + 01255930 & l5 & & l3.5 : & 15.88 & 1.77 & 2011 apr 04 & 120 & 54 ; 54 + 2mass j10511900 + 5613086 & j10511900 + 56130860 & l2 & & l0.8 & 13.24 & 1.34 & 2012 may 30 & 75 & 49 ; 49 + sdss j10515124 + 13111633 & j10515124 + 13111633 & l0 & & m9.5 : & 14.83 & 1.20 & 2013 apr 26 & 120 & 48 ; 48 + sdss j105547.29 + 080842.64 & j10554729 + 08084264 & m9 & & m8.5 & 12.55 & 1.18 & 2013 dec 12 & 120 & 48 ; 48 + 2massi j1059513 - 211308 & j10595138 - 21130819 & l1 & & l2.8 : & 14.56 & 1.35 & 2012 dec 27 & 120 & 27 ; 27 + lhs 2351 & j11061897 + 0428327 & m7 & & m7.6 & 12.33 & 1.00 & 2013 dec 12 & 120 & 4 ; 39 + 2mass j11073750 - 2759385b & j11073750 - 2759385 & & m7 & m7.0 & & & 2009 dec 29 & 120 & ; 84 + gliese 417bc & j11122567 + 35481310 & l4.5 & & l4.8 & 14.58 & 1.86 & 2007 mar 16 & 120 & 13 ; 13 + lhs 2397a & j11214909 - 1313082 & m8 & & m8.5 & & & 2011 mar 09 & 120 & ; + 2massw j1122362 - 391605 & j11223623 - 39160540 & l3 & & l3.7 : : & 15.71 & 1.83 & 2012 dec 27 & 120 & 22 ; 22 + sdss j11264703 + 5816322 & j11264703 + 5816322 & l3 & & l1.3 & 15.84 & 1.30 & 2011 apr 03 & 120 & 60 ; 60 + lehpm 2 - 333 & j11414421 - 2232204 & & m8 & m8.8 & 12.63 & 1.06 & 2006 apr 09 & 120 & 32 ; 75 + sdss j11491231 - 0153006 & j11491231 - 0153006 & l1 & & m8.5 & 14.67 & 0.96 & 2011 mar 09 & 120 & 60 ; 60 + 2mass j11533966 + 5032092 & j11533966 + 5032092 & l1 : & & l0.3 & 14.19 & 1.34 & 2008 jan 11 & 120 & 49 ; 49 + gj 3693 & j11535267 + 0659561 & m8 & & m7.0 & 11.26 & 0.99 & 2013 dec 12 & 120 & 4 ; 63 + 2mass j11544223 - 3400390 & j11544223 - 3400390 & l0 & & l0.5 & 14.20 & 1.34 & 2008 jan 09 & 120 & 38 ; 52 + 2massw j1155395 - 372735 & j11553951 - 37273499 & l2 & & l2.3 & 12.81 & 1.35 & 2012 dec 27 & 120 & 22 ; 22 + lp 851 - 346 & j11554286 - 2224586 & m7.5 & & m7.7 & 10.93 & 1.05 & 2013 dec 12 & 120 & 24 ; 24 + denis - p j1157480 - 484442 & j11574809 - 48444283 & l0.5+/-1 & & m9.4 & 14.01 & 1.21 & 2008 jan 24 & 120 & 53 ; 53 + denis - p j1159 + 0057 & j11593850 + 00572679 & l0 & & l0.4 & 14.08 & 1.27 & 2013 apr 25 & 120 & 10 ; 20 + sdss j11594072 + 5409386 & j11594072 + 5409386 & l2 & & l0.9 & 15.22 & 1.46 & 2013 apr 24 & 120 & 60 ; 60 + sdssp j120358.19 + 001550.3 & j12035811 + 00155000 & l3 & & l5.0 & 14.01 & 1.53 & 2011 apr 04 & 120 & 12 ; 12 + 2massi j1204303 + 321259 & j12043036 + 32125950 & l0 & m9 & m9.6 & 13.82 & 1.30 & 2013 apr 24 & 120 & 27 ; 27 , 30 + sdss j12061049 + 6242572 & j12061049 + 6242572 & l1 & & l2.1 & 15.65 & 1.70 & 2011 mar 09 & 120 & 60 ; 60 + denis j1206501 - 393725 & j12065011 - 3937261 & l2 & & l1.9 & 14.32 & 1.22 & 2007 mar 17 & 120 & 59 ; 59 + 2mass j12073804 - 3909050 & j12073803 - 39090500 & l2 : & & l0.9 & 14.69 & 1.44 & 2011 mar 11 & 120 & 49 ; 49 + 2massi j1213033 - 043243 & j12130335 - 04324369 & l5 & & l4.2 & 14.68 & 1.67 & 2011 apr 04 & 120 & 27 ; 27 + bri 1222 - 1222 & j12245222 - 1238352 & m9 & & m8.5 & 12.57 & 1.22 & 2006 jun 01 & 120 & 89 ; + 2mass j12312141 + 4959234 & j12312141 + 49592339 & l2 & & l3.4 & 14.62 & 1.48 & 2013 apr 24 & 120 & 44 ; 44 + 2mass j12321827 - 0951502 & j12321827 - 0951502 & l0 & & m9.5 & 13.73 & 1.17 & 2008 jan 14 & 120 & 49 ; 49 + sdss j12455566 + 4902109 & j12455566 + 4902109 & l1 & & m7.8 : & 15.95 & 0.75 & 2011 mar 09 & 120 & 60 ; 60 + 2massw j1246467 + 402715 & j12464677 + 40271500 & l4 & & l4.0 & 15.09 & 1.81 & 2011 apr 03 & 120 & 13 ; 13 + sdss j125128.43 + 624310.7 & j12512843 + 6243107 & l4 & & l4.2 & 15.49 & 1.13 & 2013 apr 24 & 120 & 54 ; 54 + 2mass j12565688 + 0146163 & j12565688 + 0146163 & l2 : & & l1.6 & 14.48 & 1.69 & 2009 jun 30 & 120 & 49 ; 49 + sdssp j125737.26 - 011336.1 & j12573726 - 0113360 & l4 & l5 & l5.8 & 15.94 & 1.82 & 2009 jun 30 & 120 & 21 ; 20 , 34 + 2massw j1300425 + 191235 & j13004255 + 1912354 & l1 & l3 & l1.7 & 12.72 & 1.09 & 2006 apr 10 & 120 & 14 ; 14 , 50 + 2mass j13015465 - 1510223 & j13015464 - 15102229 & l1 & & l0.5 & 14.54 & 1.44 & 2013 apr 26 & 120 & 49 ; 49 + gj 499c & j13054106 + 2046394 & l4 : & & l6.5 & 15.20 & 1.83 & 2009 jun 30 & 120 & 27 ; 27 + 2mass j13061727 + 3820296 & j13061726 + 38202960 & l0 & & l0.6 & 14.63 & 1.41 & 2011 apr 03 & 120 & 49 ; 49 + sdss j131142.1 + 362923.9 & j13114192 + 3629247 & l6 & l5 pec ( blue ) & l6.8 & 15.55 & 1.40 & 2012 jul 09 & 120 & 54 ; 54 , 66 + 2mass j13120700 + 3937440 & j13120700 + 3937440 & l0 : & & m8.8 & 14.14 & 1.25 & 2012 may 30 & 75 & ; + 2massi j1315309 - 264951 & j13153094 - 2649513 & l5.5 & & l6.7 & 15.20 & 1.73 & 2009 jun 30 & 120 & 79 ; 52 + 2mass j13204427 + 0409045 & j13204427 + 04090450 & l3 : : & & l2.5 & 15.25 & 1.63 & 2013 apr 24 & 120 & 49 ; 49 + denis - p j1323 - 1806 & j13233597 - 18063790 & l0 & & l0.0 & 14.90 & 1.24 & 2013 apr 26 & 120 & 10 ; 10 + 2massw j1326201 - 272937 & j13262009 - 2729370 & l5 & & l6.6 : & 15.85 & 1.99 & 2009 jun 30 & 120 & 22 ; 22 + sdss j13271521 + 0759375 & j13271521 + 0759375 & l1 & & l0.5 & 14.60 & 1.36 & 2013 apr 25 & 120 & 60 ; 60 + 2massw j1328550 + 211449 & j13285503 + 2114486 & l5 & & l4.1 & 16.19 & 1.93 & 2009 jun 30 & 120 & 11 ; 11 + sdss j133148.92 - 011651.4 & j13314893 - 01165000 & l6 & l8+/-2.5 & l6.1 & 15.46 & 1.39 & 2011 mar 09 & 120 & 20 ; 20 , 34 + sdss j13331279 + 1509566 & j13331279 + 1509566 & l0 & & m7.8 & 15.84 & 0.75 & 2011 mar 11 & 120 & 60 ; 60 + sdss j13334536 - 0216002 & j13334536 - 0216002 & l3 & & l2.2 & 15.38 & 1.53 & 2013 apr 25 & 120 & 60 ; 60 + 2mass j13364062 + 3743230 & j13364062 + 3743230 & l1 & & l0.4 & 14.41 & 1.31 & 2013 apr 26 & 120 & 44 ; 44 + 2mass j13373116 + 4938367 & j13373115 + 49383670 & l0 & & m8.9 & 13.77 & 1.19 & 2012 may 30 & 75 & 44 ; 44 + 2massw j1338261 + 414034 & j13382614 + 41403420 & l2.5 & & l2.4 & 14.22 & 1.45 & 2012 may 30 & 75 & 13 ; 13 + 2mass j13384944 + 0437315 & j13384944 + 04373150 & l1 & & l0.0 & 14.16 & 1.42 & 2012 may 30 & 75 & 49 ; 49 + 2mass j13411160 - 3052505 & j13411160 - 30525049 & l2 : : & & l2.7 : & 14.61 & 1.53 & 2012 jul 09 & 120 & 49 ; 49 + 2massw j1343167 + 394508 & j13431670 + 39450870 & l5 & & l5.3 : : & 16.16 & 2.01 & 2013 apr 24 & 120 & 13 ; 13 + sdss j13571490 - 1438520 & j13571490 - 1438520 & m7 & & m8.5 & 12.85 & 1.11 & 2012 may 30 & 75 & ; + 2mass j13595510 - 4034582 & j13595510 - 40345819 & l1 & & l3.1 : & 13.65 & 1.08 & 2013 apr 25 & 120 & 49 ; 49 + 2mass j14022235 + 0648479 & j14022235 + 0648479 & m9 & & m8.9 & 13.72 & 1.21 & 2010 jul 07 & 120 & ; + 2mass j14044495 + 4634297 & j14044495 + 4634297 & l0 : & & m9.7 & 14.34 & 1.28 & 2008 jul 13 & 120 & 44 ; 44 + sdss j14060148 + 5249309 & j14060148 + 5249309 & l0 & & m9.8 & 15.56 & 1.00 & 2011 mar 09 & 120 & ; 60 + 2mass j14090310 - 3357565 & j14090310 - 3357565 & l2 & & l1.3 & 14.25 & 1.38 & 2010 jul 07 & 120 & 52 ; 52 + 2massw j1411175 + 393636 & j14111735 + 3936363 & l1.5 & & l1.5 & 14.64 & 1.40 & 2008 jul 13 & 120 & 13 ; 13 + 2mass j14122270 + 2354100 & j14122270 + 2354100 & m9 & & m7.8 & 13.73 & 1.08 & 2009 jun 30 & 120 & ; + 2massw j1412244 + 163312 & j14122449 + 1633115 & l0.5 & & m9.9 & 13.89 & 1.37 & 2008 jul 30 & 120 & 13 ; 13 + sdss j14205830 + 2131566 & j14205830 + 2131566 & l1 & & l0.0 : & 15.12 & 1.06 & 2011 mar 11 & 120 & 60 ; 60 + 2massw 1421314 + 182740 & j14213145 + 1827407 & l0 & & m8.9 & 13.23 & 1.29 & 2008 jul 30 & 120 & 14 ; 49 + sdss j142227.25 + 221557.1 & j14222720 + 2215575 & & l6.5+/-2 & l5.4 : & 17.06 & 1.42 & 2009 jun 30 & 120 & 40 ; 40 + sdss j14225715 + 0827521 & j14225715 + 0827521 & l2 & & l2.1 & 15.10 & 1.45 & 2013 apr 25 & 120 & 60 ; 60 + 2mass j14232186 + 6154005 & j14232186 + 6154005 & & l4 & l4.5 & 16.60 & 1.32 & 2013 apr 26 & 120 & 65 ; 65 + gd 165b & j14243909 + 0917104 & l4 & l3+/-2 & l4.4 & 15.69 & 1.52 & 2009 jun 29 & 120 & 3 ; 11 , 34 + denis - p j142527.97 - 365023.4 & j14252798 - 3650229 & l3 : & l5 & l4.5 : : & 13.75 & 1.94 & 2010 jul 07 & 120 & 31 ; 49 , 31 + 2mass j14261286 + 3130394 & j14261286 + 3130394 & l4 & & l4.0 : : & 16.62 & 1.90 & 2013 apr 25 & 120 & 54 ; 54 + 2mass j14283132 + 5923354 & j14283132 + 59233546 & l4 & & l4.4 & 14.78 & 1.52 & 2012 jul 09 & 120 & 49 ; 49 + lhs 2924 & j14284313 + 3310368 & m9 v & & m9.1 & 11.99 & 1.25 & 2007 jul 04 & 120 & ; + 2massi j1430435 + 291540 & j14304358 + 2915405 & l2 & l0.5 & l1.5 & 14.27 & 1.50 & 2008 jul 13 & 120 & 30 ; 27 , 30 + sdss j14324210 + 3451427 & j14324210 + 3451427 & l1 & & l1.2 & 15.75 & 0.98 & 2011 apr 19 & 120 & 60 ; 60 + sdss j14380829 + 64083631 & j14380829 + 64083631 & l0 & & l0.2 & 12.98 & 1.33 & 2011 mar 11 & 120 & 48 ; 48 + sdss j143832.63 + 572216.9 & j14383259 + 5722168 & l5 & & l4.6 & 15.96 & 1.59 & 2010 jul 07 & 120 & 54 ; 54 + 2massw j1438549 - 130910 & j14385498 - 13091029 & l3 : & & l3.1 & 15.49 & 1.63 & 2012 jul 09 & 120 & 13 ; 13 + sdss j143933.44 + 031759.2 & j14393342 + 03175909 & l1 & & l0.7 & 15.99 & 1.18 & 2011 apr 19 & 120 & 20 ; 20 + sdss j14403025 + 12333391 & j14403025 + 12333391 & m9 & & m8.6 & 14.41 & 1.22 & 2013 mar 27 & 75 & 48 ; 48 + 2massi j1441045 + 271932 & j14410457 + 27193234 & m7 & & m7.0 & 12.99 & 1.02 & 2013 mar 27 & 75 & 44 ; 44 + g 124 - 62bc & j14413716 - 09455900 & l0.5 & & l0.4 & 14.02 & 1.36 & 2008 jul 14 & 120 & 10 ; 13 + & & & & & & & 2011 mar 09 & 120 & 10 ; 13 + g 239 - 25 b & j14422175 + 6603198 & & l0+/-1 & m8.1 & 11.51 & 1.18 & 2010 jul 07 & 120 & 35 ; 78 + ssspm 1444 - 2019 & j14442067 - 2019222 & d / sdm9 & d / sdm7 & m9.0 & 12.55 & 0.61 & 2005 mar 23 & 120 & 33 ; 45 , 75 + 2mass j14442946 + 0048530 & j14442946 + 0048530 & m9 & & m8.1 & 15.95 & 0.76 & 2009 jun 30 & 120 & ; + sdssp j144600.60 + 002452.0 & j14460060 + 00245190 & l6 & l5 & l4.2 & 15.89 & 1.96 & 2011 mar 11 & 120 & 21 ; 20 , 34 + 2massw j1449378 + 235537 & j14493784 + 2355378 & l0 & & m9.5 & 15.82 & 1.51 & 2008 jul 13 & 120 & 13 ; 13 + 2mass j14520183 + 1114590 & j14520183 + 1114590 & l2 & & l2.9 & 15.52 & 1.18 & 2013 apr 26 & 120 & 77 ; 62 + sdss j14525558 + 2723244 & j14525558 + 2723244 & l0 & & l0.2 : & 14.92 & 0.84 & 2011 mar 09 & 120 & 60 ; 60 + sdss j14532589 + 1420418 & j14532589 + 1420418 & l1 & & l4.1 : & 15.07 & 1.18 & 2013 apr 26 & 120 & 60 ; 60 + lehpm 2 - 50 & j14560172 - 2747288 & & l1 & m9.3 & 13.25 & 1.06 & 2006 mar 12 & 120 & 32 ; 75 + lhs 3003 & j1456383 - 280947 & m7 & & m7.7 & 9.970 & 1.04 & 2008 jul 29 & 120 & ; + lehpm 2 - 498 & j14565736 - 2631265 & & m8 & m7.7 & 13.56 & 0.95 & 2006 mar 12 & 120 & 32 ; 75 + pss 1458 + 2839 & j1458245 + 283958 & m8.5 & & m8.5 & 13.08 & 1.23 & 2008 jul 13 & 120 & ; + 2mass j15004572 + 4219448 & j15004572 + 4219448 & m9 & & m8.4 & 13.77 & 1.13 & 2012 may 30 & 75 & ; + tvlm 513 - 46546 & j1501081 + 225002 & m8.5 & & m9.3 & 11.87 & 1.16 & 2008 jul 30 & 120 & ; + 2mass j15101685 - 024107 & j15101685 - 024107 & m9 & & m9.3 & 12.61 & 1.27 & 2009 jun 29 & 120 & ; + sdss j15102955 + 36194699 & j15102955 + 36194699 & m9 & & m8.3 & 14.00 & 1.18 & 2013 aug 14 & 120 & 48 ; 48 + 2mass j15111091 + 4340363 & j15111091 + 4340363 & l5 & & l5.8 & 16.60 & 1.90 & 2013 apr 25 & 120 & 54 ; 54 + sdss j15124067 + 3403501 & j15124067 + 3403501 & l3 & & l0.7 & 15.04 & 1.63 & 2013 apr 25 & 120 & 60 ; 60 + 2mass j15230657 - 2347526 & j15230657 - 2347526 & & l2.5 & l0.3 & 14.20 & 1.30 & 2008 jul 29 & 120 & 47 ; 82 + sdss j153453.33 + 121949.2 & j15345324 + 12194950 & & l4+/-1.5 & l5.5 & 15.33 & 1.51 & 2012 jul 09 & 120 & 40 ; 40 + 2mass j15345700 - 1418480 & j15345700 - 1418480 & m7 & & m8.6 & 11.39 & 1.08 & 2012 may 30 & 75 & ; + denis - p j153941.96 - 052042.4 & j15394189 - 0520428 & l4 : & l2 & l4.2 & 13.92 & 1.35 & 2008 jul 14 & 120 & 31 ; 52 , 31 + lehpm 2 - 287 & j15453990 - 2255167 & & m8 & m8.2 & 13.71 & 1.09 & 2006 apr 09 & 120 & 32 ; 75 + 2mass j15474719 - 2423493 & j15474719 - 2423493 & m9 & & m9.0 & 13.97 & 1.23 & 2008 jul 29 & 120 & 49 ; 49 + sdss j154849.02 + 172235.4 & j15484911 + 17223590 & & l5 & l6.6 & 16.10 & 1.65 & 2013 aug 14 & 120 & 40 ; 40 + 2mass j15485834 - 1636018 & j15485834 - 1636018 & & l2+/-1 & m8.8 & 13.89 & 1.26 & 2010 jul 07 & 120 & 47 ; 47 + sdss j15512086 + 4329303 & j15512086 + 4329303 & l3 & & l3.1 & 15.13 & 1.50 & 2011 apr 19 & 120 & 60 ; 60 + sdss j15525232 - 0035019 & j15525232 - 0035019 & l0 & & m9.6 : & 15.99 & 0.98 & 2011 apr 19 & 120 & 60 ; 60 + 2massw j1552591 + 294849 & j15525906 + 2948485 & l0beta & l1 & l0.5 : & 13.48 & 1.46 & 2008 jul 30 & 120 & 30 ; 56 , 30 + 2massw j1553214 + 210907 & j15532142 + 2109071 & l5.5 & & l6.6 : & 16.70 & 2.02 & 2009 jun 29 & 120 & 11 ; 11 + 2massw j1555157 - 095605 & j15551573 - 0956055 & l1 & & l1.6 & 12.56 & 1.11 & 2008 jul 14 & 120 & 22 ; 22 + sdss j15564435 + 1723089 & j15564435 + 1723089 & l0 & & l0.7 & 14.67 & 1.32 & 2013 aug 14 & 120 & 60 ; 60 + 2mass j15573270 + 1752380 & j15573270 + 1752380 & m7.5 & & m8.3 & 13.54 & 1.09 & 2013 aug 14 & 120 & 44 ; 44 + 2massi j1600054 + 170832 & j16000548 + 1708328 & l1.5 & & l1.8 & 16.05 & 1.37 & 2008 jul 12 & 120 & 13 ; 13 + 2mass j16082460 + 195747 & j16082460 + 195747 & m9 & & m8.6 & 13.52 & 1.17 & 2009 jun 29 & 120 & ; + lsr 1610 - 0040 & j16102900 - 0040530 & sdl : & d / sdm6 & m7.0 : & 12.91 & 0.89 & 2004 mar 12 & 120 & 26 ; 26 , 42 + 2mass j16134550 + 1708270 & j16134550 + 1708270 & m9.5 & & m9.6 & 13.47 & 1.28 & 2009 jun 29 & 120 & ; + 2massw j1615441 + 355900 & j16154416 + 3559005 & l3 & & l3.6 & 14.54 & 1.60 & 2009 jun 30 & 120 & 13 ; 13 + 2mass j16184503 - 1321297 & j16184503 - 1321297 & l0 : & & m9.3 & 14.25 & 1.33 & 2010 jul 07 & 120 & 52 ; 52 + sdss j161928.31 + 005011.9 & j16192830 + 0050118 & l2 & & l1.3 & 14.39 & 1.20 & 2008 jul 30 & 120 & 20 ; 20 + lehpm 2 - 1973 & j16202207 - 2446025 & & m7 & m7.0 & 14.32 & 0.94 & 2006 apr 09 & 120 & 32 ; 75 + gj 618.1b & j16202614 - 0416315 & l2.5 & & l2.4 & 15.28 & 1.69 & 2008 jul 30 & 120 & 18 ; 18 + 2mass j16203450 - 2430200 & j16203456 - 2430205 & m6.5 & & m9.4 : : & 14.20 & 1.66 & 2008 jul 29 & 120 & ; + 2mass j16210822 + 2938480 & j16210822 + 2938480 & m9 & & m9.4 & 15.17 & 0.84 & 2009 jun 30 & 120 & ; + sdss j16260303 + 2113130 & j16260303 + 2113130 & l3 & & l4.7 : & 15.48 & 1.56 & 2013 apr 24 & 120 & 60 ; 60 + 2mass j16301770 - 2120010 & j16301770 - 2120010 & & m9 : & m9.4 & 14.51 & 1.32 & 2008 jul 29 & 120 & 77 ; + & & & & & & & 2008 aug 28 & 120 & 77 ; + 2mass j16304139 + 0938446 & j16304138 + 09384459 & l0 : : & & l0.4 : & 14.87 & 1.57 & 2011 apr 19 & 120 & 49 ; 49 + sdss j16311227 + 32271141 & j16311227 + 32271141 & m7 & & m8.2 & 13.15 & 0.91 & 2013 apr 26 & 120 & 48 ; 48 + sdss j163256.1 + 350507.3 & j1632561 + 3505073 & l1 & & l0.7 & 14.65 & 1.35 & 2012 jul 09 & 120 & 74 ; 74 + sdss j16351918 + 42230531 & j16351918 + 42230531 & m8 & & m8.6 & 12.88 & 1.09 & 2013 apr 26 & 120 & 48 ; 48 + sdssp j163600.79 - 003452.6 & j16360078 - 0034525 & l0 & & m9.0 & 14.59 & 1.18 & 2008 jul 30 & 120 & 12 ; 12 + 2mass j16452207 + 3004071 & j16452207 + 3004071 & l3 & & l2.3 & 15.19 & 1.60 & 2008 jul 13 & 120 & 44 ; 44 + 2mass j16452207 + 3004071 & j16452207 + 30040718 & l3 & & l3.0 & 15.19 & 1.60 & 2012 jul 09 & 120 & 44 ; 44 + 2mass j16490419 + 0444571 & j16490419 + 0444571 & m8 & & m8.1 & 12.96 & 1.08 & 2013 aug 14 & 120 & 49 ; 49 + sdss j16545079 + 3747146 & j16545079 + 3747146 & l2 & & l0.7 & 15.01 & 1.35 & 2013 apr 26 & 120 & 60 ; 60 + vb 8 & j16553529 - 0823401 & m7 v & & m7.0 & 9.780 & 0.96 & 2007 jul 04 & 120 & ; + 2mass j16573454 + 1054233 & j16573454 + 1054233 & l2 & & l1.4 & 14.15 & 1.35 & 2009 jun 29 & 120 & 49 ; 49 + wise j165842.56 + 510335.0 & j16584256 + 5103350 & & l6pec & l5.5 : & 15.06 & 1.40 & 2013 oct 23 & 120 & 69 ; 69 + sdss j16585026 + 1820006 & j16585026 + 1820006 & l0 & & l0.9 & 15.48 & 0.91 & 2010 jul 07 & 120 & ; + sdss j17031670 + 19063603 & j17031670 + 19063603 & l0 & & m8.9 & 14.92 & 1.27 & 2012 jul 09 & 120 & 48 ; 48 + 2mass j17072343 - 0558249b & j17072343 - 05582489 & & l3 & l0.0 & 12.26 & 1.23 & 2006 aug 28 & 120 & 41 ; 41 + denis j1707252 - 013809 & j17072529 - 0138093 & l0.5 & l2 & l1.0 & 14.29 & 1.22 & 2006 aug 28 & 120 & 59 ; 59 , 64 + 2massi j1707333 + 430130 & j17073334 + 4301304 & l0.5 & & m8.7 & 13.97 & 1.35 & 2008 sep 08 & 120 & 27 ; 27 + sdss j17104934 + 33232518 & j17104934 + 33232518 & l0 & & m9.4 & 15.13 & 1.05 & 2011 apr 19 & 120 & 48 ; 48 + 2mass j17111353 + 2326333 & j17111353 + 2326333 & l0 : & & l0.5 : & 14.50 & 1.44 & 2008 jul 13 & 120 & 44 ; 44 + g 203 - 50 b & j17114530 + 4029021 & & l5 + 2 - 1.5 & l5.0 & 15.00 & 1.20 & 2008 sep 09 & 120 & 86 ; 86 + & & & & & & & 2010 jul 07 & 120 & 86 ; 86 + 2mass j10511900 + 5613086 & j17170450 + 1509530 & m7 & & m7.9 & 13.61 & 1.11 & 2012 may 30 & 75 & ; + sdss j171714.10 + 652622.2 & j17171408 + 6526221 & l4 & & l6.2 & 14.95 & 1.77 & 2009 jun 30 & 120 & 20 ; 20 + sdss j17175402 + 64274503 & j17175402 + 64274503 & m8 & & m8.8 & 14.41 & 1.03 & 2013 oct 22 & 120 & 48 ; 48 + 2massi j1721039 + 334415 & j17210390 + 3344160 & l3 & l5+/-1 & l5.3 : & 13.63 & 1.14 & 2008 sep 08 & 120 & 27 ; 27 , 50 + sdss j172244.32 + 632946.8 & j17224432 + 6329470 & l0 & & l0.7 & 15.37 & 1.29 & 2009 jun 30 & 120 & 20 ; 20 + sdss j17254384 + 5325349 & j17254384 + 5325349 & l1 & & m8.0 & 15.16 & 0.95 & 2010 jul 07 & 120 & ; + 2massi j1726000 + 153819 & j17260007 + 1538190 & l3beta & & l3.6 : : & 15.67 & 2.01 & 2008 may 11 & 120 & 13 ; 56 + & & & & & & & 2008 jul 13 & 120 & 13 ; 56 + 2mass j17281134 + 0839590 & j17281134 + 0839590 & m9 pec & & m9.8 & 13.63 & 1.13 & 2009 jun 29 & 120 & ; + denis - p j1733423 - 165449 & j17334227 - 1654500 & l0.5+/-1 & & l0.9 & 13.53 & 1.18 & 2006 apr 11 & 120 & 53 ; 53 + 2massw j1743415 + 212707 & j17434148 + 2127069 & l2.5 & & l2.4 & 15.83 & 1.51 & 2009 jun 30 & 120 & 13 ; 13 + denis - p j1745346 - 164053 & j17453466 - 1640538 & l1.5+/-1 & & l1.3 & 13.65 & 1.24 & 2008 sep 08 & 120 & 53 ; 53 + 2mass j17461199 + 5034036 & j17461199 + 50340362 & l5 & & l5.7 & 15.10 & 1.57 & 2012 jul 09 & 120 & 49 ; 49 + 2mass j18000116 - 1559235 & j18000116 - 1559235 & l5.5 & & l4.3 & 13.30 & & 2012 jul 09 & 120 & 68 ; 68 + 2mass j18064570 + 2923591 & j18064570 + 2923591 & m8 : & & m8.3 & 14.20 & 1.26 & 2010 jul 07 & 120 & ; + lsr 1826 + 3014 & j18261131 + 3014201 & m8.5 & d / sdm8.5 & m9.9 & 11.66 & 0.85 & 2004 sep 09 & 120 & 23 ; 23 , 75 + sdss j183929.17 + 442438 & j18392917 + 442438 & m9 & & m8.9 & 13.43 & 1.08 & 2008 nov 04 & 120 & 48 ; 48 + denis j19013910 - 3700170 & j19013910 - 3700170 & m8 & & l0.1 : : & 14.26 & 1.96 & 2013 apr 24 & 120 & 59 ; 59 + denis - p j1909081 - 193748 & j19090821 - 1937479 & l1+/-1 & & l1.4 & 14.52 & 1.61 & 2008 sep 08 & 120 & 53 ; 53 + vb 10 & j19165762 + 0509021 & m8 v & & m8.0 & 9.910 & 1.14 & 2007 jul 04 & 120 & 90 ; 73 + 2mass j19233810 - 3308410 & j19233810 - 3308410 & m7 & & m8.5 & 13.27 & 1.04 & 2012 sep 27 & 120 & ; + denis jj1934511 - 184134 & j1934511 - 184134 & m8.5 & & m8.5 & 14.28 & 1.15 & 2013 oct 23 & 120 & 59 ; 59 + lehpm 2 - 90 & j19453495 - 2557190 & & m9 & m8.9 & 12.35 & 0.84 & 2006 sep 02 & 120 & 32 ; 75 + 2mass j19561542 - 1754252 & j19561542 - 1754252 & m8 & l0+/-1 & m8.5 & 13.75 & 1.10 & 2007 sep 16 & 120 & 47 ; 49 , 82 + 2mass j20025073 - 0521524 & j200250732 - 05215242 & l6 & & l6.8 : : & 15.32 & 1.90 & 2007 jul 04 & 120 & 44 ; 44 + denis j2013108 - 124244 & j2013108 - 124244 & l1.5 & & l0.7 & 14.52 & 1.21 & 2013 oct 22 & 120 & 59 ; 59 + 2mass j20263647 + 0439400 & j20263647 + 0439400 & m9 : & & m8.5 & 14.16 & 1.23 & 2009 jun 28 & 120 & ; + 2mass j2035203 - 311008 & j2035203 - 311008 & m7 & & m7.9 & 13.19 & 1.03 & 2010 jul 07 & 120 & ; + 2mass j20414283 - 3506442 & j20414283 - 3506442 & l2 : & & l1.9 & 14.89 & 1.49 & 2008 sep 08 & 120 & 44 ; 44 + sdss j204724.7 + 142152 & j2047247 + 142152 & m7.5 & & m8.9 & 13.04 & 1.16 & 2008 nov 03 & 120 & 48 ; 48 + lehpm 2 - 381 & j20522811 - 4758362 & m8 & m8 & m8.1 & 12.94 & 1.06 & 2004 sep 06 & 120 & 32 ; 75 + 2massi j2054358 + 151904 & j20543585 + 1519043 & l1 : & & l0.3 & 16.37 & 1.39 & 2009 jun 28 & 120 & 13 ; 13 + 2massi j2057153 + 171515 & j20571538 + 1715154 & l1.5 & & m9.9 & 15.97 & 1.47 & 2009 jun 28 & 120 & 13 ; 13 + 2mass j20575592 - 0050060 & j20575592 - 0050060 & m9 & & m9.3 & 14.97 & 1.20 & 2009 jun 29 & 120 & ; + 2mass j21075409 - 4544064 & j21075409 - 45440639 & l0 : & & l2.6 & 14.92 & 1.53 & 2013 aug 14 & 120 & 49 ; 49 + hb 2115 - 4518 & j2118317 - 450552 & m8.5 & & m8.6 & 13.43 & 1.06 & 2008 jul 14 & 120 & ; + sdss j211846.77 - 001044.6 & j21184677 - 00104469 & l1 & & l2.2 : & 16.20 & 1.13 & 2012 sep 27 & 120 & 54 ; 54 + sdss j212033.89 + 102159 & j21203389 + 102159 & m8 & & m9.0 & 13.54 & 1.12 & 2008 nov 03 & 120 & 48 ; 48 + 2mass j21233110 - 2345180 & j21233110 - 2345180 & m7.5 & & m8.2 & 13.58 & 1.04 & 2013 jul 17 & 75 & 44 ; 44 + 2mass j21263403 - 3143220 & j21263403 - 3143220 & m9 & & m9.2 & 13.47 & 1.10 & 2009 jun 28 & 120 & ; + hb 2124 - 4228 & j2127261 - 421518 & m7.5 & & m8.9 & 13.32 & 1.14 & 2008 jul 14 & 120 & ; + hb 2126 - 4459 & j2130086 - 444627 & m8.5 & & m8.2 & 14.32 & 1.15 & 2008 jul 14 & 120 & ; + 2massw j2130446 - 084520 & j21304463 - 08452049 & l1.5 & & m8.3 & 14.14 & 1.32 & 2006 sep 01 & 120 & 52 ; 52 + & & & & & & & 2006 sep 02 & 120 & 52 ; 52 + & & & & & & & 2008 jul 13 & 120 & 52 ; 52 + & & & & & & & 2008 sep 09 & 120 & 52 ; 52 + & & & & & & & 2012 sep 27 & 120 & 52 ; 52 + & & & & & & & 2012 oct 27 & 120 & 52 ; 52 + sdss j213240.36 + 102949.4 & j21324035 + 10294940 & & l4.5+/-1 & l4.8 : : & 16.59 & 1.96 & 2012 oct 27 & 120 & 40 ; 40 + sdss j213307.94 + 232159 & j21330794 + 232159 & m9.5 & & m9.6 & 13.74 & 1.18 & 2008 nov 04 & 120 & 48 ; 48 + sdss j213435.61 + 240408 & j21343561 + 240408 & m8 & & m9.2 & 13.57 & 1.17 & 2008 nov 04 & 120 & 48 ; 48 + 2mass j21371044 + 1450475 & j21371044 + 1450475 & l2 & & l0.8 & 14.13 & 1.32 & 2008 sep 08 & 120 & 49 ; 49 + denis j21391360 - 3529500 & j21391360 - 3529500 & l0 & & l2.1 : & 14.47 & 1.11 & 2013 aug 14 & 120 & 59 ; 59 + sdss j214046.55 + 011259.7 & j21404654 + 0112594 & l3 & & l4.4 & 15.89 & 1.47 & 2008 sep 08 & 120 & 20 ; 20 + 2mass j21420580 - 3101162 & j21420580 - 3101162 & l3 & & l2.9 & 15.84 & 1.88 & 2010 jul 07 & 120 & 43 ; 43 + sdss j214527.82 - 073434.2 & j21452782 - 0734342 & m9 & & m8.7 & 15.59 & 1.28 & 2009 nov 04 & 120 & 48 ; 48 + 2mass j21472764 + 0101040 & j21472764 + 0101040 & m9 & & m9.1 & 14.57 & 1.10 & 2009 jun 28 & 120 & ; + 2mass j21481628 + 4003593 & j21481633 + 4003594 & l6 & l6.5pec & l4.0 & 14.15 & 2.38 & 2005 sep 09 & 120 & 51 ; 51 + 2mass j21483083 + 0020540 & j21483083 + 0020540 & m9 & & m8.5 & 15.46 & 1.15 & 2009 jun 28 & 120 & ; + sdss j214956.55 + 060334 & j21495655 + 060334 & m9 & & m8.1 & 13.34 & 1.17 & 2008 nov 03 & 120 & 48 ; 48 + sdss j215339.77 + 295005 & j21533977 + 295005 & m9 & & m8.8 & 13.94 & 1.17 & 2008 nov 03 & 120 & 48 ; 48 + 2mass j21580457 - 1550098 & j21580456 - 15500980 & l4 : & & l4.3 : & 15.04 & 1.85 & 2006 sep 01 & 120 & 52 ; 52 + 2mass j22044198 - 0036510 & j22044198 - 0036510 & m9 & & m8.5 & 15.73 & 1.24 & 2009 jun 29 & 120 & ; + 2massw j2206450 - 421721 & j22064498 - 4217208 & l2 & & l4.2 : : & 15.56 & 1.95 & 2008 jul 14 & 120 & 13 ; 13 + grh 2208 - 2007 & j2210499 - 195224 & m7.5 & & m8.8 & 14.00 & 0.85 & 2008 jul 14 & 120 & ; + 2mass j22114470 + 6856262 & j22114470 + 6856262 & & l2 & l2.2 & 15.67 & 1.65 & 2005 sep 08 & 120 & 61 ; 61 + 2mass j22134491 - 2136079 & j22134491 - 21360789 & l0gamma & & l1.9 : & 15.38 & 1.62 & 2006 sep 02 & 120 & 44 ; 56 + & & & & & & & 2008 aug 29 & 120 & 44 ; 56 + wise j222219.93 + 302601.4 & j22221993 + 3026014 & & l9 & l6.7 : & 16.55 & 1.37 & 2013 oct 23 & 120 & 69 ; 69 + lhs 523 & j22285440 - 1325178 & m6.5 & & m7.7 & 10.77 & 0.92 & 2007 sep 16 & 120 & ; + 2maspj2234330 + 291850 & j2234331 + 291849 & m8 : & & m7.9 & 14.04 & 1.08 & 2008 jul 12 & 120 & ; + 2mass j22355013 + 1227370 & j22355013 + 1227370 & m9 & & m9.2 & 15.21 & 1.17 & 2009 jun 30 & 120 & ; + 2massi j2238074 + 435317 & j22380742 + 4353179 & l1.5 & & l0.6 & 13.84 & 1.32 & 2009 nov 07 & 120 & 27 ; 27 + & & & & & & & 2010 jul 07 & 120 & 27 ; 27 + sdss j22434553 - 08215302 & j22434553 - 08215302 & m8 & & m8.4 & 15.43 & 1.14 & 2013 jul 17 & 75 & 48 ; 48 + sdss j225003.72 + 143046.7 & j22500372 + 1430467 & m9 & & m8.7 & 14.94 & 1.22 & 2009 nov 04 & 120 & 48 ; 48 + & & & & & & & 2009 nov 07 & 120 & 48 ; 48 + 2massi j2254519 - 284025 & j22545194 - 2840253 & l0.5 & l0.5 & l0.2 & 14.13 & 1.18 & 2008 jul 14 & 120 & 27 ; 27 , 31 + sdssp j225529.09 - 003433.4 & j22552907 - 0034336 & l0 : & & m8.6 & 15.65 & 1.21 & 2008 sep 07 & 120 & 19 ; 19 + ulas j22585405 + 0113512 & j22585405 + 0113512 & m9 & & m8.8 & 13.91 & 1.00 & 2013 aug 14 & 120 & 62 ; 62 + sdss j225913.88 - 005158.2 & j22591388 - 0051581 & l2 & & l1.6 & 16.36 & 1.71 & 2009 jun 29 & 120 & 20 ; 20 + sdss j230809.9 - 313122 & j2308099 - 313122 & m7 & & m7.5 & 13.62 & 1.04 & 2008 nov 04 & 120 & 48 ; 48 + denis j2308113 - 272200 & j2308113 - 272200 & l1.5 & & l1.4 & 14.58 & 1.40 & 2013 sep 03 & 120 & 59 ; 59 + ssspm j2310 - 1759 & j23101846 - 1759090 & l0 : & l1 & m9.9 & 14.38 & 1.41 & 2008 jul 14 & 120 & 83 ; 44 , 37 + sdss j23172515 - 00543358 & j23172515 - 00543358 & m9 & & m8.7 & 15.68 & 0.94 & 2012 sep 27 & 120 & 48 ; 48 + 2mass j2320292 + 412341 & j2320292 + 412341 & l1 : & & m9.8 & 14.59 & 1.39 & 2003 aug 12 & 75 & 76 ; + & & & & & & & 2008 sep 08 & 120 & ; + 2mass j23211254 - 1326282 & j23211254 - 1326282 & & l1 & l0.4 & 14.50 & 1.36 & 2003 sep 04 & 75 & 47 ; 47 + & & & & & & & 2008 jul 12 & 120 & 47 ; 47 + & & & & & & & 2008 sep 08 & 120 & 47 ; 47 + sdss j232136.11 - 002819.1 & j23213611 - 0028191 & m9 & & m8.3 & 15.48 & 1.04 & 2009 nov 04 & 120 & 48 ; 48 + sdss j232246.84 - 313323 & j23224684 - 313323 & l0 : & & l1.6 : & 13.58 & 1.25 & 2006 aug 28 & 120 & 48 ; 48 + & & & & & & & 2008 nov 04 & 120 & 48 ; 48 + sdss j232313.4 - 024435 & j2323134 - 024435 & m8.5 & & m8.0 & 13.58 & 1.10 & 2008 nov 03 & 120 & 48 ; 48 + 2mass j23302258 - 0347189 & j23302258 - 0347189 & l1 : & & l0.5 & 14.48 & 1.35 & 2008 jul 14 & 120 & 44 ; 44 + hd 221356bc & j23310217 - 0406248 & m9+l3 & & m8.7 & 12.94 & 0.98 & 2013 dec 05 & 120 & 14 ; 46 + lspm j2331 + 4607n & j23311807 + 4607310 & & d / sdm7 & m7.4 & 15.92 & 0.74 & 2005 sep 07 & 120 & ; 75 + 2mass j23312935 + 1552220 & j23312935 + 1552220 & l0 & & m9.5 & 15.06 & 1.06 & 2009 jun 30 & 120 & ; + sdss j233224.38 - 005025 & j23322438 - 005025 & m8 & & m7.7 & 13.65 & 1.04 & 2008 nov 03 & 120 & 48 ; 48 + sdss j233350.76 - 000011.3 & j23335076 - 0000113 & m9 & & m8.6 & 15.52 & 0.98 & 2009 nov 07 & 120 & 48 ; 48 + 2mass j23335838 + 0050110 & j23335838 + 0050110 & m9 & & m9.3 & 15.01 & 1.17 & 2009 jul 01 & 120 & ; + sdss j23352642 + 0817213 & j23352642 + 0817213 & l0 & & m9.5 & 14.72 & 1.34 & 2012 oct 24 & 120 & 60 ; 60 + sdss j23371664 - 09332480 & j23371664 - 09332480 & m8 & & m7.6 & 13.41 & 1.13 & 2013 aug 14 & 120 & 48 ; 48 + 2mass j2341286 - 113335 & j2341286 - 113335 & m8 & & m7.9 & 13.55 & 1.00 & 2010 jul 07 & 120 & ; + 2mass j23440624 - 0733282 & j23440624 - 0733282 & l4.5 & & l6.0 & 14.80 & 1.57 & 2009 jun 29 & 120 & 52 ; 52 + 2mass j23453903 + 0055137 & j23453903 + 0055137 & m9 & & m9.2 & 13.77 & 1.19 & 2008 jul 14 & 120 & 49 ; 49 + sdss j234654.7 - 315353 & j2346547 - 315353 & m8 & & m8.7 & 13.28 & 1.08 & 2008 nov 04 & 120 & 48 ; 48 + 2mass j2352050 - 110043 & j2352050 - 110043 & m7 & & m8.0 & 12.84 & 1.10 & 2010 jul 07 & 120 & ; + denis - p j2353 - 0833 & j2353594 - 083331 & m8.5 & & m8.6 & 13.03 & 1.10 & 2010 jul 07 & 120 & ; + denis j23545990 - 1852210 & j23545990 - 1852210 & l2 & & l2.6 : & 14.22 & 1.39 & 2013 aug 14 & 120 & 59 ; 59 + ssspm j2356 - 3426 & j23561081 - 3426044 & m9.0 & l0.5 & m9.3 & 12.95 & 0.98 & 2007 sep 16 & 120 & 37 ; 37 + ssspm j2400 - 2008 & j23595762 - 2007394 & m9.5 & l1 & m8.9 & 14.38 & 1.13 & 2008 jul 14 & 120 & 37 ; 37 + ccccc h@xmath25o-@xmath87 & 1.14 - 1.165 & 1.26 - 1.285 & 1.15 @xmath21 m h@xmath25o & 1 + ch@xmath22-@xmath87 & 1.315 - 1.335 & 1.26 - 1.285 & 1.32 @xmath21 m ch@xmath22 & 1 + h@xmath25o-@xmath88 & 1.48 - 1.52 & 1.56 - 1.60 & 1.40 @xmath21 m h@xmath25o & 1 + ch@xmath22-@xmath88 & 1.635 - 1.675 & 1.56 - 1.60 & 1.65 @xmath21 m ch@xmath22 & 1 + h@xmath25o-@xmath89 & 1.975 - 1.995 & 2.08 - 2.10 & 1.90 @xmath21 m h@xmath25o & 1 + ch@xmath22-@xmath89 & 2.215 - 2.255 & 2.08 - 2.12 & 2.20 @xmath21 m ch@xmath22 & 1 + @xmath90 & 2.06 - 2.10 & 1.25 - 1.29 & @xmath91 color & 1 + @xmath88-dip & 1.61 - 1.64 & 1.56 - 1.59 + 1.66 - 1.69 & 1.63 @xmath21 m feh / ch@xmath22 & 2 + @xmath89-slope & 2.06 - 2.10 & 2.10 - 2.14 & @xmath89-band shape / cia h@xmath25 & 3 + @xmath87-slope & 1.27 - 1.30 & 1.30 - 1.33 & 1.28 @xmath21 m flux peak shape & 4 + @xmath87-curve & 1.04 - 1.07 + 1.26 - 1.29 & 1.14 - 1.17 & curvature across @xmath87-band & 4 + @xmath88-bump & 1.54 - 1.57 & 1.66 - 1.69 & slope across @xmath88-band peak & 4 + h@xmath25o-@xmath92 & 1.04 - 1.07 & 1.14 - 1.17 & 1.15 @xmath21 m h@xmath25o & 4 + cl spt vs. ch@xmath22-_h _ & best fit curve : @xmath93 , @xmath28 = 0.0354 . + & select points below the @xmath94 curve . + h@xmath25o-_j _ vs. ch@xmath22-_h _ & intersection of : @xmath95 and @xmath96 . + & select points on lower left corner . + h@xmath25o-_j _ vs. _ h_-bump & intersection of : @xmath97 and @xmath96 . + & select points on upper left corner . + ch@xmath22-_j _ vs. ch@xmath22-_h _ & intersection of : @xmath98 and @xmath99 . + & select points on lower left corner . + ch@xmath22-_j _ vs. _ h_-bump & intersection of : @xmath100 , @xmath101 , and @xmath102 . + & select points on upper left corner . + ch@xmath22-_h _ vs. _ j_-slope & intersection of : @xmath103 , @xmath104 , and @xmath105 . + & select points on upper left corner . + ch@xmath22-_h _ vs. _ j_-curve & best fit curve : @xmath106 , @xmath28 = 0.088 . + & select points above the @xmath107 curve , up to ch@xmath22-_h _ = 1.03 . + ch@xmath22-_h _ vs. _ h_-bump & best fit curve : @xmath108 , @xmath28 = 0.013 . + & select points below the @xmath94 curve , down to _ h_-bump = 0.92 . j_-slope vs. _ h_-dip & intersection of @xmath109 and @xmath104 . + & select points on lower right corner . + _ j_-slope vs. _ h_-bump & intersection of : @xmath110 and @xmath102 . + & select points on upper right corner . + _ k_-slope vs. h@xmath25o-_y _ & best fit curve : @xmath111 , @xmath28 = 0.064 . + & select points above the @xmath107 curve and between _ k_-slope = 0.93 - 0.96 . + _ j_-curve vs. _ h_-bump & best fit curve : @xmath112 , @xmath28 = 0.048 . + & select points above the @xmath107 and greater than _ j_-curve = 2.00 and _ h_-bump = 0.92 . + lccrcccl 2mass j02361794 + 0048548 & l5.0@xmath480.6 & t1.9@xmath481.1 & @xmath113 & 1.05@xmath480.48 & 1.80@xmath480.55 & 1 & + sdss j093113.23 + 280227.1 & l1.4@xmath480.1 & t2.6@xmath480.9 & @xmath114 & 2.22@xmath480.23 & 2.74@xmath480.25 & 1 & v , b + 2mass j13114227 + 3629235 & l4.8@xmath480.6 & t4.1@xmath482.7 & @xmath114 & 2.19@xmath481.02 & 3.14@xmath481.33 & 5 & v + 2mass j13411160 - 30525049 & l1.2@xmath480.3 & t6.3@xmath481.0 & @xmath115 & 3.28@xmath480.53 & 4.82@xmath480.63 & 1 & v + 2mass j14532589 + 1420418 & l1.1@xmath480.0 & t6.0@xmath481.1 & @xmath114 & 3.27@xmath480.46 & 4.42@xmath480.63 & 1 & v , ( b ) + 2mass j20261584 - 2943124 & l0.1@xmath480.5 & t5.8@xmath481.0 & @xmath114 & 3.42@xmath480.40 & 4.82@xmath480.57 & 4 & v + 2mass j02060879 + 22355930 & l5.1@xmath480.5 & t3.2@xmath482.3 & @xmath116 & 1.61@xmath480.89 & 2.36@xmath481.25 & 1 & + denis - p j04272708 - 1127143 & m7.4@xmath480.2 & t5.1@xmath481.5 & @xmath117 & 4.13@xmath480.62 & 4.98@xmath480.84 & 1 & + 2mass j10365305 - 3441380 & l5.2@xmath480.4 & t1.4@xmath480.4 & @xmath114 & 0.51@xmath480.32 & 1.41@xmath480.24 & 1 & + 2mass j10595138 - 2113082 & l0.6@xmath480.4 & t3.4@xmath481.3 & @xmath114 & 2.58@xmath480.32 & 3.30@xmath480.64 & 1 & v + sdss j142227.20 + 221557.5 & l4.2@xmath480.6 & t4.1@xmath482.3 & @xmath118 & 2.36@xmath480.78 & 3.22@xmath481.18 & 1 & b + wise j16235970 - 0508114 & l0.6@xmath480.3 & t6.0@xmath480.8 & @xmath114 & 3.39@xmath480.40 & 4.80@xmath480.53 & 1 & + 2mass j17072529 - 0138093 & l0.7@xmath480.5 & t4.3@xmath482.0 & @xmath113 & 2.87@xmath480.75 & 3.75@xmath481.01 & 1 & + 2massi j1711457 + 223204 & l1.5@xmath480.6 & t2.5@xmath481.0 & @xmath114 & 1.20@xmath480.40 & 3.08@xmath480.64 & 6 & v + 2mass j11181292 - 0856106 & l1.4@xmath480.7 & t2.3@xmath482.3 & @xmath119 & 1.50@xmath480.82 & 2.42@xmath481.11 & & b + sdss j141624.09 + 134826.7 & l4.4@xmath481.1 & t3.9@xmath481.4 & @xmath114 & 2.12@xmath480.42 & 2.99@xmath480.67 & & b + 2mass j15150083 + 4847416 & l5.0@xmath480.6 & t2.7@xmath481.9 & @xmath119 & 1.32@xmath480.61 & 2.17@xmath480.84 & 1 & b + 2mass j17114559 + 4028578 & l4.4@xmath480.3 & t2.7@xmath480.8 & @xmath114 & 1.60@xmath480.18 & 2.32@xmath480.33 & & b + 2mass j03205965 + 1854233 & m7.8@xmath480.1 & t6.0@xmath481.5 & @xmath120 & 4.43@xmath480.73 & 5.47@xmath480.92 & & v + 2mass j03264453 + 1919309 & m8.5@xmath480.0 & t6.7@xmath480.8 & @xmath121 & 4.66@xmath480.50 & 5.91@xmath480.58 & & v + 2mass j03303847 - 2348463 & m7.7@xmath480.3 & t5.6@xmath481.6 & @xmath122 & 3.97@xmath480.85 & 4.58@xmath481.05 & & + 2mass j03301720 + 3505001 & m7.7@xmath480.5 & t5.3@xmath481.7 & @xmath123 & 4.27@xmath480.81 & 5.08@xmath481.03 & & + 2mass j03440891 + 0111249 & l0.6@xmath480.5 & t4.8@xmath482.1 & @xmath124 & 3.34@xmath480.88 & 4.14@xmath481.19 & & + 2mass j04024315 + 1730136 & m7.5@xmath480.2 & t5.4@xmath481.7 & @xmath125 & 4.59@xmath480.78 & 5.29@xmath481.01 & & + 2mass j04430581 - 3202090 & l4.5@xmath480.3 & t1.7@xmath481.0 & @xmath126 & 1.41@xmath480.34 & 2.05@xmath480.39 & 1 & b + 2mass j08433328 + 1024435 & l0.9@xmath480.3 & t4.9@xmath482.0 & @xmath127 & 3.01@xmath480.75 & 3.99@xmath481.03 & & + 2mass j08475148 + 0138110 & l2.0@xmath480.7 & t5.7@xmath482.1 & @xmath128 & 3.19@xmath480.74 & 4.61@xmath481.07 & & + 2mass j14232186 + 6154005 & l1.9@xmath480.8 & t4.6@xmath481.9 & @xmath129 & 2.63@xmath480.67 & 3.78@xmath480.97 & 3 & v + 2mass j14493784 + 2355378 & m9.4@xmath480.3 & t6.2@xmath481.7 & @xmath130 & 4.22@xmath480.69 & 5.39@xmath481.00 & 2 & v + 2mass j15412408 + 5425598 & m7.6@xmath480.3 & t5.4@xmath481.6 & @xmath123 & 4.53@xmath480.75 & 5.22@xmath480.99 & & + 2mass j16403561 + 2922225 & m8.1@xmath480.7 & t5.3@xmath481.7 & @xmath130 & 4.38@xmath480.77 & 5.09@xmath481.00 & & + 2mass j17175402 + 64274503 & m8.5@xmath480.1 & t4.6@xmath481.8 & @xmath131 & 4.18@xmath480.70 & 4.93@xmath481.02 & & + 2mass j19064847 + 4011068 & l0.0@xmath480.4 & t5.9@xmath481.8 & @xmath132 & 3.68@xmath480.75 & 4.85@xmath481.01 & & + 2mass j20472471 + 1421526 & m8.4@xmath480.2 & t5.7@xmath481.6 & @xmath133 & 4.10@xmath480.70 & 5.13@xmath480.95 & & v + 2mass j23311807 + 4607310 & m7.5@xmath480.0 & t5.7@xmath481.5 & @xmath134 & 4.60@xmath480.75 & 5.30@xmath480.95 & & +

mathematical modeling is a prominent tool used to better understand complex mechanical or biological systems @xcite . a common problem that arises when developing a model of a biological or mechanical system is that some of its parameters are unknown . this is especially important when those parameters have special meaning but can not be directly measured . thus a natural question arises : can all , or at least some , of the model s parameters be estimated indirectly and _ uniquely _ from observations of the system s input and output ? this is the question of _ structural identifiability_. sometimes the uniqueness holds only within a certain range . in this case , we say that a system is only _ locally _ structurally identifiable . there are numerous reasons why one would be interested in establishing identifiability . structural identifiability is a necessary condition for the practical or numerical identifiability problem , which involves parameter estimation with real , and often noisy , data . the unobservable biologically meaningful parameters of a model can only be determined ( or approximated ) if the model is structurally identifiable . moreover , optimization schemes can not be employed reliably since they will find difficulties when trying to estimate unidentifiable parameters @xcite . the concept of structural identifiability was introduced for the first time in the work of bellman and strm @xcite . since then , numerous techniques have been developed to analyze the identifiability of linear and nonlinear systems with and without controls @xcite ; see also @xcite for a review of different approaches . viscoelastic mechanical models that utilize springs and dashpots in various configurations have been widely used in numerous areas of research including material sciences @xcite , computer graphics @xcite , and biomedical engineering to describe mechanical properties of biological systems @xcite . to achieve a desirable response , networks with different numbers of springs and dashpots in various configurations have been constructed . for example , it is well - known that the simplest models of viscoelastic materials such as voigt ( spring and dashpot in parallel ) or maxwell ( spring and dashpot in series ) do not offer satisfactory representation of the nature of real materials @xcite . thus more complicated configurations are usually constructed and analyzed @xcite . in this paper we investigate the identifiability problem of viscoelastic models represented by an arbitrarily complex spring - dashpot network . although there exist numerous methods that can determine the type of identifiability of a system of ordinary differential equations , generally they are difficult to apply . our results will show in a remarkably simple way how to verify whether the studied model is ( locally or globally ) structurally identifiable . in case it is unidentifiable , our method provides an explanation why this is the case and how to reformulate the problem . moreover , the existing methods usually allow to establish the identifiability only _ a posteriori _ , i.e. after concrete systems have been established thus , we also introduce `` identifiability tables '' , which allow not only to check but also to construct an arbitrarily complex _ identifiable _ spring - dashpot network . a particular motivation for this work comes from cardiovascular modeling @xcite , although the results of this paper can be applied to any viscoelastic modeling approach . _ arterial wall . _ changing blood pressure causes periodic expansion and contraction of the arterial wall ( see fig . [ fig : cv ] ) . it is well - known that the stress - strain curves of the artery walls exhibit hysteresis , which is understood to be a consequence of the fact that the wall is viscoelastic . another manifestation of the viscoelasticity of the arterial tissue is the stress relaxation experiments under constant stretch ( strain ) . spring - dashpot ( s - d ) networks are often used in order to describe the biomechanical properties of the arterial tissue @xcite . identifiable networks can be determined using the results of this paper ( see theorem [ thm : parameqcoeff ] ) . _ neural activity . _ it is common to use the spring - dashpot network to describe the neural firing of various sensors ( e.g. muscle spindle , baroreceptors ) , see @xcite . typically one assumes that the firing activity is proportional to the strain sensed by a spring connected in series with a spring - dashpot network , which represents a local integration of the nerve endings to the arterial wall ( see fig . [ fig : cv ] ) . then the arterial wall and neural activity models are combined . although separately each model is structurally identifiable , there is no guarantee that the resulting viscoelastic structure is identifiable . thus , using our results given in theorem [ thm : localmt ] , we can establish whether the combined viscoelastic model is identifiable , and if not , what needs to be modified . after reviewing basic concepts of viscoelasticity of systems , we present and discuss our main results related to local and global structural identifiability of such systems . finally , we illustrate our results with a number of examples from the literature . the ideal linear elastic material follows hooke s law @xmath0 , where @xmath1 is a young s modulus ( or a spring constant ) , which describes the relationship between the stress @xmath2 and the strain @xmath3 . analogously , the relation @xmath4 describes the viscous material , where @xmath5 and @xmath6 is a viscous constant @xcite . in the basic linear viscoelasticity theory , the elastic and viscous elements are combined . in this work , we shall be concerned with the problem of identifiability of networks of springs and dashpots that are essentially one - dimensional . the elements can be combined either in series or in parallel . in order to obtain the relationship between the total stress ( force ) @xmath2 and the total strain ( extension ) @xmath3 for a given spring - dashpot network , we use two fundamental rules . for two viscoelastic elements connected in series , the stress is the same in both elements , but the total strain is the sum of individual strains on each element . on the other hand , for elements connected in parallel , the strain is the same for both elements , but the total stress is the sum of individual stresses on each element . now we consider concrete viscoelastic networks , starting with the simplest configurations . [ ex : m ] the series combination of a spring , denoted by its constant @xmath1 , and a dashpot , denoted by its constant @xmath6 , is known as a maxwell element ( see fig . [ fig : simple](a ) ) . since the elements are connected in series , the stress @xmath2 is the same on both elements and the total strain @xmath3 is the sum of strains @xmath7 and @xmath8 corresponding to the spring and dashpot , respectively . now , the relationship between the total strain and stress for this system is @xmath9 [ ex : v ] another simple example is the voigt element ( also known as kelvin or kelvin - voigt ) given in fig . [ fig : simple](b ) . following the steps outlined in the previous example , we obtain the @xmath10 relationship @xmath11 [ ex : burgers ] in our third example we consider a particularly popular four - element model , represented by a maxwell element combined in series with a voigt element , and known as the burgers model ( fig . [ fig : simple](c ) ) . denote by subscript @xmath12 and @xmath13 the spring and viscous constants of the maxwell and voigt elements , respectively . note that the stress @xmath2 is the same on all three elements connected in series ( voigt , spring and dashpot ) . eliminating the corresponding local strains , we obtain the following relationship @xmath14\dot\sigma+\frac{e_m e_v}{\eta_m\eta_v}\sigma.\ ] ] first note ( cf . examples [ ex : m ] , [ ex : v ] , and [ ex : burgers ] ) that for any configuration of springs @xmath15 and dashpots @xmath16 , the total strain stress relationship can always be written as the following @xmath17-st order linear ordinary differential equation @xmath18 where the coefficients @xmath19 and @xmath20 are functions of the spring and dashpot constants . the precise value of @xmath21 and the forms of @xmath22 and @xmath23 will depend on the particular structure of the spring - dashpot model . equation [ eq : ode ] is known as the _ constitutive equation_. in the context of spring - dashpot networks , identifiability concerns whether or not it is possible to recover the unknown parameters ( @xmath24 and @xmath25 ) of the system from the governing equation of the model , given only the total stress @xmath2 and total strain @xmath3 . in other words , we assume that we know the stress and the strain at the bounding nodes only and ask if it is possible to determine the unknown parameters ( @xmath24 and @xmath25 ) . in order to uniquely fix the coefficients of the constitutive equation , we require that be _ normalized _ so that the leading term ( in @xmath2 or @xmath3 , depending on the situation ) is monic . thus , letting the @xmath26 non - monic coefficients of be represented by the vector @xmath27 , we have the following formal definition of identifiability . [ defn : id ] let @xmath28 be a function @xmath29 , where @xmath30 is the parameter space . the model is _ globally identifiable _ from @xmath28 if and only if the map @xmath28 is one - to - one . the model is _ locally identifiable _ from @xmath28 if and only if the map @xmath28 is finite - to - one . the model is _ unidentifiable _ from @xmath28 if and only if the map @xmath28 is infinite - to - one . note that local identifiability is equivalent to saying that around each point in parameter space there exists a neighborhood on which the function @xmath28 is one - to - one . for example , for the burgers model considered in example [ ex : burgers ] , the coefficient function @xmath31 is defined as @xmath32 technically speaking , in this paper we will consider the slightly weaker notion of _ generic global identifiability _ ( or generic local identifiability , or generic unidentifiability ) , where _ generic _ means that the property holds almost everywhere . we will omit the use of the term generic when speaking of identifiability . definition [ defn : id ] implies that if there are more parameters than non - monic coefficients , then the system must be unidentifiable . thus , a necessary condition for structural identifiability is that the number of parameters @xmath33 ( elements of the network ) is less than or equal to the number of non - monic coefficients in the constitutive equation . we will soon show that the number of non - monic coefficients is bounded by the number of parameters in spring - dashpot networks . thus , in this case , a necessary condition for structural identifiability is that the number of parameters and non - monic coefficients are equal . we will prove that , remarkably , in the case of viscoelastic models represented by a spring - dashpot network , the converse to this statement holds as well . [ local identifiability ] [ thm : parameqcoeff ] a viscoelastic model represented by a spring - dashpot network is locally identifiable if and only if the number of non - monic coefficients of the corresponding constitutive equation equals the total number of its moduli @xmath34 and viscosity parameters @xmath35 . note that although the constitutive equation is a linear differential equation , its coefficients considered as functions of spring and viscous constants are not linear functions of the parameters ( see ) . thus , theorem [ thm : parameqcoeff ] allows to reduce the difficult problem of checking one - to - one or finite - to - one behavior of nonlinear functions to simply counting the number of parameters ( springs and dashpots ) and non - monic coefficients of the constitutive equation and asking whether the two numbers are equal . the positive answer implies local identifiability , whereas a negative answer implies unidentifiability . consider , for example , the maxwell and voigt elements , and the burgers model . we note that the constitutive equations , , and for all three models are already in the normalized form . now , simply by counting the number of parameters and the non - monic coefficients of the constitutive equations , we see that the two are equal for each model . thus , by the above theorem , all three models are locally structurally identifiable . now we examine when combining two identifiable models results also in an identifiable model . this will allow us to construct arbitrarily complex and identifiable spring - dashpot networks . we start with an observation , which we prove in the following section , related to the possible form of any differential equation that describes a spring - dashpot network . [ prop : types ] every spring - dashpot network , given by equation , has one of the four possible types @xmath36 recall that @xmath21 is the highest derivative of the stress component @xmath37 , which appears in the total strain - stress equation . now we illustrate the different types of networks defined in the above proposition by considering the simplest elements . for a spring , given by @xmath38 , we have @xmath39 ( only @xmath2 appears in the constitutive equation , but none of its derivatives ) , @xmath40 , and @xmath41 . therefore a spring is of type a. note that for a dashpot , which is given by @xmath42 , we also have @xmath39 but @xmath43 and @xmath44 . thus , according to notation given in proposition [ prop : types ] , a dashpot is of type b. for the voigt element , given by , we have @xmath39 as well as @xmath45 and @xmath46 ( that is @xmath47 ) . we conclude that it is of type c. finally , a maxwell element is given by . note that here @xmath48 ( since @xmath49 appears in ) , @xmath50 , @xmath51 , and @xmath52 . thus a maxwell element is of type d. once the constitutive equation has been determined for a given spring - dashpot network , it is very easy to establish the type that it belongs to . unfortunately , @xmath21 does not always have a physical significance . the value of @xmath21 is determined by the specific network and can not be easily related to the number of springs and dashpots as we will illustrate later on . [ thm : localmt ] consider two locally identifiable spring - dashpot systems @xmath53 and @xmath54 of one of the four types @xmath55 , @xmath56 , @xmath57 , @xmath58 . then the new model resulting in joining @xmath53 and @xmath54 either in parallel or in series is of the type indicated by the identifiability tables ( table [ thetable ] ) . the letter _ u _ indicates that the network is unidentifiable , otherwise it is identifiable of the given type . there are several ways one could use the above theorem . one way is to establish the local identifiability of a given spring - dashpot network . contrary to our similar result given in theorem [ thm : parameqcoeff ] , this can be done without actually calculating the constitutive equation . we will show how to apply theorem [ thm : localmt ] to establish structural identifiability after first introducing some notation . given any two spring - dashpot models @xmath59 and @xmath60 , we use the following notation @xmath61 and @xmath62 to denote respectively the parallel and series combination of @xmath59 and @xmath60 . let @xmath63 denote the function that takes a spring and dashpot model @xmath59 and outputs its type ( @xmath64 ) if it is locally identifiable , and @xmath65 if it is unidentifiable . to apply @xmath63 to a complicated model built up from springs and dashpots using series and parallel connections , we replace any springs and dashpots with their respective types @xmath55 and @xmath56 as well as the operations @xmath66 and @xmath67 with @xmath68 and @xmath69 , respectively . then we apply the operations in the identifiability tables ( see table [ thetable ] ) . when connecting two identifiable spring - dashpot networks of one of the types @xmath55 , @xmath56 , @xmath57 , @xmath58 , or an unidentifiable @xmath65 either in series or in parallel , the above tables establish the type of the resulting identifiable system . if the resulting structure is unidentifiable it is indicated by @xmath65 . for example , a parallel connection of two networks of types @xmath55 and @xmath58 gives rise to an identifiable network of type @xmath55 ( see ( a ) ) , but the series connection results in an unidentifiable structure ( see ( b ) ) . [ thetable ] note that the maxwell model shown in fig . [ fig : simple](a ) can be symbolically written as @xmath70 in this formula , we simply replace the spring and the dashpot with @xmath55 and @xmath56 , respectively , as well as the operations @xmath66 and @xmath67 with @xmath68 and @xmath69 , respectively , to obtain @xmath71 thus we conclude that the maxwell model is locally identifiable and is of type @xmath58 . [ ex : burgers ] similarly , the burgers model shown in fig . [ fig : simple](c ) can be symbolically written as @xmath72 to check the local identifiability , we find @xmath73 and use table [ thetable ] to obtain @xmath74 we conclude that the burgers model is locally identifiable and of type @xmath58 . in the next example we show how we can easily establish local structural identifiability of a more complicated network . [ ex : turski ] consider a viscoelastic material studied in @xcite and represented by a spring - dashpot network shown in fig . [ fig : turski](a ) . it can be symbolically represented by @xmath75\land e_4.\ ] ] again , we can verify the local identifiability of the above model using table [ thetable ] and obtain @xmath76\odot a\\ & = \big[(((\overbrace{(c\odot a)}^a\odot b)\oplus b)\odot a)\oplus b\big]\odot a\\ & = \big[((\overbrace{(a\odot b)}^d\oplus b)\odot a)\oplus b\big]\odot a=\ldots = d\\ \end{aligned}\ ] ] this simple computation confirms that the model is locally structurally identifiable . our method can also verify if a network is unidentifiable , providing the reason for the lack of its identifiability . consider the following example . consider a viscoelastic model used in @xcite and shown in fig . [ fig : turski](b ) . using the notation previously introduced , it can symbolically be written as @xmath77\land e_5\land \eta_5.\ ] ] now applying table [ thetable ] , we obtain @xmath78\odot \overbrace{a\odot b}^d\\ & = \big[(\overbrace{(d\oplus d)}^d\odot d)\oplus d\big]\odot d\\ & = \big[\overbrace{(d\odot d)}^{u}\oplus d\big]\odot d=\emph{u}. \end{aligned}\ ] ] whenever table [ thetable ] indicates @xmath65 ( i.e. the corresponding substructure is unidentifiable ) , this inevitably leads to the whole model being unidentifiable . moreover , our method can also explain what is the reason for the lack of identifiability . in this example the situation is simple : joining in series a maxwell element ( type d ) with a generalized maxwell model leads to an unidentifiable network . so far we have considered only _ local _ identifiability of mechanical systems . now we complete the presentation and discussion of our results by introducing a criterium , which establishes when a given network is _ globally _ structurally identifiable . [ thm : global ] a viscoelastic model represented by a spring - dashpot network is globally identifiable if and only if it is locally identifiable and the network is constructed by adding either in parallel or in series at the bounding nodes exactly one basic element ( spring or dashpot ) at a time . note that the network given in fig . [ fig : turski](a ) and considered in example [ ex : turski ] was deemed locally structurally identifiable . we note that it can be constructed by adding just one element at a time and therefore it is _ globally _ structurally identifiable . similarly , all the simple models shown in fig . [ fig : simple ] can also by constructed adding only one element at a time , and since they are locally identifiable , we conclude that they are also globally structurally identifiable . now consider a model which is locally , but not globally , structurally identifiable . consider a generalized kelvin - voigt model shown fig . [ fig : turski](c ) and used in @xcite ) in the context of cardiovascular modeling . it can be symbolically represented by @xmath79 thus the local identifiability can be checked by computing @xmath80 we immediately conclude that the network is locally identifiable . in order to verify whether it is also globally identifiable , note that this network _ can not _ be constructed by adding only one element at a time . thus the system is only locally , but not globally , identifiable . however , in this case the non - global identifiability arises from merely permuting the parameters among the three voigt elements . in this section , we prove the main results from the previous section . to do this requires a careful analysis of the structure of the constitutive equation after combining a pair of systems in series or in parallel . let @xmath53 and @xmath54 be spring - dashpot models whose respective constitutive equations are @xmath81 and @xmath82 , where @xmath83 represent linear differential operators . we can write the differential operators ( in general form ) as : @xmath84 table [ shapestable ] shows that there are restrictions on the values of the @xmath85 and @xmath86 , e.g. the differential order of the lowest order term in @xmath2 is always zero and the differential order of the lowest order term in @xmath3 is zero or one , but we leave the operators in general form for simplicity . we now show the form of the resulting constitutive equation after combining these systems in series or in parallel , in terms of these differential operators . in what follows , we will treat the differential operators @xmath83 as polynomial functions in the variable @xmath87 . for example , @xmath88 can be thought of as a polynomial @xmath89 . suppose that @xmath90 is a series connection of models @xmath53 and @xmath54 , whose constitutive equations are @xmath91 and @xmath92 , respectively . then the stresses ( @xmath2 ) are the same for the two systems while the strains ( @xmath3 ) are added . if @xmath88 and @xmath93 are relatively prime , then the constitutive equation of @xmath59 is : @xmath94 we assume that @xmath95 , so that the constitutive equation is monic . if @xmath88 and @xmath93 have a common factor , then the constitutive equation of @xmath59 is obtained by dividing ( [ eq : series ] ) by the greatest common divisor of @xmath88 and @xmath93 . suppose that @xmath96 is a parallel connection of models @xmath53 and @xmath54 , whose constitutive equations are @xmath91 and @xmath92 , respectively . then the strains ( @xmath3 ) are the same for the two systems while the stresses ( @xmath2 ) are added . if @xmath97 and @xmath98 are relatively prime , then the constitutive equation is : @xmath99 we assume that @xmath100 , so that the constitutive equation is monic . if @xmath97 and @xmath98 have a common factor , then the constitutive equation is obtained by dividing ( [ eq : parallel ] ) by the greatest common divisor of @xmath97 and @xmath98 . now we prove proposition [ prop : types ] , that is , we show that every spring - dashpot network , given by equation , has one of the four possible types displayed in table [ shapestable ] , which are defined by the _ shapes _ of the linear operators @xmath83 acting on @xmath3 and @xmath2 . we make this notion precise : the _ shape _ of a linear operator @xmath83 is a pair of numbers , written @xmath101 $ ] , where @xmath85 is the highest differential order and @xmath86 is the lowest different order . . * possible types of constitutive equations * [ cols="<,<,<,<,<",options="header " , ] two systems of types a , b , c , or d are combined in parallel , where in the first system @xmath102 and in the second system @xmath103 . [ paralleltable ] we note that if a type b or d is combined in series with a type b or d , then @xmath88 and @xmath93 have a common factor ( since both lacked a constant term ) , so the equation @xmath104 is divided by @xmath105 to arrive at the shapes listed in the table . in addition to the type of equation that results after combining two equations of types @xmath106 , we have in tables [ seriestable ] and [ paralleltable ] the resulting identifiability properties of each equation , which we will obtain in the next section . note that definition [ defn : id ] implies that if there are more parameters than non - monic coefficients , then the system must be unidentifiable . the tables show that the number of non - monic coefficients is bounded by the number of parameters , thus a necessary condition for identifiability is that the number of parameters equals the number of non - monic coefficients in the constitutive equation . in the next section , we show that this is also a sufficient condition . consider a spring - dashpot system @xmath59 whose final step connection is a series connection of two systems @xmath53 and @xmath54 , i.e. @xmath90 . since the number of non - monic coefficients in any spring - dashpot model is always less than or equal to the number of parameters in that model , we know that a necessary condition for this system to be locally identifiable is that @xmath53 and @xmath54 are both locally identifiable . let @xmath91 be the constitutive equation for @xmath53 and @xmath92 be the constitutive equation for @xmath54 . each of the operators @xmath107 , @xmath108 , @xmath109 , and @xmath110 will have a fixed shape determined by the structure of @xmath53 and @xmath54 . assuming that @xmath53 and @xmath54 are locally identifiable , we can choose parameters in each of the models @xmath53 and @xmath54 so that the coefficients of these constitutive equations are arbitrary numbers . thus , deciding identifiability of this system amounts to determining whether the map that takes the pair of equations @xmath111 to the constitutive equation @xmath112 , where @xmath113 , @xmath114 , @xmath115 , and @xmath116 ( cf . ) , for the system @xmath59 is finite - to - one or not . the same reasoning works _ mutatis mutandis _ for parallel connections , where we now concern ourselves with the map from the pair of equations @xmath111 with generic coefficients to the constitutive equation for @xmath117 given in . to make the above , intuitive , statements precise we introduce the following definition . the _ shape factorization problem _ for a quadruple of shapes @xmath118 , [ n_{2 } , m_{2 } ] , [ n_{3 } , m_{3 } ] , [ n_{4 } , m_{4}])\ ] ] is the following problem : for a generic pair of polynomials @xmath119 with @xmath120 monic such that the @xmath121 $ ] and @xmath122 $ ] , do there exist finitely many quadruples of polynomials @xmath123 with @xmath124 $ ] , @xmath107 and @xmath109 are monic , and such that @xmath125 and @xmath126 ? a quadruple of shapes @xmath127 is said to be _ good _ if the shape factorization problem for that quadruple has a positive solution . since the above definition introduces one of the key concepts of the paper , in the following example we shall further illustrate the meaning of the shape factorization problem . suppose that our quadruple @xmath128 , [ n_{2 } , m_{2 } ] , [ n_{3 } , m_{3 } ] , [ n_{4 } , m_{4 } ] ) = ( [ 2,0 ] , [ 2,0 ] , [ 3,0],[2,0])\ ] ] which is a special case of joining models of types @xmath55 and @xmath57 in series . the shape factorization problem in this case asks the following question : let @xmath119 be a generic pair of polynomials where @xmath120 and @xmath129 are degree @xmath130 polynomials with nonzero constant term and @xmath120 is monic : @xmath131 do there exist finitely many polynomials @xmath132 such that @xmath125 and @xmath133 ? or to say it another way , for generic values of @xmath134 and @xmath135 , does the system of @xmath136 equations in @xmath136 unknowns : @xmath137 have only finitely many solutions ? the language of shape factorization problems and the remarks in the preceding paragraphs allow us to reduce the local identifiability problem for a spring - dashpot system to determining whether a certain quadruple is a good quadruple . let @xmath90 be a spring - dashpot model joined in series from @xmath53 and @xmath54 , where @xmath53 has constitutive equation @xmath138 of shapes @xmath139 $ ] and @xmath140 $ ] , respectively , and @xmath54 has constitutive equation @xmath141 of shapes @xmath142 $ ] and @xmath143 $ ] , respectively . then the model @xmath59 is locally identifiable if and only if 1 . @xmath53 and @xmath54 are locally identifiable , and 2 . @xmath144 , [ n_{2 } , m_{2 } ] , [ n_{3 } , m_{3 } ] , [ n_{4 } , m_{4}])$ ] is a good quadruple . similarly , if @xmath96 is a spring - dashpot model joined in parallel from @xmath53 and @xmath54 , then @xmath59 is locally identifiable if and only if 1 . @xmath53 and @xmath54 are locally identifiable , and 2 . @xmath145 , [ n_{1 } , m_{1 } ] , [ n_{4 } , m_{4 } ] , [ n_{3 } , m_{3}])$ ] is a good quadruple . so what remains to show is that , for the shapes that arise in spring - dashpot models , whether a quadruple of shapes is a good quadruple only depends on the types ( @xmath146 or @xmath58 ) of the systems being combined . the proof of this statement will occupy the rest of this section . let @xmath120 and @xmath129 be two polynomials . note that for given fixed shapes , @xmath139 $ ] and @xmath142 $ ] , there are at most finitely many factorizations @xmath125 , where @xmath107 has shape @xmath139 $ ] and @xmath109 has shape @xmath142 $ ] and both are monic . this is because there are at most finitely many ways to factorize a monic polynomial into monic factors . once we fix one of these finitely many choices for @xmath107 and @xmath109 , the equation @xmath147 is a linear system in the ( unknown ) coefficients of @xmath108 and @xmath110 . for a polynomial @xmath148 of shape @xmath149 $ ] , we can write the coefficients of @xmath120 as a vector , which we denote @xmath150 : = \begin{pmatrix } f_{n } \\ \vdots \\ f_{m } \end{pmatrix}.\ ] ] let @xmath83 have shape @xmath101 $ ] , as defined in equation . the vector of coefficients of @xmath151 can be written as the result of a matrix vector product as : @xmath152= \begin{pmatrix } a_{n_1 } & 0 & \cdots & 0 \\ \vdots & a_{n_1 } & \cdots & 0 \\ a_{m_1 } & \vdots & \cdots & \vdots \\ 0 & a_{m_1 } & \cdots & 0 \\ \vdots & 0 & \cdots & a_{n_1 } \\ \vdots & \vdots & \cdots & \vdots \\ 0 & 0 & \cdots & a_{m_1 } \end{pmatrix } \begin{pmatrix } e_{n_4 } \\ \vdots \\ e_{m_4 } \end{pmatrix}.\ ] ] we will refer to this product as @xmath153 $ ] , where @xmath154 is a @xmath155 by @xmath156 matrix . likewise , the coefficients of @xmath157 can be written as the result of a matrix vector product as : @xmath158= \begin{pmatrix } c_{n_3 } & 0 & \cdots & 0 \\ \vdots & c_{n_3 } & \cdots & 0 \\ c_{m_3 } & \vdots & \cdots & \vdots \\ 0 & c_{m_3 } & \cdots & 0 \\ \vdots & 0 & \cdots & c_{n_3 } \\ \vdots & \vdots & \cdots & \vdots \\ 0 & 0 & \cdots & c_{m_3 } \end{pmatrix } \begin{pmatrix } b_{n_2 } \\ \vdots \\ b_{m_2 } \end{pmatrix}.\ ] ] we will refer to this product as @xmath159 $ ] , where @xmath160 is a @xmath161 by @xmath162 matrix . then we call the _ matrix factored form _ of @xmath163 $ ] the expression : @xmath164+h[l_4],\ ] ] where the matrices @xmath165 and @xmath166 are the matrices @xmath160 and @xmath154 padded with rows of zeros so that coefficients corresponding to monomials of the same degree appear in the same row . this makes @xmath167 a @xmath168 by @xmath169 matrix . we can now state a criteria for determining if the shape factorization problem has finitely many solutions : [ prop : abinvert ] the quadruple @xmath144 , [ n_{2 } , m_{2 } ] , [ n_{3 } , m_{3 } ] , [ n_{4 } , m_{4}])$ ] is a good quadruple if and only if the matrix @xmath167 is generically invertible . we can write the shape factorization problem of type @xmath144 , [ n_{2 } , m_{2 } ] , [ n_{3 } , m_{3 } ] , [ n_{4 } , m_{4}])$ ] in matrix factored form as @xmath170+h[l_4]=[g]$ ] ( see ) , so that @xmath171.\ ] ] this system has a unique solution if and only if @xmath167 is generically invertible , i.e. invertible for a generic choice of parameter values . [ ex : quadruple ] suppose that our quadruple @xmath144 , [ n_{2 } , m_{2 } ] , [ n_{3 } , m_{3 } ] , [ n_{4 } , m_{4}])$ ] is @xmath172 , [ 2,0 ] , [ 3,0],[2,0])$ ] , which is a special case of joining models of types @xmath55 and @xmath57 in series . the resulting matrix @xmath167 is the matrix @xmath173 we now determine when this matrix @xmath167 is generically invertible , i.e. square and full rank . sylvester matrix _ associated to two polynomials @xmath174 and @xmath175 is the @xmath176 by @xmath176 matrix that has the coefficients of @xmath177 repeated @xmath21 times as columns and the coefficients of @xmath178 repeated @xmath12 times as columns in the following way : @xmath179 @xmath180 the determinant of the sylvester matrix of the two polynomials @xmath181 and @xmath182 is the _ resultant _ , which is zero if and only if the two polynomials have a common root . in particular , for generic polynomials @xmath181 and @xmath182 , the sylvester matrix is invertible ( * ? ? ? * chapter 3 ) . we will use the sylvester matrix in the following way . we will show that there are submatrices of @xmath167 that correspond to the sylvester matrix associated to @xmath88 and @xmath93 . [ prop : squarefullrank ] if the matrix @xmath167 is square , then it is generically invertible . we claim that the columns of @xmath167 can be ordered so that the resulting matrix has the shape @xmath183 where @xmath184 is the sylvester matrix associated to the nonzero coefficients of @xmath107 and @xmath109 . note that this means that we might shift the coefficients down if necessary so there are no extraneous zero terms of low degree ( i.e. if the shape is @xmath185 $ ] with @xmath186 ) . the matrix @xmath187 is a square lower triangular matrix with nonzero entries on the diagonal , and @xmath188 is a square upper triangular matrix with nonzero entries on the diagonal . this will prove that @xmath167 is invertible , since its determinant will be the product of the determinants of @xmath184 , @xmath187 and @xmath188 , all of which are nonzero . to prove that claim requires a careful case analysis . the number of columns of @xmath167 is @xmath189 and the number of rows is @xmath190 . without loss of generality , we can assume that the maximum is attained by @xmath191 . we need to distinguish between the two cases where the minimum is attained by @xmath192 and by @xmath193 . * case 1 : @xmath194 = m_{1 } + m_{4}$ ] . * since @xmath167 is a square matrix , this implies that @xmath195 . in this case we group the columns of @xmath167 in the following order . 1 . the first @xmath196 columns of @xmath165 2 . then the next @xmath197 columns of @xmath165 3 . then all @xmath198 columns of @xmath166 4 . then the remaining @xmath199 columns of @xmath165 . this choice has the property that the middle two blocks of columns together have the desired form , since we have chosen to start including columns from @xmath165 and @xmath166 precisely when they both have nonzero entries in the same rows , and stopping the formation of these when they stop having nonzero entries in the same rows , which has the correct form . note we have used all columns of @xmath165 since @xmath200 * case 2 : @xmath194 = m_{2 } + m_{3}$ ] . * note that since @xmath167 is square , this implies that @xmath201 . in this case , we do not need to reorder the columns to obtain the desired form . we mention how to block the columns to obtain the desired form . 1 . the first @xmath196 columns of @xmath165 2 . then the next @xmath197 columns of @xmath165 3 . then the first @xmath202 columns of @xmath166 4 . then the remaining @xmath203 columns of @xmath166 . note that we have the desired number of columns from the second and third blocks , and we have chosen them so that that those columns have nonzero entries at exactly the same rows . furthermore , we have used all columns of @xmath165 since @xmath204 and all columns of @xmath166 since @xmath205 we can rewrite the matrix in example [ ex : quadruple ] as @xmath206 here the the @xmath207 matrix in the lower righthand corner is the sylvester matrix , the matrix @xmath187 is the @xmath208 matrix in the upper lefthand corner , and the matrix @xmath188 is the empty matrix . [ proof of theorem [ thm : parameqcoeff ] ] we will show that if the number of parameters equals the number of non - monic coefficients , then the matrix @xmath167 is square . by propositions [ prop : abinvert ] and [ prop : squarefullrank ] , this will imply that the model is locally identifiable . let @xmath90 be a spring - dashpot model joined in series from @xmath53 and @xmath54 , where @xmath53 has constitutive equation @xmath138 of shapes @xmath139 $ ] and @xmath140 $ ] , respectively , and @xmath54 has constitutive equation @xmath141 of shapes @xmath142 $ ] and @xmath143 $ ] , respectively . by induction , we can assume that the number of parameters equals the number of non - monic coefficients for the systems @xmath53 and @xmath54 , i.e. there are @xmath209 parameters in the first and @xmath210 in the second . assume the number of parameters equals the number of non - monic coefficients in this full system , i.e. @xmath211 subtracting @xmath212 from both sides , we get that @xmath213 from the definition of @xmath167 , this means the number of rows equals the number of columns , so that @xmath167 is square . the argument for the parallel extension is identical and is omitted . [ proof of theorem [ thm : localmt ] ] theorem [ thm : parameqcoeff ] shows that the model is locally identifiable if and only if the number of parameters equals the number of non - monic coefficients . thus the identifiability properties of the @xmath214 cases in tables [ seriestable ] and [ paralleltable ] are determined by checking if the numbers in the columns corresponding to the number of parameters and the number of non - monic coefficients are equal . we now determine necessary and sufficient conditions for global identifiability . let @xmath90 be a spring - dashpot model joined in series from @xmath53 and @xmath54 , where @xmath53 has constitutive equation @xmath138 of shapes @xmath139 $ ] and @xmath140 $ ] , respectively , and @xmath54 has constitutive equation @xmath141 of shapes @xmath142 $ ] and @xmath143 $ ] , respectively . then the model @xmath59 is globally identifiable if and only if 1 . @xmath53 and @xmath54 are globally identifiable , 2 . the shape factorization problem for the quadruple @xmath144 , [ n_{2 } , m_{2 } ] , [ n_{3 } , m_{3 } ] , [ n_{4 } , m_{4}])$ ] generically has a unique solution . similarly , if @xmath96 is a spring - dashpot model joined in parallel from @xmath53 and @xmath54 , then @xmath59 is globally identifiable if and only if 1 . @xmath53 and @xmath54 are globally identifiable , and 2 . the shape factorization problem for the quadruple @xmath145 , [ n_{1 } , m_{1 } ] , [ n_{4 } , m_{4 } ] , [ n_{3 } , m_{3}])$ ] generically has a unique solution . we handle the case of series extensions , parallel extensions being identical . let @xmath90 . clearly , @xmath53 and @xmath54 must be globally identifiable otherwise we could give two sets of parameters yielding the same constitutive equation for @xmath53 , which could then be combined with parameters for @xmath54 to get two sets of parameters for @xmath59 yielding the same constitutive equation . now if the shape factorization problem has a unique solution , there is a unique way to take the constitutive equation for @xmath59 and solve for the constitutive equations for @xmath53 and @xmath54 , since @xmath53 and @xmath54 are globally identifiable , there is a unique way to solve for parameters of those models giving a unique solution for parameters for @xmath59 . conversely , if there were multiple solutions to the shape factorization problem , then by global identifiability of @xmath53 and @xmath54 , we could solve all the way back to get multiple parameter choices for the same parameter choice for @xmath59 . note that in our analysis of the shape factorization problem in the previous section , we saw that once @xmath107 and @xmath109 are chosen among all their finitely many values , when the model is locally identifiable there is a unique way to then construct @xmath108 and @xmath110 . hence , the shape factorization problem has a unique solution when there is a unique way to factor @xmath125 . this happens if and only if either @xmath215 or @xmath216 , otherwise , generically , we can exchange roots of @xmath107 and @xmath109 giving multiple solutions . [ cor : globidconstzero ] suppose that @xmath90 is globally identifiable . then either @xmath53 or @xmath54 must have been one of a spring , a dashpot , or a maxwell model . suppose that @xmath96 is globally identifiable . then either @xmath53 or @xmath54 must have been one of a spring , a dashpot , or a voigt model . the four models given by the spring , dashpot , voigt , and maxwell elements are the only four locally identifiable models that have the property that at least one of the differential operators in its constitutive equation has exactly one term . this can be seen by analyzing the four types ( a , b , c , d ) and looking at all possibilities that arise on combining two equations . once both operators do not have a single term , no model combined from such a model can have an operator with a single term . the three choices for the series connection ( spring , a dashpot , or a maxwell model ) are the three of four models that put a differential operator with a single term in the correct place so there could be a unique solution to the shape factorization probelm . similarly for the parallel connection . [ proof of theorem [ thm : global ] ] clearly a globally identifiable model is locally identifiable . by corollary [ cor : globidconstzero ] , we must be able to construct such a globally identifiable model by adding at each step either a spring , dashpot , maxwell or voigt element at each step , but when adding a maxwell element it must be used in series and when using a voigt element it must have been added in parallel . however , adding a maxwell element in series can be achieved by adding a spring and then a dashpot both in series . similarly , adding a voigt element in parallel can be achieved by adding a spring and then a dashpot both in parallel . hence , we can work only adding springs or dashpots at each step . adam mahdi was partially supported by the vpr project under nih - nigms grant # 1p50gm094503 - 01a0 sub - award to ncsu . nicolette meshkat was partially supported by the david and lucille packard foundation . seth sullivant was partially supported by the david and lucille packard foundation and the us national science foundation ( dms 0954865 ) . nichols ww ( 2011 ) mcdonald s blood flow in arteries : theoretical , experimental and clinical principles . london : hodder arnold , 6th ed . / wilmer w. nichols , michael f. orourke , charalambos vlachopoulos . edition .

one of the key open questions in the pursuit of a quantum theory of gravity is that of understanding its behaviour in time - dependent scenarios . in the context of string theory this has recently led to intensive investigation of asymptotically de sitter spacetimes to see what the prospects are for developing a de sitter / cft correspondence @xcite . another line of approach has been to construct simple time - dependent solutions that provide ( at least to leading order ) consistent backgrounds for string theory . this has been carried out in asymptotically flat spacetimes , and recently extended to include the asymptotically anti de sitter ( ads ) case . this latter situation is of interest since the ads / cft correspondence conjecture could be employed to relate the time - dependence to the behaviour of the non - perturbative field theory dual . in so doing one might be able to avoid ( or at least ameliorate ) difficulties that arise in understanding the behaviour of string theory on time - dependent backgrounds . in this paper we extend the search for time - dependent backgrounds to asymptotically locally flat / ads spacetimes . such spacetimes have a form of magnetic mass referred to as a nut charge . their euclideanized solutions have been a subject of study in recent years @xcite , in part because in the asymptotically locally ads case they are relevant to 2 + 1 dimensional `` exotic '' conformal field theories that live on the world volume of m2-branes ( and closely related theories ) , after placing them on squashed three spheres . we consider here the multiple analytic continuation of such spacetimes to form time - dependent bubble spacetimes with nut charge . such `` nutty bubbles '' form interesting time - dependent backgrounds that are asymptotically locally flat / ads , extending the variety of time - dependent cases previously studied @xcite . the stability of these time - dependent bubble solutions has also recently been investigated @xcite . we consider in this paper the different ( even - dimensional ) taub - nut - ads spacetimes and study their different analytic continuations . after a brief overview of bubble solutions obtained from spherically symmetric spacetimes in section 2 , we go on in section 3 to find three bubble solutions in the four dimensional case , all time - dependent . two of them are obtained from analytic continuation of taub - nut - ads spacetime and the last one is inspired from the first two bubble solutions . as all of these ` nutty ' bubbles arefour dimensional , no issue of classical gravitational instability arises in contrast to the situation considered in ref.@xcite . we then compute the boundary stress - energies associated with such solutions in section 4 . in section 5 we consider double analytic continuations of four dimensional topological nut - charged spacetimes . we find here some new static bubble - type spacetimes , and an interesting new time - dependent solution . we find that the presence of a nut charge induces an ergoregion , in which timelike curves must have non - zero momenta in a compact direction . in section 6 we consider double analytic continuations of four dimensional nut - charged spacetimes with a flat subspace and find a new static bubble - type spacetime that is a nut - charged generalization of the ads soliton , as well as an interesting new time - dependent solution : it is asymptotically ads , yet the bubble evolution is milne - like throughout . similar to topological nut spacetimes , the presence of a nut charge induces an ergoregion in the static case . in section 7 we search for well - behaved bubble solutions with nut charge in more than four dimensions . our search proves fruitless : we find in more than four dimensions that double ( or multiple ) analytic continuations give us spacetimes with at least two timelike coordinates . we then study the four dimensional kerr - nut spacetimes @xcite in section 8 and find that their asymptotic behaviour at late times is similar to that of the zero - nut charge case @xcite , and so such spacetimes can be candidate time - dependent backgrounds for string theory . however the presence of a nut charge can induce quantitatively different ( and sometimes even asymmetric ) particle creation and temporal evolution of the bubble . in section 9 we consider the ads generalizations of such spacetimes and find that they are beset with the same quantum instability problems as their non - rotating counterparts . we conclude with some brief comments about generalizations to de sitter spacetimes . the general form of a four - dimensional static , spherically symmetric metric can be written as @xmath3 where @xmath4 and @xmath5 are functions that are determined by the einstein equations . a bubble solution is obtained by a double analytic continuation in the time coordinate ( @xmath6 ) and in some other combination of coordinates on the @xmath7-section , effectively changing it into a two - dimensional de sitter spacetime . three distinct possibilities arise from this procedure @xmath8 that are characterized by the topology of the two - dimensional de sitter spacetime at every fixed @xmath9 . locally these are all equivalent , being related to one another by the coordinate transformations @xmath10 globally they are all equivalent provided the coordinates @xmath11 are all unwrapped . however if any or all of them are identified , then the de sitter spacetime corresponds three distinct evolutions for the circle geometry they describe . for the metric ( [ bubds1 ] ) the de sitter evolution is that of a circle with exponentially large initial radius as @xmath12 that exponentially shrinks with proper time to a minimal value and then expands again the worldsheet is like a candlestick . for ( [ bubds2 ] ) the circle begins with zero radius at @xmath13 and then expands exponentially like a bowl . for ( [ bubds3 ] ) the circle begins with infinitesimally small initial radius at @xmath14 and exponentially expands with increasing proper time like a trumpet . the bubble is located at the largest positive root @xmath15 of @xmath4 , assumed to be at least as great as the largest positive root of @xmath5 ( otherwise there is a wormhole geometry ) . the coordinate @xmath16 can either be unwrapped or identified with period @xmath17 unless @xmath18 , in which @xmath16 must have period @xmath19 . the radial variable is then restricted to the range @xmath20 and the solutions ( [ bubds1]-[bubds3 ] ) have a @xmath21 symmetry . the spacetime is geodesically complete for each . at any fixed time , @xmath22 corresponds to the circle of minimal circumference , referred to as the bubble . this circumference exponentially contracts to @xmath23 after which time it exponentially expands for the solution ( [ bubds1 ] ) . for ( [ bubds2 ] ) this circumference has zero size at @xmath13 after which it expands exponentially . this spacetime could matched smoothly at @xmath13 onto any spacetime with a hypersurface of matching extrinsic curvature and 3-metric ( including its time - reversed counterpart ) , and so can be regarded as describing the nucleation of a bubble ( or its collapse if @xmath24 ) . for ( [ bubds3 ] ) the circumference exponentially expands from infinitesimally small values beginning at @xmath25 . it is not time - reversal invariant , and so has a counterpart in which the bubble undergoes exponential contraction . it is straightforward to show that any null curve in spacetimes for which the bubble is expanding has @xmath26 at late times , where @xmath27 is any pair @xmath28 , and the overdot refers to a derivative with respect to proper time . hence observers at different values of @xmath29 will eventually lose causal contact since at late times any null geodesic can only traverse exponentially vanishing changes in @xmath29 . null geodesics at fixed @xmath29 obey the equation @xmath30 where @xmath31 is the conserved momentum in the @xmath16-direction and @xmath32 is the conserved energy . ( [ nullfixphi ] ) is an equation for a particle in an effective potential @xmath33 @xmath34 , with all allowed motions having @xmath35 . if @xmath18 and the spacetime is asymptotically flat , then null geodesics oscillate between some minimal and maximal values of @xmath36 , and so by appropriate choices of @xmath37 observers at any two differing values of @xmath36 can be causally connected . however if @xmath38 then the effective potential diverges at @xmath22 , and so there will always be some region @xmath39 which is not in causal contact with observers in regions @xmath40 . if the spacetime is not asymptotically flat then the effective potential may become positive for some @xmath41 , in which case observers in this region are causally disconnected from the region @xmath42 . since @xmath43 it is always possible for observers at different values of @xmath44 to remain in causal contact at fixed @xmath36 , and so any two observers in causal contact at differing values of @xmath36 will also be in causal contact if they are also at differing values of @xmath16 . the above results are generalizable to higher - dimensional spacetimes @xcite . the same classical stability and quantum instability properties hold for all of these bubbles as for the ones analyzed in ref . the ( 3 + 1)-dimensional taub - nut - ads solution is @xcite : @xmath45 where @xmath46 where @xmath47 is the nonvanishing nut charge , @xmath48 is related to the cosmological constant @xmath49 by the relation @xmath50 . its euclidean version is found by setting @xmath51 : @xmath52 where @xmath53 the coordinate @xmath16 parametrizes a circle , which is fibered over the sphere parametrized by @xmath54 the geometry of a constant-@xmath36 surface is that of a hopf fibration of @xmath55 over @xmath7 . to avoid singularities the periodicity of @xmath16 must simultaneously be @xmath56 and @xmath57 , which forces one of two conditions : @xmath58 or @xmath59 depending on the value of the largest root @xmath15 of @xmath60 . the solution ( [ nutmass ] ) is referred to as the nut solution the fixed point of the @xmath61 isometry is 0-dimensional . for the special case @xmath62 and @xmath63 , it is pure ads . solution ( [ boltmass ] ) is referred to as the bolt solution the fixed point of the @xmath64 isometry is 2-dimensional , as is the case for a black hole . we can obtain bubble spacetimes by forming several other analytic continuations under the constraint that there be only one timelike direction . these metrics are all generalizations of ( [ bubds1]-[bubds3 ] ) : @xmath65 @xmath66 @xmath67 and all are exact solutions to the einstein equations . each is nonsingular and geodesically complete . for large @xmath36 they all can easily be shown to approach ads@xmath68 if @xmath69 . the geometry of sections of fixed @xmath36 at each constant time slice is that of a twisted torus @xmath70 using the notation of the previous section , at each constant time slice @xmath71 , the twisted torus has one fundamental circle respectively parametrized by a coordinate @xmath72 the other twisted circle is respectively parameterized by a shifted coordinate @xmath73 . the torus becomes exponentially more twisted with time . if @xmath74 the geometry of fixed @xmath36 sections is @xmath75ds@xmath76 the quartic function @xmath4 in eq . ( [ ftnlo ] ) has two real roots , one at @xmath77 and one at @xmath78 respectively . without loss of generality we can locate the bubble at the positive root of @xmath4 , since bubble solutions exist for any values of the parameter @xmath79 , positive , zero or negative ( unlike the spherical bubbles of the previous section ) . the penrose diagrams of the asymptotically flat bubble spacetimes obtained from ( [ tnbub1])-([tnbub3 ] ) are given in figures ( [ fig1])-([fig3 ] ) . the penrose diagrams of the asymptotically ads bubble spacetimes ( [ tnbub1])-([tnbub3 ] ) are similar to ( [ fig1])-([fig3 ] ) , by changing the diagonal and off - diagonal lines in these figures to a cylinder which encloses the bubble . the bubble is located on the surface of the shaded region at @xmath80 , where @xmath15 is the root of @xmath81 in the asymptotically flat case ( @xmath82 ) we obtain from ( [ ftnlo ] ) @xmath83 for the location of the bubble . if @xmath84 the function @xmath4 monotonically approaches unity as @xmath85 , whereas if @xmath86 it has a local maximum . in the asymptotically ads case @xmath4 has a local maximum and minimum outside of the bubble if @xmath87 ; otherwise it monotonically increases . the only restriction on the periodicity of @xmath16 is that it has the value@xmath88 which equals @xmath89 in the asymptotically flat case . using the notation of the previous section , it is straightforward to check that observers at different values of @xmath29 eventually fall out of causal contact , and that observers at fixed @xmath29 but differing values of @xmath90 can remain in causal contact in both the asymptotically flat and ads cases . consider next the double analytic continuation @xmath91 applied to the metric ( [ tnlo ] ) : @xmath92 this metric is simply a local patch of one of the solutions ( [ bubds1]-[bubds3 ] ) ; for example under the coordinate transformation @xmath93 \nonumber\end{aligned}\ ] ] the metric ( [ tnbub1 ] ) is recovered . the lapse function is @xmath94 and so is non - vanishing everywhere except @xmath95 . in the coordinates ( [ tnb1 ] ) the metric has a killing vector @xmath96 and so describes a static bubble , not evolving in time . the norm of this killing vector is @xmath97\sin ^{2}\theta \label{timecoeff}\ ] ] and becomes spacelike unless @xmath98 where @xmath99 \label{thetalimits}\ ] ] taking @xmath100 . these limits define an ` ergocone ' for the spacetime , in which @xmath101 ( @xmath102 ) for all timelike curves in the region @xmath103 ( @xmath104 ) . for @xmath105 ( or @xmath106 ) the angle ( [ thetalimits ] ) becomes vanishingly small at large @xmath36 , and the ergocone shrinks to a pair of lines on the @xmath107 axis where the lapse function vanishes . observers using the coordinates ( [ tnb1 ] ) can , on a surface fixed @xmath36 ( with respective spacelike / timelike unit normals @xmath108 and @xmath109 ) , associate a quasilocal mass density with this spacetime @xmath110 where we have used the counterterm method @xcite to compute the mass density . we therefore obtain a total quasilocal mass @xmath111 of @xmath112 where the integration over @xmath113 is from @xmath114 to @xmath115 and then from @xmath116 to @xmath117 * * we note that if @xmath118 the total mass can be either positive or negative , a situation reminiscent of one that can occur for topological black holes @xcite . if @xmath119 is finite then this surface penetrates the ergocone , where @xmath120 is restricted to the interval @xmath121 at large @xmath36 . since the lapse function is positive in this region , we can perform a similar calculation ; the mass density is again given by ( [ massdens ] ) and we obtain @xmath122 where @xmath15 is the largest positive root of @xmath4 . somewhat counterintuitively , there is no angular momentum associated with the killing vector @xmath123 in either case , despite the presence of an ergoregion . it is straightforward to show that both the extrinsic curvature and counterterm contributions to the angular momentum density are identically zero . interpretation of the preceding results must be taken with some care . since at fixed @xmath124 the metric ( [ tnb1 ] ) describes a torus , observers at @xmath125 are separated from observers at @xmath126 by a line of vanishing lapse , and so their surface does not actually enclose the bubble . however observers on one side of this boundary ( e.g. @xmath127 ) could infer the above results by making use of the @xmath128 symmetry of their metric and assuming that a corresponding set of observers existed on the other side that were making similar calculations / measurements . the large-@xmath36 boundary stress - energy tensor for the bubble solution ( [ tnbub1 ] ) can be calculated by the well known counterterm method , @xmath129 as one can see , @xmath130 which means that the bubble solution has a negative mass and @xmath131 which means there exists a negative pressure in the @xmath16 direction . we see that the presence of a nut charge enhances these respective values in the next - to - leading - order term . the boundary energy momentum tensor components for the bubble solution ( [ tnbub2 ] ) also are given by the same relations ( [ tnb3stress ] ) changing coordinates to @xmath132 and @xmath133 in the last equation of ( [ tnb3stress ] ) . similarly , the components for the bubble solution ( [ tnbub3 ] ) also are given in ( [ tnb3stress ] ) changing coordinates @xmath134 and sinh@xmath135 in the last equation . we can compare the above boundary stress - energy tensor components of the bubble solution with the components of boundary energy momentum tensor for the taub - nut - ads spacetime ( [ tnlo ] ) . the non - vanishing components of the latter are given by : @xmath136 as a consistency check , we see that in special case of @xmath74 for which ( [ tnlo ] ) reduces to the 4-dimensional schwarzschild - ads black hole , the covariant energy - momentum tensor components of taub - nut - ads spacetime ( [ tnlo ] ) reduce to the same as obtained in @xcite . the equivalence of the non - vanishing boundary energy - momentum tensor of the bubble solution with the components of boundary energy momentum tensor for the taub - nut - ads spacetime ( [ tnlo ] ) is a direct consequence of analytic continuation . so the boundary cft quantities related to the energy - momentum tensor ( like the central charge or casimir energy ) are the same for the two nutty bubble spacetimes and the taub - nut - ads spacetime . such an equivalence was observed for the boundary energy - momentum tensors of bubble solutions derived from analytic continuation of the five - dimensional schwarzschild - ads metric and schwarzschild - ads metric @xcite . although the boundary theory is not known for the bubble metrics ( [ tnbub1])-([tnbub3 ] ) ( in contrast to the 5d - schwarzschild bubble @xcite , for which the boundary theory is @xmath137 sym living on @xmath75ds@xmath138 ) , since the boundary is conformally flat we can use the following relation for the stress tensor @xcite : @xmath139 where @xmath140and @xmath141 are conserved quantities constructed from the curvatures via @xmath142 and @xmath143 is the traceless part . for the boundary metric of the bubble ( [ tnbub1 ] ) , the functions @xmath144and @xmath141 are : @xmath145 @xmath146 by fixing the constants @xmath147 and @xmath148 from the boundary theory , the result of ( [ s ] ) is in agreement with ( [ tnb3stress ] ) in the limit @xmath149 , since the two constants must vanish due to absence of the conformal anomaly in the odd dimensional boundary theory . consequently the above comparison of the stress tensor ( [ tnb3stress ] ) to ( [ s ] ) does not result in any non - trivial connection between them . in other words , in the limit @xmath85 , all of the stress tensor components ( [ tnb3stress ] ) vanish and since the boundary conformal anomaly vanishes in odd dimensions , so the boundary stress tensor ( [ s ] ) also vanishes . these bubbles have in four - dimensional euclidean space the general form @xmath150 where @xmath151 there are three distinct analytic continuations : @xmath152 @xmath153 @xmath154 and these all satisfy the einstein equations . the @xmath113-coordinate is no longer periodic and takes on all real values . these metrics can , after a coordinate transformation be written in the form @xmath155 where @xmath156 for the respective cases ( [ top1]-[top3 ] ) . they are all locally equivalent to each other under a coordinate transformation , and so it is tempting to conclude that they are globally equivalent since the @xmath157-coordinate is not periodic . however at fixed @xmath158 the @xmath159 section describes ( for @xmath106 ) a two - dimensional anti de sitter spacetime . this spacetime can have non - trivial identifications @xcite and so can describe either a two - dimensional black hole @xcite as in eqs . ( [ top1],[top3 ] ) ( the latter being a special case of a massless extremal black hole ) or else standard ads@xmath160 , as in eq . ( [ top2 ] ) . another clear difference between this case and the previous cases is that the quartic function @xmath161 in ( [ ftopnut ] ) can have four real roots , which we label @xmath162 in increasing order of size . from this perspective these metrics can represent two different situations . first , for @xmath163 ( or @xmath164 ) , these spacetime represent static bubbles . the coordinate @xmath36 has the range @xmath165 ( or alternatively @xmath166 ) , and the @xmath113-coordinate takes on all possible values . at large @xmath167 it is straightforward to show that these metrics are asymptotically ads . both invariants @xmath168 and @xmath169 are finite everywhere and the spacetime has no singularities . since @xmath170 is not periodic , the bubble does not close back on itself in this sense it is not really a bubble , but it should be a valid background for string theory . in this case each of the metrics contains an ergoregion given by the expression @xmath171 the structure of the ergoregion depends upon the value of @xmath172 and @xmath173 . consider first the metric ( [ top3 ] ) , ( @xmath174 ) for which the ergoregion is independent of @xmath170 . if @xmath175 , then there will be an ergoregion everywhere except for a very narrow strip of infinite length near the bubble where @xmath176 , with @xmath177 . the strip slightly broadens as @xmath79 increases , as can be seen from figure ( [ fig31 ] ) , where the function @xmath178versus @xmath36 is plotted . as @xmath179 from above @xmath180 gets larger , diverging at @xmath181 , at which the ergoregion vanishes . once @xmath182 there is no ergoregion . for the metric ( [ top2 ] ) , with @xmath183 , there will be no ergoregion for @xmath184 . for @xmath185 @xmath186 , the ergoregion is confined to the right half of the @xmath187 plane bounded by a curve with a vertical asymptote at @xmath188 , and a horizontal asymptote at @xmath189 . this curve has a mirror image about @xmath190 and the ergoregion for @xmath191 has a mirror image for negative @xmath170 . a typical ergoregion is plotted in figure ( [ fig32 ] ) for special values of @xmath192 in this figure the vertical axis is @xmath170 and horizontal axis is @xmath193 for the metric ( [ top1 ] ) , with @xmath194 , the ergoregion is confined to a strip outside the event horizon at @xmath195 bounded by a curve and its mirror image in @xmath170 that asymptote for large @xmath36 at @xmath196 for @xmath182 . this curve smoothly intersects the location of the bubble at the event horizon . a typical ergoregion is plotted in figure ( [ fig33 ] ) for special values of @xmath197 as @xmath179 from below , the strip becomes increasingly wide , and at @xmath198 the boundary curve asymptotes to @xmath199 . for @xmath185 @xmath186 , the boundary of the ergoregion has a vertical asymptote at @xmath188 , and covers the remainder of the @xmath187 plane for larger values of @xmath36 . if @xmath200 has 4 real roots , a second possibility emerges : the coordinate @xmath36 can have the range @xmath201 , and can be transformed to a periodic variable . removal of conical singularities at @xmath202 is satisfied by requiring @xmath203 , which is @xmath204 or alternatively , if @xmath205 @xmath206 \label{toproots2}\ ] ] since for the function @xmath200 we must have @xmath207 . ( [ toproots2 ] ) has no solutions since its right hand side is always negative . however if @xmath208 then conical singularities can be removed . if @xmath209 then @xmath62 , whereas if @xmath210 we can attempt to obtain a new spacetime by setting @xmath211 , @xmath212@xmath213 and then taking the limit as @xmath214 . this is possible only if @xmath106 , since a rescaling of the coordinates @xmath215 by factors of @xmath216 yields a singularity in the metric . however if @xmath106 one of the roots becomes @xmath217 at which a singularity is located . if @xmath62 then @xmath218 in order that @xmath200 be real . the metric can be written as * * @xmath219 where @xmath220 $ ] , @xmath16 has the period @xmath221 , and @xmath222 strictly speaking this is not a bubble spacetime : instead the @xmath223 section is the surface of a deformed 2-sphere . the ergoregion is now located in a band about the equator at @xmath224 , whose specific boundary curves are complicated functions of @xmath172 . if @xmath174 this band is very narrow and exists only for a very small range of @xmath47 , and it remains constant everywhere . however if @xmath225 the band will change shape as @xmath170 increases . note that if we set @xmath226 then we must have at least one conical singularity , since the condition @xmath227 ca nt be satisfied unless @xmath106 , in which case there is again a singularity . the black hole ( @xmath194 ) becomes a compact region , surrounded by a deformed ` ergosphere ' for small values of @xmath170 . finally , we can continue to the metric @xmath228 which locally can be transformed to one of the metrics ( [ top1]-[top3 ] ) via a change of coordinates ; it can be regarded as the extension of the @xmath229 case if @xmath230 is not periodic . however if the @xmath230-coordinate in ( [ topnutperiodic ] ) is identified then these metrics are not globally equivalent , yielding new bubble solutions . if @xmath231 ( or @xmath164 ) , the metric ( [ topnutperiodic ] ) describe a bubble spacetime in which the bubble expands from zero size to the finite size @xmath232 and then contracts back to zero size again . all spacetime events are causally connected to one another , and near the initial expansion ( or final contraction ) the scale factor is linear in time and the spacetime behaves like a milne universe . if @xmath201 then the metric ( [ topnutperiodic ] ) describes a compact time - dependent 3-geometry of a twisted circle over a 2-sphere . from the previous analysis , conical singularities are avoided only if either @xmath62 or @xmath106 . in each case the spacetime oscillates , with the @xmath55 represented by the @xmath230 coordinate repeatedly expanding and contracting . there are again no event horizons . these spacetimes are not really bubbles because the @xmath9 section is a compact space . however they are candidate time - dependent backgrounds . since all of the topological backgrounds ( [ top1])-([top3 ] ) and ( [ topnutperiodic ] ) are four dimensional they are stable , similar to the other nutty bubbles ( [ tnbub1])-([tnbub3 ] ) . we close this section by noticing that in the case of ( [ topnutperiodic ] ) , null geodesics relevant for s - waves behaviour satisfy : @xmath233 the integration involves the square root of a quartic function and so finding a closed - form relation between coordinates @xmath36 and @xmath157 for a null geodesic is quite problematic . the relation between these two coordinates is essential to obtain a relation between phase differences of two geodesics on @xmath234 and @xmath235 . this relation can be used as an evidence for particle creation . we shall not consider this complicated problem here . however in the case of kerr - nut bubbles , the integration can be done and a simple relation is obtained for null geodesics , eq . ( [ trnull ] ) , as we shall later consider . another class of nut - charged spacetimes has the metric @xmath236 where @xmath237 has only two real roots , and the @xmath238 section describes a flat two - dimensional space , possibly with identifications . there are two distinct double analytic continuations : @xmath239 where @xmath240 . the @xmath241 section is now a two - dimensional milne / rindler spacetime , which can also be written as @xmath242 which is a nut - charged generalization of the ads soliton @xcite ; we shall continue to refer to it as a bubble . without loss of generality the bubble ( [ nutsoliton ] ) can be located at the positive root of @xmath243 since @xmath79 can take on all real values . the cases @xmath240 are locally equivalent under a coordinate transformation , but will be globally distinct if @xmath230 is periodically identified . when @xmath244 the bubble undergoes a milne - type evolution . at any fixed @xmath245 the geometry of the bubble is that of a twisted torus , with the coordinate @xmath230 parametrizing one circle , and with the other twisted circle parameterized by the shifted coordinate @xmath246 . * * turning next to the causal structure of the bubble ( [ nutrind ] ) , we consider two observers at different coordinates @xmath247 in the background of bubble ( [ nutrind ] ) , a null curve satisfies @xmath248 and so @xmath249 as @xmath250 , we have @xmath251 , and so observers at two points with different @xmath230 can communicate to each other for arbitrarily large @xmath252 . the penrose diagram of the bubble spacetime ( [ nutrind ] ) with @xmath244 is given in figure ( [ fig34 ] ) . any radial null ray leaving the surface of the bubble reaches infinity in a finite amount of coordinate time @xmath253 , and there is always an infinite amount of proper distance between the bubble and spatial infinity.the bubble is located on the surface of the shaded region at @xmath22 , where @xmath15 is the root of @xmath254 when @xmath194 the geometry is that of a static bubble with an ergoregion . writing the metric ( [ nutrind ] ) for this case as @xmath255 we see that the lapse function is @xmath256 and so vanishes at the rindler horizon @xmath190 . the ergoregion is described by @xmath257 and at large @xmath36 includes the entire @xmath187 plane except for a strip bounded by a pair of curves opposite the @xmath36-axis that asymptote to @xmath258 . as @xmath36 decreases , this strip widens , becoming infinitely wide at the bubble edge . the ergoregion is similar to that given in fig . ( [ fig32 ] ) , except that the vertical asymptotes are at the bubble edge @xmath259 . using the same coordinates as ( [ nutrinderg ] ) , we have @xmath260 and so there is no restriction as to the maximum change of coordinate @xmath170 for points on the null curve as @xmath261 . so observers at any two points with different values of @xmath170 can communicate with each other . in this section we consider the possibility of extending our nut - charged bubble solutions to higher dimensions . the six dimensional taub - nut - ads is a fibration over @xmath262 @xcite : @xmath263 if @xmath264 is the largest root of @xmath265 , then singularities in ( [ tneuc ] ) are avoided by choosing the mass parameter and the function @xmath265 by @xmath266 and @xmath267 the coordinate @xmath268 must be periodic with the period @xmath269 in performing a single analytic continuation from the euclidean metric ( [ tneuc ] ) we are forced to introduce two timelike directions unless we analytically continue @xmath16 . the reason is that there is a single nut charge @xmath270 and all analytic continuations in one of the @xmath7 sections forces @xmath271 . this in turn forces an additional analytic continuation in the other @xmath7 section , yielding two timelike directions . for example performing the following analytic continuation @xmath272 on the metric ( [ tneuc ] ) yields @xmath273 where @xmath274 \label{vtnb1}\ ] ] to find the structure of the bubble , we set @xmath275 where @xmath276 then the induced metric on the bubble is @xmath277 \label{nut1}\ ] ] note that the angle @xmath278 will be restricted to within a certain range in order for @xmath279 to be a timelike killing vector . another continuation is @xmath280 and this gives @xmath281 with the induced bubble metric : @xmath282 \label{nut3}\ ] ] similarly , the following continuation : @xmath283 gives @xmath284 with the induced bubble metric : @xmath285 \label{nut4}\ ] ] which again has two timelike directions . similar considerations apply for all higher - dimensional taub - nut - ads solutions have @xmath286 fibrations over a product of @xmath7 s . for example the 8-dimensional solutions has a @xmath286 fibration over @xmath287 @xcite : @xmath288 where @xmath289 in this case , the mass parameter @xmath79 takes the value @xmath290 to have a nut solution . the coordinate @xmath268 has a period @xmath291 , to avoid a conical singularity at @xmath292 similarly , the ten dimensional taub - nut - ads solution is a result of a @xmath286 fibration over @xmath293 @xmath294 @xmath295 where for a nut solution , the mass parameter and function @xmath4 are given by : @xmath296 and @xmath297 the coordinate @xmath268 must has a period @xmath298 , to avoid a conical singularity at @xmath292 the form for a general fibration over @xmath299 is the same . all analytic continuations in the @xmath7 sections force the introduction of more than one timelike direction because the nut parameter must be analytically continued to avoid a complex metric . another possibility is a over @xmath300 . in six dimensions this is @xcite : @xmath301 where @xmath147 is given by @xmath302 and @xmath303 is the metric over the @xmath300 , which is given by @xmath304 the function @xmath265 in the metric is the same as function ( [ vtneuc ] ) , appearing in the taub - nut - ads fibered over @xmath305 in the limit @xmath306 the metric ( [ tn6cp2 ] ) approaches the metric considered in @xcite . the situation here is somewhat different in that we have not found any analytic continuation of the coordinates of the @xmath307 section that yields a single timelike direction . further higher - dimensional generalizations are products of @xmath7 and @xmath308 and the same considerations apply , namely more than one timelike direction appears after analytic continuation unless @xmath16 is the coordinate continued . we consider next rotating nut solutions in four dimensions . such solutions are given by the class of kerr - nut metrics , which have the ( euclidean ) form @xcite @xmath309^{2}+\sin ^{2}\theta \lbrack ad\chi -(r^{2}-n^{2}-a^{2})d\phi ] ^{2}}{[r^{2}-(n+a\cos \theta ) ^{2 } ] } \\ + [ r^{2}-(n+a\cos \theta ) ^{2}](\frac{dr^{2}}{v(r)}+d\theta ^{2 } ) \end{array } \label{kn}\ ] ] where @xmath310 applying the following analytic continuations @xmath311 on the metric ( [ kn ] ) yields : @xmath312^{2}+\cosh ^{2}t[ad\chi -(r^{2}+n^{2}-a^{2})d\phi ] ^{2}}{[r^{2}+(n - a\sinh t)^{2 } ] } \\ + [ r^{2}+(n - a\sinh t)^{2}](\frac{dr^{2}}{\widetilde{v}(r)}-dt^{2 } ) \end{array } \label{kn1}\ ] ] where the function @xmath313 @xmath314is given by : @xmath315 the largest root of @xmath313 is @xmath316 and so the induced metric on the bubble is @xmath317dt^{2}+\frac{\cosh ^{2}td% \tilde{\chi}^{2}}{[\widetilde{r}_{+}^{2}+(n - a\sinh t)^{2 } ] } \label{kn1i}\ ] ] where @xmath318 . for all values of @xmath319 the coefficient of time is negative definite . so , unlike the case of kerr - ads @xcite , the induced bubble metric obtained through the analytic continuation of the kerr - nut metric in four dimension has a well defined behaviour . in the special case of @xmath320 ( [ kn1i ] ) reduces to the kerr bubble obtained in @xcite . for @xmath321 the bubble evolves like ds@xmath160 . this situation is similar to the case of schwarzschild and kerr bubbles @xcite . however the presence of a nut charge causes the time evolution to become asymmetric , with the bubble approaching a minimal radius and then later a maximal radius near @xmath322 ( or the time - reversal of this , depending on the signs of the parameters ) . for very large @xmath319 the bubble metric becomes @xmath323 which is independent of the nut charge . in this limit , the circle parametrized by @xmath324 remains at a fixed radius , so we do not have any time evolution for the bubble . so when the bubble evolves like ds@xmath325 the effect of nut charge is to shift the value of the coordinate @xmath157 to satisfy the above mentioned inequality . on the other hand , for large values of @xmath319 the nut charge does not have any physical effect on the induced bubble metric ( [ kn1i ] ) , and the spacetime evolves like a milne universe at late times . the penrose diagram of the kerr - nut bubble ( [ kn1i ] ) is given in figure ( [ fig4 ] ) . for the four dimensional kerr - nut bubble ( [ kn1 ] ) , the null geodesics relevant for the s - waves satisfy : @xmath326 integrating the above equation , we find the relation @xmath327 where @xmath147 is a constant labeling the different null geodesics and @xmath328refers to outgoing and incoming modes . an s - wave at @xmath234 is given by a purely positive frequency @xmath329 , where @xmath330 here @xmath331 and @xmath332 are the minkowski coordinates in the time and radial directions at infinity . for this s - wave , the phase difference between two geodesics at @xmath333 and @xmath334 on @xmath335 i.e. @xmath336can be related to the phase difference traced back to @xmath337 and @xmath338on @xmath339 where @xmath340 the bubble metric in the @xmath341section for very large @xmath36 ( such that @xmath342 ) and very large @xmath319 ( @xmath343 ) , is given by @xmath344 where in terms of the original coordinates @xmath345 using relation ( [ trnull ] ) , we find that the future coordinate @xmath346 of the null geodesic is @xmath347 similarly , the past coordinate @xmath348 is @xmath349 as with the @xmath106 case @xcite , we can find @xmath350 between two geodesics . we solve the equation ( [ vv ] ) for @xmath351 in terms of @xmath352 ( taking geodesic @xmath353 as the reference geodesic ) and then substitute the result in eq.([uu ] ) . this gives @xmath354 indicating that a positive frequency s - wave at future can not continue back to a purely positive frequency wave at past and so this means that we have a particle creation in bubble spacetime ( [ kn1 ] ) . from the above obtained result , one can calculate the bogoliubov coefficients ( which determines the extent of particle creation ) , the total number of particles and other qualitative results about the particle creation . another analytic continuation of the metric ( [ kn ] ) is obtained through @xmath355 which yields : @xmath356^{2}+\sinh ^{2}t[ad\chi -(r^{2}+n^{2}+a^{2})d\phi ] ^{2}}{[r^{2}+(n+a\cosh t)^{2 } ] } \\ + [ r^{2}+(n+a\cosh t)^{2}](\frac{dr^{2}}{v^{\ast } ( r)}-dt^{2 } ) \end{array } \label{kn2}\ ] ] where the function @xmath357 is given by : @xmath358 in this case @xmath359 @xmath360 , the induced metric on the bubble is given by : @xmath361dt^{2}+\frac{\sinh ^{2}td% \widehat{\chi } ^{2}}{[r_{+}^{\ast 2}+(n+a\cosh t)^{2 } ] } \label{kn2i}\ ] ] where @xmath362 so again we have a bubble solution . for @xmath363 the bubble evolves like ds@xmath160 , now with symmetric time evolution . however in this case there are two possible time - evolution scenarios . if @xmath47 and @xmath364 are both of the same sign , the bubble has only one minimal vanishing radius . if @xmath47 and @xmath365 have opposite sign then the bubble will expand to a maximum , shrink to zero , expand again to the same maximum before shrinking once again . for large @xmath319 the bubble metric becomes @xmath366 again , independent of the nut charge . hence in this limit , the circle parametrized by @xmath367 remains at a fixed radius , so we do not have any time evolution for the bubble . as with the last case , when the bubble evolves like ds@xmath325 the effect of the nut charge is to shift the value of coordinate @xmath157 to satisfy the above mentioned inequality . on the other hand , for large values of coordinate @xmath319 the nut charge does not have any physical effect on the induced bubble metric ( [ kn2i ] ) , whose late - time evolution is also milne - like . a typical penrose diagram of the kerr - nut bubble ( [ kn2i ] ) , when @xmath47 and @xmath365 have the same sign , is given by figure ( [ fig5 ] ) . also , by considering an s - wave in the bubble background ( [ kn2 ] ) , we can show that particle creation can occur . we can describe the causal structure of kerr - nut bubbles ( [ kn1 ] ) and ( [ kn2 ] ) as follows . if we consider two observers at different coordinates @xmath368 they are causally connected even for times @xmath369 to see this , we consider a null curve which satisfies @xmath370^{2}+\sinh ^{2}t[a\stackrel{\cdot } { \chi } -(r^{2}+n^{2}+a^{2})% \stackrel{\cdot } { \phi } ] ^{2}}{[r^{2}+(n+a\cosh t)^{2 } ] } & & \nonumber \\ + [ r^{2}+(n+a\cosh t)^{2}](\frac{\stackrel{\cdot } { r}^{2}}{v_{0}(r)}-% \stackrel{\cdot } { t}^{2 } ) & = & 0\end{aligned}\ ] ] where @xmath371 or @xmath357 and the overdot represents derivative with respect to an affine parameter labeling the curve . since we assumed @xmath372 we conclude that @xmath373 hence , for points on the null curve , when time coordinate @xmath374 , we do not have any restriction on the maximum change in the coordinate @xmath36 . so , two observers at two points with different values of @xmath36 can communicate with each other . the two killing vectors @xmath375 and @xmath376 yield two constants of the motion @xmath377 we can use the equation ( [ com ] ) to eliminate @xmath378 and @xmath379 from the null geodesics . in this case , for a fixed time @xmath319 the null geodesic equation reduces to that of motion of a particle in an effective potential : @xmath380 where the effective potential is : @xmath381 so geodesics exist in the region for which @xmath382 for example , on a constant time slice @xmath383 if we set @xmath384 the effective potential is plotted in figure ( [ fig6 ] ) . it is infinite at @xmath385 for @xmath386 the potential is positive definite and so a geodesic can not penetrate into this region . the four dimensional euclidean kerr - nut - ads spacetime is given by @xcite : @xmath387^{2}+h(\theta ) \sin ^{2}\theta \lbrack ad\chi -(r^{2}-n^{2}-a^{2})d\phi ] ^{2}}{\sigma ^{4}[r^{2}-(n+a\cos \theta ) ^{2 } ] } \\ + [ r^{2}-(n+a\cos \theta ) ^{2}](\frac{dr^{2}}{v(r)}+\frac{d\theta ^{2}}{% h(\theta ) } ) \end{array } \label{kerrnut}\ ] ] where the functions @xmath265 and @xmath388 are given by the following expressions : @xmath389 and @xmath390 there is a bolt solution @xmath391 given by @xmath392 , with @xmath393 ; noanalogue of the pure nut solution exists for this case @xcite . to avoid conical singularities in different sections of the spacetime , we must impose some constraints . for example , avoiding conical singularities in the @xmath394 section implies @xmath395 the period of the coordinate @xmath16 is @xmath396 where the left and right hand sides of the equality come from avoiding conical singularities in the @xmath397 section and ensuring regularity of the metric along the @xmath398-axis . in the previous equality , @xmath399 and @xmath400 is the real root of @xmath401 applying the following analytic continuations @xmath402 on the metric ( [ kerrnut ] ) yields : @xmath403^{2}+% \widetilde{h}(t)\cosh ^{2}t[ad\chi -(r^{2}+n^{2}-a^{2})d\phi ] ^{2}}{\sigma ^{4}[r^{2}+(n - a\sinh t)^{2 } ] } \\ + [ r^{2}+(n - a\sinh t)^{2}](\frac{dr^{2}}{\widetilde{v}(r)}-\frac{dt^{2}}{% \widetilde{h}(t ) } ) \end{array } \label{kerrnut1}\ ] ] where the functions @xmath313 and @xmath404 are given by : @xmath405 and @xmath406 at @xmath407 ( where @xmath408 ) , and for the coordinate @xmath157 in the range of @xmath409 @xmath410 the function @xmath411 is positive definite , and the induced metric on the bubble is given by : @xmath317\frac{dt^{2}}{\widetilde{% h}(t)}+\frac{\widetilde{h}(t)\cosh ^{2}t[ad\chi -(\widetilde{r}% _ { + } ^{2}+n^{2}-a^{2})d\phi ] ^{2}}{\sigma ^{4}[\widetilde{r}% _ { + } ^{2}+(n - a\sinh t)^{2 } ] } \label{kerrnut1i}\ ] ] where the critical coordinate @xmath412 is : @xmath413 on the other hand , for the range @xmath414 the function @xmath404 is negative definite and so the induced metric ( [ kerrnut1i ] ) does not describe a well defined bubble . in fact , the situation is similar to the induced bubble metrics obtained through analytic continuation of the kerr - ads metric in four dimensions @xcite . in that case the induced metric also changes signature at some critical time . the result of @xcite is a special case of our result when @xmath106 . another special case of the metric ( [ kerrnut ] ) is when @xmath415 in this case , the functions @xmath416 and from relation ( [ thetacriticalkerrnut ] ) , we have @xmath417 so we have a well behaved induced bubble metric @xmath418 \label{nuti}\ ] ] where we derived this form of the induced metric in section one . the other analytic continuation of the metric ( [ kerrnut ] ) is obtained through @xmath419 which yields @xmath356^{2}+h^{\ast } ( t)\sinh ^{2}t[ad\chi -(r^{2}+n^{2}+a^{2})d\phi ] ^{2}}{\sigma ^{\ast 4}[r^{2}+(n+a\cosh t)^{2 } ] } \\ + [ r^{2}+(n+a\cosh t)^{2}](\frac{dr^{2}}{v^{\ast } ( r)}-\frac{dt^{2}}{h^{\ast } ( t ) } ) \end{array } \label{kerrnut2}\ ] ] where the functions @xmath357 and @xmath420 are given by @xmath421 and @xmath422 and @xmath423 in this case at point @xmath359 ( for some values of parameters @xmath424 , we have @xmath425 ) , and for the coordinate @xmath157 in the range of @xmath409 @xmath426 the function @xmath420 is positive definite , and the induced metric on the bubble is given by : @xmath361\frac{dt^{2}}{h^{\ast } ( t)}+% \frac{h^{\ast } ( t)\sinh ^{2}t[ad\chi -(r_{+}^{\ast 2}+n^{2}+a^{2})d\phi ] ^{2}% } { \sigma ^{\ast 4}[r_{+}^{\ast 2}+(n+a\cosh t)^{2 } ] } \label{kerrnut2i}\ ] ] where the critical coordinate @xmath427 is now @xmath428 on the other hand , for the range @xmath429 the function @xmath430 is negative definite and so the induced metric ( [ kerrnut2i ] ) does not describe a well defined bubble . our consideration of bubble spacetimes with nut charge has yielded several new candidate four - dimensional time - dependent backgrounds for string theory whose properties merit further exploration . we have found that the presence of a nut charge can considerably modify the evolution of the spacetime , particularly in the topological and rotating cases . the nut charge induces the presence of an ergoregion in the latter case , although for the spacetimes in which this occurs the bubble is not compact . the bubble ( [ nutrind ] ) is particularly intriguing , since its evolution is milne - like throughout . in higher dimensions the presence of a nut charge appears to forbid the existence of any bubble spacetime , since the double analytic continuation yields more than one timelike direction . we close with a few comments about de sitter spacetime . all of the above results in the ads case can , by the simple continuation @xmath431 be changed to taub - nut - ds bubbles . for example the tn - de sitter space can be obtained from ( [ tnlo ] ) , and is given by : where @xmath433 these spacetimes are expected to be better behaved , at least in the rotating case , since the functions @xmath434 and @xmath420 in eqs . ( [ kerrnut1 ] ) and ( [ hkerrnut2 ] ) respectively need never vanish if the parameters are appropriately chosen . we leave a study of this case for future work . 99 a. strominger , jhep * 0110 * ( 2001 ) 034 , _ hep - th/0106113 _ ; e. witten , _ hep - th/0106109 _ ; m. li , jhep * 0204 * ( 2002 ) 005 , _ hep - th/0106148 _ ; d. klemm , _ nucl.phys . _ * b625 * ( 2002 ) 295 , _ hep - th/0106247 _ ; v. balasubramanian , p. horava and d. minic , jhep * 0105 * ( 2001 ) 043 , _ hep - th/0103171 _ ; a. chamblin and n. d. lambert , _ phys . * b508 * ( 2001 ) 369 , _ hep - th/0107031 _ ; y. gao , _ hep - th/0107067 _ ; j. bros , h. epstein and u. moschella , _ phys . * d65 * ( 2002 ) 084012 , _ hep - th/0107091 _ ; s. nojiri and s. d. odintsov , jhep * 0112 * ( 2001 ) 033 , _ hep - th/0107134 _ ; e. halyo , _ hep - th/0107169 _ ; i. sachs and s. n. solodukhin , _ phys . rev . _ * d64 * ( 2001 ) 124023 , _ hep - th/0107173 _ ; a. j. tolley and n. turok , _ hep - th/0108119 _ ; s. nojiri and s. d. odintsov , _ phys . lett . _ * b519 * ( 2001 ) 145 , _ hep - th/0106191 _ ; c. m. hull , jhep * 0111 * ( 2001 ) 012 , _ hep - th/0109213 _ ; s. cacciatori and d. klemm , _ class . * 19 * ( 2002 ) 579 , _ hep - th/0110031 _ ; b. mcinnes , _ nucl . phys_. * b627 * ( 2002 ) 311 , _ hep - th/0110062 _ ; s. nojiri and s. d. odintsov , _ phys . lett . _ * b523 * ( 2001 ) 165 , _ hep - th/0110064 _ ; a. strominger , jhep * 0111 * ( 2001 ) 049 , _ hep - th/0110087 _ ; v. balasubramanian , j. de boer and d. minic , _ phys . rev . _ * d65 * ( 2002 ) 123508 , _ hep - th/0110108 _ ; y. s. myung , _ mod . * a16 * ( 2001 ) 2353 , _ hep - th/0110123 _ ; b. g. carneiro da cunha , _ phys . rev . _ * d65 * ( 2002 ) 104025 , _ hep - th/0110169 _ ; r. cai , y. s. myung , y. zhang , _ phys . _ * d65 * ( 2002 ) 084019 , _ hep - th/0110234 _ ; u. h. danielsson , jhep * 0203 * ( 2002 ) 020 , _ hep - th/0110265 _ ; s. ogushi , _ mod . lett . _ * a17 * ( 2002 ) 51 , _ hep - th/0111008 _ ; a. c. petkou and g. siopsis , jhep * 0202 * ( 2002 ) 045 , _ hep - th/0111085 _ ; r. cai , _ phys . _ * b525 * ( 2002 ) 331 , _ hep - th/0111093 _ ; t. shiromizu , d. ida and t. torii , jhep * 11 * ( 2001 ) 010 , _ hep - th/0109057 _ ; a. m. ghezelbash and r. b. mann , jhep * 0201 * ( 2002 ) 005 , _ hep - th/0111217 _ ; m. h. dehghani , _ phys . _ * d65 * ( 2002 ) 104003 , _ hep - th/0112002_.

an empirical tenet in the study of accreting compact objects is that their maximum radiative luminosity is about eddington luminosity , that is the limit where the inward - directed gravitational force is balanced by the outward - directed radiation force . for spherical accretion of fully ionized hydrogen , the eddington limit can be written as @xcite @xmath5 where @xmath6 is the thomson scattering cross section , @xmath7 is the proton mass , and @xmath8 is the black hole ( bh ) mass . the x - ray luminosity is generally used as a good proxy for the bolometric luminosity for accreting neutron stars and stellar mass bhs . the eddington argument proved very valuable in the early stages of x - ray astrophysics , to show that most galactic x - ray binaries are neutron stars @xcite . later on , dynamical mass measurements confirmed the existence of a population of galactic bhs with masses @xmath9@xmath10 ( * ? ? ? * for a review ) : they also satisfy the eddington limit , with peak luminosities @xmath11 erg s@xmath12 . the _ einstein _ satellite revealed for the first time a small number of non - nuclear x - ray sources in nearby galaxies ( particularly , in star - forming galaxies ) with apparent luminosities @xmath13 @xcite . however , it was not immediately clear whether these sources could be a distinct class from galactic stellar mass bh candidates , which were just being discovered around the same years . _ s relatively crude spatial resolution allowed for the possibility of source confusion in compact star - forming regions ; lack of long - term monitoring made it hard to distinguish between persistently luminous sources and transient events , _ e.g. _ , young supernovae ( sne ) missed by optical observers . in the following decade , as nearby galaxies were repeatedly observed with x - ray satellites such as _ rosat _ and _ asca _ , with superior spectral coverage and resolution , it became clear that some of those non - nuclear sources could not be sne , and had apparent luminosities well above the eddington limit of typical galactic stellar mass bhs . for example , the `` most likely interpretation '' of the brightest non - nuclear source in the spiral galaxy ngc1365 was `` an ultra - powerful x - ray binary , with either a highly super - eddington low - mass black hole or a very massive black hole '' with @xmath14@xmath15 , `` which may pose a challenge for stellar evolution models '' @xcite ; a third scenario , beamed emission , was also proposed for other sources @xcite . thirteen years later , distinguishing between those three alternative scenarios remains the unsolved fundamental question for this class of sources , despite the wealth of new discoveries from _ chandra _ , _ xmm - newton _ , _ suzaku _ and _ swift _ , intensive modeling of their x - ray spectral and timing properties , and integrating studies of their multiband counterparts . for a while , different names have been in use to identify this class of accreting bhs : extraluminous x - ray binaries @xcite , superluminous x - ray sources @xcite , intermediate - luminosity x - ray objects @xcite . in recent years , the community consensus has settled on ultraluminous x - ray sources ( ulxs ) , adopting the terminology first used by japanese _ asca _ teams @xcite . a handful ulxs are observed to exhibit strong x - ray variability at time scales as short as minutes @xcite , and many others have shown random variation at long time scales from days , weeks to years @xcite , confirming their compact nature . they are 10100 times more luminous than galactic black hole binaries ( bhbs ) , whose peak luminosity is about @xmath16 . as the eddington limit is proportional to the bh mass , this suggests that ulxs may contain bhs at least 10100 times more massive than galactic ones which are typically of 10 @xmath1 , and thus represents a new class of bhs with masses in the range of @xmath3@xmath4 @xmath1 , intermediate between stellar mass and supermassive bhs . however , some models suggest that the eddington limit could be violated by a factor up to @xmath1710 for stellar mass bhs @xcite , and the x - ray luminosity may be overestimated due to anisotropic emission caused by geometric beaming expected at high accretion rate @xcite . in these circumstances , intermediate mass black holes ( imbhs ) are not required to interpret ulxs . a combination of super - eddington and mildly beamed emission from stellar mass bhs could account for ulxs with an apparent luminosity up to @xmath17@xmath18 @xcite . some objects other than bhbs may appear like ulxs if their timing or multiwavelength properties are unavailable . young sne , especially of type iin , could be as luminous as @xmath19 in the x - ray band @xcite . their x - ray spectrum is usually thermal and soft , and will not show chaotic variability . young x - ray pulsars could theoretically be more luminous than @xmath16 but the number of such sources appears to be very small @xcite . in general , these objects can be distinguished from accreting sources with variability @xcite , but it is possible that a sn leaves behind a compact remnant which becomes x - ray luminous when a fallback disk forms @xcite and mixed properties may be seen . in a survey of ulxs in nearby galaxies with archival _ xmm - newton _ data , it was found that the x - ray spectra of a few sources can be decomposed into a power - law component plus a thermal plasma , and in one of them x - ray variability was detected , suggesting that they may be recent sne with an accreting compact core @xcite . background active galactic nuclei ( agn ) that happen to lie behind a nearby galaxy could be misidentified as ulxs in the galaxy . their cosmological location could be revealed if redshift is measured from optical lines ( _ e.g. _ , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? the fraction of background objects is usually high in massive , elliptical galaxies ( _ e.g. _ , * ? ? ? * ) . many ulxs are found to be spatially associated with nebulae , young star clusters , or globular clusters ; these are most likely located in the galaxy rather than foreground or background objects . even if they are identified as bhbs , their nature is still not unique in terms of the companion star ( low mass or high mass ) , the mode of the mass transfer ( roche - lobe overflow or wind - fed ) , or the mass of the accretor ( stellar mass or intermediate mass ) . for example , @xcite found that ulxs in young , star forming galaxies are much more variable than in old , elliptical galaxies ( associated with globular clusters ) . no matter what their nature is , ulxs are of great interest in astrophysics . if they are imbhs , their formation mechanism could be rather different from typical stellar mass bhs ; extreme environment , more massive and/or less abundant progenitors are likely required @xcite . imbhs could have played an important role in the formation of supermassive bhs in the early universe and be important targets to search for gravitational wave emission @xcite . instead , if they are stellar mass bhs , it is interesting to investigate why they are so distinct from their galactic cousins . they may be good candidates for the study of the accretion physics at near or super eddington accretion rates @xcite , and they may shed light on binary stellar evolution channels not seen in our galaxy . ulxs are usually defined as non - nuclear , point - like objects which at least once have been observed at an apparent isotropic x - ray luminosity higher than that of stellar mass galactic bhs : typically , @xmath20 erg s@xmath12 in the 0.310 kev band . some authors ( _ e.g. _ * ? ? ? * ) prefer to use @xmath21 ergs s@xmath12 , that is the eddington limit of the heaviest stellar bh ( @xmath22 ) expected at normal metallicity @xcite , to exclude a large number of `` faint '' ulxs that should be similar to normal galactic bhbs . such an empirical definition may have ambiguities and also include young x - ray pulsars , sne , and sn remnants as mentioned above . the consensus in the ulx community is to restrict the definition of ulxs to accreting bhs . therefore , we define ulxs as _ non - nuclear accreting bhs with peak luminosities inferred assuming isotropic emission above the eddington limit of normal stellar mass bhs_. this definition is confined to compact objects that are most likely accreting from a companion star in a binary system , but more exotic possibilities , for example isolated imbhs accreting via the bondi process from the surrounding medium are not excluded a priori . observationaly , variability information is required to define a ulx . recently , many ulxs are found to have a mechanical power output as high as in radiation @xcite . in one of them , the mechanical output is dominant while the radiative output is way below the ulx threshold @xcite . this source is apparently not a ulx , but speculated to be an offset ulx or quiescent ulx ( see section [ sec : bub ] for details ) . we thus define them as `` ultrapowerful sources ( upss ) '' for completeness , to standard for the same population of sources perhaps in a different state or viewing angle . a _ rosat_/hri catalog of luminous , non - nuclear x - ray sources by @xcite proved to be very important for bringing ulxs to the forefront of research , and for emphasizing the possible observational link with theoretically - predicted intermediate mass bhs , with masses @xmath23 . larger and more systematic x - ray source catalogs based on re - analysis of archival _ rosat_/hri data soon followed : in particular , by @xcite , by @xcite , and by @xcite , which included 106 sources above @xmath16 erg s@xmath12 . the first xmm - newton catalog was done by @xcite . catalog with 154 sources in 82 galaxies was produced by @xcite , and 226 objects were included in the catalog by @xcite , compiled from mixed literature sources . such works allowed the first population studies of these objects as a well recognized group , and in relation to the properties of their host galaxies . they also provided the rationale for longer , targeted observations with _ chandra _ and _ xmm - newton_. today , the largest catalog @xcite is based on the 2xmm serendipitous survey , and contains 470 ulxs , of which 367 not listed in previous compilations . updated catalogs based on _ chandra _ observations have been compiled by @xcite and @xcite . there are three basic ways to explain non - nuclear systems with apparent accretion luminosities @xmath24 erg s@xmath12 , using only standard accretion physics already included in @xcite s model . we can either increase the mass of the bh ( thus , increasing its eddington limit ) , or attribute the enhanced brightness to beamed emission , or require a mass accretion rate high enough to allow the source to exceed the eddington limit by a factor of a few . it is also possible that the most luminous ulxs may be explained by a combination of all three factors . assuming solar abundances , the _ apparent _ luminosity . ] of an accreting bh is : @xmath25 @xcite , where @xmath26 is the beaming factor , and @xmath27 is the dimensionless accretion rate at large radii ( typically , the mass transfer rate from the donor star ) , that is normalized to the eddington accretion rate , typical of efficient disk accretion ; hence , a dimensionless accretion rate @xmath28 produces a luminosity @xmath29 . other authors prefer to define @xmath30 , in which case the eddington limit is reached at @xmath31 . ] : @xmath32 . let us examine the main tenets of these alternative or complementary scenarios . _ strong beaming _ ( @xmath33 ) a decade ago , accreting bhs with strongly beamed emission were already well known in extragalactic astronomy ( bl lac objects and blazars : * ? ? ? * ) , and at least one galactic bh ( ss 433 ) was known to have semi - relativistic jets , massive outflows and likely anisotropic emission . naturally , a possible explanation for the early sample of _ rosat _ ulxs was that they were microblazars , or ss 433-like objects seen face - on @xcite . the relativistic beaming scenario soon ran into several fatal difficulties . it predicts that for every ulxs , there should also be a large number of lower - luminosity beamed sources : @xmath34 sources with apparent @xmath35 erg s@xmath12 for every ulx at @xmath36 erg s@xmath12 ; but there is no evidence of this population @xcite . instead , empirical luminosity functions show only @xmath9@xmath37 sources more luminous than @xmath16 erg s@xmath12 for every ulx at @xmath19 erg s@xmath12 @xcite , consistent with the higher end of the high - mass x - ray binary luminosity function . there is also no multiband evidence of the large number of jet sources beamed in other directions , required by this scenario @xcite . conversely , the presence of photoionized bubbles around several ulxs ( for example holmberg ii x-1 : @xcite ; ngc5408 x-1 : @xcite ) requires quasi - isotropic x - ray emission with luminosities @xmath38 erg s@xmath12 . finally , the majority of ulxs do not have radio counterparts and/or rapid x - ray variability , expected for relativistic beaming . for all those reasons , relativistic beaming is now ruled out as a general scenario . _ mild beaming and/or super - eddington accretion _ ( @xmath39 , @xmath40 ) when the accretion rates exceeds the eddington limit , radiatively - driven outflows are launched from the inner part of the disk , inside the spherization radius where the disk becomes geometrically thick @xcite . it was suggested @xcite that the wind may produce a funnel wall that scatters and collimates the emission along the axis perpendicular to the disk plane . the apparent high luminosity of ulxs is then a combination of this mild collimation ( with @xmath41 , @xcite ) and mild super - eddington emission ( @xmath42 ) . in this scenario , accretion rates @xmath31@xmath43 are enough to explain most or all ulxs up to luminosities @xmath44 erg s@xmath12 with bh masses @xmath45 . recent numerical radiation - magneto - hydrodynamical simulations @xcite supported and advanced earlier analytical results ( slim disk models ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and the radiation - hydrodynamical simulations of , among others , @xcite . an important result of the simulations is that even a moderately super - critical mass supply @xmath46 can produce a total luminosity @xmath47 ( in rough agreement with the standard analytical scaling used earlier ) , and an apparent luminosity @xmath48 for face - on observers , due to mild collimation in the magnetized outflow @xcite . an alternative explanation for ulxs relies on super - critical accretion but does not require collimation of the emerging radiation . it was suggested that radiation pressure - dominated accretion disks have strong density inhomogeneities due to the development of photon - bubble instabilities @xcite . radiation would escape from such disk regions at rates @xmath49 times above the classical eddington limit for stellar mass bhs . _ quasi - isotropic eddington luminosity _ ( @xmath50 , @xmath51 ) although some models and simulations offer us compelling reasons to ignore the eddington limit as a constraint for the ulx interpretation , some authors remain wary of this scenario . they point out that the eddington limit works well for galactic bhs and neutron stars . more significantly , the isotropic bolometric luminosity of the whole population of @xmath52 quasars at redshift @xmath53 in the sloan digitized sky survey is strongly eddington - limited at all bh masses , with only a very small fraction of sources between 1 and 3 @xmath54 @xcite . perhaps like the bumblebee , the eddington limit should not fly , but does . thus , it is argued that the null hypothesis for ulxs should be that their luminosities are @xmath55 , despite the difference on the gas reservoir between agn and binary systems , and their radiative anisotropy is only a factor of @xmath56 ( a standard accretion disk has a beaming factor @xmath57 in the absence of any other collimation ) . these conditions are possible to explain most ulxs up to a few @xmath58 with accreting bhs of masses @xmath59 , which is the theoretical upper limit for a direct collapse of a single stellar core in the current universe ( see section [ sec : bhmass ] ) . sources more luminous than that may require imbhs . for example , hlx-1 in eso243 - 49 , the most luminous ulx found to date , has a peak bolometric luminosity @xmath60 erg s@xmath12 @xcite . if it is eddington limited , a black hole mass of at least @xmath61 is inferred . even if super - eddington emission is allowed , an imbh is required . _ quasi - isotropic sub - eddington luminosity _ ( @xmath62 , @xmath63 ) finally , it is possible that ulxs are accreting at sub - eddington accretion rates with nearly isotropic emission or disk beaming , which requires the compact objects to be imbhs ( @xmath64 ) . then we expect to see similar state transition and x - ray spectral properties from both ulxs and galactic bhbs , scaled only by the mass . the emission scenarios outlined above require three different ranges of bh masses , with different formation mechanisms . _ ordinary stellar mass bhs _ ( @xmath65 ) the first advantage of explaining ulxs with ordinary stellar mass black holes ( smbhs ) is that we already have direct evidence of their existence in the local universe . there are more than 20 bhs identified in x - ray binaries with dynamically measured masses @xmath9@xmath10 ( for reviews see * ? ? ? * ; * ? ? ? * ) . in the milky way , the most massive smbhs are those in grs 1915@xmath66105 ( @xmath67 ; * ? ? ? * ) and in cygnus x-1 ( @xmath68 ; * ? ? ? a similarly massive bh is found in m33 x-7 ( @xmath69 @xmath1 ; * ? ? ? an accreting binary system in the nearby starburst galaxy ic 10 @xcite contains a bh with a mass of @xmath70@xmath71 ( the mass uncertainty is mainly due to the unconstrained companion mass and inclination angle of the binary plane ) . the second advantage is that they are the easiest bhs to form : standard stellar models predict smbhs as the endpoint of individual massive star evolution , for a wide range of stellar masses and metallicities . thus , the strictest constraints to smbh models for ulxs come not from the bh formation process , but from the extreme super - eddington accretion , beaming and outflows required by these models . _ massive stellar bhs _ ( @xmath72 ) the mass of a compact remnant depends both on the initial mass of the stellar progenitor , and on how much mass is lost through stellar winds during the stellar lifetime @xcite . radiatively - driven winds from massive stars are strongly dependent on their metallicity @xcite . therefore , metallicity plays an important role in stellar evolution and determines the maximal mass of a bh that can be formed via core collapse . @xcite argued that the collapse of a single star can form a bh up to @xmath73 at solar metallicity , @xmath74 at 0.3 times solar metallicity , and @xmath75 at 0.01 times solar metallicity ; consistent results ( upper bh mass limit @xmath76 at low metallicities ) that are obtained by @xcite and @xcite . these results can well explain the inferred bh masses in galactic x - ray binaries ( average metallicity close to the solar value ) and in the ic 10 source ( metallicity @xmath77 solar ) . in summary , current stellar evolution models predict a bh remnant distribution peaked around @xmath78 , with a tail extending up to @xmath59 . as the formation of stellar bhs more massive than @xmath79 may require rarer , more extreme conditions , we will refer to them as `` massive stellar black holes ( msbhs ) '' . there is another interesting physical difference between smbh and msbh formation , in addition to different progenitor mass and metallicity : smbhs are probably formed through a sn explosion and subsequent rapid fallback ; msbhs are probably formed from direct core collapse without an explosion @xcite . this is because for the more massive objects , the gravitational energy released during core collapse is not enough to unbind and expel the stellar envelope . at low metallicities , the threshold for direct collapse is an initial stellar mass @xmath80 @xcite ; at solar metallicity , even the most massive stars may lose enough mass in winds and explode as sne . _ intermediate mass bhs _ ( @xmath81@xmath4 @xmath1 ) to produce an imbh of @xmath82@xmath83 , the core collapse of an isolated star in the current epoch is not a viable process . this is not because of the lack of very massive stars ( in fact , several stars with initial masses @xmath84@xmath85 may exist in the lmc ; * ? ? ? * ) , but because helium cores more massive than @xmath86 ( corresponding to initial stellar masses @xmath87 at zero metallicities ) undergo an electron / positron pair - instability explosion that leads to the complete disruption of the star , without bh formation @xcite . however , a bh may form again if the helium core is more massive than @xmath88 , corresponding to initial stellar masses @xmath89 at zero metallicity @xcite . it has been suggested that metal - free population iii stars formed in the very early universe could reach masses of a few hundred @xmath90 @xcite , above the pair - instability limit , and thus may have collapsed into imbhs @xcite . in young and dense star clusters , dynamical friction could lead to massive stars sinking towards the center and undergoing runaway collisions and mergers on timescales @xmath91 yr . numerical simulations suggest that such a process may result in the formation of a huge star , which may subsequently collapse into an imbh @xcite . dynamical evidence for imbhs with masses @xmath92 has been proposed for a few globular clusters , _ e.g. _ , g1 in m31 @xcite , although the issue is still controversial @xcite . @xcite suggested that in massive and/or compact globular clusters , a central seed smbh may grow by up to a factor of 100 via accretion of gas lost by the first generation of cluster stars in their red - giant phase . imbhs may also wander in the halo of major galaxies , after tidal stripping of merging satellite dwarfs that contained nuclear bhs @xcite ; this is currently the most plausible explanations for the most extreme ulx found to date , hlx-1 in eso243@xmath9349 . to summarize , three types of non - nuclear bhs could be powering ulxs ( table [ tab : bh ] ) . cclc type & mass & progenitor & @xmath94 + smbh & @xmath1710 @xmath1 & normal star & @xmath56 + msbh & @xmath95100 @xmath1 & low - z star & @xmath96 + & & pop iii & + imbh & @xmath3@xmath4 @xmath1 & cluster core & @xmath97 + & & stripped nucleus & + assuming that ulxs are binary systems with a bh receiving matter directly from an individual donor star , we can try to fit them into the traditional classification of galactic x - ray sources , as either high - mass or low - mass x - ray binaries ( hmxbs and lmxbs , respectively ) . hmxbs have a massive ob donor star with a characteristic age @xmath98 yrs ; lmxbs have a low - mass donor ( often in the subgiant or giant phase ) with ages @xmath99 yrs ( _ e.g. _ * ? ? ? hmxbs are mostly found in spiral and irr galaxies , with current or recent star formation , while lmxbs populate old spheroidals and globular clusters . ulxs are found both in ellipticals and in spiral / irregular galaxies ; however , the most luminous ones are found predominantly in star - forming galaxies . specifically , more than two thirds of ulxs found in ellipticals are in the lowest luminosity range , with @xmath100 erg s@xmath12 @xcite , and can easily be explained as the upper end of the lmxb distribution , with bh masses @xmath101 . it was even suggested @xcite that all ulx candidates in ellipticals with apparent @xmath102 erg s@xmath12 could be dismissed as background or foreground sources . today , it is generally accepted that there are a few genuine ulxs in ellipticals at luminosities @xmath103@xmath104 erg s@xmath12 , but none has been found above @xmath19 erg s@xmath12 . in spirals , by comparison , one third of ulxs have luminosities @xmath105@xmath106 erg s@xmath12 , and about 10% have luminosities @xmath107 erg s@xmath12 @xcite . both early - type ( hubble types 0 to 4 ) and late - type ( hubble types 5 to 9 ) spirals may contain ulxs , with similar probability , within the current statistical error of the samples ; @xcite find slightly more ulxs in early - type spirals , while @xcite find slightly more in late types , due to the different spatial resolutions of the instruments . furthermore , the fact that a luminous ulx is found in a spiral galaxy does not imply that it has an ob donor star . many ulxs in spirals are not located in regions ( within a few 100s of pc ) with evidence of current or very recent star formation @xcite . using the optical color distribution in a sample of face - on host galaxies , @xcite found that the brightest ulxs turn on @xmath0@xmath108 myr after the end of star formation . characteristic ages @xmath109 myr for the surrounding stellar population were also found for example in the case of ngc1313 x-1 and x-2 @xcite , ic 342 x-1 @xcite , ngc4559 x-1 @xcite , and ngc 1073 ixo 5 @xcite . @xcite normalized the radial distance of all ulx candidates in their sample to the deprojected @xmath110 diameter of their host galaxies . they found that the ulx surface density profile is strongly peaked towards the center . the profile can be fitted with a sersic function , with a best - fit index consistent with a de vaucouleurs profile ( @xmath111 ) , and an exponential scale - radius @xmath112 . this finding suggests that the ulx population is intrinsically related to the stellar population of their host galaxies . less centrally concentrated distributions would be expected if most ulxs were primordial halo relics such as population iii remnants , or nuclear bhs of accreted satellite galaxies . a _ chandra _ sample of ulxs in 58 face - on star - forming galaxies within 15 mpc , studied in details by @xcite , does not show statistically significant associations between ulxs and young , massive ( @xmath113 ) star clusters , within @xmath114 regions . the same conclusion was found by @xcite for ulxs in the antennae galaxies . this is inconsistent with formation scenarios based on runaway core collapse and mergers of o stars inside a compact , young stellar cluster , leading to the possible formation of an imbh of a few hundred solar masses via direct collapse of the resulting superstar @xcite . in fact , most ulx host galaxies do not even have stellar clusters sufficiently massive and compact to satisfy the requirements for runaway core collapse . in theory , imbhs could be formed inside clusters that have since dispersed . however , the evaporation timescale of such clusters would be too long to explain the observed association of many ulxs with young ( @xmath115 myr ) stellar populations . here , we further quantify the association between ulx populations and host galaxy properties . @xcite studied 107 ulxs in a complete sample of 127 nearby galaxies ( @xmath116 mpc ) , selected to be above the completeness limit of both the uppsala galaxy catalog and the _ infrared astronomical satellite _ ( _ iras _ ) survey . the total star formation rate ( sfr ) of the galaxies in the sample is @xmath117 yr@xmath12 based on the _ iras _ data , consistent with other measurements of star formation based on the h@xmath118 and [ o@xmath119ii ] indicators @xcite . this , in turn , corresponds to an sfr density @xmath120 yr@xmath12 mpc@xmath121 in the local universe . finding @xmath122 ulxs in that volume , after correcting for incompleteness and background interlopers , implies that we expect , on average , @xmath123 ulxs for a sfr of @xmath124 yr@xmath12 , and 1 ulx in a volume of @xmath125 mpc@xmath121 . this is consistent with the hmxb luminosity function in local star - forming galaxies : @xcite find @xmath34 sources with @xmath126 erg s@xmath12 ( corresponding to @xmath127 erg s@xmath12 in the ulx definition used in @xcite ) in a sample of galaxies with a total sfr @xmath128 yr@xmath12 . the average specific frequency in star - forming galaxies , in the local universe , is 1 ulx per @xmath129 , for galaxies with dynamical mass of @xmath130 @xcite . this rate decreases for increasing galaxy mass , scaling as @xmath131 @xcite . in elliptical galaxies , the specific rate is 1 ulx per @xmath132 @xcite . thus , it appears that low - mass spirals and irregulars are more efficient at producing ulxs , per unit mass . this is true at least down to galaxy masses @xmath133 . there may be two reasons why low - mass galaxies contain more ulxs per unit mass @xcite . one is that their sfr per unit mass is higher , so they are more efficient at producing ulxs . the other is that they have lower metal abundances , which may favor the formation of heavier stellar bhs from the collapse of massive o stars @xcite . however , it is suggested by @xcite that the metallicity does not affect the ulx production strongly after comparing two pairs of colliding galaxies . finally , it is often noted that many luminous ulxs reside in colliding or strongly interacting galaxies : the antennae @xcite , the cartwheel @xcite , the mice @xcite , ngc4485/90 @xcite , ngc7714/15 @xcite , ngc3256 @xcite , and a few more arp galaxies . this may simply be due to the fact that colliding galaxies have the largest sfr , and the number of ulxs scales with the recent sfr . the cumulative luminosity function ( lf ) of a population of accreting sources ( also known as @xmath134@xmath135 curve ) is expected to have a break or cut - off at the eddington luminosity of the most massive objects in the population . therefore , in the absence of direct mass measurements , the lf can be used as a rough indicator of the maximum mass . in fact , things are not so simple . the lf of x - ray binaries in our galaxy and in other nearby galaxies does not show any features at @xmath136 erg s@xmath12 , eddington limit of the neutron star population . moreover , the location of the upper cutoff may be a function both of mass ( via the eddington limit ) and of the age of the population , since the most luminous sources tend to have a shorter lifetime @xcite . thus , careful modeling is required to interpret a lf . matches the observed distribution better than an unbroken power - law , although the issue remains controversial ( _ cf . at the low - energy end , it also matches the slope and normalization of the hmxb distribution below @xmath16 @xcite . [ fig : lf ] ] with this caveat , we can use the ulx lf to test the scenario that ulxs in star - forming galaxies are mostly the high - luminosity tail of hmxbs ( plus some additional lmxbs ) , and ulxs in ellipticals are the high - luminosity tail of lmxbs . chandra _ and _ xmm - newton _ surveys of discrete x - ray sources in nearby galaxies have produced consistent results in agreement with this interpretation @xcite . in summary , they have all found that the cumulative lf of young x - ray populations in star - forming galaxies is a simple power - law with a slope @xmath137@xmath138 , while the cumulative lf of old x - ray populations in ellipticals and spiral bulges has a slope of @xmath139 . thus , short - lived hmxb - like sources dominate the high - luminosity end of the distribution . the normalization of the hmxb lf is proportional to the sfr ; the normalization of the lmxb lf is proportional to the stellar mass @xcite . what remains controversial is the behavior of the lf for the young ulx population above an x - ray luminosity @xmath38 erg s@xmath12 . from their _ chandra _ survey , @xcite find that a pure power - law @xmath140 provides a very poor fit to the data above that threshold ( figure [ fig : lf ] ) . a cut - off power - law model with @xmath141 provides an acceptable fit , with @xmath142 ( as required by the low - luminosity end of the lf ) and @xmath143 erg s@xmath12 . a very similar result was also found by @xcite , @xcite , and @xcite . the cut - off power - law model does not preclude the possibility of finding ulxs above @xmath144 erg s@xmath12 , but they would be increasingly rarer . it predicts @xcite 1 ulx with @xmath145 erg s@xmath12 within a radius of @xmath146@xmath147 mpc , and 1 ulx with @xmath148 erg s@xmath12 within a radius of @xmath149@xmath150 mpc . if the luminosity of a few ulxs recently discovered by @xcite is confirmed to be @xmath44 erg s@xmath12 , their existence would not be consistent with the cut - off model of @xcite , and may suggest a physically different population of accretors @xcite . and , of course , so does hlx-1 in eso243@xmath9349 with its peak outburst luminosity @xmath60 erg s@xmath12 @xcite . in contrast with those findings , the _ xmm - newton _ survey of @xcite does not see any turnover or cut - off in the cumulative lf above @xmath19 erg s@xmath12 , but their data can not rule out the presence of a similar break . at this stage , the discrepant result may be caused by a different binning technique , or by small - number statistics . both the @xcite and @xcite complete subsamples have only 5 sources with @xmath151 erg s@xmath12 out of a total of @xmath152 ulxs . a few other ulxs above that threshold are known , but can not simply be added to the lf because they do not form part of a complete sample . thanks to the large effective area of _ xmm - newton _ , combined with _ chandra _ observations , the spectral and timing properties of ulxs in the @xmath153@xmath37 kev band have been widely investigated and modeled over the last decade , even though the physical interpretation is still controversial . most of the x - ray studies have focused on bright ulxs ( @xmath154 ) in nearby galaxies ( @xmath155 mpc ) that allows spectral and timing analysis at high signal to noise ratios . in the next two sections , we review x - ray spectral and timing properties of bright ulxs and discuss their physical interpretation in comparisons with well - defined emission properties of galactic bhbs . phenomenologically , the energy spectra of ulxs in the @xmath153@xmath37 kev band can be divided into two groups : those that are consistent with a simple power - law , and those that are more complex @xcite . the latter mainly refers to a mild broad curvature ( convex shape ) over the whole band , a break or steepening above @xmath103 kev , or a soft excess below @xmath103 kev . a typical ulx spectral shape of the latter is sketched in figure [ fig : spec ] . many ulx spectra are well fitted by a single , absorbed power - law model . the photon index has a broad distribution , peaked at @xmath1562.0 @xcite but with a few much harder ( down to @xmath157 ) as well as much softer sources ( up to @xmath158 ) . the distribution does not appear to be bimodal ; there is no evidence of a gap between a `` hard state '' and a `` soft state '' . @xcite found that the power - law photon index tends to be harder at increasing luminosity ; a selection of ulxs with high luminosity ( @xmath159 erg s@xmath12 ) and hard power - law spectrum ( @xmath160 ) is also presented in @xcite . some of the hard sources with multiple observations show strong flux variability ( sometimes by an order of magnitude ) but constant photon index @xcite . similar behavior has been seen from galactic bhbs in the hard state @xcite . the x - ray luminosity in the hard state is usually @xmath161 times eddington @xcite , but during hard - to - soft state transitions the hard state can reach a luminosity up to 30% of eddington @xcite . thus , if hard power - law ulxs belong to the same canonical hard state defined in galactic bhs , their bh masses must be @xmath162 @xcite . a key signature of the hard state is the presence of compact , continuous radio jets . so far , the only possible detection of such jets is from ic 342 x-1 , with _ vla _ observations @xcite . using the fundamental plane of accreting bhs , @xcite estimate a compact object mass @xmath163@xmath164 . further observations are needed to spectrally constrain the nature of the radio core in this system . m82 x-1 is another good candidate for hard state ulxs . this source displays low frequency quasi - periodic oscillations ( qpos ) @xcite above a band - limited broad noise component , and its x - ray spectrum is a featureless power - law with a photon index @xmath165 @xcite . we note that m82 x-1 may be one of the very few ulxs that change their spectral state during outbursts , switching from a hard to a thermal state ( see section [ sec : curve ] ) . the hard spectrum may come from an advection - dominated accretion flow @xcite , a luminous hot accretion flow @xcite , a classical hot corona above the accretion disk @xcite , or from a combination of synchrotron and inverse - compton emission at the base of the jet @xcite . the first mechanism is limited to low accretion rates , but the latters can in principle operate also near the eddington luminosity and be relevant to ulxs @xcite . thus , an alternative interpretation favored by some authors is that hard power - law ulxs are not in the canonical low / hard state ; in other words , that the ulx hard state is not confined to luminosities @xmath166 eddington , and does not switch to a softer , disk - dominated thermal state as the accretion rate increases @xcite . we could call it a persistent high / hard state ( unknown in galactic bhbs ) , which can extend all the way to the eddington luminosity and smoothly morph into the very high state or other super - eddington accretion states . at the opposite end of the broad photon - index distribution , several ulxs with @xmath167 could be classified in the steep power - law state of galactic bhbs , typically found at near - eddington luminosities @xcite . the fact that there are no clear gaps , but instead a continuous distribution in photon index and an overlapping luminosity distribution between hard and soft ulxs may be an indication that the two regimes are smoothly connected . in conclusion , with longer x - ray observations and increasing photon statistics , we have reached a stage where we can identify and model a number of subtle spectral features on top of the simple power - law for many ulxs . in the rest of this section , we discuss a variety of spectral models that account for such features , and their physical implications . -free ) or a warm , thick comptonization model can adequately fit the hard curvature ; in the plot , a @xmath168-free model with @xmath169 kev and @xmath170 has an almost identical spectral shape to a comptt model with @xmath171 kev and @xmath172 , which is shifted vertically for clarity . for comparison , two diskbb models are plotted with @xmath173 and 2.5 kev , respectively . they are more curved and can not fit the energy spectrum over the 0.3 - 10 kev band . [ fig : spec ] ] in many ulxs , excessive residuals at low energies ( @xmath174 kev ) would be seen if one makes a power - law fit at high energies ; an additional disk component could adequately fit the excesses and the inferred disk inner temperature is around 0.10.4 kev . this cool , soft excess was first identified by @xcite in ngc 5408 x-1 ; @xcite showed the emission was point like and applied a disk plus power - law model . similar phenomena were then found in ngc 1313 x-1 and x-2 @xcite , and later on in many other ulxs @xcite . such a fit is analogous to the empirical model for galactic bhbs except that the disk is cooler and more luminous in ulxs than in smbhs at their high state @xcite . for a standard accretion disk @xcite extending to the last stable orbit around a bh , the inner disk temperature @xmath175 is a function of bh mass @xmath8 and total luminosity @xmath176 : @xmath177 where @xmath118 depends on the bh spin ( @xmath178 for a schwarzschild bh , @xmath179 for an extreme kerr bh ) , @xmath180 is a corrective factor that takes into account the no - torque condition at the innermost orbit , and @xmath181 is the spectral hardening factor @xcite . given the same spinning and correction factors , the bh mass is scaled with the disk inner temperature and luminosity as a function of @xmath182 thus , the low - temperature , high - luminosity characters of the assumed disk suggest that the emission arises from accretion onto imbhs . for comparison , the typical disk inner temperature is about 1 kev at a luminosity of @xmath183 for smbhs about 10 @xmath1 . the characteristic temperature for ulxs is usually around 0.10.4 kev and the luminosity is some @xmath16@xmath19 , suggestive of a bh mass of @xmath184 @xmath4 @xmath1 . however , the interpretation of soft excesses as emission from cool accretion disk has confronted challenges . @xcite argued that soft excesses could also be soft deficits depending on the energy band where the power - law is fitted . @xcite found similar soft excesses in both ulxs and less luminous x - ray sources , implying that they are not a unique signature of imbhs in a statistical sense . it is suggested that supercritical accretion may produce massive outflows , which are optically thick with soft , thermal emission at similar temperature , being another possible origin for the detected soft excesses @xcite . other authors @xcite have suggested that the soft excess does come from the disk surface , but much further away than the innermost stable orbit , where the disk may be obscured or replaced by an outflow or a scattering corona ( section [ sec : curve ] ) . in that case , the characteristic radius of the thermally - emitting disk region can not be used to infer the bh mass . repeated observations of an individual ulx may help test the cool disk model @xcite . if the disk touches the innermost stable orbit around the bh , one expects to see the disk luminosity vary to the 4th power of the inner temperature , @xmath185 , as well known in the thermal state of galactic bhbs @xcite . if , instead , the observed inner disk radius is the boundary between an outer standard disk and an inner corona or outflow , we expect that such transition will move further and further out at higher accretion rates when super - eddington . in the simplest outflow scenario , @xmath186 @xcite . this implies @xcite @xmath187 constant for a viscously heated outer disk and @xmath188 for a strongly x - ray - irradiated outer disk ( as is more likely the case in ulxs ) . most other plausible relations between @xmath189 and @xmath27 predict negative correlations between the disk luminosity and temperature , with slopes between @xmath190 and @xmath191 @xcite . a negative correlation between the luminosity and temperature is clearly detected for at least ngc 1313 x-2 and ic 342 x-1 @xcite . from a sample of 8 ulxs , @xcite found that the luminosity of the soft excesses scales with the temperature as @xmath192 , roughly consistent with the model prediction that the soft emission arises from a photosphere due to strong outflows from a supercritical disk and its temperature is determined by the spherization radius @xcite . on the contrary , a positive correlation is seen from ngc5204 x-1 ; however , it has a shallower slope @xmath193 @xcite . @xcite suggested , based on a disk atmosphere model , that the hardening correction factor becomes larger at increasing luminosity , and explains the apparent shallower slope in galactic bhs such as xte j1550@xmath93564 and h1743@xmath93322 . applying such a varying hardening correction , one can recover the @xmath194 relation for ngc5204 x-1 , which would favor the imbh scenario . however , this result is not robust yet : @xcite found that the positive correlation disappears or even becomes negative if one chooses a different lower energy bound for the spectral fitting . also , the disk contributes only @xmath195 of the 0.310 kev luminosity , inconsistent with the thermal dominant state . instead , if ulxs were imbhs , we would expect to see many of them in the canonical thermal state dominated by the soft disk component . in summary , using the disk parameters inferred from x - ray spectra to determine the bh mass is based on assumptions that have proved either incorrect or problematic for most sources . it was noted by @xcite and @xcite that high energy curvature / steepening is a common feature in high - quality ulx spectra . generally , the break / steepening could be fitted by a comptonization model with relatively low energy electrons and high optical depth @xcite . a few ulx spectra are mildly convex , formally consistent with a standard disk model . we thus discuss the physical implications of these models . with _ asca _ observations , a hot multicolor accretion disk model was found to fit the spectra of a number of ulxs ( _ e.g. _ * ? ? ? for those with relatively low temperatures ( @xmath196 kev ) , they could be explained by accretion onto smbhs or msbhs @xcite . the high temperature and luminosity may be explained by large bh spin @xcite , but this does not work for extreme cases where relatively small bh mass is required , which in turn is inconsistent with the observed high luminosity . for _ xmm - newton _ spectra , a single hot disk with absorption is usually unable to fit the data ; an additional component is required and dominates in the low energy band @xcite , and the disk component is not the dominant part @xcite . the presence of a soft component in addition to the hot disk is unusual in bhbs and does not have a simple physical interpretation . in general , the hot disk model is problematic . the standard disk emission is best known in the thermal state in bhbs , in which the x - ray spectrum is dominated by emission from a standard accretion disk ; plus , the x - ray variability is low , with no or weak qpos @xcite . so far , the thermal state may have been identified only in one case , m82 x-1 , constrained using both spectral and timing information . with simultaneous _ chandra _ and _ xmm - newton _ observations of m82 , @xcite found surprisingly that the previously known qpos in the source disappeared . the light curve was no longer highly variable and the power spectrum was consistent with that of white noise . the energy spectrum also changed dramatically from a straight power - law to a disk - like spectrum . fitting with a standard accretion disk model , the inferred luminosity and temperature from three repeated observations are consistent with a 4th power relation . all results are well consistent with that expected for the thermal state . the monitoring data from _ rxte _ indicate that these _ chandra _ and _ xmm - newton _ observations were made during the source outbursts , suggesting that m82 x-1 usually stays in the hard state and could transition to the thermal state during outbursts . fitting with a relativistic disk model , the bh is estimated to be a fast spinning imbh of 200800 @xmath1 . there are a few more sources @xcite which have shown spectral variability possibly consistent with a @xmath194 relation if fitted with a hot standard disk model . another group of five nearby ulxs consistent with the thermal state spectrum , based on x - ray spectral modeling , was discussed by @xcite . these sources all lie at the lower end of the ulx luminosity regime , and could be explained by kerr msbhs in the thermal state . they do not stand for the majority of ulxs , in particular those with a typical luminosity around @xmath19 . also , lack of fast timing information hampers a firm identification . therefore , it is generally accepted that the canonical thermal state is rare in ulxs @xcite . the slim disk can radiate near or moderately above the eddington limit estimated from spherical accretion @xcite . this super - eddington disk is thus a good candidate for ulx emission @xcite . when advective energy transport dominates over radiative cooling in an accretion disk , the radial temperature profile changes from @xmath197 for a standard disk to @xmath198 for a slim disk @xcite . an extended multicolor disk model called the @xmath168-free model was introduced , with the radial dependence @xmath168 on temperature ( @xmath199 ) being an additional free parameter @xcite , and used to explain ulx emission @xcite . @xcite investigated the energy spectra of four ulxs , ngc 5204 x-1 , ngc 4559 x-7/x-10 , and ngc 1313 x-2 , and found that they can be well fitted by the @xmath168-free model . interestingly , the best - fit @xmath168 value was found to be around 0.5 , consistent with that for a slim disk . in the luminosity vs. disk temperature diagram , the four ulxs and galactic bhbs appear to be in line with the same @xmath194 relation , suggesting that ulxs may be bhs of similar or slightly higher masses ( smbhs or msbhs ) but in the slim disk state . another successful test of the @xmath168-free or slim disk model was for m33 x-8 , which is marginally classified as a ulx but a good candidate for deep investigation owing to its relatively small distance ( _ e.g. , _ * ? ? ? * ; * ? ? ? * ) . @xcite first identified that an extra power - law component is required in addition to the @xmath168-free disk to fit the source spectrum . the best - fit @xmath168 value ( @xmath200 ) suggests notable advection in the disk , close to the slim disk nature rather than a standard one . these indicate that the slim disk state may be a natural extension of the canonical thermal state at supercritical accretion . this is consistent with the radiation - magnetohydrodynamic simulations that the hard , thermal and slim states could be produced under a single model with different accretion rates @xcite . however , the @xmath168-free model encounters difficulties in both observation and theory . in some sources , the disk inner temperature was found too high to be physically reasonable when fitted with the @xmath168-free model @xcite . if photon trapping is taken into account , the radial temperature profile is expected to be significantly flatter than @xmath201 within the trapping radius @xcite , in contrast to the spectrum predicted without considering photon trapping @xcite . also , numerical simulations suggest that dense , warm outflows are always associated with supercritical accretion @xcite , which would comptonize and obscure the emission from the underlying disk , see details in the following subsection . in galactic bhbs , high energy spectral curvature is also detected , but usually occurs at tens to hundreds of kev . it is interpreted as comptonization in a thin corona with temperatures typical of @xmath3 kev . however , the spectral curvature in ulxs suggests an electron temperature of a few kev and a scattering optical depth significantly higher than unity , indicative of a warm and thick corona . due to the presence of such a thick corona , the emission from the disk would be distorted and the previous interpretation , a simple summation of a cool disk and a warm corona @xcite , needs be modified consequently . @xcite revisited best quality _ xmm - newton _ data and concluded that two common features simultaneously exist in the energy spectrum of ulxs : soft excesses below 2 kev and high energy curvature above 2 kev . this is not frequently seen in the spectra of galactic bhbs when they are thought to accrete at sub - eddington level . thus , they suggested that ulxs represent a new state distinct from the four well - known states ( quiescent , hard , thermal , and steep power - law ) in galactic bhbs @xcite , and call it ultraluminous state . please refer to @xcite for a more detailed review of this state . the galactic bhb that comes closest to this state is xte j1550@xmath93564 at the peak of its 1998 september outburst . a coupled disk - corona model @xcite was proposed to explain such state , in which the disk and corona are not independent but share the total gravitational energy release . the un - scattered outer disk is responsible for the soft excesses , while the comptonized inner disk produces the hard , curved tail . the bh mass , estimated from the unmasked temperature and radius of the underlying disk , lies in the regime of smbhs or msbhs . numerical simulations of supercritical accretion indicates that radiation pressure would power strong outflows above the disk in the inner region @xcite . the wind is optically thick and warm ( a few kev ) , and would strongly comptonize the disk emission . at high accretion rate , the disk becomes thick and would collimate the radiation along the symmetric axis causing geometric , mild beaming with an apparent isotropic luminosity over the eddington limit by a factor up to @xmath1710 . this picture is consistent with the proposed ultraluminous state . observationally , from the spectral evolution of m33 x-8 , @xcite suggested that the outflow is launched at near - eddington luminosity , and the radius at which it is launched moves further out with increasing luminosity . if the comptonizing region is the fast outflow , rather than a static corona , it is possible that bulk motion @xcite can be as or more important than thermal comptonization , depending on the characteristics of the outflow @xcite . observationally , ulx spectra that can be formally fitted with thermal comptonization models can be equally well fitted with bulk motion models ( _ e.g. _ * ? ? ? the only difference is in the interpretation of the characteristic energy of the upscattering electrons . as noted by @xcite , simple physical considerations , comparisons with other astrophysical classes of outflows , and numerical simulations all suggest that super - eddington ulx outflows have a characteristic speed @xmath202 escape velocity from the launching radius , corresponding to a bulk kinetic energy @xmath203 kev for the outflowing electrons . this is similar to the characteristic energy of the spectral breaks . most ulxs and bhbs have considerable radiation output above 2 kev , while supersoft sources ( ssss ) have most of their energy released below 2 kev . canonical ssss are speculated to be powered by steady nuclear burning on the surface of white dwarfs @xcite . a number of supersoft ulxs have been discovered , whose spectrum is dominated by a cool blackbody component of tens to a hundred ev , being a distinct class from the majority of ulxs . these include m101 ulx-1 @xcite , antennae x-13 @xcite , m81 uls1 @xcite , ngc 4631 x1 @xcite , and ngc 247 ulx @xcite . at long timescales , most of these sources have shown dramatic variability with a luminosity change by a factor up to @xmath184 but with consistently low temperatures . a weak power - law tail is significantly detected in at least one source @xcite , but none of the supersoft ulxs has become a power - law dominated ulx or vice versa . fast timing noise was detected in two of them , m101 ulx-1 @xcite and ngc 247 ulx @xcite , down to a timescale around 100 s , indicating that they are not agn . the nature of these peculiar sources is very uncertain . they are unlikely powered by white dwarfs , because they are too luminous ( a few times @xmath16 ) and some of them are too hot ( @xmath204 ev ) . they are also unlikely to be due to cool disks around imbhs , because such disks can not have a huge variability with a constant temperature . one possible scenario is that their emission is produced by massive outflows from supercritical accreting sources viewed at a high inclination angle @xcite . a unification model ( supercritical accretion with thick outflows , see figure [ fig : model ] ) may be able to explain both normal , broad - band ulxs and less luminous , supersoft ulxs with different viewing angles . so far , we have briefly reviewed a variety of scenarios that can account for the observed ulx spectra ( see also figure [ fig : spec ] ) . in figure [ fig : model ] , we sketch the physical ingredients behind the most common combination of such models , which may be successfully used to explain the multiband spectrum of most ulxs . ] the study of long term spectral variability of ulxs mainly relies on pointed observations with _ chandra _ or _ xmm - newton _ , or quasi - regular snapshots with the x - ray telescope on board _ swift_. in addition , _ rxte _ has been used to monitor the brightest ( in terms of observed flux ) ulx in m82 . the most interesting science with these observations , in addition to the search for orbital periods , is perhaps to find state transitions known in galactic bhbs . canonical transitions between the low / hard and high / soft states were reported in ic 342 x-1 and x-2 with _ asca _ observations @xcite and also in holmberg ix x-1 with multiple missions @xcite . however , almost no such canonical state transitions have been reported in the _ chandra _ and _ xmm - newton _ era . in contrast , transitions between soft and hard states were found but not in a way similar to galactic sources @xcite . a major difference between ulxs and galactic bhbs is that most ulxs are persistent sources and have been bright since their discovery decades ago , while the latter are usually transients especially accreting from a low mass companion with roche lobe overflow . _ swift _ monitoring @xcite of bright ulxs have revealed flaring activity with a luminosity change by a factor of @xmath155 on time scales of days to months . in particular , @xcite found holmberg ix x-1 remains in the hard state as its flux varies , and @xcite found the x - ray spectral state is not correlated with the luminosity in holmberg ii x-1 . also , as mentioned previously in section [ sec : curve ] , @xcite found m82 x-1 may have changed from the hard state to thermal state during outbursts . there is one exceptional case . eso243@xmath9349 hlx-1 shows a low / hard to high / soft jump reminiscent of the canonical state transition @xcite . it also displays outbursts in a `` fast rise exponential decay '' pattern , very similar to those seen in galactic bhbs @xcite . however , the authors argued that the observed variability is unlikely triggered by disk instability due to inconsistent timescale . the power spectrum of galactic bhbs can be decomposed into broad components ( power - law red noise , band - limited noise , flat - top or zero - centered broad lorentzian , etc . ) and narrow components ( qpos ) . although we still know little about the physical origin of these temporal features , their phenomenology has proved useful in defining and identifying accretion states @xcite and possibly in determining bh masses via model independent calibrations ( _ e.g. _ * ? ? ? an important model - independent finding of @xcite , based on the study of sample of 16 bright ulxs , is that those sources can be divided into two groups defined by their short - term x - ray variability , even though they have similar energy spectra and luminosities . one group ( including among others ngc 5204 x-1 ) shows weak or absent variability on timescales at least as large as 100s . the other group ( including among others ngc 5408 x-1 ) has similar variability levels and power spectra as luminous galactic bhbs and agn in the observed frequency bandpass ( @xmath205@xmath206 hz ) . it is still unclear what suppresses short - term variability in one class but not the other . in galactic bhbs , high variability is often associated to non - thermal emission with steady jets ; low variability to thermal emission with no jets @xcite . in the following , we select four ulxs in which interesting structures have been found in their power spectra , shown in figure [ fig : psd ] . _ m82 x-1 , also known as m82 x41.4 + 60 _ @xcite first discovered low frequency qpos around 54 mhz from the m82 galaxy with both _ xmm - newton _ and _ rxte _ observations . by fitting the _ xmm - newton _ images to its point spread function with known source positions from _ chandra _ , @xcite identified that m82 x-1 is the source that produced the qpos , and interestingly , it was not the brightest x - ray source in m82 at the time of qpo discovery . a following _ xmm - newton _ observation revealed that the qpos have changed the central frequency from 54 mhz to 114 mhz , above a flat - top component with a low frequency break at 34 mhz @xcite . @xcite also found that the qpos changed frequency from 107 mhz to 120 mhz during the second observation , and a plausible harmonic ratio of 1:2:3 for qpo frequencies with _ these qpos , however , were undetected with three _ xmm - newton _ observations of similar depth , suggestive of a spectral state transition @xcite . _ ngc 5408 x-1 _ with a short _ xmm - newton _ observation , @xcite found the source is highly variable , and its power spectrum has a break at 2.5 mhz . @xcite reported the identification of narrow qpos around 20 mhz from the source , and confirmed the presence of the frequency break . the qpos are narrow and may show a pair with a 4:3 frequency ratio . a longer _ xmm - newton _ observation later detected the qpos at about 10 mhz and the energy spectral parameters of the source also changed @xcite . the qpo frequency was found to scale with the disk flux and also the spectral index of the power - law component . these correlations are analogous to that seen in galactic bhbs , suggesting that this ulx is in an emission state similar to sub - eddington sources . the qpo properties ( narrow , strong , with a flat - top continuum ) and the temporal - spectral evolution patterns suggest that the qpos are of type c and the compact object in ngc 5408 x-1 is an imbh . however , as the qpo frequency is not proportional to the break frequency of the continuum as expected , the identification as of type c was questioned and so was the imbh scenario @xcite ; they proposed that the source is super - eddington and the turbulence in the comptonization region gives rise to the observed variability . other authors have also rejected the imbh argument @xcite , pointing out that neither lower qpo frequencies nor larger , cooler disks can be used as indicators of higher bh masses in ulxs ; both findings may simply suggest that the visible inner radius of the disk has receded much further than the innermost stable orbit , and has been covered or replaced by a thick comptonizing region or outflow in the inner region . besides , @xcite also found that the source shows a linear correlation between the rms amplitude and flux , similar to that seen in bright bhbs and agn . _ m82 x42.3 + 59 _ this is a transient ulx in m82 @xcite . it could not be detected by _ chandra _ in its low state , but sometimes , was even brighter than m82 x-1 , being the most luminous source in m82 @xcite . with _ chandra _ observations , and also confirmed by a simultaneous _ xmm - newton _ observation once , @xcite discovered low frequency qpos around 3 - 4 mhz in its power spectrum . these qpos are wide and strong , appear above white noise , and can only be detected when the source is brighter than @xmath19 . this is consistent with a type a or b classification , but is inconsistent with the properties of type c qpos . _ ngc 6946 x-1 _ this source has the highest short - term variability found in a ulx , with a fractional rms amplitude of 60% integrated in the frequency range of 1 - 100 mhz @xcite . its power spectrum shows a flat - top continuum that breaks at about 3 mhz with possible quasi - periodic oscillations ( qpos ) near 8.5 mhz . ] the characteristic frequency alone usually is not a direct indicator of the bh mass , as these noise features can vary dramatically in a single source . however , correlations between the frequency and spectral parameters have been found , and the correlation patterns are thought to scale with the bh mass . the most common qpos seen in bhbs are of type c ( _ cf . the frequency of type c qpos are found to vary over a wide range . for example , the lowest detected frequencies in grs 1915 + 105 are around 1 mhz @xcite , even lower than the frequencies of all qpos detected in ulxs . a positive correlation between the qpo frequency and disk flux has been found for type c qpos in xte j1550@xmath93564 , gro j1655@xmath9340 and h 1743@xmath93322 @xcite . the qpo frequency also scales with the photon index of the power - law component in the energy spectrum ( _ e.g. _ * ? ? ? * ; * ? ? ? @xcite suggested that the qpo frequency scales inversely with the bh mass at a fixed power - law photon index , and thus one can infer the mass given spectral index @xcite . on the contrary , type a and b qpos show in a relatively narrow frequency range . the broad continuum seems to be a more reliable determinant of bh mass , as it is detected in both bhbs and agn and the scaling has been tested over a wide frequency and mass range . smbhs and agn appear to populate a `` variability plane '' defined by the bh mass , accretion rate , and @xmath207 , which is the frequency of the lower high - frequency lorentzian component in the power spectrum @xcite . limited by sensitivity , @xmath207 can not be detected in ulxs . based on a correlation between @xmath207 and the frequency of type c qpos , @xcite extended the variability plane to type c qpos that are visible in ulxs . the bh mass can be inferred from the x - ray luminosity ( here we adopt the radiatively efficient accretion case ) and type c qpo frequency as @xmath208 where the inequality is due to the unknown bolometric correction ( @xmath209 ) . although the `` variability plane '' is tested for a wide range of bh masses , the correlation between @xmath207 and @xmath210 has not been tested in ulxs and even the identification as of type c is questioned at least for ngc 5408 x-1 @xcite . attempts to scale bh mass with the variability amplitude ( rms ) have made great success in agn @xcite . however , it is found that ulxs are significantly less variable than agn , and a simple extrapolation could not be obtained @xcite . based on the correlations mentioned above , the bh masses inferred from the characteristic frequencies are all about @xmath184@xmath4 @xmath1 , indicating that ulxs contain imbhs . however , application of these mass scaling methods requires strong caveats . there are at least three problems . first , we are not sure whether the timing features detected in ulxs are the same type found to scale with bh masses in galactic bhbs ; for example , we are not sure whether the qpos in ngc 5408 x-1 are of type c @xcite . for other types of timing features ( _ e.g. _ , type a and b qpos ) , it is still not known whether their frequency is proportional to the bh mass @xcite . second , frequency - luminosity - mass relations have been calibrated and tested for galactic bhbs and agn in sub - eddington accretion , that is when the inner edge of the disk is constant or moves towards the innermost stable orbit as the accretion rate increases ( thus , increasing qpo frequencies ) . there is no empirical evidence that the scaling goes the same way in ulxs , which may be super - eddington systems where the visible inner edge of the disk recedes from the innermost stable orbit for increasing accretion rates ( thus , reducing qpo frequencies ) . the third problem is that we do not have enough signal - to - noise ratio to detect high - frequency qpos , which severely limits possibility of direct comparisons with galactic bhbs . in this section the `` optical counterpart '' refers to the point - like optical source that is spatially associated with the ulx ; sources with resolved structure , such as optical and radio nebulae , will be discussed in the next section . the optical emission from the counterpart could arise from the donor star or outer accretion disk , or both . it provides interesting information regarding the binary evolution history , nature of the donor star , disk geometry , and mode of mass transfer , and would also help constrain the bh mass . several groups have modeled x - ray irradiation in order to distinguish between the contributions of the outer disk and the donor star , and constrain the intrinsic stellar properties . @xcite applied their irradiation model to the photometric data for seven ulxs ; they suggested that donor stars are consistent with spectral type b or later , and the bh masses are in the range of @xmath0@xmath211 . @xcite computed evolution tracks of ulx counterparts on the color - magnitude diagram , assuming a massive companion orbiting an smbh or a msbh ; they applied their models to ngc1313 x-2 , suggesting a bh mass @xmath212@xmath211 . in addition to reproducing the observed colors and luminosity ( after accounting for irradiation and disk contribution ) , a donor star in a ulx must be able to supply a mass accretion rate @xmath213 yr@xmath12 , which requires roche lobe overflow , and to do it over a sufficiently long period of time to explain the observed number of ulxs per galaxy . binary population synthesis models have been developed and used to address this problem @xcite . modeling the binary evolution requires calculations of the mass transfer rate , the corresponding change in orbital separation , and the subsequent fate of the transferred mass . this in turn depends critically on whether we impose a strict eddington limit on the mass transfer rate that can be accepted by the bh ( with the rest being expelled in a wind ) , or instead we assume that the bh can accrete at higher rates . thus , for the same kind of donor star , the outcome in terms of x - ray luminosity and lifetime is very different depending on whether the primary is an eddington - limited stellar bh , a non - eddington - limited stellar bh , or an imbh . @xcite considered three characteristic families of binary evolution tracks , with different bh mass distributions : @xmath214@xmath10 , @xmath214@xmath215 , and a single - mass population of @xmath216 bhs . they built probability diagrams in different slices of parameter space ( optical color - magnitude diagram ; orbital period versus potential x - ray luminosity ; x - ray luminosity versus age ) , representing the total time spent by all the systems in those regions of parameter space . their main conclusion is that the most probable ulx system parameters correspond to high - mass donors ( @xmath217 ) , with orbital period between 1 and 10 days and characteristic age @xmath218 yr . in terms of bh mass , they found that both the @xmath216 bh population , and that with non - eddington - limited bhs up @xmath219 were consistent with the optical and x - ray observations and with the observed number of ulxs per galaxy ; the @xmath220@xmath211 bh mass range has not been tested yet . thanks to the sub - arcsecond resolution of _ chandra _ and _ hst _ , the optical counterparts of more than a dozen ulxs have been identified . these are done mainly by aligning _ chandra _ and _ hst _ images using objects with emission on both , to reduce the _ chandra _ x - ray position error from 0.6 to about 0.20.4 . @xcite collected all available data in the _ hst _ archive to investigate the optical properties of 13 ulxs , which can be summarized as follows . a. the x - ray to optical flux ratios and optical colors are consistent with that of lmxbs , but unlike hmxbs or agn . this indicates that the optical emission from ulxs is mainly due to reprocessing of x - rays on the outer accretion disk . as a consequence , one can not infer the spectral type of the companion star using optical colors directly . b. the broad band optical spectrum can be generally fitted with a power - law model ( @xmath221 ) . the distribution of the spectral index @xmath118 peaks between 1 and 2 , which is also consistent with disk irradiation models . c. the counterpart has unique optical properties within its neighborhood . nearby sources usually show a blackbody spectrum instead of a power - law spectrum . d. the optical emitting region is estimated to be large , of the order of @xmath222 cm , indicative of long binary periods ( @xmath109 days ) for smbh or msbh primaries , or shorter periods for imbhs . e. sources with multiple observations usually show strong , random variability on both magnitudes and colors . there are a few outliers showing peculiar properties . the optical spectra of ngc 2403 x-1 and m83 ixo 82 have a slope consistent with that of the standard disk emission , @xmath223 . m101 ulx-1 and m81 uls1 are more similar to hmxbs than lmxbs in terms of x - ray to optical flux ratios . ic 342 x-1 may show a blackbody - like spectrum instead of a power - law . more details can be found in @xcite . there have been a lot of observational efforts to determine the nature of the companion stars in nearby ulxs . so far , ngc 7793 p13 is the only ulx in which emission from the companion star is directly seen beyond reasonable doubt @xcite ; detection of narrow emission lines identifies the companion star as a late - b type supergiant with a mass @xmath79 . orbital periods may have been detected in a few sources . assuming roche - lobe overflow , the companion density can be simply determined from the binary period , @xmath224 g cm@xmath121 @xcite . m82 x-1 is the strongest case of a positive detection ; a monitoring program with _ rxte _ revealed a steady period of about 62 day @xcite , corresponding to a companion density of @xmath226 g cm@xmath121 . @xcite discovered a 115 day x - ray period from ngc 5408 x-1 with _ swift _ monitoring , and inferred a similar companion density of @xmath227 g cm@xmath121 ; however , the interpretation of that modulation as a binary period has been disputed @xcite . these claimed long periods imply an evolved donor star . possible detections include a 6.1 day period from ngc 1313 x-2 with _ hst _ observations @xcite ; however , low significance of the period was claimed by other groups @xcite . m82 x-1 is located near or within a super star - cluster mcg-11 with an age of 7 - 12 myr @xcite . ngc 1073 ixo 5 appears near a ring of star formation of 816 myr old @xcite . ngc 1313 x-2 is found near a star cluster with an age of 20 myr @xcite . ic 342 x-1 lies in a field with relatively old stellar objects , mostly older than 10 myr @xcite . the stellar field around ngc 1313 x-1 is at least 30 myr old @xcite . holmberg ix x-1 is suggested to be part of a loose cluster with an age @xmath115 myr @xcite . all those findings suggest that ulxs are in relatively old stellar environments with ages of at least 10 myr . an exceptional case is provided by a transient ulx recently discovered in m83 @xcite . it is located in an interarm region with a stellar population older than @xmath228 yr . _ hst _ observations when the system was probably in quiescence constrain the donor star to be @xmath229 . instead , _ observations during the recent outburst revealed a bright , blue counterpart with absolute @xmath230 magnitude @xmath231 , due to the irradiated outer disk and/or irradiated donor star . identification of the optical counterparts to ulxs is the first step for a potential dynamical mass measurement . @xcite found that the broad he ii @xmath232 line from ngc 1313 x-2 changed its velocity by about 380 km s@xmath12 between two observations . if this is due to binary motion of the disk , an upper limit on the bh mass of 50 @xmath1 can be placed . there have been a few attempts to obtain the mass function of a few ulxs using emission - line velocity curves . however , the lines were found to vary randomly instead of showing an ellipsoidal modulation ; thus , they did not provide dynamical constraints on the bh masses @xcite . the orbital period is a key parameter for dynamical mass measurement . with current facilities , it may be more feasible to search for orbital periods in x - rays and then confirm them in optical for extragalactic sources , like the cases of m33 x-7 and ic 10 x-1 . we naturally expect ulxs to ionize the surrounding interstellar medium ( see * ? ? ? * ; * ? ? ? * ) for the basic principles of x - ray ionization . the main difference between uv ionized h@xmath119ii regions and x - ray ionized nebulae is the lack , in the latter , of a sharp boundary ( strmgren sphere ) between the ionized and neutral plasma @xcite . instead , there is an extended warm region where weakly ionized atoms coexist with collisionally excited neutral species . as a result , an x - ray ionized nebula is characterized by high - ionization emission lines such as he@xmath119ii @xmath2334686 and [ o@xmath119iv ] at 25.89 @xmath234 m in the inner regions , and low - ionization forbidden lines such as [ o@xmath119i ] @xmath2336300 in the outer regions . in particular , he@xmath119ii @xmath2334686 acts as a photon counter of the ionizing flux between 54 ev and @xmath235 ev @xcite . by comparing the true source flux inferred from the he@xmath119ii @xmath2334686 line with the apparent flux measured at energies @xmath236 ev with _ chandra _ or _ xmm - newton _ , we can test whether the x - ray emission is isotropic or beamed . the n159f nebula , powered by the smbh lmc x-1 , is the best - known example of a large - scale ( @xmath0 pc ) x - ray ionized nebula in galactic systems @xcite . @xcite opened a new field with their systematic search for similar nebulae around ulxs . they found a spectacular case around the ulx in m81 group dwarf galaxy holmberg ii . the holmberg ii x-1 ionized nebula ( now known as `` foot nebula '' ) has a he@xmath119ii @xmath2334686 luminosity @xmath237@xmath238 erg s@xmath12 @xcite , 30 times higher than the lmc x-1 nebula , which implies an average in the nebula , estimated as @xmath239 yr @xcite . ] ionizing x - ray flux @xmath240@xmath241 erg s@xmath12 . this result has been confirmed by further optical / infrared spectroscopic studies @xcite . on the other hand , the apparent x - ray luminosity of holmberg ii x-1 has been seen to vary over the years between @xmath17 a few @xmath16 erg s@xmath12 and @xmath242 erg s@xmath12 @xcite . the consistency between the two independent measurements gives the strongest evidence to date that beaming is negligible ( @xmath243 ) for this ulx ; this favors models with relatively large bh masses ( section [ sec : bh ] ) . similar arguments have also been applied to the photoionization - dominated nebula around ngc 5408 x-1 . from the flux in high - ionization lines such as he@xmath119ii @xmath2334686 and [ ne@xmath119v ] @xmath2333426 , @xcite inferred a lower limit on the isotropic x - ray luminosity @xmath244 erg s@xmath12 , only a factor of 3 lower than the directly measured x - ray luminosity . as in holmberg ii x-1 , this suggests that beaming factor is @xmath245 . more examples of x - ray ionized nebulae in ulxs are reported by @xcite , @xcite , and @xcite , although in those cases the interpretation is more model dependent , because the observed line spectra are a mix of photo - ionized and shock - ionized plasma emission . + + it is often hard to distinguish between x - ray photoionization and shock excitation , for ulx nebulae just as it is generally difficult to separate the contribution of a starburst nucleus and/or a low - luminosity agn in a liner . diagnostic diagrams with several line ratios are computed with shock and photoionization codes such as mappings iii @xcite and cloudy @xcite . shock - ionized plasma is characterized by high ratios of low - excitation lines such as [ o@xmath119i ] @xmath246 [ s@xmath119ii ] @xmath247,@xmath248 and [ n@xmath119ii ] @xmath249,@xmath250 over h@xmath118 . for shock velocities @xmath251@xmath252 km s@xmath12 , the high - ionization , forbidden transitions [ o@xmath119iii ] @xmath253,@xmath254 are strong ( [ o@xmath119iii ] @xmath233 5007/h@xmath255 for @xmath256 km s@xmath12 ) , while he@xmath119ii @xmath2334686 indicates @xmath257 km s@xmath12 @xcite . the luminosity in the h@xmath258 line is only a function of the total shock power @xmath259 and shock velocity @xcite ; for typical speeds @xmath260@xmath261 km s@xmath12 , @xmath262@xmath263 . supershells were first found around ngc 6946 mf 16 @xcite and holmberg ix x-1 @xcite , and were initially speculated to be powered by unusually energetic sn remnants . the x - ray emission could not be from the nebulae , because of its variability @xcite . following the pioneering work of @xcite , several more huge shock - ionized nebulae have been identified around ulxs ( `` ulx bubbles '' , see figure [ fig : bubble ] for examples ) , with characteristic diameter @xmath264@xmath265 pc , expansion velocity @xmath260@xmath266 km s@xmath12 , characteristic age @xmath267@xmath206 myr , mechanical power @xmath268@xmath19 erg s@xmath12 ( independently determined from the bubble expansion speed and from the h@xmath258 luminosity ; @xcite ) and energy content @xmath269 erg @xcite . the two best - studied examples are the bubbles around ngc 1313 x-2 and holmberg ix x-1 @xcite . ulx bubbles resemble ordinary sn remnants , but are an order of magnitude larger and 10 to 100 times more luminous . they can not come from ( hypothetical ) single hypernova explosions : first , because the surrounding stellar populations are already too old ( typically @xmath270 yr ) and without o stars ; secondly , because it is unlikely that a lower - mass binary companion that survived the hypothetical hypernova explosion could evolve to fill its roche lobe and feed the ulx already within @xmath271 myr @xcite . invoking 10 to 100 ordinary sn explosions over the last 1 myr would require a massive star cluster with @xmath272 a few @xmath273 ( _ e.g. _ , from starburst99 ; @xcite ) . however , none of those ulxs is associated with a massive star cluster : at most , they are inside ordinary ob associations with age @xmath0@xmath108 myr and total stellar mass @xmath92 . the only plausible conclusion is that ulx bubbles are inflated by a continuous jet or outflow from the ulx , with a power comparable to the radiative output @xcite . jet - inflated bubbles around accreting bhs are found in our galaxy , too . the best - known example is the w50 nebula around ss433 @xcite : a spherical shell with two lateral extensions , the ears , that are inflated by interaction of the jets with the interstellar medium . the characteristic diameter along the major axis is @xmath152 pc . both the jet power and the radiative luminosity are @xmath13 erg s@xmath12 . indeed , it is suggested that ss433 would appear as a beamed , super - eddington ulx if it was seen face - on @xcite . another local example is the 5-pc bubble inflated by the smbh cygnus x-1 @xcite : the shock velocity determined from optical emission line ratios is @xmath260@xmath274 km s@xmath12 ; the required jet power @xmath275 erg s@xmath12 is @xmath276@xmath206 times the x - ray luminosity . the ulx bubbles discovered so far have a quasi - spherical appearance , with no direct signature of relativistic , collimated jets ( as in the galactic source ss433 ) . the first extra - galactic super - bubble with collimated jets was recently found in the outskirts of the spiral galaxy ngc7793 @xcite . this source ( figure [ fig : bubble ] ) , known as s26 , has : x - ray and optical hot spots where the jet and counter - jet interact with the interstellar medium ; an optical nebula with a size of @xmath277 pc , expanding at @xmath278 km s@xmath12 ; an associated x - ray and radio nebula with bright radio lobes . the long - term - average jet power @xmath17 a few @xmath19 erg s@xmath12 @xcite , over a characteristic age @xmath279 yrs . the x - ray luminosity of the central bh is only @xmath280 erg s@xmath12 . this might be because the bh emission is highly obscured and we only see a scattered component ; or because it is beamed away from us ; or because the bh is a transient , currently in a radiative - faint state , but may appear as a ulx at other times . whatever the reason for the current radiative dimness , it is important to note that ngc7793 s26 has had collimated jets with a long - term average power at least of order of the eddington power ( and probably much greater ) of a stellar bh ( smbh or msbh ) . in other words , it is a jet - dominated system at super - eddington mass accretion rates . there is no physical principle that prevents mechanical power to exceed the eddington luminosity . however , the fact that it is collimated in a large - scale jet may be surprising . the canonical view @xcite of bh accretion states based mostly on the behavior of galactic transients is that collimated , steady jets are found at @xmath281 , when accretion is radiatively inefficient ( low / hard state ) . whether and how steady , collimated jets can be launched during super - eddington accretion phases remains an unsolved theoretical problem . a possible solution comes from the radiation - magnetohydrodynamic simulations of @xcite , who find that a jet can be accelerated up to mildly relativistic speed by radiation pressure , and collimated by the lorentz force of a magnetic tower made of wound - up toroidal magnetic field lines . the magnetic structure is itself kept collimated by the geometrically thick accretion flow that dominates in the innermost region when @xmath282 . an ultrapowerful ( rather than ultraluminous ) system such as ngc7793 s26 is the local - universe analog of a recently discovered class of quasars , dominated by mechanical power @xcite . in those quasars , the jet power is up to 25 times higher than the bolometric radiative luminosity ; in some of those systems , the jet power is near or above the eddington power . mechanically - dominated , radiatively inefficient quasars in the early universe may grow much faster than eddington - limited sources , and would provide strong mechanical rather than radiative feedback onto the surrounding gas . thus , understanding the inflow / outflow / radiation processes in ulxs has wider astrophysical relevance . moreover , radiative and mechanical feedback from hmxbs and ulxs themselves may have been an additional , important source of heating and reionization of the intergalactic medium at high redshifts @xcite . some ulx bubbles ( from both the photoionized and shock - ionized variety ) are associated to extended radio emission . the best - studied examples are holmberg ii x-1 @xcite and ngc5408 x-1 @xcite . both radio nebulae are dominated by optically thin synchrotron emission , somewhat similar but much larger ( @xmath283 pc and @xmath284 pc , respectively ) and more luminous than typical radio sn remnants . they are also an order of magnitude more luminous than the radio nebula around ss433 @xcite , which well matches the relative values of radiative and mechanical core power in those sources . the total power carried by synchrotron - emitting electrons has been estimated to be between @xmath17 a few @xmath285 erg s@xmath12 and @xmath286 erg s@xmath12 , depending on the very uncertain lower - energy cut - off of the distribution . in any case , both nebulae are much more energetic than w50 around ss433 ( @xmath17 a few @xmath287 erg s@xmath12 : @xcite ) or than the extended lobes of typical galactic microquasars such as grs1758@xmath93258 ( @xmath288 erg s@xmath12 ; @xcite ) . ngc7793 s26 also has a large radio nebula @xcite , interpreted as a combination of flat - spectrum free - free emission ( mostly in the central region ) and steep - spectrum synchrotron emission from the radio hot spots and lobes . the energy stored in relativistic electrons is @xmath17 a few @xmath289 erg s@xmath12 . the energy content in relativistic electrons for typical ulx nebulae is only @xmath290@xmath291 of the core power in particular , of the mechanical power when it can be reliably estimated from bubble size or optical line fluxes integrated over the characteristic age of the bubble . the main reason is that most of the input energy is transferred to non - relativistic electrons , protons and nuclei @xcite . the fraction of energy in non - radiating particles to that in relativistic electrons in a system such as ngc7793 s26 is similar to the values found by @xcite for a sample of agn , seyfert and giant elliptical galaxies . for the same jet power , agn , ulxs and galactic microquasars may be radio louder or radio quieter depending on the amount of plasma entrained by the jets . we speculate that an interesting test may come from comparisons of the radio loudness and lobe energy in low / hard state microquasars , probably dominated by magnetically accelerated jets @xcite , and in ulxs , possibly dominated by radiation - pressure accelerated jets / outflows @xcite , with more entrainment . there is no doubt that our understanding of the phenomenology of ulxs has greatly advanced in the past decade thanks to _ chandra _ , _ xmm - newton _ , and many other space- and ground - based facilities . however , the key physical issue is still unresolved : that is , whether they are powered by imbhs or normal stellar bhs . a few sources have emerged as strong imbh candidates : in particular , eso243@xmath9349 hlx-1 and m82 x-1 . on the other hand , for the majority of ulxs , there is no strong theoretical need nor compelling observational evidence for imbhs . in the absence of direct mass - function measurements from phase - resolved optical spectroscopy , we still have to rely on x - ray spectral and timing modeling and other indirect clues . reviewing the results that we have collected here , we suggest the following conservative statement : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ulxs are a diverse population ; msbhs with moderate super - eddington accretion seem to be the easiest solution to account for most sources up to luminosities @xmath17 a few @xmath19 erg s@xmath12 ; strong beaming ( @xmath292 ) can be ruled out for the majority of ulxs ; imbhs are preferred in a few exceptional cases . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ some of the following kinds of observations may provide further breakthroughs : a. measurements of the fast timing behavior down to a time scale of sub - seconds may probe the inner part of the inflow : ideally , we would search for possible high frequency features ( breaks and qpos ) that are found in galactic bhs at frequencies @xmath82 hz . this would require a new generation of x - ray telescopes with large effective area . b. determining the relative contribution of thermal emission and comptonization component is a key test for competing accretion models . for that , multilayer - coated x - ray telescopes with good sensitivity up to a few tens of kev are needed . c. detection of , or a more solid upper limit to broad iron k lines may constrain the properties of the inner disk . d. high - resolution x - ray spectroscopy with gratings or microcalorimeters may constrain the metal abundance and ionization state of the inflow , the mass and energy of the outflow , and probe the absorption from the surrounding interstellar medium . e. the ulx sample within @xmath293 mpc is almost complete , and it does not contain enough sources to permit statistical studies of the luminosity distribution above @xmath294 a few @xmath58 erg s@xmath12 . a homogeneous all - sky survey , complete for ulxs up to @xmath212 mpc would allow us to determine whether or not the luminosity distribution extends unbroken up to at least @xmath18 erg s@xmath12 . such survey will hopefully be successfully conducted by _ erosita _ ; f. repeated snapshot monitoring of nearby ulxs over the next few years will extend their long - term lightcurves enough to prove that their duty cycles and transient behavior are different from those of galactic bhs . this can be easily done with smaller x - ray missions ; g. phase - resolved spectroscopy would lead to the measurement of radial velocity curves and mass functions . attempts to do so with 8-m class telescopes have so far been inconclusive but have suggested that it will be easily done with future 30-m class telescopes ; h. integral - field spectroscopic studies of star forming galaxies , as well as low - frequency radio studies , may reveal fainter , older ulx bubbles or lobes , still expanding into the interstellar medium but no longer energized by a central source . this would constrain the duration of the ulx active phase ; i. searching for compact radio jets down to a sensitivity of @xmath295jy will constrain the bh mass , in relation to other classes of accreting bhs in the fundamental plane . this may be done in the near future with the _ evla _ and with _ ska _ pathfinders , and in the more distant future with the ska itself . in conclusion , we gather from this list that there is more than one way to skin a cat . many fundamental aspects of ulx physics will be understood with instruments already planned or under construction . thus , we anticipate a fruitful decade of ulx studies beyond the _ chandra _ and _ xmm - newton _ era . we thank andy fabian , sean farrell , luigi foschini , fabien gris , phil kaaret , kip kuntz , manfred pakull , tim roberts , mat servillat , tod strohmayer , doug swartz , dany vanbeveren , and luca zampieri for helpful discussions and comments . hf acknowledges funding support from the national natural science foundation of china under grant no . 10903004 and 10978001 , the 973 program of china under grant 2009cb824800 , the foundation for the author of national excellent doctoral dissertation of china under grant 200935 , the tsinghua university initiative scientific research program , and the program for new century excellent talents in university . rs acknowledges support from a curtin university senior research fellowship , and hospitality at the mullard space science laboratory , tsinghua university and the university of sydney during part of this work . , f. , pakull , m. w. , soria , r. , & motch , c. 2009 , in american institute of physics conference series , vol . 1126 , american institute of physics conference series , ed . j. rodriguez & p. ferrando , 201203 , m. w. , & gris , f. 2008 , in american institute of physics conference series , vol . 1010 , a population explosion : the nature & evolution of x - ray binaries in diverse environments , ed . r. m. bandyopadhyay , s. wachter , d. gelino , & c. r. gelino , 303307 , a. h. , galache , j. , kalogara , v. , linden , t. , kilgard , r. , zezas , a. , wolter , a. , & trinchieri , g. 2010 , in bulletin of the american astronomical society , vol . 36 , american astronomical society meeting abstracts # 215

inflation proved to be very successful as a theory of early universe and structure formation @xcite which is strongly supported by cosmic observations @xcite . the simplest models of inflation are based on a scalar field minimally coupled to gravity with a potential flat enough to support an extended period of inflation . despite its observational successes there is no deep theoretical understanding of inflation and the nature of the inflation field . there have been many attempts during past decade to embed inflation within the context of string theory , for a review see @xcite and references therein . brane inflation is an interesting realization of inflation from string theory @xcite . in its original form , it contained a pair of d3 and anti d3 branes moving in the calabi - yau ( cy ) compactification . the inflaton field is the radial distance between them so in this sense the inflaton field has a geometric interpretation in string theory . inflation ends when the distance between the brane and the anti - brane reaches the string length scale where a tachyon develops in the open strings spectrum stretched between the pair . inflation ends soon after tachyon formation and the energy stored in branes tensions are released into closed string modes @xcite . however , it was soon realized that the potential between the pair of brane and anti - brane is too steep to allow a long enough period of slow - roll inflation . to flatten the potential , it was suggested to put the pair of brane and anti - brane inside a warped throat @xcite , where the potential between d3 and @xmath0 is warped down as in randall - sundrum scenario @xcite . as a novel feature of brane inflation the question of branes annihilation and particles creation during inflation was studied in @xcite . in this picture two stacks of branes , located at different places inside the throat , are moving in the slow - roll limit towards the bottom of the throat where anti - branes are located . the stack closer to the tip will be annihilated during inflation transferring its energy into closed string modes . the second stage of inflation is driven by the remaining stack of branes until it is annihilated by the remaining anti - branes at the bottom of the throat . the process of fields annihilation @xcite and particle creations @xcite can have interesting observational consequences on cmb . here we generalize the results in @xcite to the case where the mobile stacks of branes are moving ultra relativistically as in dbi inflation @xcite . as in @xcite , the stack closer to the tip is annihilated by the background anti - branes resulting in particles creation during inflation while the second stage of inflation is driven by the remaining stack . the rest of the paper is organized as follows . in section [ setup ] we present our set up and background solutions for dbi inflation . in section [ perturbation ] we study the curvature perturbations and present our matching conditions . in section [ power ] we obtain the power spectrum of curvature perturbations and the spectral index . discussions and conclusions are presented in section [ conclusion ] . in this section we present our set up . as explained above we consider two stacks containing @xmath1 and @xmath2 coincident branes . the positions of the stacks are represented by functions @xmath3 and @xmath4 inside the throat . there are @xmath5 anti - branes located at the bottom of the throat , @xmath6 . we assume that the first stack is closer to the tip of the throat so it is annihilated during inflation transferring its energy into closed strings modes . the second stage of inflation is driven by the remaining stack of @xmath2 branes . for a schematic view , see * fig . * [ throat1 ] and * fig . * [ throat2 ] . and @xmath2 coincident branes inside the throat . the stacks are moving relativistically towards the bottom of the throat where there are @xmath7 anti - branes . the figure is borrowed from @xcite , width=192 ] [ throat1 ] . the stack closer to the tip of the throat is annihilated during inflation transferring its energy into radiation . the second stage of inflation is driven by the remaining stack containing @xmath2 branes . inflation ends when this stack is annihilated by the remaining @xmath2 anti - branes at the bottom of the throat . the figure is borrowed from @xcite . , width=432 ] [ throat2 ] as usual the metric of the throat is [ metric ] ds^2=h^-(r)g_dx^dx^+h^(r)(dr^2+r^2d_5 ^ 2 ) , where @xmath8 is the radial coordinate , @xmath9 represents the internal five - dimensional azimuthal directions which we may take to be an @xmath10 and @xmath11 is the warp factor [ warp - factor ] h(r)= . here @xmath12 is the ads length scale of the throat which is created by @xmath13 coincident background branes located at @xmath14 ( for a review see @xcite ) , [ l ] l^4 = 4g_s n ^2 , where @xmath15 is the perturbative string coupling and @xmath16 where @xmath17 is the string length scale . the mobile stacks of @xmath1 and @xmath2 branes are located at positions @xmath18 and @xmath19 as probe branes . in order for our probe branes approximation to be valid , we need @xmath20 . in addition there are @xmath7 @xmath0 branes at the bottom of the throat @xmath6 . once the distance between the stack of @xmath1 branes and these anti - branes reaches at the order of @xmath17 the open strings stretched between them become tachyonic , indicating an instability in the system . shortly after tachyon formation , the stack of @xmath1 branes is annihilated by @xmath1 anti - branes at the bottom of the throat . consequently , the energy stored in @xmath1 branes and anti - branes tension , which is @xmath21 , is transferred into closed strings modes . for simplicity , we assume that these closed strings modes are in the form of massless particles ( radiation ) . of course a combinations of massive and massless particles would be created . however , once these particles are created they will be diluted quickly by the background inflation so for our practical purposes , there are not much differences between massive and massless particle creations . the goal of this work is to examine the effects of branes annihilation and particles creation on the remaining field which drives the second stage of inflation . in @xcite this was studied for the case where the stacks are moving slowly . here we generalize those analysis to our case at hand where the stacks are moving ultra relativistically @xcite . the dirac - born - infeld ( dbi ) action for the stacks of @xmath1 and @xmath2 branes moving relativistically in the background of eq . ( [ metric ] ) is [ action ] s=_i=1 ^ 2 p_i t_3 d^4x , where @xmath22 is the d3-brane tension with @xmath23 . we also have added the unknown contributions @xmath24 and @xmath25 from the back - reactions of the background fluxes , branes and kahler modulus stabilization on the dynamics of mobile branes @xcite . as usual , we shall continue from the phenomenological point of views and parametrize these potentials to be suitable for inflation @xcite . to proceed further , we simplify the dbi action into the following form which is more suitable for our analysis [ action2 ] s=_i=1 ^ 2d^4x , where @xmath26 , @xmath27 and @xmath28 . this form of multiple field dbi inflation has been studied extensively in the past @xcite . now we are ready to promote the system into a cosmological set up . the background four - dimensional frw metric is ds^2 = -dt^2 + a(t)^2d^2 , where @xmath29 is the scale factor . the friedmann equation and the energy conservation equations are [ back1 ] 3h^2= , + 3h(+p ) = 0 , where @xmath30 is the hubble expansion rate . the energy density @xmath31 and the pressure @xmath32 are given by = _ i , p=_i . here the lorentz factor for each field , @xmath33 , is defined by _ finally , the background klein - gordon equation for each field is [ eqm phi ] _ i+3h_i^-2_i+_i^2-+_i^-3 ( v_i+ ) = 0 . so far we have not specified the form of the potentials , @xmath34 . as explained above , these potentials arise from the back - reactions of the fluxes , branes and kahler modulus stabilization to the mobile branes . it is a non - trivial question in string theory as how one can calculate these back - reactions @xcite concretely . however , to be specific , we consider the phenomenological approach where the potentials have the form @xmath35 . we shall also keep the masses @xmath36 undetermined except that their magnitudes should be such that they give the desired number of e - foldings and fit the cmb observations such as the cobe normalization . we are interested in the limit where the branes are moving ultra relativistically inside the throat so one can not expand the square root in eq . ( [ action2 ] ) perturbatively in terms of @xmath37 . this corresponds to the case where @xmath38 and the stacks of branes are approaching the speed limit @xmath39 . in this limit ( [ eqm phi ] ) reduces to [ eqm phi app ] _ i+_i^2-=0 , which can be verified to have the speed limit solution @xmath40 . one can plug this solution back into eq . ( [ eqm phi ] ) to find the next leading correction and [ phi - back ] _ i(t ) ( 1- ) . to get this solution , we have neglected the variation of @xmath41 during inflation to leading order . to obtain the solution for the scale factor @xmath29 we require that the energy density is dominated by the potential energy so the friedmann equation ( [ back1 ] ) reduces to @xmath42 . below we find the background solutions for both inflationary stages , before the collision and after the collision . before collision both stacks contribute to the energy density and the friedmann equation can easily be solved to give [ dot - a ] = , where the dimensionless number @xmath43 is defined by [ lambda-]_- ( _ i ) ^1/2 . working with the number of e - foldings @xmath44 as the clock , @xmath45 , one obtains [ n ] n(t)= _ - , where @xmath46 is the time at the start of inflation when @xmath47 . define @xmath48 the time when the first branes collision take place and the stack of @xmath1 branes is annihilated by @xmath1 background anti - branes . similarly , define @xmath49 as the corresponding fields values at the time of collision . combining eqs . ( [ n ] ) and ( [ phi - back ] ) yield [ phiic ] _ ic _ i in e^- , where @xmath50 are the corresponding fields values at the start of inflation . one can find an approximate formula for the time of branes collision as follows . as mentioned before , the time of branes annihilation is when the physical distance between the stack of @xmath1 branes and anti - branes located at the bottom of the throat @xmath6 reaches the string length scale @xmath17 . starting with the metric ( [ metric ] ) the physical distance between the first stack and the anti - branes , @xmath51 , is calculated to be = _ setting @xmath52 and using eq . ( [ phiic ] ) the onset of branes collision , @xmath48 , is obtained to be [ nc ] n_c & & _ - ( - ) + & & _ - ( - ( 4 g_s n)^-1/4 ) , where to get the final relation eq . ( [ l ] ) has been used . here we defined @xmath53 as the values of the corresponding fields at the position of the anti - branes . correspondingly , the position of @xmath54 at the time of collision is obtained to be _ 2c _ 2 in e^- . also it is instructive to calculate @xmath33 . using eqs . ( [ phi - back ] ) and ( [ dot - a ] ) in the expression of @xmath33 one finds _ this indicates that as inflation proceeds , @xmath33 increases like @xmath55 . after the first stack is annihilated , the second stage of inflation is driven by the remaining stack of @xmath2 branes and the scale factor is given by @xmath56 where now [ lambda+ ] _ + ( ) ^1/2 . defining the total number of e - foldings by @xmath57 , with @xmath58 to solve the flatness and the horizon problem , one has n_t - n_c , where @xmath59 is the time of end of inflation when the second stack is annihilated by the remaining @xmath2 anti - branes . as before , this happens when the physical distance between the second stack and the anti - branes reaches the string scale @xmath17 which can be used to determine the value of @xmath54 at the end of inflation @xmath60 . similarly , one can show [ nt ] n_t & & n_c + _ + + & & ( 1- ) n_c + _ + , one should arrange the background parameters such as @xmath61 , @xmath62 and @xmath50 such that large enough @xmath57 can be obtained to solve the flatness and the horizon problem . correspondingly , the lorentz factor after the collision is _ 2^+ . one can find a measure of the magnitudes of @xmath62 as follows . for dbi inflation to match the cobe normalization , one requires a large background charges such that @xmath63 @xcite so one can safely neglect this term in eqs . ( [ nt ] ) and ( [ nc ] ) . assuming that there is not exponential hierarchy between @xmath64 and @xmath65 one concludes that @xmath66 . however , by keeping the ratio @xmath67 exponentially large in the light of @xcite , one can lower @xmath68 by a factor of ten or so . in our analysis below , we shall take take @xmath69 . as shown in @xcite significant amount of non - gaussianities can be created in dbi inflation where the branes are moving ultra - relativistically . the non - gaussianity parameter @xmath70 is related to the lorentz factor via @xmath71 . to satisfy the wmap constraints on @xmath70 one requires that @xmath72 @xcite . having studied our inflationary background in some details , now we consider the perturbations in this background . as discussed in @xcite a complete treatment of perturbations in our set up with branes annihilation and particles creation is a non - trivial task . the process of field annihilation and particles creation are determined by the dynamics of open string tachyon condensation . in the presence of background branes and fluxes this is a non - trivial phenomena @xcite . we shall instead take the phenomenological approach and impose some physically reasonable approximations to bypass these difficulties . before the collision we have two scalar fields @xmath73 and @xmath54 corresponding to two stacks of branes moving relativistically . after the collision , the energy associated with field @xmath73 is transferred into closed strings modes , which we approximate them as massless particles . the process of field annihilation and closed strings formation takes some time scale , @xmath74 . in our set up where the collision happens at the bottom of the throat @xmath74 is given by the inverse of the warped string mass scale , that is @xmath75 where @xmath76 is the string theory mass scale . note that the factor @xmath77 in @xmath74 is in the light of randall - sundrum idea @xcite . in our effective four - dimensional approach where @xmath78 , we can take the process of field annihilation and particles creation as instantaneous and set @xmath79 for practical purposes . this approximations is certainly true for large scale perturbations . to be slightly more specific , we can borrow the results of @xcite . as mentioned before , after brane and anti - brane annihilation their energy is transferred into closed strings modes . subsequently , these modes cascade down to graviton zero mode and the kaluza - klein ( kk ) excitations . this happens in a time scale controlled by the inverse of the warped sting mass scale , much shorter than @xmath80 . subsequently , the massive kk modes decay into lighter modes . because of the warp factor , the wave function of these modes are highly peaked at the bottom of the throat where the annihilation takes place . their effects on the remaining stack of @xmath2 branes at position @xmath19 may be negligible in the limit where the stack of @xmath2 branes is far away from the tip of the throat . even in the instantaneous brane annihilation approximation it is not clear within string theory set up as how to turn off field @xmath73 and turn on radiation . to bypass this shortcoming we follow the practical approach employed in @xcite . we assume that field @xmath73 is subdominant in the background energy density , corresponding to @xmath81 . after @xmath73 is annihilated , we assume that not only its energy is transferred into radiation energy density , @xmath82 , but also the perturbations carried by @xmath83 are transferred into perturbations in radiation , @xmath84 . we also assume that the perturbations @xmath84 , before significantly diluted by the background expansion , do not interfere with @xmath85 after collision . this can be considered as a higher order back - reaction , perhaps controlled by the ratio @xmath86 , which we take to be small . also this back - reaction may be negligible as long as the separation between two stacks of branes are large enough inside the throat . within these approximations we do not need to follow the perturbations carried by @xmath83 before the collision and the perturbations carried by @xmath84 after the collision . in these approximations we shall follow only the perturbations @xmath85 before and after the collision . this may not be such a bad approximation because physical quantities , such as curvature perturbations , are calculated at the end of inflation when only field @xmath54 and its perturbations are present . since we do not follow @xmath83 and @xmath84 , our treatments of perturbations eventually reduce to those of single field model where only @xmath85 is counted before and after the branes collision . however , the effect of @xmath73 and radiation is kept in the background so that is how @xmath85 feels their presences . with these approximations we start the analysis of perturbations in our set up . the metric perturbations in the conformal newtonian gauge where there is no anisotropy in stress energy tensor is ds^2=-(1 + 2 ) dt^2+(1 - 2)a^2(t)_ijdx^idx^j , where @xmath87 is the bardeen potential which in this gauge coincides with the newtonian potential . defining the curvature perturbations @xmath88 on comoving surface @xmath89 via [ r - eq ] _ 2h+ , and introducing the sasaki - mukhanov variables @xmath90 [ z - def ] v_2=z_2 r_2 , z_2 , one obtains the following standard second order differential equation @xcite [ v eq ] v_2^+ ( k^2 c_s ^2 - ) v_2=0 . here the prime denotes the derivative with respect to the conformal time @xmath91 and @xmath92 is the sound speed associated with @xmath54 perturbations defined by c_s = _ 2 ^ -1 = . as mentioned before , we work in the ultra relativistic limit where @xmath93 corresponding to a very small sound speed @xmath94 . this has interesting consequences for non - gaussianities @xcite . to solve eq . ( [ v eq ] ) we need to find the form of @xmath95 . plugging different components of @xmath95 from the background specified in the previous section , one obtains z_2^t^ _ + 2 ^- . this in turn yields = ( 2 + ) ^-2 , where to get the final approximation the relation @xmath96 is used as explained in previous section . consequently , one can cast eq . ( [ v eq ] ) into more conventional form [ v2-eq ] v^_2_+(k^2c_s^2- ) v_2_=0 , where the indices @xmath97 are given by [ mu - index ] _ + + 3 _ . we have defined the `` slow - roll '' parameter @xmath98 for the later convenience . as argued before , @xmath99 . one should solve eq . ( [ v2-eq ] ) for both @xmath100 and @xmath101 and glue the solutions via the matching conditions . in the limit where we can neglect the running of @xmath102 the solutions of eq . ( [ v2-eq ] ) as usual are given in terms of hankel functions @xmath103 and @xmath104 . at the early stage of inflation when all physically relevant modes are inside the sound horizon , corresponding to @xmath105 , the solutions are given in terms of positive frequency modes @xmath106 created from vacuum . this initial condition eliminates @xmath104 and the solution of eq . ( [ v2-eq ] ) during the first inflationary stage is [ v- ] v_2_-=e^i(_-+1/2)/2h^(1)__-(-c_sk ) . after the collision , due to particles creation , the system is not in vacuum and both solutions @xmath103 and @xmath104 are permitted . we parametrize the solution after the collision in terms of the coefficients @xmath107 and @xmath108 via [ v+ ] v_2_+=e^i(_++1/2)/2 . in the absence of any brane collision and particles creation @xmath109 , @xmath110 and the two solutions ( [ v- ] ) and ( [ v+ ] ) coincides . our job below is to find the coefficients @xmath107 and @xmath108 and read off the final value of curvature perturbations as a function of @xmath107 and @xmath108 . for this purpose , we need to impose our matching prescriptions , joining eqs . ( [ v- ] ) and ( [ v+ ] ) at the time of collision @xmath111 . imposing the matching condition is the most important step in our analysis of perturbations . there are standard prescriptions in literature for imposing the cosmological matching conditions in systems where there is a sudden jump in equation of state @xcite . more specifically , in models where the matching conditions are imposed on a comoving hyper - surface one requires that both the extrinsic and the intrinsic curvatures to be continuos across this hyper - surface . these result in the following matching conditions @xcite [ match0 ] [ ] _ -^+ = [ r ] _ -^+ = 0 , where @xmath112 is the gauge invariant bardeen potential and @xmath113 is the comoving curvature perturbations . in principle we can also impose these matching conditions in our system . however , as mentioned at the start of this section , we do not have a theoretical control on the dynamics of tachyon condensation , fields annihilation and particles creation . these shortcomings in turn result in a lack of knowledge of @xmath112 and @xmath113 after the collision . to bypass these shortcomings we have chosen the phenomenological approach where only the perturbations of @xmath54 are followed before and after the collision . in this approximation our model is effectively a single field scenario with the extra modification that the effects of particles creation are kept at the background level . this in turn results in different indices for hankel functions , @xmath114 . in this view , our method is very similar to the method used in @xcite where the effects of phase transitions in a multiple field model is translated into a sudden violation of slow - roll parameters in an effective single field model . therefore in our studies here , as in @xcite , one expects that the matching conditions ( [ match0 ] ) are simplified into [ match1 ] [ v_2]_-^+ = [ v_2]_-^+ = 0 , i.e. both @xmath90 and @xmath115 are continuous across the surface of branes collision . using the matching conditions ( [ match1 ] ) in eqs . ( [ v- ] ) and ( [ v+ ] ) one obtains [ alpha - beta0 ] & = & -e^i + & = & e^i , where the derivatives are with respect to the arguments of the hankel functions and x - c_s k _ c , ( _ - - _ + ) . we are in a position to calculate the curvature power spectrum @xmath116 at the end of inflation . the curvature power spectrum is defined via @xmath117 since only the field @xmath54 is present at the end of inflation we only need the power spectrum of @xmath88 . using eqs . ( [ r - eq ] ) and ( [ z - def ] ) one obtains _ 2 = . at the later stage of inflation , when the mode of interest is outside of the sound horizon and @xmath118 , one can approximate @xmath119 plugging this into the power spectrum definition yields [ pr ] _ = _ _ 0 the curvature power spectrum in the absence of branes collision @xmath120 observationally @xmath121 is fixed by the cobe normalization , @xmath122 , when the mode corresponding to the current hubble radius leaves the horizon at about 60 e - folds before the end of inflation with @xmath123 . one can use the cobe normalization to fix a combination of parameters . specifically , using eqs . ( [ phi - back ] ) and ( [ lambda+ ] ) we have _ _ 0 . combined with eq . ( [ lambda+ ] ) the cobe normalization yields @xmath124 . assuming that @xmath125 , this yields @xmath126 . this is the infamous fine - tuning problem associated with the dbi inflation @xcite . correspondingly @xmath127 . to calculate the background curvature perturbations spectral index , @xmath128 , we note that at the time of horizon crossing @xmath129 . using this identity in ( [ pr0 ] ) yields [ nr0 ] n_r^0 - 1 ( ) ( _ + ^2 ) , where to get the final result the relations @xmath130 and @xmath131 have been used from the background . this indicates that @xmath128 is nearly scale invariant up to @xmath132 @xcite . ) with @xmath133 and @xmath134 . for large scales , @xmath135 , it has a monotonic behavior as indicated by eq . ( [ alpha - beta - sup ] ) whereas for small scales , @xmath136 , it has an oscillatory behavior approaching the unity . the transition takes place for intermediate scales , @xmath137 , modes which leave the horizon around the time of branes annihilation . , title="fig:",width=302 ] ) with @xmath133 and @xmath134 . for large scales , @xmath135 , it has a monotonic behavior as indicated by eq . ( [ alpha - beta - sup ] ) whereas for small scales , @xmath136 , it has an oscillatory behavior approaching the unity . the transition takes place for intermediate scales , @xmath137 , modes which leave the horizon around the time of branes annihilation . , title="fig:",width=302 ] [ transfer - fig ] the effects of branes collision and particles creation in the power spectrum eq . ( [ pr ] ) are encoded in the transfer function @xmath138 . in the absence of any feature , @xmath109 so we obtain the standard single field dbi inflation @xcite . below we would like to investigate the behavior of the transfer function @xmath138 for different modes . using the explicit formulae of @xmath107 and @xmath108 given in eq . ( [ alpha - beta0 ] ) one obtains [ alpha - beta ] | -|^2 = x^2 . as mentioned before we have defined @xmath139 where @xmath140 represents the critical mode which leaves the sound horizon at the time of branes collision during inflation . one can check that in the absence of any feature , corresponding to @xmath141 , one obtains @xmath142 as expected . as mentioned previously , the effects of branes annihilation and particles creations are such that @xmath143 where @xmath144 . the spectral index is also modified in the presence of the transfer function as follows [ nr ] n_r -1 = n_r^0 -1 + x | -|^2 . now let us look at the shape of the transfer function for different length scales . in * fig . * [ transfer - fig ] we have plotted the transfer function . as a measure of the length scales of the perturbations , we note that @xmath145 corresponds to the critical mode @xmath146 which leaves the sound horizon right at the time of branes annihilation . then depending on whether the mode of interest leaves the sound horizon before the collision or after the collision one respectively has @xmath135 or @xmath136 . first consider the super - horizon scales , @xmath135 . this corresponds to modes which leave the sound horizon before the collision . using the small argument limit of the bessel functions we obtain [ alpha - beta - sup ] | -|^2 ( _ -+ _ + ) ^2 ( ) ^2 . our background is such that @xmath147 due to branes collision and energy transfer into radiation so @xmath148 . as a result the transfer function scales monotonically for super - horizon modes . correspondingly , the change in spectral index is obtained to be [ nr - sup ] n_r n_r - n_r^0 2 . we see from eq . ( [ nr - sup ] ) that the spectral index for the super - horizon scales become more blue - tilted as compared to the background spectral index . this can also be seen in * fig . * [ nr - fig ] . . for super - horizon modes , @xmath135 , our analytical formula eq . ( [ nr ] ) gives a very accurate prediction for @xmath149 . for small scales , @xmath136 , the oscillatory behavior of @xmath149 is clearly seen as suggested by eq . ( [ alpha - beta ] ) where @xmath149 approaches its background value @xmath128 denoted by the dotted line . the transition takes place around @xmath137 for modes which leave the horizon around the time of branes annihilation . the parameters are as in * fig . * [ transfer - fig],width=302 ] . [ nr - fig ] now consider the sub - horizon modes , corresponding to @xmath150 . these are the modes which are deep inside the sound horizon at the time of branes collision . physically one expects these modes to live in their local minkowski background and should not be affected by branes collision and particles creation . indeed , using the large argument limit of the bessel functions @xmath151 and @xmath152 for @xmath153 we find [ alpha - beta - sub ] | -|^2 1 + o ( x^-2 ) . as expected the transfer function approaches unity so @xmath154 . however , both the transfer function and the spectral index are superimposed by oscillatory modulations with decaying amplitude @xcite . the most interesting effects of branes annihilation and particles creation is on the intermediate scales , @xmath155 . these are the modes which leave the sound horizon around the time of branes collision . in figures [ transfer - fig ] and [ nr - fig ] we have plotted the transfer function and the spectral index . as can be seen , for super - horizon scales the transfer function is monotonically increasing whereas for small scales it oscillates around unity . correspondingly , the spectral index for super - horizon scales is well approximated by eq . ( [ nr - sup ] ) whereas for small scales it oscillates around @xmath156 . the non - trivial transition , happening at the intermediate scales , can be seen in both figures . as discussed before , the effects of branes annihilation and particles creation result in @xmath157 . so far in the perturbations analysis we kept @xmath97 and @xmath158 undetermined . now let us specify @xmath159 in terms of our background model parameters . starting with eqs . ( [ lambda- ] ) , ( [ lambda+ ] ) and ( [ mu - index ] ) we obtain = _ + - _ - & & - + & & . to get the final result we used the approximation that @xmath160 . physically , this is motivated from the assumption that the energy density associated with stack 1 is subdominant compared to the energy density stored in stack 2 so the energy transferred into radiation is much smaller than the background potential energy density . this assumption is necessary for the consistency of our matching condition . we have neglected the interference of perturbations @xmath83 and @xmath84 on the perturbations @xmath85 . as a consequence , we have imposed the simple matching condition ( [ match1 ] ) . for this approximation to be valid , we require that @xmath161 or equivalently @xmath162 . we do not have good theoretical control on the masses @xmath163 and @xmath164 which were originated from the back - reactions of kahler moduli stabilization and background fluxes on the mobile branes . as a simple ansatz , we may take @xmath165 . this corresponds to the case that the back - reactions imposed from the background compactification on a single brane is independent of its position inside the throat . this may not be an unreasonable assumption . with this approximation , the consistency of our approximation requires @xmath166 . also we are working in the probe brane approximations so the mobile stacks of branes should not deform the background ads geometry significantly . for this to be the case we also need to impose @xmath167 . as an example , taking @xmath168 and @xmath169 results in @xmath170 . one may also compare @xmath171 induced from branes collision to the @xmath172 contributions to the background spectral index in eq . ( [ nr0 ] ) . we have @xmath173 . for a reasonable value of @xmath174 which was used in our numerical examples in figures [ transfer - fig ] and [ nr - fig ] we have @xmath175 or @xmath176 . this indicates that the contributions of the branes collision to the spectral index is much bigger than the sub - leading @xmath172 contributions to the background @xmath177 . observationally the result @xmath178 is disfavored @xcite noting that the background spectral index is already nearly scale invariant as seen in eq . ( [ nr0 ] ) . to remedy this problem , it may be necessary to embed the simple model of ultra relativistic dbi inflation here to a generic setup of brane inflation where a period of slow - roll brane inflation takes place before dbi inflation starts . this can help to obtain a red spectral index for large scales whereas for smaller scales the spectral index can be blue as we have here . in this work we have studied the effects of branes annihilation and particles creation in models of dbi inflation where the branes are moving ultra relativistically . in a typical string theory compactification the process of branes and anti - branes annihilation is a generic phenomenon . it is therefore an interesting question how these collisions can affect the dynamics of inflation which may also be detectable in the sky . the process of branes annihilation is determined by the dynamics of open string tachyon condensation . in a complicated background , such as the warped throat with background fluxes , this is a non - trivial process . the time - scale of tachyon condensation and branes annihilation is given by the inverse of the warped string mass scale , @xmath179 . in our effective field theory approach the hubble expansion rate is considerably smaller than this mass scale so for practical purposes we can take the process of branes annihilation and particles creation to be instantaneous . in our set up the first inflationary stage is a two - field dbi inflation model . after the first stack of branes is annihilated with the background anti - branes a burst of closed strings modes are created . these particles are rapidly diluted by inflation and we eventually end up with a single field dbi inflation system . in order to connect the perturbations in the second inflationary stage to the perturbations generated before the collision we have to use proper matching conditions . here are the major simplifications imposed in our analysis . it is assumed that the energy in field @xmath73 , which is annihilated during inflation , is sub - dominant compared to energy stored in field @xmath54 which carries the second stage of inflation . it is assumed that not only the background energy density in @xmath180 is transferred into radiation @xmath82 but also the perturbations carried by @xmath83 are also transferred into perturbations in radiation , @xmath84 . with these simplifications we can follow the perturbations in @xmath54 before and after the collision using the simple matching conditions ( [ match1 ] ) . although we did not take into account the perturbations @xmath83 and @xmath84 in our analysis but the effects of branes annihilation and particles creation are present at the level of background . in our analysis this translated into @xmath159 to be non - zero . interestingly , this is very similar to model studied in @xcite where the feature is due to a sudden change in slow - roll parameters . in @xcite the sudden change in slow - roll parameters was motivated from a phase transition @xcite during hybrid inflation model @xcite . it is very interesting that our model , motivated from a completely different setup , shares the same results . we have calculated the curvature perturbations power spectrum . the effects of branes annihilation are encoded in the transfer function @xmath138 . for super - horizon scales , modes which are outside the sound horizon at the time of collision , the transfer function scales monotonically . consequently , the spectral index is more blue compared to the background spectral index . on the other hand , for small scales , modes which are inside the sound horizon at the time of collision , the transfer function converges to unity with decaying superimposed oscillatory modulations . the non - trivial transitions in transfer function happen for modes which leave the sound horizon around the time of collision . the superimposed oscillatory modulations can have interesting observational consequences on cmb analysis @xcite . furthermore , there are strong constraints from non - gaussianity bounds on standard dbi inflation @xcite . it would be interesting to calculate the level of non - gaussianities produced in our setup to put further constraints on the model parameters . our results here should be compared with the results obtained in @xcite where similar problem was studied but in the context of slow - roll brane inflation . in @xcite , like here , it was assumed that the perturbations @xmath83 are transferred into @xmath84 . however , in @xcite the matching conditions ( [ match0 ] ) have been used instead of the simple matching conditions ( [ match1 ] ) . as a result , the transfer function for large and intermediate scales in @xcite share the same behavior as in our case here . however , in @xcite ( see also @xcite ) , the transfer function for small scales has a constant oscillatory modulation whereas in our case the amplitude of oscillations decays for these modes . for physical reasons , one expects that small scales should be blind to the process of branes annihilation and particles creation so the transfer function is expected to approach unity for these modes . to remedy this problem several methods were put forward in @xcite . our approach to use the matching conditions ( [ match1 ] ) may be interpreted as a pragmatic approach towards the methods speculated in @xcite . to simplify our analysis we considered the model where the branes are moving ultra relativistically . in practice one expects that a successful brane inflation model consists both stages of slow - roll and fast - roll brane inflation . it is possible that during early stage of inflation the branes are moving slowly so one may get a few dozens of e - foldings during the slow - roll regime @xcite . as the branes are moving towards the bottom of the throat they reach the speed limit and one should use the dbi inflation methods as we used in this work . the process of branes annihilation and particles creation can happen either during slow - roll regime , as in @xcite , during the ultra relativistic fast - roll limit , as in our case here , or in between these two stages . in the third case , one can not perform analytical studies and a numerical investigation is necessary . it would be interesting to study the effects of branes annihilation in a generic model of brane inflation containing both stages of slow - roll inflation and dbi inflation . we would like to thank f. arroja , d. battefeld , t. battefeld , n. khosravi , l. mcallister , m.h . namjoo , m. sasaki and g. shiu for useful discussions and correspondences . would like to thank yukawa institute for theoretical physics ( yitp ) for the hospitalities during the activities gravity and cosmology 2010 " and ykis2010 symposium : cosmology the next generation " when this work was in progress . a. h. guth , `` the inflationary universe : a possible solution to the horizon and flatness problems , '' phys . d * 23 * , 347 ( 1981 ) ; + a. d. linde , `` a new inflationary universe scenario : a possible solution of the horizon , flatness , homogeneity , isotropy and primordial monopole problems , '' phys . b * 108 * , 389 ( 1982 ) ; + a. albrecht and p. j. steinhardt , `` cosmology for grand unified theories with radiatively induced symmetry breaking , '' phys . * 48 * , 1220 ( 1982 ) . e. komatsu _ et al . _ , `` seven - year wilkinson microwave anisotropy probe ( wmap ) observations : cosmological interpretation , '' arxiv:1001.4538 [ astro-ph.co ] . s. h. henry tye , `` brane inflation : string theory viewed from the cosmos , '' lect . notes phys . * 737 * , 949 ( 2008 ) [ arxiv : hep - th/0610221 ] . j. m. cline , `` string cosmology , '' arxiv : hep - th/0612129 . c. p. burgess , `` lectures on cosmic inflation and its potential stringy realizations , '' pos * p2gc * , 008 ( 2006 ) [ class . quant . * 24 * , s795 ( 2007 ) ] [ arxiv:0708.2865 [ hep - th ] ] . l. mcallister and e. silverstein , `` string cosmology : a review , '' gen . grav . * 40 * , 565 ( 2008 ) [ arxiv:0710.2951 [ hep - th ] ] . d. baumann and l. mcallister , `` advances in inflation in string theory , '' arxiv:0901.0265 [ hep - th ] . a. mazumdar and j. rocher , `` particle physics models of inflation and curvaton scenarios , '' arxiv:1001.0993 [ hep - ph ] . g. dvali and s .- h.h . tye , `` brane inflation '' , phys . lett . * b450 * ( 1999 ) 72 , hep - ph/9812483 . s. h. s. alexander , `` inflation from d - anti - d brane annihilation , '' phys . d * 65 * , 023507 ( 2002 ) [ arxiv : hep - th/0105032 ] . c. p. burgess , m. majumdar , d. nolte , f. quevedo , g. rajesh and r. j. zhang , jhep * 07 * ( 2001 ) 047 , hep - th/0105204 . s. b. giddings , s. kachru and j. polchinski , `` hierarchies from fluxes in string compactifications , '' phys . d * 66 * , 106006 ( 2002 ) , hep - th/0105097 . k. dasgupta , g. rajesh and s. sethi , `` m theory , orientifolds and g - flux , '' jhep * 9908 * , 023 ( 1999 ) [ arxiv : hep - th/9908088 ] . d. baumann , a. dymarsky , i. r. klebanov , j. m. maldacena , l. p. mcallister and a. murugan , `` on d3-brane potentials in compactifications with fluxes and wrapped d - branes , '' jhep * 0611 * , 031 ( 2006 ) [ arxiv : hep - th/0607050 ] . c. p. burgess , j. m. cline , k. dasgupta and h. firouzjahi , `` uplifting and inflation with d3 branes , '' jhep * 0703 * , 027 ( 2007 ) [ arxiv : hep - th/0610320 ] . d. baumann , a. dymarsky , i. r. klebanov and l. mcallister , `` towards an explicit model of d - brane inflation , '' jcap * 0801 * , 024 ( 2008 ) [ arxiv:0706.0360 [ hep - th ] ] . f. chen and h. firouzjahi , `` dynamics of d3-d7 brane inflation in throats , '' jhep * 0811 * , 017 ( 2008 ) [ arxiv:0807.2817 [ hep - th ] ] . j. m. cline , l. hoi and b. underwood , `` dynamical fine tuning in brane inflation , '' jhep * 0906 * , 078 ( 2009 ) [ arxiv:0902.0339 [ hep - th ] ] . d. battefeld , t. battefeld , h. firouzjahi and n. khosravi , `` brane annihilations during inflation , '' jcap * 1007 * , 009 ( 2010 ) [ arxiv:1004.1417 [ hep - th ] ] . d. battefeld , t. battefeld and a. c. davis , `` staggered multi - field inflation , '' jcap * 0810 * , 032 ( 2008 ) [ arxiv:0806.1953 [ hep - th ] ] . t. battefeld , `` exposition to staggered multi - field inflation , '' nucl . * 192 - 193 * , 128 ( 2009 ) [ arxiv:0809.3242 [ astro - ph ] ] . n. barnaby , z. huang , l. kofman and d. pogosyan , `` cosmological fluctuations from infra - red cascading during inflation , '' phys . d * 80 * , 043501 ( 2009 ) [ arxiv:0902.0615 [ hep - th ] ] . n. barnaby and z. huang , `` particle production during inflation : observational constraints and signatures , '' phys . d * 80 * , 126018 ( 2009 ) [ arxiv:0909.0751 [ astro-ph.co ] ] . n. barnaby , `` on features and nongaussianity from inflationary particle production , '' arxiv:1006.4615 [ astro-ph.co ] ; n. barnaby , `` nongaussianity from particle production during inflation , '' adv . astron . * 2010 * , 156180 ( 2010 ) [ arxiv:1010.5507 [ astro-ph.co ] ] . d. green , b. horn , l. senatore and e. silverstein , `` trapped inflation , '' phys . rev . d * 80 * , 063533 ( 2009 ) [ arxiv:0902.1006 [ hep - th ] ] . m. alishahiha , e. silverstein and d. tong , `` dbi in the sky , '' phys . d * 70 * , 123505 ( 2004 ) [ arxiv : hep - th/0404084 ] . c. p. herzog , i. r. klebanov and p. ouyang , `` remarks on the warped deformed conifold , '' arxiv : hep - th/0108101 . x. chen , `` multi - throat brane inflation , '' phys . d * 71 * , 063506 ( 2005 ) [ arxiv : hep - th/0408084 ] . s. e. shandera and s. h. tye , `` observing brane inflation , '' jcap * 0605 * , 007 ( 2006 ) [ arxiv : hep - th/0601099 ] . r. bean , x. chen , h. peiris and j. xu , `` comparing infrared dirac - born - infeld brane inflation to observations , '' phys . rev . d * 77 * , 023527 ( 2008 ) [ arxiv:0710.1812 [ hep - th ] ] . r. bean , s. e. shandera , s. h. henry tye and j. xu , `` comparing brane inflation to wmap , '' jcap * 0705 * , 004 ( 2007 ) [ arxiv : hep - th/0702107 ] . x. chen , m. x. huang , s. kachru and g. shiu , `` observational signatures and non - gaussianities of general single field inflation , '' jcap * 0701 * , 002 ( 2007 ) [ arxiv : hep - th/0605045 ] . s. thomas and j. ward , `` ir inflation from multiple branes , '' phys . d * 76 * , 023509 ( 2007 ) [ arxiv : hep - th/0702229 ] . i. huston , j. e. lidsey , s. thomas and j. ward , `` gravitational wave constraints on multi - brane inflation , '' jcap * 0805 * , 016 ( 2008 ) [ arxiv:0802.0398 [ hep - th ] ] . y. f. cai and w. xue , `` n - flation from multiple dbi type actions , '' phys . b * 680 * , 395 ( 2009 ) [ arxiv:0809.4134 [ hep - th ] ] ; y. f. cai and h. y. xia , `` inflation with multiple sound speeds : a model of multiple dbi type actions and non - gaussianities , '' phys . b * 677 * , 226 ( 2009 ) [ arxiv:0904.0062 [ hep - th ] ] ; d. langlois , s. renaux - petel , d. a. steer and t. tanaka , `` primordial fluctuations and non - gaussianities in multi - field dbi inflation , '' phys . rev . * 101 * , 061301 ( 2008 ) [ arxiv:0804.3139 [ hep - th ] ] ; d. langlois , s. renaux - petel , d. a. steer and t. tanaka , `` primordial perturbations and non - gaussianities in dbi and general multi - field inflation , '' phys . d * 78 * , 063523 ( 2008 ) [ arxiv:0806.0336 [ hep - th ] ] . s. renaux - petel , `` combined local and equilateral non - gaussianities from multifield dbi inflation , '' jcap * 0910 * , 012 ( 2009 ) [ arxiv:0907.2476 [ hep - th ] ] . m. x. huang , g. shiu and b. underwood , `` multifield dbi inflation and non - gaussianities , '' phys . d * 77 * , 023511 ( 2008 ) [ arxiv:0709.3299 [ hep - th ] ] . s. mizuno , f. arroja and k. koyama , `` on the full trispectrum in multi - field dbi inflation , '' phys . d * 80 * , 083517 ( 2009 ) [ arxiv:0907.2439 [ hep - th ] ] ; s. mizuno , f. arroja , k. koyama and t. tanaka , `` lorentz boost and non - gaussianity in multi - field dbi - inflation , '' phys . d * 80 * , 023530 ( 2009 ) [ arxiv:0905.4557 [ hep - th ] ] . n. t. jones , l. leblond and s. h. h. tye , `` adding a brane to the brane anti - brane action in bsft , '' jhep * 0310 * , 002 ( 2003 ) [ arxiv : hep - th/0307086 ] . n. t. jones and s. h. h. tye , `` an improved brane anti - brane action from boundary superstring field theory and multi - vortex solutions , '' jhep * 0301 * , 012 ( 2003 ) [ arxiv : hep - th/0211180 ] . n. barnaby , c. p. burgess , j. m. cline , `` warped reheating in brane - antibrane inflation , '' jcap * 0504 * , 007 ( 2005 ) . [ hep - th/0412040 ] . j. garriga and v. f. mukhanov , `` perturbations in k - inflation , '' phys . b * 458 * , 219 ( 1999 ) [ arxiv : hep - th/9904176 ] . n. deruelle and v. f. mukhanov , `` on matching conditions for cosmological perturbations , '' phys . d * 52 * , 5549 ( 1995 ) [ arxiv : gr - qc/9503050 ] . j. martin and d. j. schwarz , `` the influence of cosmological transitions on the evolution of density perturbations , '' phys . rev . d * 57 * , 3302 ( 1998 ) [ arxiv : gr - qc/9704049 ] . i. zaballa and m. sasaki , `` boosted perturbations at the end of inflation , '' arxiv:0911.2069 [ astro-ph.co ] . d. h. lyth , k. a. malik , m. sasaki and i. zaballa , `` forming sub - horizon black holes at the end of inflation , '' jcap * 0601 * , 011 ( 2006 ) [ arxiv : astro - ph/0510647 ] . i. zaballa , a. m. green , k. a. malik and m. sasaki , `` constraints on the primordial curvature perturbation from primordial black holes , '' jcap * 0703 * , 010 ( 2007 ) [ arxiv : astro - ph/0612379 ] . m. joy , v. sahni and a. a. starobinsky , `` a new universal local feature in the inflationary perturbation spectrum , '' phys . d * 77 * , 023514 ( 2008 ) [ arxiv:0711.1585 [ astro - ph ] ] . d. baumann and l. mcallister , `` a microscopic limit on gravitational waves from d - brane inflation , '' phys . d * 75 * , 123508 ( 2007 ) [ arxiv : hep - th/0610285 ] . a. e. romano and m. sasaki , `` effects of particle production during inflation , '' phys . rev . d * 78 * , 103522 ( 2008 ) [ arxiv:0809.5142 [ gr - qc ] ] . m. zarei , `` short distance physics and initial state effects on the cmb power spectrum and cosmological constant , '' phys . d * 78 * , 123502 ( 2008 ) [ arxiv:0809.4312 [ hep - th ] ] . m. joy , a. shafieloo , v. sahni and a. a. starobinsky , `` is a step in the primordial spectral index favored by cmb data ? * 0906 * , 028 ( 2009 ) [ arxiv:0807.3334 [ astro - ph ] ] . a. a. abolhasani , h. firouzjahi and m. h. namjoo , `` curvature perturbations and non - gaussianities from waterfall phase transition during inflation , '' arxiv:1010.6292 [ astro-ph.co ] . a. d. linde , `` hybrid inflation , '' phys . d * 49 * , 748 ( 1994 ) [ arxiv : astro - ph/9307002 ] . e. j. copeland , a. r. liddle , d. h. lyth , e. d. stewart and d. wands , `` false vacuum inflation with einstein gravity , '' phys . d * 49 * , 6410 ( 1994 ) [ arxiv : astro - ph/9401011 ] . j. garcia - bellido , a. d. linde and d. wands , `` density perturbations and black hole formation in hybrid inflation , '' phys . d * 54 * , 6040 ( 1996 ) [ arxiv : astro - ph/9605094 ] . r. flauger , l. mcallister , e. pajer , a. westphal and g. xu , `` oscillations in the cmb from axion monodromy inflation , '' arxiv:0907.2916 [ hep - th ] . t. biswas , a. mazumdar and a. shafieloo , `` wiggles in the cosmic microwave background radiation : echoes from non - singular cyclic - inflation , '' arxiv:1003.3206 [ hep - th ] . r. flauger and e. pajer , `` resonant non - gaussianity , '' arxiv:1002.0833 [ hep - th ] .

optical parametric oscillators are one of the most interesting and well characterized devices in nonlinear quantum optics . novel discoveries made with them include demonstrations of large amounts of squeezing@xcite , significant quantum intensity correlations@xcite together with a quadrature correlation measurement that provided the first experimental demonstration @xcite of the original epr paradox . practical applications include their use as highly efficient and tunable frequency converters . in the present paper , we focus on the optimum below - threshold squeezing results , which determine the limits to squeezing obtained near the critical point , where nonlinear corrections start to dominate . in a companion paper@xcite , the related question of critical fluctuations at threshold is treated . the theory of quantum squeezing in the linear parametric oscillator is now well - developed@xcite-@xcite . excellent agreement between theory and experiment is obtained@xcite-@xcite , in the region below threshold . however , the usual theory is linearized , and therefore can not be used in the near - threshold region where the squeezing is largest . the drawback with linearized theories is that they predict that zero quantum noise levels are achievable at threshold . this is clearly unrealistic , since ( by the heisenberg uncertainty principle ) it necessarily requires an infinite energy in the conjugate mode . more significantly , this would imply an infinite amount of phase information - which is also impossible , since the coherent pump which drives the parametric oscillator can only supply a finite quantity of phase information . while present experiments are limited by technical noise from approaching the critical point too closely , it is reasonable to expect that progress in integrated optics will lead to more stable , highly miniaturized devices which could well operate at the quantum limit , even near threshold . accordingly , there have been a number of investigations as to the ultimate limits to the squeezing spectrum of a parametric amplifier / oscillator . this has often involved using keldysh diagrams or wyld - keldysh techniques @xcite-@xcite to extend the linear theory @xcite-@xcite , using a many - body theory analog of feynman diagrams . the two - mode hilbert space involved in these problems typically has a minimum dimension @xmath0 , even with only @xmath1 photons , and therefore would be difficult to solve using other methods that involve number - state expansions - either using a direct solution of the master equation , or stochastic wave - function methods . the hamiltonian matrix would have @xmath2 coefficients , unless simplified , with a density matrix of similar size . more typical experimental photon numbers have at least @xmath3 , with a corresponding density matrix dimension of @xmath4 - which appears completely inaccessible with number state techniques . another drawback of number - state techniques is that they usually do not permit analytic approximations , which can give more physical insight . we therefore treat these questions using the coherent - state positive p - representation @xcite , combined with an expansion technique valid below the critical point . results are also verified by the use of direct numerical stochastic equation simulations . we find an @xmath5 scaling law for the optimal squeezing predicted by plimak and walls@xcite is obtained here as well , but with a different spectrum , owing to the use of more systematic expansion techniques that result from using the positive p - representation method . our analytic results for optimal squeezing , which occurs below the critical threshold , give excellent agreement with accurate numerical simulations for the same parameter values . however , even larger noise reductions are predicted to occur simply by reducing the losses of the second - harmonic , in which case the @xmath5 scaling law no longer holds . in a companion paper we consider the related problem of the critical region , where the narrow - band squeezing is less than optimal due to the effects of critical fluctuations . we also compare the above results with a semiclassical approach , that is , a truncated wigner phase - space equation . this equation corresponds to a classical theory with added vacuum fluctuations . a comparison between the positive p - representation ( fully quantum mechanical ) and semiclassical theories permits us to see how far one can go and what is the limitation of this extended classical point of view . we find that the nonlinear corrections in the semiclassical theory are in strong disagreement with the full quantum theory far below threshold , but agree near threshold . this tells us that the semiclassical theory works surprisingly well in the threshold region , indicating that the large quantum fluctuations near threshold have a rather classical character . the model considered here is the degenerate parametric oscillator . the system of interest is an idealized interferometer , which is resonant at two frequencies , @xmath6 and @xmath7 . it is externally driven at the larger of the two frequencies . both frequencies are damped due to cavity losses . down conversion of the pump photons to resonant sub - harmonic mode photons occurs due to a @xmath8 nonlinearity present inside the cavity . the heisenberg picture hamiltonian that describes this open system@xcite is @xmath9 where the intra - cavity or system hamiltonian is given by : @xmath10 here @xmath11 represents the external driving field at frequency @xmath12 . the term @xmath13 describes the free evolution of the extra - cavity modes that are the loss - reservoirs of the cavity . the term @xmath14 is the coupling parameter due to a @xmath8 nonlinear medium internal to the cavity , and @xmath15 are reservoir operators that create and destroy photons in the loss - reservoir coupled to the internal mode of frequency @xmath16 . next , we wish to consider an interaction picture , obtained with the definition that @xmath17 in other words , the operators will evolve according to the relevant mode frequency , while the states evolve according to the rest of the system hamiltonian . the interaction hamiltonian used here then reduces to the the standard one@xcite for a non - degenerate , single - mode parametric amplifier or oscillator : @xmath18 + \frac{i\chi } { 2}\left [ \widehat{a}_{2}\widehat{a}_{1}^{\dagger2 } -\widehat{a}_{2}^{\dagger } \widehat{a}_{1}^{2}\right ] \ , \ , .\ ] ] here @xmath19 are now time - independent operators representing the fundamental and second - harmonic modes respectively . for simplicity , we have chosen the field mode - functions so that @xmath11 , @xmath14 are real . using standard techniques @xcite to eliminate the heat bath , we obtain the following master equation for the reduced density operator of the system in the interaction picture : @xmath20 + \gamma _ { 1}\left ( 2\widehat{a}_{1}\widehat{\rho } \widehat{a}_{1}^{\dagger } -\widehat{a}_{1}^{\dagger } \widehat{a}_{1}\widehat{\rho } -\widehat{\rho } \widehat{a}_{1}^{\dagger } \widehat{a}_{1}\right ) \nonumber \\ & + & \gamma _ { 2}\left ( 2\widehat{a}_{2}\widehat{\rho } \widehat{a}_{2}^{\dagger } -\widehat{a}_{2}^{\dagger } \widehat{a}_{2}\widehat{\rho } -\widehat{\rho } \widehat{a}_{2}^{\dagger } \widehat{a}_{2}\right ) \ , \ , , \label{a5 } \end{aligned}\ ] ] where @xmath21 are the internal mode amplitude damping rates and we assume that @xmath22 , where @xmath23 are the mean numbers of thermal photons in the input reservoir modes . using reservoir theory , it is possible to identify the coherent driving field with a corresponding input photon flux from an external coherent laser , with @xmath24 in @xmath25 , where @xmath26 is the input coupler decay rate . for optimum performance , we will assume that @xmath27 , and similarly for the fundamental mode - which will be assumed to only decay through its output coupling mirror . if these conditions are not satisfied , then the coupling efficiency and maximum squeezing are reduced . at this point , we note that in the classical limit , the system has the well - known classical equations of intra - cavity parametric oscillation , where we define : @xmath28 , and hence obtain , in the interaction picture : @xmath29 \ , , \nonumber \\ \frac{d\alpha _ { 2}}{dt } & = & \left [ -\gamma _ { 2}\alpha _ { 2}+{\mathcal{e}}-\frac{1}{2}\chi \alpha _ { 1}^{2}\right ] \end{aligned}\ ] ] these equations are valid in the limit of large photon number . they are obtained by the use of a classical decorrelation in which all operator products are assumed to factorize , so that @xmath30 , and @xmath31 . the solution of these equations is immediate classically , and has the property that there is a phase transition at the critical driving field of : @xmath32 . for driving fields below this value , one has : @xmath33 for fields above this value , the signal field @xmath34 is bistable , with : @xmath35 the intra - cavity photon number at the critical point is @xmath36 . classically , there are only second - harmonic photons present at this driving field , and the input photon flux is @xmath37 . however , a squeezed field - with finite intensity - is actually emitted as well . this is not taken into account in the classical theory . in order to treat the full quantum evolution , we now turn to the methods of operator representation theory . these techniques can be used to transform the density matrix equations of motion to c - number fokker - planck or stochastic equations . in the positive p - representation , the density matrix is expanded in terms of multi - mode coherent state vectors @xmath38 : @xmath39 following standard procedures , the assumption of vanishing boundary terms allows the master equation to be re - written as a fokker - planck equation in @xmath40 and hence as a stochastic equation@xcite with real noise . the assumption of vanishing boundary terms is critical to this procedure , and we note here that this is generally valid when the ratio of nonlinearity to damping is small@xcite , ( i.e. @xmath41 ) . the stochastic procedure is best regarded as being generally an asymptotic procedure , valid for small @xmath42 - in which case the boundary terms are exponentially suppressed . we check this assumption numerically here as well , and point out that the required ratio of nonlinearity to damping is extremely well - satisfied in current experiments , where the ratio is typically 10@xmath43 or less . further analysis of this problem has been given elsewhere@xcite . given this assumption , the following stochastic equations are obtained from ( [ a5 ] ) and ( [ a6 ] ) , for any driving field @xmath11 , that is , either below or above threshold : @xmath44 dt+\sqrt{\chi \alpha _ { 2}}dw_{1}(t)\ , \ , , \nonumber \\ d\alpha _ { 1}^{+ } & = & \left [ -\gamma _ { 1}\alpha ^{+}_{1}+\chi \alpha _ { 1}\alpha _ { 2}^{+}\right ] dt+\sqrt{\chi \alpha _ { 2}^{+}}dw_{2}(t)\ , \ , , \nonumber \\ d\alpha _ { 2 } & = & \left [ -\gamma _ { 2}\alpha _ { 2}+{\mathcal{e}}-\frac{1}{2}\chi \alpha _ { 1}^{2}\right ] dt\ , \ , , \nonumber \\ d\alpha _ { 2}^{+ } & = & \left [ -\gamma _ { 2}\alpha _ { 2}^{+}+{\mathcal{e}}-\frac{1}{2}\chi \alpha _ { 1}^{+2}\right ] dt\ , \ , .\end{aligned}\ ] ] the stochastic correlations are given by : @xmath45 this means that @xmath46 represent two real gaussian and uncorrelated stochastic processes , and the amplitude of the stochastic fluctuations that act on the signal mode are dependent on the pump field dynamics . our derivation is formally based on the it stochastic calculus . however , in this case , either it or stratonovic stochastic calculus gives identical results@xcite . we can also write a c - number phase space equation using an approximate form of the wigner representation@xcite , which is equivalent to stochastic electrodynamics . the characteristic function of the wigner representation is written as @xmath47 and the wigner distribution can be written as fourier transform of the characteristic function @xmath48 in the wigner representation , the phase space equation that corresponds to the master equation ( [ a5 ] ) is @xmath49 if we truncate the third derivative of the phase space equation we get a genuine fokker - planck type equation with positive definite diffusion constant . this can be mapped into the following it stochastic differential coupled equations ( for simplicity we let @xmath50 , as before ) . @xmath51 dt+\sqrt{\gamma _ { 1}}\ ; dw_{1}(t)\ , \ , , \nonumber \\ d\alpha _ { 1}^ { * } & = & \left [ -\gamma _ { 1}\alpha _ { 1}^{*}+\chi { \alpha } _ { 1}\alpha _ { 2}^{*}\right ] dt+\sqrt{\gamma _ { 1}}\ ; dw_{1}^{*}(t)\ , \ , , \nonumber \\ d{\alpha } _ { 2 } & = & \left [ -\gamma _ { 2}{\alpha } _ { 2}-\frac{\chi } { 2}{\alpha } _ { 1}^{2}+{\mathcal{e}}\right ] dt+\sqrt{\gamma _ { 2}}\ ; dw_{2}(t)\ , \ , , \nonumber \\ d\alpha _ { 2}^ { * } & = & \left [ -\gamma _ { 2}\alpha _ { 2}^{*}-\frac{\chi } { 2}\alpha _ { 1}^{*2}+{\mathcal{e}}\right ] dt+\sqrt{\gamma _ { 2}}\ ; dw_{2}^{*}(t)\ , \ , .\end{aligned}\ ] ] here @xmath46 is now a _ complex _ gaussian white noise whose mean and variance are given by @xmath52 the above equation is identical to the equation derived in positive p - representation when one discards the noise terms . this corresponds to the nonlinear classical equation for the opo system . the main difference between the two sets of equations is the noise terms . in the semiclassical theory the noise is universal for all modes and comes from to the vacuum fluctuations , while in the positive - p equation the pump has a noiseless amplitude and the signal noise comes from the nonlinear coupling . however , we note that the wigner equation after truncation is no longer completely equivalent to quantum mechanics , since it always leads to a positive wigner function - thus , not all quantum states can be represented . the positive - p stochastic method directly reproduces the normally ordered correlations and moments , while the wigner representation reproduces the symmetrically ordered moments . we also have to distinguish the internal and external operator moments , since measurements are normally performed on output fields that are external to the cavity . the technique for treating external field spectra was introduced by yurke@xcite , and by collett and gardiner@xcite . these external field measurements are obtained from the input - output relations of : @xmath53 where @xmath54 and @xmath55 are the input and output photon fields respectively , evaluated at the output - coupling mirror . the most efficient transport of squeezing is obtained if we assume that all the signal losses occur through the output coupler , so that @xmath56 . we will assume this to be the case . the crucial quadrature variables of the system have the definitions : @xmath57 there are also corresponding external quadrature field variables , defined as : @xmath58 similarly , we can define c - number stochastic quadrature variables within the relevant representations , thus giving : @xmath59 of especial interest is @xmath60 since this is the low - noise , squeezed quadrature . here we note that the instantaneous correlation functions of the intra - cavity field operators are called the moments . typically , they are not easily measurable , when compared to output moments or spectra , but they are useful in that they provide a check on the accuracy of the calculation of measurable spectra . the squeezing in terms of the intra - cavity quadrature variances corresponds to an instantaneous measurement of the field moments . if such a measurements were possible , it would include contributions from all frequencies . for measurements averaged over a long time @xmath61 , it is the low frequency part of the spectrum that is the relevant quantity , and we shall focus on this , as it usually determines the maximum squeezing possible . the output measured spectral variance @xmath62 of a general quadrature @xmath63 can be written @xmath64 where the fluctuations @xmath65 are defined as @xmath66 and the frequency argument denotes a fourier transform : @xmath67 since the p - representation is normally ordered , it automatically provides the normally - ordered moments : @xmath68 also , the + p spectral correlations correspond to the normally ordered , time - ordered operator correlations of the measured fields . we therefore define fourier components of the normalized quadratures as : @xmath69 this leads to the following well - known result for the general squeezing spectrum , as measured in an external homodyne detection scheme : @xmath70 note that vacuum ( input ) field terms do not contribute directly to this spectrum , as they have a vanishing normally - ordered spectrum , and are not correlated with the coherent amplitudes in the + p representation . in the case of the wigner representation , the correlations and moments are given with symmetric ordering . thus , for example , @xmath71 _ { + } /2\right\rangle = 1/2 $ ] in the vacuum state . the normally - ordered internal field moments are easily calculated , by using equal - time commutators to change the ordering from symmetric to normally - ordered : @xmath72 similarly , the normally - ordered squeezing spectrum , as measured in an external homodyne detection scheme is : @xmath73 it is essential here to include the vacuum field contributions from reflected input fields , as these are correlated with the internal wigner amplitudes , and hence have a significant contribution to the spectrum . in fact , these input fields can be shown to correspond directly to the noise terms in the relevant wigner equations , leading to the identification : @xmath74 where @xmath75 is a c - number amplitude corresponding ( in the wigner representation ) to the quantum vacuum input field . the fundamental property of the wigner function is that the ensemble average of any polynomial of the random variable @xmath76 and @xmath77 weighted by the wigner density exactly corresponds to the hilbert - space expectation of the corresponding symmetrized product of the annihilation and creation operators . therefore , the truncated theory with a positive wigner function can be viewed as equivalent to a hidden variable theory , since one can obtain quadrature fluctuation predictions by following an essentially classical prescription ; in which even the noise terms have a classical interpretation as corresponding a form of zero - point fluctuation . this can not be equivalent to quantum mechanics in general , but may provide similar results to quantum mechanics under some circumstances . next we wish to rescale the equations . this has the merit of showing explicitly how a small noise expansion can permit us to use a type of perturbation theory whose zero - th order solution is the classical solution , rather than the feynman approach where the zero - th order solution is the free - particle case . in order to show this systematically , a formal perturbation expansion in powers of @xmath78 is now introduced , where the scaling parameter @xmath78 is given by : @xmath79 where @xmath80 is the threshold pump photon number , and a dimensionless decay ratio , @xmath81 is introduced . an equivalent definition is : @xmath82 this clearly determines the ratio of nonlinear to linear rates of change . next , we introduce a scaled time @xmath83 and a dimensionless driving field @xmath84 , so that the equations can be expressed in terms of the three dimensionless parameters @xmath85 . finally , we expand the scaled coordinates in a power series in @xmath78 , to give : @xmath86 the expansion given here has the property that the zero - th order term corresponds to the large classical fields of order @xmath87 , while the first order term corresponds to the quantum fluctuations of order @xmath88 , and the higher order terms correspond to nonlinear corrections to the quantum fluctuations , of order @xmath78 and greater . for a given fundamental decay rate @xmath89 , the expansion coefficient @xmath90 is inversely proportional to the input photon flux required to obtain the threshold condition . thus , the smaller @xmath90 is , the larger the required input field . here we will first be interested in the analysis of the steady state moments . subsequently we will calculate the spectral correlations of the solutions using fourier transforms of the calculation done in the time domain . the equations for the quadrature variables in the positive p - representation are @xmath91 dt+\sqrt{\frac{\chi } { 2}}\ ; \left [ \sqrt{x_{2}+iy_{2}}dw_{1}(t)+\sqrt{x_{2}^{\dagger } -iy_{2}^{\dagger } } dw_{2}(t)\right ] \ ; \ ; , \nonumber \\ dy_{1 } & = & \left [ -\gamma _ { 1}y_{1}+\frac{\chi } { 2}\left ( x_{1}y_{2}-x_{2}y_{1}\right ) \right ] dt - i\sqrt{\frac{\chi } { 2}}\ ; \left [ \sqrt{x_{2}+iy_{2}}dw_{1}(t)-\sqrt{x_{2}^{\dagger } -iy_{2}^{\dagger } } dw_{2}(t)\right ] \ ; \ ; , \nonumber \\ dx_{2 } & = & \left [ -\gamma _ { 2}x_{2}-\frac{\chi } { 4}\left ( x_{1}^{2}-y_{1}^{2}\right ) + 2{\mathcal{e}}\right ] dt\nonumber \\ dy_{2 } & = & \left [ -\gamma _ { 2}y_{2}-\frac{\chi } { 2}x_{1}y_{1}\right ] dt\ ; \ ; .\end{aligned}\ ] ] the stochastic equations are now solved by the technique of matching powers of @xmath78 in the corresponding time - evolution equations . this technique can be analyzed diagrammatically , and so can be termed the ` stochastic diagram ' method@xcite . the zero - th order solution is : @xmath92 d\tau \ , \ , , \nonumber \\ dy_{1}^{\left ( 0\right ) } & = & \left [ -y_{1}^{\left ( 0\right ) } + \frac{1}{2}\left ( x_{1}^{\left ( 0\right ) } x_{2}^{\left ( 0\right ) } -x_{2}^{\left ( 0\right ) } y_{1}^{\left ( 0\right ) } \right ) \right ] d\tau \ , \ , , \nonumber \\ dx_{2}^{\left ( 0\right ) } & = & -\gamma _ { r}\left [ x_{2}^{\left ( 0\right ) } + \frac{1}{2}\left ( x_{1}^{\left ( 0\right ) } x_{1}^{\left ( 0\right ) } -y_{1}^{\left ( 0\right ) } y_{1}^{\left ( 0\right ) } \right ) -2\mu \right ] d\tau \ , \ , , \nonumber \\ dy_{2}^{\left ( 0\right ) } & = & -\gamma _ { r}\left [ y_{2}^{\left ( 0\right ) } + x_{1}^{\left ( 0\right ) } y_{1}^{\left ( 0\right ) } \right ] d\tau \ , \ , .\end{aligned}\ ] ] these equations are the classical nonlinear equations for the cavity , expressed in terms of the quadrature amplitudes of dimensionless scaled fields . the steady - state solution below threshold is well - known , and is given by : @xmath93 with no loss of generality , we can set all odd orders of @xmath94 , and all even orders of @xmath95 to zero , since one can set these to zero initially , and these orders do not change in time . to first order , the equations are given by : @xmath96 where , @xmath97 . these equations are the ones that are normally used to predict squeezing . they are non - classical , but correspond to a very simple form of linear , non - classical fluctuation which has a gaussian quasi - probability distribution . in other words , if no higher - order terms existed , the result would be an ideal squeezed state in the sub - harmonic , together with an ideal coherent state in the pump . of more interest to the present paper , is the behaviour to the next order . this is the first order where nonlinear corrections to ideal squeezed - state behaviour will occur . we find the following : @xmath98 d\tau \ , \ , , \nonumber \\ dy_{2}^{\left ( 2\right ) } & = & -\gamma _ { r}\left [ y_{2}^{\left ( 2\right ) } + x_{1}^{\left ( 1\right ) } y_{1}^{\left ( 1\right ) } \right ] d\tau \ , \ , .\label{8 } \end{aligned}\ ] ] while we do not wish to include any effects beyond the first nonlinear corrections , it is not possible to consistently neglect the third - order in perturbation theory . this is because the first non - trivial correlations arise in terms like @xmath99 ^{2}\right\rangle , $ ] which have the same formal order as terms of the type @xmath100 therefore , to obtain a consistent expansion for the correlations that are of interest , we must compute the third - order terms as well . these satisfy the following equations : @xmath101 d\tau + \nonumber \\ & + & \frac{1}{2\sqrt{2\mu } } \left [ x_{2}^{\left ( 2\right ) } dw_{x}(\tau ) + iy_{2}^{\left ( 2\right ) } dw_{y}(\tau ) \right ] \ , \ , , \nonumber \\ dy_{1}^{\left ( 3\right ) } & = & \left [ -\left ( 1+\mu \right ) y_{1}^{\left ( 3\right ) } + \frac{1}{2}\left ( x_{1}^{\left ( 1\right ) } y_{2}^{\left ( 2\right ) } -x_{2}^{\left ( 2\right ) } y_{1}^{\left ( 1\right ) } \right ) \right ] d\tau \nonumber \\ & + & \frac{1}{2\sqrt{2\mu } } \left [ y_{2}^{\left ( 2\right ) } dw_{x}(\tau ) -ix_{2}^{\left ( 2\right ) } dw_{y}(\tau ) \right ] \ , \ , .\label{10 } \end{aligned}\ ] ] the equations of this order have a non - trivial noise term , which depends on the second order pump quadrature solution . we now wish to calculate the operator moments . to proceed further , we use it calculus to derive stochastic equations for quantities of interest , which in the present calculation are @xmath102 and @xmath103 . these equations contain quantities involving variables lower down in the hierarchy , as well as terms generated from the noise correlations . finally , we compute the the steady state averages of the quantities of interest , so that the noise terms vanish . in the present case , this yields , @xmath104 \ , \ , , \nonumber \\ \left\langle x_{1}^{\left ( 1\right ) } y_{1}^{\left ( 1\right ) } y_{2}^{\left ( 2\right ) } \right\rangle & = & \frac{\gamma _ { r}}{\left ( \gamma _ { r}+2\right ) } \left ( \frac{\mu ^{2}}{1-\mu ^{2}}\right ) \ , \ , .\end{aligned}\ ] ] the first quantity above is related to the depletion of the pump that supplies the energy for the sub - harmonic mode . the follow two quantities are the squeezed and enhanced quadratures normally obtained in the linearized theory , while the fourth one is the first correction to the linearized calculation . the last one is the steady state triple quadrature correlation . this quantity has been suggested previously as a way to test quantum mechanics against a local hidden variable theory @xcite . the steady state intra - cavity squeezed quadrature fluctuations are obtained as : @xmath105 \ , \ , .\end{aligned}\ ] ] the intra - cavity squeezing quadrature near threshold is not perfectly squeezed , as the nonlinear correction is divergent near this point . this is shown in fig . ( [ sqmom ] ) . it is clear that the nonlinear corrections to the overall moment scale as @xmath106 , and hence only give large corrections extremely close to threshold , with @xmath107 . considerations related to optimal squeezing will be treated later , in the frequency domain . we can scale the quadratures variables in in semiclassical theory in the same way as before . firstly , the equations for the quadratures are : @xmath91 dt+\sqrt{\gamma _ { 1}}\ ; \left [ dw_{1}(t)+dw_{1}^{*}(t)\right ] \ ; \ ; , \nonumber \\ dy_{1 } & = & \left [ -\gamma _ { 1}y_{1}+\frac{\chi } { 2}\left ( x_{1}y_{2}-x_{2}y_{1}\right ) \right ] dt - i\sqrt{\gamma _ { 1}}\left [ dw_{1}(t)+dw_{1}^{*}(t)\right ] \ ; \ ; , \nonumber \\ dx_{2 } & = & \left [ -\gamma _ { 2}x_{2}-\frac{\chi } { 4}\left ( x_{1}^{2}-y_{1}^{2}\right ) + 2{\mathcal{e}}\right ] dt+\sqrt{\gamma _ { 2}}\ ; \left [ dw_{2}(t)+dw_{2}^{*}(t)\right ] \ ; \ ; , \nonumber \\ dy_{2 } & = & \left [ -\gamma _ { 2}y_{2}-\frac{\chi } { 2}x_{1}y_{1}\right ] dt - i\sqrt{\gamma _ { 2}}\ ; \left [ dw_{2}(t)-dw_{2}^{*}(t)\right ] \ ; \ ; .\end{aligned}\ ] ] in the new scaled time , the correlation function of the noise terms is : @xmath108 where we have written the wiener increment as @xmath109 . next , we redefine the white noise that drives the stochastic semiclassical equations as @xmath110 } { \sqrt{2}}\ , \ , , \nonumber \\ dw_{y1(2)}(\tau ) & = & \frac{\left [ dw_{1(2)}(\tau ) -dw_{1(2)}^{*}(\tau ) \right ] } { i\sqrt{2}}\ , \ , .\end{aligned}\ ] ] the dimensionless driving field @xmath111 is introduced as before , and the wiener increments @xmath112 have the same properties as defined in ( [ a8 ] ) , except for changing @xmath113 to the dimensionless scaled time @xmath114 . next , we use the same technique of matching the powers of g in the corresponding time - evolution equations . the zero - th order equations are : @xmath115 d\tau \ , \ , , \nonumber \\ dy_{1}^{(0 ) } & = & \left [ -y_{1}^{(0)}+\frac{1}{2}\left ( x_{1}^{(0)}y_{2}^{(0)}-x_{2}^{(0)}y_{1}^{(0)}\right ) \right ] d\tau \ , \ , , \nonumber \\ dx_{2}^{(0 ) } & = & -\gamma _ { r}\left [ x_{2}^{(0)}+\frac{1}{2}\left ( { x_{1}^{(0)}}{x_{1}^{(0)}}-{y_{1}^{(0)}}{y_{1}^{(0)}}\right ) -2\mu \right ] d\tau \ , \ , , \nonumber \\ dy_{2}^{(0 ) } & = & -\gamma _ { r}\left [ y_{2}^{(0)}+x_{1}^{(0)}y_{1}^{(0)}\right ] d\tau \ , \ , .\end{aligned}\ ] ] as in the positive - p case , the steady state solution below threshold is given by : @xmath116 to first order , the equations are given by : @xmath117 while the zero - th order equations are essentially classical , in this first order set the noise appears as a quantum effect . this is still a linear approximation , as all nonlinear corrections come from the next orders . the second order equations are : @xmath118 d\tau \ , \ , , \nonumber \\ dy_{1}^{(2 ) } & = & \left [ -\left ( 1+\mu \right ) y_{1}^{(2)}+\frac{1}{2}\left ( x_{1}^{(1)}y_{2}^{(1)}-x_{2}^{(1)}y_{1}^{(1)}\right ) \right ] d\tau \ , \ , , \nonumber \\ dx_{2}^{(2 ) } & = & -\gamma _ { r}\left [ x_{2}^{(2)}+\frac{1}{2}\left ( { x_{1}^{(1)}}{x_{1}^{(1)}}-{y_{1}^{(1)}}{y_{1}^{(1)}}\right ) \right ] d\tau \ , \ , , \nonumber \\ dy_{2}^{(2 ) } & = & -\gamma _ { r}\left [ y_{2}^{(2)}+x_{1}^{(1)}y_{1}^{(1)}\right ] d\tau \ , \ , .\end{aligned}\ ] ] we need to go beyond this order in perturbation theory to compute the first nonlinear corrections . the third order equations are : @xmath119 d\tau \ , \ , , \nonumber \\ dy_{1}^{(3 ) } & = & \left [ -\left ( 1+\mu \right ) y_{1}^{(3)}+\frac{1}{2}\left ( x_{1}^{(1)}y_{2}^{(2)}+x_{1}^{(2)}y_{2}^{(1)}-x_{2}^{(1)}y_{1}^{(2)}-x_{2}^{(2)}y_{1}^{(1)}\right ) \right ] d\tau \ , \ , , \nonumber \\ dx_{2}^{(3 ) } & = & -\gamma _ { r}\left [ x_{2}^{(3)}+\left ( { x_{1}^{(1)}}{x_{1}^{(2)}}-{y_{1}^{(1)}}{y_{1}^{(2)}}\right ) \right ] d\tau \ , \ , , \nonumber \\ \frac{dy_{2}^{(3)}}{d\tau } & = & -\gamma _ { r}\left [ y_{2}^{(3)}+x_{1}^{(1)}y_{1}^{(2)}+x_{1}^{(2)}y_{1}^{(1)}\right ] d\tau \ , \ , .\end{aligned}\ ] ] the steady state averages of the quantities of interest can now be using the truncated wigner distribution , therefore obtaining the symmetrically ordered correlation functions : @xmath120}\ , \ , , \nonumber \\ \langle y_{1}^{(1)}y_{1}^{(3)}\rangle & = & \frac{\mu } { 4(1+\mu ) ( 1-\mu ^{2})}\left [ \frac{-\gamma _ { r}}{\gamma _ { r}+2}+\frac{\gamma _ { r}(2-\mu ) + 2(1+\mu ) } { ( 1+\mu ) [ \gamma _ { r}+2(1+\mu ) ] } \right ] \ , \ , , \nonumber \\ \langle x_{1}^{(1)}y_{1}^{(1)}y_{2}^{(2)}\rangle & = & -\left ( \frac{\gamma _ { r}}{\gamma _ { r}+2}\right ) \left ( \frac{1}{1-\mu ^{2}}\right ) \ , \ , , \nonumber \\ \langle x_{1}^{(1)}y_{1}^{(1)}y_{2}^{(2)}\rangle + \langle x_{1}^{(2)}y_{1}^{(1)}y_{2}^{(1)}\rangle & = & \left ( \frac{2\gamma _ { r}}{\gamma _ { r}+2}\right ) \left ( \frac{1}{1-\mu ^{2}}\right ) \ , \ , .\end{aligned}\ ] ] the main difference in these calculation compared with the positive - p result , appears in the nonlinear correction for the sub - harmonic squeezed quadrature . up to second order in @xmath78 we have @xmath121 \nonumber \\ & = & \frac{1}{1+\mu } + \nonumber \\ & + & \frac{g^{2}}{2(1+\mu ) ( 1-\mu ^{2})}\left [ \frac{\gamma _ { r}}{\gamma _ { r}+2}+\frac{\gamma _ { r}(1 + 2\mu -2\mu ^{2})+2\mu ( 1+\mu ) } { ( 1+\mu ) [ \gamma _ { r}+2(1+\mu ) ] } \right ] \ , \ , .\end{aligned}\ ] ] the similarities and disagreement between this result and the positive - p expression for the same quantity deserve further comments given in the conclusion section . in particular , we notice that while the linear term agrees , the nonlinear term is not in agreement well below threshold . this comparison is shown in fig([nlsqmom ] ) , which compares the nonlinear parts of the moment in the two representations . just below threshold both theories give nonlinear corrections which are essentially identical . there is also good agreement in the limit of @xmath122 , where @xmath123 ; but for @xmath124 and driving fields below threshold , there is substantial disagreement in the nonlinear corrections to the squeezing between the two representations . this can be attributed to the neglect of third - order quantum correlations in the truncated wigner representation , which results in the appearance of nonlinear squeezing effects even in the limit of zero driving field . such effects are due to the semiclassical vacuum inputs , which do not appear in the positive p - representation . next , we proceed to analyze the problem in the frequency space by taking the fourier decomposition of the fields in order to understand the role of the first nonlinear correction in the squeezing spectrum . it is important to stress that most of the measurements performed are done in fourier space . the nonlinear corrections to the spectrum have a strikingly different behaviour to the case of the squeezing moments . the reason for this is that the nonlinear corrections are due to low - frequency , narrow - band critical fluctuations . these have a very small effect on the moments , which correspond to an integral of the spectrum over all frequencies , unless extremely close to threshold . however , they can have a very large and disruptive effect on the very important zero frequency component of the squeezing spectrum , where the quantum noise is at its lowest level . the spectrum can be calculated directly from the fourier transform of the stochastic equations . we also represent the white noise that drives the stochastic equations by its fourier transform @xmath125 , where the spectral moments of the stochastic processes are : @xmath126 it is also useful to introduce a standard convolution notation , where : @xmath127(\omega ) = \int \frac{d\omega ^{\prime } } { \sqrt{2\pi } } a(\omega ^{\prime } ) b(\omega -\omega ^{\prime } ) \ , \ , .\ ] ] the stochastic equations may now be rewritten in the frequency domain as - * first order : @xmath128 * second order : @xmath129 \left ( \omega \right ) } { 2\left ( i\omega + \gamma _ { r}\right ) } \ , \ , , \nonumber \\ { \tilde{y}}_{2}^{\left ( 2\right ) } \left ( \omega \right ) & = & -\frac{\gamma _ { r}\left [ \tilde{x}_{1}^{\left ( 1\right ) } \star \tilde{y}_{1}^{\left ( 1\right ) } \right ] \left ( \omega \right ) } { \left ( i\omega + \gamma _ { r}\right ) } \ , \ , .\label{19 } \end{aligned}\ ] ] * third order : @xmath130 \left ( \omega \right ) } { 2\left ( i\omega + 1-\mu \right ) } \ , \ , , \nonumber \\ { \tilde{y}}_{1}^{\left ( 3\right ) } \left ( \omega \right ) & = & \frac{\left [ \tilde{y}_{2}^{\left ( 2\right ) } \star \left ( \tilde{x}_{1}^{\left ( 1\right ) } + \xi _ { x}/\sqrt{2\mu } \right ) -\tilde{x}_{2}^{\left ( 2\right ) } \star \left ( \tilde{y}_{1}^{\left ( 1\right ) } + i\xi _ { y}/\sqrt{2\mu } \right ) \right ] \left ( \omega \right ) } { 2\left ( i\omega + 1+\mu \right ) } \ , \ , .\end{aligned}\ ] ] we now calculate the spectrum of the squeezed field @xmath131 , which is given by @xmath132 . thus , we obtain @xmath133\right\rangle + \cdots \end{aligned}\ ] ] the contribution from the first order perturbation theory is the usual linearized squeezing result , given in this case by : @xmath134 } \ , \ , .\ ] ] similarly , the complementary ( unsqueezed ) spectrum is : @xmath135 } \ , \ , .\ ] ] also , we can obtain the next order contribution to the squeezing , by calculating @xmath136 . to check the results , we can compare with the moment calculations , since : @xmath137 using these results , we find that the internal spectrum of the squeezed quadrature , to this order , is given by @xmath138 and the squeezing spectrum is calculated to be : @xmath139^{2}}\nonumber \\ & \times & \left [ \frac{(\omega ^{2}+1-\mu ^{2})}{2\mu \gamma _ { r}(1-\mu ^{2})}+\frac{(1-\mu + \gamma _ { r})(1+\mu ) -\omega ^{2}}{(1-\mu ) [ \omega ^{2}+(1-\mu + \gamma _ { r})^{2}]}\right . \nonumber \\ & & \left . -\frac{(1+\mu + \gamma _ { r})(1+\mu ) -\omega ^{2}}{(1+\mu ) [ \omega ^{2}+(1+\mu + \gamma _ { r})^{2}]}\right ] \ , \ , .\end{aligned}\ ] ] the corresponding external squeezing spectrum is then : @xmath140^{2}}\nonumber \\ & \times & \left [ \frac{(\omega ^{2}+1-\mu ^{2})}{2\mu \gamma _ { r}(1-\mu ^{2})}+\frac{(1-\mu + \gamma _ { r})(1+\mu ) -\omega ^{2}}{(1-\mu ) [ \omega ^{2}+(1-\mu + \gamma _ { r})^{2}]}\right . \nonumber \\ & & \left . -\frac{(1+\mu + \gamma _ { r})(1+\mu ) -\omega ^{2}}{(1+\mu ) [ \omega ^{2}+(1+\mu + \gamma _ { r})^{2}]}\right ] \ , \ , .\end{aligned}\ ] ] this equation gives the complete linear and nonlinear squeezing spectrum , including all the nonlinear correction terms that contribute to order @xmath90 or @xmath141 . an illustration of the behaviour of the total spectrum is given in fig ( [ totalspec ] ) . fig ( [ nlspec ] ) shows how the nonlinear contribution changes with driving field , giving just the portion of the spectrum proportional to @xmath90 . next , we can calculate the triple spectral correlations , giving as in the moment calculations : @xmath142 solving for @xmath143 , we have @xmath144(\omega _ { 3})\right\rangle } { \left ( i\omega _ { 3}+\gamma _ { r}\right ) } \ , \ , .\end{aligned}\ ] ] substituting from the first order spectrum , the final result to this order is obtained to be : @xmath145 \left [ \omega _ { 2}^{2}+\left ( 1+\mu \right ) ^{2}\right ] } \ , \ , .\ ] ] to check this result , we can evaluate moments : @xmath146 on integrating , we obtain the same result as in our moment calculation , given above . we will now compare these results with the corresponding results calculated in the semiclassical theory . some differences between them could be an interesting test comparing quantum mechanical predictions with a hidden variable theory . again , the spectral correlations are calculated from the fourier transform of the stochastic equations . in the frequency domain , the equations are written as * first order @xmath147 * second order @xmath148 ( \omega ) } { 2\left ( i\omega + 1-\mu \right ) } \ , \ , , \nonumber \\ \tilde{y}_{1}^{(2)}(\omega ) & = & \frac{\left [ \tilde{x}_{1}^{(1)}\star \tilde{y}_{2}^{(1)}-\tilde{y}_{1}^{(1)}\star \tilde{x}_{2}^{(1)}\right ] ( \omega ) } { 2\left ( i\omega + 1+\mu \right ) } \ , \ , , \nonumber \\ \tilde{x}_{2}^{(2)}(\omega ) & = & -\frac{\gamma _ { r}\left [ \tilde{x}_{1}^{(1)}\star \tilde{x}_{1}^{(1)}-\tilde{y}_{1}^{(1)}\star \tilde{y}_{1}^{(1)}\right ] ( \omega ) } { 2\left ( i\omega + \gamma _ { r}\right ) } \ , \ , , \nonumber \\ \tilde{y}_{2}^{(2)}(\omega ) & = & -\frac{\gamma _ { r}\left [ \tilde{x}_{1}^{(1)}\star \tilde{y}_{1}^{(1)}\right ] ( \omega ) } { \left ( i\omega + \gamma _ { r}\right ) } \ , \ , .\end{aligned}\ ] ] * third order ( sub - harmonic field ) @xmath149 ( \omega ) } { 2\left ( i\omega + 1-\mu \right ) } \ , \ , , \nonumber \\ \tilde{y}_{1}^{(3)}(\omega ) & = & \frac{\left [ \tilde{x}_{1}^{(1)}\star \tilde{y}_{2}^{(2)}+\tilde{x}_{1}^{(2)}\star \tilde{y}_{2}^{(1)}-\tilde{y}_{1}^{(2)}\star \tilde{x}_{2}^{(1)}-\tilde{y}_{1}^{(1)}\star \tilde{x}_{2}^{(2)}\right ] ( \omega ) } { 2\left ( i\omega + 1+\mu \right ) } \ , \ , .\end{aligned}\ ] ] the spectrum of the fields are given , for instance for the squeezed quadrature @xmath131 , by @xmath150 the first order perturbation theory generates the usual linearized squeezed result as in quantum theory @xmath151 and , similarly , for the amplified fluctuation quadrature @xmath152 and for the pump quadratures , there is no first order squeezing @xmath153 the next contribution to the squeezing field quadrature is : @xmath154 } + \right . \nonumber \\ & + & \left . \frac{1+\mu + \gamma _ { r}}{(1+\mu ) \left [ \omega _ { 1}^{2}+(1+\mu + \gamma _ { r})^{2}\right ] } \right\ } \ , \ , , \end{aligned}\ ] ] and @xmath155 ^{2}}\nonumber \\ & \times & \left\ { -\frac{(1+\mu ) ( 1-\mu + \gamma _ { r})-\omega _ { 1}^{2}}{(1-\mu ) \left [ \omega _ { 1}^{2}+(1-\mu + \gamma _ { r})^{2}\right ] } \right . \nonumber \\ & & \left . + \frac{(1+\mu ) ( 1+\mu + \gamma _ { r})-\omega _ { 1}^{2}}{(1+\mu ) \left [ \omega _ { 1}^{2}+(1+\mu + \gamma _ { r})^{2}\right ] } + \frac{(1+\mu ) } { \gamma _ { r}(1-\mu ^{2})}\right\ } \ , \ , .\end{aligned}\ ] ] and the internal ( symmetrically ordered ) squeezing spectrum is @xmath156 ^{2}}\left\ { \frac{2\mu ( 1+\mu ) } { \gamma _ { r}(1-\mu ^{2})}\right . \nonumber \\ & + & \left . \frac{(1-\mu + \gamma _ { r})\omega ^{2}+\left [ ( 1+\mu ) ^{2}+2\mu ( 1+\mu ) \right ] ( 1+\mu + \gamma _ { r})}{(1+\mu ) \left [ \omega ^{2}+(1+\mu + \gamma _ { r})^{2}\right ] } \right . \nonumber \\ & + & \left . \frac{(1+\mu + \gamma _ { r})\omega ^{2}+(1-\mu ^{2})(1-\mu + \gamma _ { r})}{(1-\mu ) \left [ \omega ^{2}+(1-\mu + \gamma _ { r})^{2}\right ] } \right\ } \ , \ , .\end{aligned}\ ] ] of greater interest is the external squeezing spectrum , which is obtained by including both internal fields and the correlated reflected vacuum noise terms : @xmath157 ^{2}}\left\ { \frac{\mu ( 1+\omega ^{2}-\mu ^{2})}{\gamma _ { r}(1-\mu ^{2})}\right . \nonumber \\ & + & \left . \frac{\left [ ( 1-\mu ) ( 1-\mu + \gamma _ { r})-2\mu ^{2}\right ] \omega ^{2}+\left[1-\mu + \gamma _ { r}\right]\left [ 1+\mu + \mu ^{2}+\mu ^{3}\right ] } { ( 1-\mu ) \left [ \omega ^{2}+(1-\mu + \gamma _ { r})^{2}\right ] } \right . \nonumber \\ & + & \left . \frac{\left [ ( 1+\mu ) ( 1+\mu + \gamma _ { r})+2\mu ^{2}\right ] \omega ^{2}+\left[1+\mu + \gamma _ { r}\right]\left [ 1 + 3\mu + \mu ^{2}-\mu ^{3}\right ] } { ( 1+\mu ) \left [ \omega ^{2}+(1+\mu + \gamma _ { r})^{2}\right ] } \right . \ , \ , .\end{aligned}\ ] ] this semi - classical spectrum is quite different from positive - p calculation when @xmath158 but gives a compatible result near threshold , that is in the limit @xmath159 . a detailed comparison of the zero - frequency behaviour is shown in fig ( [ nl2pwspec ] ) . this means that even when the pump is off , semiclassical theory gives a distorted vacuum spectrum due to the presence of the nonlinear crystal . this happens because in this theory the vacuum fluctuations are taken as real , and then two vacuum modes can interact inside the crystal as real fields . in the limit of @xmath122 , the two spectra become compatible again , as the semiclassical theory decouples the second - harmonic mode from its vacuum input in this limit . in the case of threshold fluctuations , we can interpret the agreement as due to the large photon numbers involved - which means that the truncation approximation used for the semiclassical calculation is more reliable . it is interesting to evaluate the squeezing or low - noise quantum correlations in the limit of zero frequency , that is in the resonance regime which is generally the frequency of maximum squeezing . we obtain from the positive - p result : @xmath160}\right ] \ , \ , .\ ] ] near threshold , where @xmath161 , we can set @xmath162 , and expand in powers of @xmath163 . minimizing this result with respect to @xmath164 , we find that , to leading order in @xmath78 , the optimal driving field is the solution to the following equation:@xmath165 this is a quintic equation , but it has simple closed form solutions in two limits , depending on whether @xmath166 or @xmath167 . in the first case , the variance can be rewritten as : @xmath168 \ , \ , .\ ] ] minimizing this result with respect to @xmath164 , we find that the minimum level of internal fluctuations occurs in a narrow frequency range near @xmath169 , at a driving field just below threshold , with @xmath170 so that : @xmath171 to leading order in @xmath78 , the corresponding spectral variance is : @xmath172 this result of an @xmath5 scaling confirms an approximate calculation of plimak and walls @xcite , although the self - consistent method used by these authors makes it difficult to obtain the relevant driving field . the physics of this is clearly that the onset of critical fluctuations starts to spoil the noise - reduction even before the critical point is reached at @xmath173 . for example , with @xmath174 and @xmath175 , we find that @xmath176 , or about @xmath177 below shot noise , as predicted from the analytic theory . this can also be seen from the way that the third order term includes contributions from the critical fluctuations in @xmath178 . a direct calculation from the full spectrum shows that this is a true minimum for all frequencies , even including @xmath179 . however , the situation clearly changes as @xmath122 , in which case much greater levels of spectral noise - reduction are possible . this is plotted in fig ( [ opt2dpspec ] ) : analytically , this limit gives the following result , provided that @xmath180 : @xmath181 \ , \ , .\ ] ] minimizing this result with respect to @xmath164 , we find that the minimum level of internal fluctuations occurs at a driving field very close to threshold , with : @xmath182 the corresponding variance is therefore : @xmath183 this result can be much smaller than predicted by the calculation of plimak and walls @xcite , since the damping ratio can be reduced ( at least in principle ) to an arbitrarily low level - although still bounded below by @xmath90 , in order for perturbation theory to be applicable , so that we do not expect to obtain @xmath184 . of course , there are experimental limitations on this , due to absorption losses in the nonlinear medium at short wavelengths . thus , for example , with the same value of @xmath185 as previously , but with @xmath186 , we find that the minimum spectral noise is predicted to occur at a driving field of @xmath187 , with a squeezing variance of @xmath188 , or about @xmath189 below shot noise - about @xmath190 lower than before . this operating regime also has the property that the optimum frequency of noise reduction moves away from zero frequency as the driving field is increased above the optimum value , towards threshold . at slightly higher driving fields than the optimum point , a bifurcation to a spectrum with two minima occurs , although with similar levels of noise reduction , as shown in fig ( [ tot2dp01spec ] ) . in this regime the results of the perturbation theory need to be checked by a full simulation of the stochastic equations . we have carried this out ( see next section ) , and find that the full simulations do agree very well with the analytic predictions , even with this small damping ratio . the value of the nonlinear correction to the spectrum of the squeezed quadrature @xmath191 can be worked out from a full numerical simulation@xcite of the relevant nonlinear stochastic equations . the optimal squeezing in the zero frequency part of the squeezing spectrum is predicted to scale as @xmath5 with roughly equal values of decay rates . for the simulations , we chose values of @xmath192 , @xmath193 . the simulations used a total dimensionless time - interval of @xmath194 . to ensure equilibrium , only the last @xmath195 time units were utilized in the fourier transforms . time steps of @xmath196 and @xmath197 were compared to ensure convergence . the algorithmic technique is described elsewhere @xcite , and uses a semi - implicit central partial difference technique . to obtain the small nonlinear corrections near the optimum squeezing , we simulated the difference between the linear and nonlinear forms of the stochastic equation , in order to minimize sampling errors . it was also useful to initialize the @xmath198 quadratures with a gaussian ensemble close to the known steady - state variance , in order to reduce the time taken to achieve equilibrium . typically , the relative error in the correlations due to finite step - size was around @xmath199 with these step - sizes . for these parameters the optimal driving field is predicted to occur at @xmath175 , or approximately @xmath200 of the critical intensity . we used @xmath201 trajectories to improve the relative error due to sampling with a finite trajectory population , giving relative sampling errors of less than @xmath202 the calculated squeezing moment from the sde simulations was : @xmath203 . this is in excellent agreement with the below - threshold expansion , which gives @xmath204 , as this is just outside the critical region . we find that the spectral predictions are also well verified by the simulations . these resulted in a value for the nonlinear correction to the zero - frequency spectrum , of @xmath205 . by comparison , the analytic theory , worked to fourth order in @xmath78 , gives the prediction that @xmath206 . the residual difference of about @xmath207 - which is significant compared to sampling error - can be attributed to the fact that there are higher order corrections that are not included in the analytic theory , and these are more significant in the zero - frequency spectrum than they are in the moment calculation . ( [ otsim9 ] ) shows the detailed results of the simulation . in the analytic theory , we found that a smaller decay rate for the second - harmonic is predicted to yield a better squeezing optimum , as a function of driving field . in fig . ( [ optsim01 ] ) we verify this to be the case , by carrying out a full numerical simulation for @xmath208 , @xmath209 , with @xmath210 , time steps of @xmath211 and @xmath196 for error - checking , and @xmath212 trajectories . the results show that the analytic predictions and numerical simulations are almost indistinguishable in this regime . the sampling error was relatively larger , possibly due to the fact that the absolute noise levels are lower here . the agreement indicates that the perturbation theory is an excellent approximation to the full nonlinear equations with these parameters . we have calculated the nonlinear quantum fluctuations in a parametric oscillator below the classical threshold , using a nonlinear stochastic positive - p theory , with both asymptotic approximations and a numerical technique . there is excellent agreement between numerical and analytic calculations . corresponding results for the keldysh diagram method require a summation over infinite sets of diagrams , in order to fully include the reservoirs . the advantage of the present method is due to the fact that the coherent state basis is a more natural basis set for an open system , since it allows the damping reservoirs to be treated non - perturbatively . optimal squeezing in the output spectra corresponding to these moments were estimated . we found that the best squeezing in the zero frequency part of the squeezing spectrum scales like @xmath5 just below threshold , _ provided the two fields have similar damping rates . _ in other words , at the true critical threshold - where the linear squeezing is optimized - the nonlinear corrections are too large to give the lowest overall zero - frequency squeezing . instead , one should operate below the critical point to optimize the spectral squeezing . using an entirely different method , a calculation by plimak and walls @xcite also predicted that the optimum zero frequency squeezing spectrum scales like @xmath5 , or equivalently , as @xmath213 for a given input flux @xmath214 . our general scaling results agree with theirs , except with a different spectrum . we attribute the difference to the systematic + p stochastic diagram procedure used here to calculate the spectrum , rather than the feynman diagram method - which involves additional approximations . we also found a new regime in which the lower limit to the spectral noise - reduction depends on the decay rate of the second - harmonic field , which can be reduced to an arbitrarily low level . this has a reasonable physical interpretation , since the second - harmonic losses are essentially parasitic losses , which do not contribute to the desired squeezing output . the ultimate limit to squeezing in this regime is set by even higher order terms in perturbation theory . we conjecture that optimization of both the driving field and the relative decay rate may result in a final squeezing variance scaling as @xmath215 . a calculation with the truncated wigner method , or semiclassical technique , was also carried out . well below threshold , we found that while the linear terms agreed with full quantum calculation , nonlinear corrections and higher order correlations tended to disagree , especially for high second - harmonic losses . however , near the critical point , the situation changed . here , even though the dominant terms are nonlinear , we found excellent agreement between the two methods . l. a. wu , h. j. kimble , j. l. hall , h. wu , phys . 57 * , 2520 ( 1986 ) . a. heidmann , r. j. horowicz , s. reynaud , e. giacobino , c. fabre , g. camy , phys . lett . * 59 * , 2555 ( 1987 ) . z. y. ou , s. f. pereira , h. j. kimble , phys . lett . * 68 * , 3663 ( 1992 ) . p. d. drummond , k. dechoum and s. chaturvedi , submitted to phys . a. p. d. drummond , k. j. mcneil and d. f. walls , optica acta * 27 * , 321 - 335 ( 1980 ) ; p. d. drummond , k. j. mcneil and d. f. walls : optica acta 28 , 211 - 225 ( 1981 ) . b. yurke , phys . a * 32 * , 300 ( 1985 ) . c. w. gardiner and m. j. collett , phys . a * 31 * , 3761 ( 1985 ) ; m. j. collett and d. f. walls : phys . a * 32 * , 2887 ( 1985 ) . p. kinsler and p. d. drummond , phys . rev a * 43 * , 6194 ( 1991 ) . p. d. drummond and p. kinsler , quantum and semiclass . opt * 7 * , 727 ( 1995 ) ; p. kinsler and p.d . drummond , phys . a * 52 * , 783 ( 1995 ) . j. schwinger , j. math . phys . * 2 * , 407 ( 1961 ) . l. v. keldysh , sov . jetp * 20 * , 1018 ( 1965 ) . h. w. wyld jr , ann * 14 * , 143 ( 1961 ) . c. j. mertens , t. a. b. kennedy and s. swain , phys . lett . * 71 * , 2014 ( 1993 ) ; c. j. mertens , t. a. b. kennedy and s. swain , phys . rev . a * 48 * , 2374 ( 1993 ) ; c. j. mertens , j. m. hasty , h. h. roark iii , d. nowakowski and t. a. b. kennedy , phys . rev . a * 52 * , 742 ( 1995 ) ; c. j. mertens and t. a. b. kennedy : phys . a * 53 * , 3497 ( 1996 ) . l. i. plimak and d.f . walls , phys . a * 50 * , 2627 ( 1994 ) . o. veits and m. fleischhauer , phys . a * 52 * , r4344 ( 1995 ) ; phys . rev a * 55 * , 3059 ( 1997 ) . s. chaturvedi , p. d. drummond and d. f. walls , j. phys . a * 10 * , l187-l192 ( 1977 ) ; p. d. drummond and c. w. gardiner , j. phys . a * 13 * , 2353 ( 1980 ) . h. j. carmichael , _ statistical methods in quantum optics 1 _ , ( springer , berlin , 1999 ) . l. arnold , _ stochastic differential equations : theory and applications _ , ( john wiley and sons , new york , 1974 ) ; c. w. gardiner , _ handbook of stochastic methods _ ( springer , berlin , 1983 ) . a. gilchrist , c. w. gardiner and p. d. drummond , phys . a**55 * * , 3014 - 3032 ( 1997 ) . s. chaturvedi and p. d. drummond : eur . j. b**8 * * , 251 ( 1999 ) . s. chaturvedi and p. d. drummond : physical review a**55 * * , 912 ( 1997 ) ; d. t. pope , p. d. drummond , w. j. munro , phys . rev . a * 62 * , 042108 ( 2000 ) . d. drummond : phys . rev . a**33 * * , 4462 ( 1986 ) . h. j. carmichael , j. s. satchell , s. sarkar : phys . rev . * a34 * , 3166 ( 1986 ) . p. d. drummond , i. k. mortimer : j. comput . * 93 * , 144 ( 1991 ) .

this paper is devoted to the survival analysis of a system subject to competing failure modes within an external stressing environment . the external environment is assumed to stress the system at random and isolated times according to a random shock model . such a model can represent external demands e.g. , which put some stress on the system at their arrivals . shock models have been the subject of an extensive literature . following @xcite , shock models may be classified into different categories , according to whether arrival times and shock magnitudes are correlated @xcite , or independent @xcite , and according to the assumption put on the shocks arrival process : homogeneous or non - homogeneous poisson process @xcite , renewal process @xcite or non - stationary pure birth process @xcite . according to the influence of shocks on the system , shock models may be further classified into three types : extreme shock models ( a shock can cause the system immediate failure , see @xcite ) , cumulative shock models ( a shock increases some intrinsic characteristic of the system such as its deterioration , failure rate , age , number of already endured shocks , ... , see @xcite ) and mixed shock models ( a shock can either cause the system immediate failure or increase some intrinsic characteristic , see @xcite ) . more references can be found in @xcite ; @xcite ; @xcite ; @xcite . in this paper , a shock model with mixed effects is considered , where the occurrence of shocks is classically modelled through a non - homogeneous poisson process and where each shock may result in the system immediate failure through a bernoulli trial , independent of the system intrinsic behaviour . apart from shocks , the system suffers from competing soft and sudden failures , where soft failures refer to the reaching of some given threshold for the degradation level , and sudden failures to accidental failures , characterized by a failure rate . a possible representation corresponds to a two - component series system , where the first component is subject to soft failures and the second one to sudden failure . the system may hence fail through three different competing modes : traumatic failure due to a fatal shock ; soft failure ; sudden failure . each non fatal shock induces some increase of both deterioration level and failure , with possible dependence in - between . in the oldest literature , most models considered one single possible type of failures for the system : e.g. soft failures in @xcite or traumatic failures due to shocks in @xcite , both in the multivariate setting . more recently , different models have been developed , which consider two different types of failures . for instance , competition between soft and sudden failures is studied in @xcite . several application cases are proposed in the paper ( see also references therein ) . an industrial example is also provided in @xcite , which studies the reliability of an aircraft engine . note that @xcite assumes soft and sudden failures to be independent whereas the stressing environment of the present paper makes them dependent . competition between soft failures and occurrence of a traumatic event are considered in degradation - threshold - shock models ( dts - models , denomination of @xcite ) which have been proposed by @xcite and further studied by @xcite . a case study is provided in @xcite for the analysis of fatigue crack growth , where the effects of shocks on the degradation are put into evidence . note that , contrary to the present paper , dts - models consider some possible influence of the deterioration level of the system on the shocks arrival rate . however , the first shock of a dts - model is always fatal to the system ( leading to a single shock possibly endured by the system ) , whereas successive non fatal shocks are here envisioned . we could not find any paper which takes into account three competing failure modes , as in the present paper . however , based on the case studies from the previous literature , one can think that our model can reflect lots of systems subject to competing soft and hard failures , within a stressing environment . for instance , one can think of an aerial cabin hooked to a cable : the supporting cable is deteriorating due to corrosion and fatigue ; the linking pulley is subject to sudden failures ( and maybe also to deterioration ) ; both cable and pulley endure shocks at each cabin travel , which increase jointly their respective deterioration level and failure rate . the present paper consider several kinds of dependence between the three competing failure modes : at each non fatal shock , the increase of deterioration and failure rate is simultaneous . this induces a first type of dependence between soft and sudden failures . each shock may induce a failure , either because the shock is fatal , or because the deterioration is suddenly increased beyond the threshold level . this induces a second type of dependence between soft and traumatic failures , which may be simultaneous . also , some possible dependence is envisioned between the increments of failure rate and of deterioration at each non fatal shock . this induces a third type of dependence between soft and sudden failures . finally , following @xcite , the probability for a shock to be fatal depends on the shock arrival time , which induces a last type of dependence . up to our knowledge , all these kinds of dependence have not been yet considered altogether and , as will be seen all along the text , this model enlarges several ones from the previous literature . the paper is organized as follows : the model is specified in section [ model ] . the system reliability is computed through different methods in section 3 . sufficient conditions are provided in section 4 * * * * for the system lifetime to be new better than used . the influence of various parameters of the shock environment on the lifetime is studied in section 5 . numerical experiments are proposed in section [ section num ] and concluding remarks end the paper in section [ concl ] . to make the model clear , a two - component series system is considered , where the first component is subject to sudden failure and the second one to soft failure . this two - unit system is just a representation for the competing soft and sudden failure modes , with no restriction . in the ideal condition ( for example in a laboratory environment ) , the lifetime of the first component is characterized by its intrinsic hazard rate @xmath0 , @xmath1 while the second one is subject to some accumulative deterioration modeled by an increasing stochastic process @xmath2 ( e.g. a gamma process ) . the second component fails once its deterioration level exceeds a failure threshold @xmath3 . the lifetimes of the two components are made dependent by their common stressing environment . this environment is modelled by a random shock process , where the shocks arrive according to a non - homogeneous poisson process @xmath4 with intensity @xmath5 ( or cumulated intensity @xmath6 ) . to avoid useless technical details , we assume that @xmath7 for all @xmath8 . more generally , one might consider that @xmath7 only for @xmath9 greater than some @xmath10 , which would mean that the shocks would arrive only after time @xmath11 . the points of the poisson process are denoted by @xmath12 , ... , @xmath13 , ... with @xmath14 almost surely . a shock at time @xmath9 may cause the system immediate failure ( fatal shock ) with probability @xmath15 $ ] , which depends on the age @xmath9 of the system at the shock arrival . a shock at time @xmath9 is non fatal with probability @xmath16 . a non fatal shock at time @xmath17 increases the deterioration of both components in a different way : * for the first component , its hazard rate is increased of a non negative random amount @xmath18 , * for the second one , its accumulated deterioration is increased of a non negative random amount @xmath19 . the random vectors @xmath20 , @xmath21 , @xmath22 , ... are assumed to be independent and identically distributed ( i.i.d . ) with common distribution @xmath23 , and independent of the shocks arrival times @xmath24 ( and hence independent of the poisson process @xmath4 ) . at each shock , the increments @xmath18 and @xmath19 are possibly dependent . when subscript @xmath25 is unnecessary , we drop it and set @xmath26 to be a generic copy of @xmath20 . for @xmath27 , the distribution of @xmath28 is denoted by @xmath29 . we set @xmath30 to be the bivariate compound poisson process defined by @xmath31 with@xmath32where @xmath33 . the processes @xmath34 and @xmath2 are assumed to be independent . provided that the system is functioning up to time @xmath9 , the random variables @xmath35 and @xmath36 stand for the cumulated increments on @xmath37 $ ] of the failure rate of the first component and of the deterioration of the second component due to the external environment , respectively . setting @xmath38 to be the @xmath39field generated by @xmath40 and provided that the system is still up at time @xmath9 , the conditional hazard rate of the first component given @xmath41 is@xmath42 and the conditional deterioration of the second component given @xmath41 is@xmath43 to make it clearer , we introduce @xmath44 , @xmath45 to be the lifetime of the @xmath46 component under the external environment , without taking into account the possibility of fatal shocks for the system . to simplify the writing , we denote by @xmath47 the conditional expectation @xmath48 for any measurable set @xmath49 , where @xmath50 stands for the indicator function ( @xmath51 if @xmath52 , 0 elsewhere ) . we then have:@xmath53 where @xmath54 is the cumulated intrinsic failure rate of the first component and where @xmath55 stands for the cumulative distribution function ( c.d.f . ) of @xmath56 . ( recall that @xmath57 is independent of @xmath41 ) . we now let @xmath58 to be the time to the first fatal shock for the system with @xmath59 and we assume that the bernoulli trials ( fatal shocks or not ) which happen at each shock arrival are independent one with each other , and that they depend on @xmath41 only through the @xmath60 s , that is : @xmath61 } \left ( t_{i}\right ) = \prod_{i=1}^{n_{t}}q\left ( t_{i}\right ) \label{tau3}\ ] ] where @xmath62 . the system failure is induced either by a fatal shock or by a component failure ( soft or sudden failure ) , whatever arrives first . the lifetime of the system hence is @xmath63 we finally make the additional assumption that @xmath64 , @xmath65 and @xmath58 are conditionally independent given @xmath66@xmath67 to sum up , the whole model is specified by : * @xmath68 the ( generic ) random increments in failure rate ( first component , @xmath69 ) and deterioration ( second component , @xmath70 ) , * @xmath71 the intensity of the non - homogeneous poisson process , * @xmath72 the intrinsic failure rate of the first component ( sudden failure ) , * @xmath73 the intrinsic deterioration of the second component ( soft failure ) , * @xmath74 the probability for a shock at time @xmath9 to be fatal at the system level , ( plus some independence assumptions ) . by taking special cases for these five ingredients , we can see that our model extents some well - known models from the literature . for instance , taking @xmath75 ( no fatal shocks ) , @xmath76 constant , @xmath77 and @xmath78 all @xmath1 ( no second component ) , one gets the `` stochastic failure model in random environment '' from @xcite . taking @xmath77 and @xmath78 all @xmath1 ( one single component ) , one gets the `` stochastic survival model for a system under randomly variable environment '' from @xcite . taking @xmath78 all @xmath1 , @xmath79 , the model resumes to a classical extreme shock model ( one single component ) , where system failures are only due to shocks arriving according to a non - homogeneous poisson process , with probability @xmath80 for a shock to be fatal ( and @xmath81 to be harmless ) . this model is interpreted as the brown - proschan model by @xcite ; see also @xcite where various properties of the model are explored . taking @xmath75 ( no fatal shocks ) , @xmath82 ( one single component ) , @xmath83 ( homogeneous poisson process ) , @xmath78 all @xmath1 ( no intrinsic deterioration for the second - and single - component ) , one gets the `` cumulative damage threshold model '' from ( * ? ? ? * section 4 , case of i.i.d . damage increments ) . taking @xmath84 ( constant ) , @xmath85 ( one single component ) , @xmath78 all @xmath1 , one gets the `` cumulative damage model with two kinds of shocks '' from ( * ? ? ? * case of i.i.d . damage increments ) . taking @xmath86 , @xmath70 exponentially distributed , @xmath87 , all @xmath1 , one gets the model from subsection * 3.b * in @xcite . all these models are summed up in table [ table1 ] . note that we do not pretend at any exhaustibility and our model will include lots of other previous models which are not provided here . [ c]ccbrown - proschan model from @xcite & @xmath88 , + & @xmath78 , @xmath89 + deterministic boundary in @xcite & @xmath90 , @xmath87 + & @xmath70 exponentially distributed + @xcite & @xmath91 , @xmath92 is a constant , + & @xmath77 , @xmath78 , @xmath89 + @xcite & @xmath77 , @xmath78 , @xmath93 + cumulative damage threshold models & @xmath91 , @xmath77 , @xmath94 is a constant + in section 4 of @xcite & @xmath95 , @xmath78 , @xmath96 + @xcite & @xmath97 is a constant , @xmath98 , + & @xmath78 , @xmath89 + the objective of this section is to calculate the reliability @xmath99 of the system at time @xmath9 , with@xmath100 where we recall that the system lifetime @xmath101 is defined by @xmath102 . a first way to compute @xmath99 is to use classical monte - carlo simulations and to simulate a large number of independent histories for the system up to time @xmath9 ( * method 1 * ) . this method will serve as a comparison tool in the numerical experimentations in section [ section num ] . this method requires the simulation of a random variable with conditional hazard rate @xmath103 ( see algorithm [ algo ] in section [ section num ] for details ) and may imply long computational times for the system reliability . we provide below a few alternate methods which may be quicker and also easier to implement . [ method 2][prop1]the reliability is given by @xmath104 where @xmath105 is provided by @xmath106 and where @xmath107 due to the conditional independence of @xmath64 , @xmath65 and @xmath58 given @xmath41 , we have : @xmath108 using @xmath109 , we get:@xmath110 which provides @xmath111 and next @xmath112 , due to @xmath113 and @xmath114 based on the previous result , the only point to get the reliability @xmath115 is to compute @xmath116 . the remaining of the section is hence devoted to the computation of @xmath116 . starting from @xmath111 ( or @xmath117 ) , a possibility is to compute @xmath116 through monte - carlo simulations of @xmath118 and @xmath30 , which is simpler and quicker than simulating trajectories of the system according to the initial model . this method is called * method 2 * in section [ section num ] . following @xcite and @xcite , one may also use a series expansion of @xmath116 , as provided by the following proposition . [ method 3 , general case][prop2 ] @xmath119 where@xmath120 and @xmath121 are i.i.d . random variables with probability density function ( p.d.f . ) @xmath122}(x)$ ] and independent of @xmath123 . conditioning on the poisson process @xmath124 , we have:@xmath125 with@xmath126 now , given that @xmath127 , the conditional joint distribution of @xmath128 is the same as the joint distribution of the order statistics @xmath129 of @xmath130 i.i.d . random variables @xmath131 with p.d.f . @xmath132}(x)$ ] ( see @xcite e.g. ) . using the fact that @xmath133 is independent of @xmath124 , we get:@xmath134 noting that the expression within the expectation is invariant through permutation of the @xmath135 s , we derive that:@xmath136 which provides the result . based on the previous result , one can see that our model is equivalent to a classical shock model , where the shocks arrive according to a non - homogeneous poisson process with intensity @xmath137 and with conditional probability of survival at time @xmath9 equal to @xmath138 , given that there has been @xmath130 shocks up to time @xmath9 . [ method 3 , independent case][cor1]in the special case where @xmath69 and @xmath139 are independent , we get:@xmath140 with @xmath141 and @xmath142 where @xmath143 stands for the convolution operator , @xmath144 , all @xmath145 , and @xmath146 stands for the laplace transform of the distribution @xmath147 of @xmath148 , with@xmath149 starting from @xmath150 and using the independence of all @xmath151 s , @xmath152 s and @xmath135 s , and the identical distributions of all @xmath151 s and of all @xmath135 s , we get:@xmath153 with@xmath154 substituting this expression into eq . @xmath155 and next into eq . @xmath150 provides the result . [ ex1]let @xmath139 be an exponentially distributed with mean @xmath156 and @xmath148 an independent random variable . in that case , @xmath157 is gamma distributed with parameter @xmath158 . this provides@xmath159 using @xmath160 and eq . @xmath161 , we get:@xmath162 taking @xmath87 for all @xmath1 , @xmath163 and @xmath164 as a special case , we get@xmath165 and , for @xmath166@xmath167 using successive integrations by parts for the last integral . this last expression is the result of theorem 2 in @xcite , which hence appears as a special case of the previous results . from proposition [ prop2 ] and corollary [ cor1 ] , one may derive the following approximation for @xmath168 . [ method 3 , approximation][cor method3]for @xmath169 , let @xmath170 be defined by @xmath171 in the general case , and by @xmath172 in case @xmath148 and @xmath139 are independent , where @xmath138 and @xmath173 are provided by @xmath150 and @xmath174 , respectively . then , for all @xmath1 , the sequence @xmath175 increases to the limit @xmath116 when @xmath176 and for all @xmath169 , we have @xmath177 where@xmath178 @xmath179 is defined by @xmath180 and @xmath181 is poisson distributed with mean @xmath179 . we just look at the general case . from proposition , we have @xmath182 due to @xmath183 , we get:@xmath184 based on the proof of corollary [ cor1 ] . this provides:@xmath185 and the result . the previous proposition provides numerical bounds for @xmath116 , which may be adjusted as tight as necessary , taking @xmath186 large enough . also , the required number of terms in the truncated series is given , to get a specified precision . this method is quite adapted as soon as it is possible to compute the @xmath138 s ( or the @xmath187 s ) . this mostly requires the distribution of @xmath188 to be known in full form , which is the case e.g. when the @xmath189 s are constant or gamma distributed . an example is provided in example [ ex1 ] , in the special case of an exponential distribution . in the most general case , the computation of the @xmath138 s ( or of the @xmath173 s ) may be as difficult as the initial problem of computing @xmath168 , so that the previous method is not always adapted . we finally present another method based on laplace transform , which does not suffer from the same restriction . [ method 4][prop3]we have:@xmath190 or equivalently @xmath191 where @xmath192 is provided by its laplace transform @xmath193 with @xmath194 the bivariate laplace transform of the distribution @xmath195 of @xmath68@xmath196 and @xmath197 , all @xmath198.in the special case where @xmath148 and @xmath139 are independent , @xmath199 may be simplified into:@xmath200 where @xmath179 is provided by @xmath180 and where @xmath201 is the univariate laplace transform of @xmath202 . remembering that @xmath203 , we get from @xmath117 that:@xmath204 ~dl\nonumber\\ & = \mathbb{e}\left [ \left ( \int_{0}^{\infty}e^{-sl}f_{g_{t}}\left ( l - a_{t}^{(2)}\right ) ~dl\right ) e^{-\sum_{i=1}^{n_{t}}v_{i}^{(1)}(t - t_{i})}\prod_{i=1}^{n_{t}}q(t_{i})\right ] \label{lap fi}\ ] ] with@xmath205 because @xmath57 is non negative . setting @xmath206 , we obtain @xmath207 substituting this expression into @xmath208 provides:@xmath209 with@xmath210 \\ & = \mathbb{e}\left [ e^{-\sum_{i=1}^{n_{t}}\left ( v_{i}^{(1)}(t - t_{i})+sv_{i}^{(2)}-\ln q(t_{i})\right ) } \right ] .\end{aligned}\ ] ] because @xmath211 is equivalent to @xmath212 , we get:@xmath213 with@xmath214 noting that the sequence @xmath215 are the points of a poisson random measure @xmath216 with intensity @xmath217 , the function @xmath218 may be interpreted as a laplace functional with respect of @xmath219 @xmath220 the formula for laplace functionals of poisson random measures ( * ? ? ? * theorem 2.9 ) next provides:@xmath221 with@xmath222 substituting @xmath223 by its expression @xmath224 , we get:@xmath225 ( t).\end{aligned}\ ] ] substituting this expression into @xmath226 and next into @xmath227 provides @xmath228 with @xmath199 given by @xmath229 . equation @xmath230 is a direct consequence . finally , in case @xmath148 and @xmath231 are independent , we have:@xmath232 and@xmath233 which ends this proof . based on the previous result , one can compute @xmath116 by inverting its laplace transform with respect of @xmath3 . looking at equation @xmath234 , the key point is the inversion of the laplace transform @xmath199 . we next provide an example where the inversion is possible in full form . in the most general case , this can be done numerically using some laplace inversion software . let @xmath95 , @xmath78 ( all @xmath1 ) , @xmath94 and @xmath97 constant , @xmath235 identically exponentially distributed with mean @xmath156 ( so that @xmath148 and @xmath70 are completely dependent ) . then:@xmath236 where @xmath237 stands for the dirac mass at @xmath238 . we easily get:@xmath239 and@xmath240 for @xmath241 , we have @xmath242 where @xmath243 inverting the laplace transform @xmath199 , we obtain : @xmath244 as @xmath245 , we get the following full form for the reliability:@xmath246 where @xmath247 is the cumulative distribution function of a gamma distributed random variable with parameter @xmath248 . in the special case where @xmath148 and @xmath139 are independent , the laplace inversion of @xmath199 is reduced to inverting @xmath249 , or equivalently to inverting @xmath250 , where @xmath251 is a constant . this is hence easier than in the most general case of correlated @xmath69 and @xmath139 . to sum up the section , we have at our disposal four different methods for computing the reliability : * method 1 * ( direct mc simulations ) : : the main drawbacks of this method are that it suffers from long computation times and that its implementation is less direct than for the other methods . * method 2 * ( computing @xmath168 through formula @xmath117 and mc simulations ) : : this method is much quicker and much easier to implement than method 1 . besides , it is always possible to use it . however , method 3 ( when possible ) and method 4 are quicker . * method 3 * ( truncated series expansion + control of the truncation error through corollary [ cor method3 ] ) : : this method provides very good results as soon as the @xmath138 s ( or the @xmath173 s ) are available in full form . * method 4 * ( laplace transform inversion ) : : this method is the best when it is possible to inverse the laplace transform @xmath252 in full form . numerical laplace inversion also provides quite good results . let us recall that a random variable @xmath253 ( or @xmath254 ) is new better than used ( nbu ) if @xmath255 all @xmath256 . we here provide sufficient conditions under which @xmath101 is nbu . [ propnbu]assume that the intrinsic lifetimes of both components are nbu , which means that : @xmath257 where the second condition is true as soon as @xmath2 is a univariate non negative lvy process . then , @xmath101 is nbu if one among the two following conditions is satisfied : 1 . @xmath97 is non increasing and @xmath94 is constant , 2 . @xmath97 is constant and @xmath258 is super - additive ( @xmath259 , all @xmath260 ) . let us first note that , in case @xmath2 is a univariate non negative lvy process , we have:@xmath261 due to the independent and homogenous increments of @xmath2 for the third line . assumption @xmath262 is hence true . starting again from @xmath263 and based on the nbu assumption @xmath264 , it is sufficient to show that @xmath116 is nbu . now : @xmath265 due to the second nbu assumption @xmath262 . under each of the two provided conditions , @xmath97 is non increasing so that @xmath266 , all @xmath267 . using @xmath268 and splitting the exponential and the product into two parts , one gets:@xmath269 .\label{edfi}\ ] ] setting @xmath270 then @xmath271 are points of the poisson process @xmath272 with admits @xmath273 for intensity . equation @xmath274 now writes:@xmath275 , \end{aligned}\ ] ] or equivalently:@xmath276 .\end{aligned}\ ] ] as @xmath277 is independent on @xmath278 and as the @xmath279 s are i.i.d . and independent on @xmath278 , one gets:@xmath280 with@xmath281 \\ b_{n } & = \mathbb{e}\left [ f_{g_{s}}\left ( l-\sum_{j=1}^{n_{s}^{\left ( t\right ) } } v_{j+n}^{(2)}\right ) e^{-\sum_{j=1}^{n_{s}^{\left ( t\right ) } } v_{j+n}^{(1)}(s - t_{j}^{\left ( t\right ) } ) } \prod_{j=1}^{n_{s}^{\left ( t\right ) } } q(t_{j}^{\left ( t\right ) } ) \right ] \text{.}\ ] ] noting that @xmath282 is independent on @xmath130 ( @xmath283 , all @xmath284 ) and that @xmath285 , we finally have@xmath286 where @xmath287 .\ ] ] the point now is to prove that @xmath288 under the two different assumptions . 1 . if @xmath94 is a constant , then @xmath289 is identically distributed as @xmath290 . we hence have@xmath291 and the result is clear . 2 . if @xmath97 is constant , then:@xmath292 , \ ] ] where @xmath293 for @xmath294 . moreover , the respective cumulated intensities of @xmath290 and @xmath295 are @xmath296 and @xmath297 , with@xmath298 due to the super - additivity of @xmath296 . we derive from ( * ? ? ? * theorem 6.b.40 , example 6.b.41 ) that @xmath299 for all @xmath300 , where @xmath301 stand for the standard stochastic order . as @xmath302 is non decreasing with respect to each @xmath303 , we get that:@xmath304 \geq\mathbb{e}\left [ f_{n}\left ( \left ( t_{i}^{(t)}\right ) _ { i=1}^{n}\right ) \right]\ ] ] for each @xmath305 . setting @xmath306 , we derive by lebesgue s dominated convergence theorem that@xmath307 = \phi_{s}\left ( l\right ) \geq \lim\limits_{n\rightarrow+\infty}\mathbb{e}\left [ f_{n}\left ( \left ( t_{i}^{(t)}\right ) _ { i=1}^{n}\right ) \right ] = \phi_{s}^{\left ( t\right ) } \left ( l\right ) , \ ] ] which achieves the proof . the conditions of the previous theorem means that 1 . the probability for a shock to be non fatal decreases with time ( and @xmath94 is constant ) , 2 . the cumulated rate of shocks arrivals is larger at time @xmath9 than at time @xmath308 ( and @xmath97 is constant ) . such conditions hence mean that the environment is more and more stressing , or that it is more stressing after a while than at the beginning . such conditions are quite natural . we here study the influence on the lifetime @xmath101 of different parameters of the stressing environment : we study the influence of probability @xmath309 , of the dependence between @xmath148 and @xmath139 and of the cumulated intensity function @xmath258 . we hence look at the influence on the lifetime @xmath101 of all characteristics of the stressing environment which make the components dependent . the influence of @xmath310 is straightforward . we mention it for sake of completeness . let us consider two different systems with identical parameters except from @xmath310 ( first system ) and @xmath311 ( second system ) and such that @xmath312 for all @xmath313 . then , adding a tilde ( @xmath314 ) to any quantity referring to the second system , we directly get from @xmath111 that @xmath315 or equivalently that @xmath101 is smaller than @xmath316 in the sense of the standard stochastic order @xmath317@xmath318 as expected , the lifetime @xmath101 is hence increasing with the probability @xmath310 for a shock to be non fatal . we here study the influence of the dependence between the two marginal increments @xmath148 and @xmath139 on the lifetime @xmath101 . to measure the dependence level between @xmath148 and @xmath139 , we use the lower ( or upper ) orthant order , where we recall that @xmath319 is said to be smaller than @xmath320 in the lower orthant order ( @xmath321 ) if @xmath322 or equivalently if @xmath323 [ prop_dependence]let us consider two different systems , with identical parameters except from @xmath324 ( first system ) and @xmath325 ( second system ) . as previously , a tilde ( @xmath314 ) is added to any quantity referring to the second system . assume that @xmath326 . then @xmath101 is smaller than @xmath316 in the sense of the standard stochastic order @xmath327 . the point is to show that @xmath328 . starting from @xmath117 and conditioning on @xmath329 , we have @xmath330 where @xmath331 for all @xmath332 , all @xmath333 , with @xmath334 based on the independence between @xmath324 and @xmath124 , it is sufficient to show that @xmath335 all @xmath332 , all @xmath333 to get @xmath328 . as @xmath336 is non decreasing in @xmath337 and @xmath338 , we first know from ( * ? ? ? * theorem 6.g.3 . ) that @xmath339 all @xmath340 . as @xmath341 and @xmath342 are two sequences of independent random vectors , we derive from the same theorem that @xmath343 as both functions @xmath344 and @xmath345 are non decreasing , we derive ( same theorem ) that @xmath335 which achieves this proof . the previous result shows that the more dependent @xmath148 and @xmath139 are , the larger the system lifetime is . we finally study how the lifetime of the system depends on the frequency of shocks . [ influence_lambda]let us consider two different systems , with identical parameters except from @xmath258 ( first system ) and @xmath346 ( second system ) . as previously , a tilde ( @xmath314 ) is added to any quantity referring to the second system . assume that @xmath347 , and that @xmath97 is non decreasing . then @xmath101 is smaller than @xmath316 in the sense of the standard stochastic order @xmath327 . using a similar method as for the proof of theorem [ propnbu ] , we can write @xmath116 as @xmath348 where @xmath349 is non decreasing with respect to each @xmath303 , all @xmath340 . as @xmath347 , we derive in the same way that @xmath350 for all @xmath300 , which allows to conclude . this result is very natural . the more frequent the shocks occur , the shorter the lifetime is . as already mentioned in subsection [ rel ] , a first possibility for computing the system reliability @xmath99 is to use classical monte - carlo ( mc ) simulations ( method 1 ) and simulate a large number of independent histories for the system up to time @xmath9 . we here provide the algorithm that we have used , considering the case where @xmath351 , all @xmath1 and where @xmath148 is almost surely positive ( @xmath352 ) ( not essential assumptions ) . [ algo]repeat @xmath216 times with @xmath216 large enough : 1 . simulate @xmath57 with given distribution . 2 . simulate @xmath353 according to the poisson distribution with parameter @xmath6 . 3 . simulate @xmath353 i.i.d . random variables @xmath354 with p.d.f . @xmath122}(x)$ ] . the shock arrival times are given by @xmath355 where @xmath356 is the order statistics of @xmath357 . 4 . simulate @xmath135 as the result of a bernoulli trial between a fatal ( 0 ) and a non fatal shock ( 1 ) at time @xmath17 , with probability @xmath358 for a shock to be non fatal , all @xmath359 . simulate @xmath353 i.i.d . random vectors @xmath360 , @xmath359 according to distribution @xmath195 . 6 . simulate the lifetime @xmath361 of the first component with conditional hazard rate @xmath362 given @xmath363 . with that aim , setting @xmath364 ( see @xmath365 ) , @xmath366 is a one - to - one increasing function from @xmath367 into @xmath368 and , for @xmath369 with @xmath370 , we have:@xmath371 it is then known that , if @xmath372 is uniformly distributed on @xmath373 $ ] , then @xmath374 is identically distributed as @xmath361 , see ( * ? ? ? * proposition 1.20 ) . 7 . compute @xmath375 where @xmath376 refers to the @xmath377th mc history , with @xmath378 . at the end of the algorithm , symbol @xmath379 stands for the realization of a bernoulli trial @xmath380 between an up ( 1 ) or down ( 0 ) system at time @xmath9 , with probability @xmath381 for the system to be up . the reliability @xmath381 is then classically approximated by the empirical mean @xmath382 of the @xmath379 s and a 95% asymptotic confidence interval is computed . method 2 is based on mc simulations of trajectories of @xmath383 and of @xmath124 ( see section [ rel ] ) . for both methods 1 & 2 , mc simulations are based on @xmath384 histories . methods 3 & 4 are described in section [ rel ] . the four methods are compared on a few specific examples . in all these examples , we compare the reliability at time @xmath385 ( @xmath386 ) and we suppose that the shock are due to a homogeneous poisson process with parameter @xmath387 , @xmath388 , @xmath95 and @xmath2 is a null process . all other parameters are provided in table [ table4 ] , where @xmath389 means that the random variable @xmath390 is exponentially distributed with mean 1 . [ c]cccccinput & @xmath391 & method & @xmath386 & 95 % ci + @xmath392 & & 1 & 0.5196 & [ 0.5181 0.5212 ] + @xmath393 , @xmath394 & independence & 2 & 0.5195 & [ 0.5184 0.5205 ] + & & 3 , 4 & 0.5198 & + @xmath395 & & 1 & 0.5049 & [ 0.5033 0.5064 ] + @xmath396 & complete dependence & 2 & 0.5049 & [ 0.5039 0.5059 ] + & & 4 & 0.5054 & + @xmath392,@xmath397 & & 1 & 0.4809 & [ 0.4793 0.4824 ] + @xmath92 , @xmath398 independent & dependence & 2 & 0.4813 & [ 0.4801 0.4825 ] + @xmath393 , @xmath399 & & & & + methods 1 and 2 may be used in any case . method 2 is more effective than method 1 ( shorter c.p.u . time and tighter 95% confidence interval - ic - ) . method 3 is more practical when @xmath92 and @xmath139 are independent with some specific distribution . method 4 is adapted to the case where @xmath92 and @xmath139 are dependent . we here illustrate several properties from a numerical point of view on a few examples . examples parameters are provided in table [ table3 ] . [ c]ccccccccc & @xmath3 & @xmath400 & @xmath57 & @xmath401 & @xmath94 & @xmath92 & @xmath139 & @xmath402 + ex.[ex9 ] & - & 0 & 0 & - & 1 & @xmath403 & @xmath403 & independent + ex.[ex8 ] & 2 & 0 & 0 & 1 & 1 & @xmath403 & @xmath403 & - + ex.[ex6 ] & 2 & 0 & 0 & @xmath404 & - & 1 & @xmath403 & independent + [ ex9 ] this example illustrates the nbu property of the lifetime when @xmath94 is constant and @xmath392 is non increasing , see theorem [ propnbu ] . taking @xmath388 , fig . [ fig7_1 ] indeed shows that the remaining lifetime of a system with age @xmath405 is stochastically smaller than the lifetime of a new system . on the contrary , when @xmath94 is still constant but @xmath406 is non decreasing , the remaining lifetime of a system with age @xmath405 is not comparable with that of a new system , see fig . [ fig7_2 ] with @xmath407 . so the nbu property does not hold anymore in that case . , nbu case ] , not nbu case ] [ ex8]we here consider two extreme cases for the dependency between @xmath148 and @xmath139 : independent or completely dependent ( here @xmath408 ) . the reliability in the completely dependent case is always greater than in the independent case ( fig . [ fig6 ] ) . this result is coherent with proposition [ prop_dependence ] . and @xmath139 , example [ ex8 ] ] [ ex6]this example shows the monotony of the reliability with respect to the intensity of the poisson process @xmath409 when @xmath97 is increasing , see proposition [ influence_lambda ] . the more frequently the shocks occur , the lower the reliability is ( fig . [ fig3 ] ) . , example [ ex6 ] ] we here proposed a random shock model with competing failure modes , which enlarges several models from the previous literature . the model takes into account different types of dependence between competing failures modes , where the dependence is induced by a common external shock environment . the reliability has been calculated by several different methods and conditions have been provided under which the system lifetime is new better than used . due to this ageing property , it might be of interest to propose and study some maintenance policy to enlarge the system lifetime . several versions might be proposed , according to the available information . also , the influence of the characteristics of the stressing environment on the lifetime @xmath101 has been studied . as expected , we saw that the lifetime was stochastically increasing with the probability @xmath310 for a shock to be non fatal . besides , and that result was not necessarily so clear at first sight , we saw that the lifetime was also stochastically increasing with the dependence between the two marginal shock sizes . finally , in case of a non decreasing function @xmath310 , we saw that the lifetime was stochastically decreasing with the cumulated frequency of shocks . this means that the more frequent the shocks occur , the shorter the lifetime is . this result is natural but our proof is limited to the case of a non decreasing function @xmath310 . in the special case of @xcite , the survival function of @xmath101 is however given by @xmath410 and it is easy to check that if @xmath411 ( stronger assumption than @xmath347 ) then @xmath412 , without any special condition on @xmath310 . so , the stochastic monotonicity of the lifetime with respect of the ( cumulated ? ) frequency of shocks might be true under a more general setting than in the present paper , without assuming any monotonicity condition on @xmath309 . we however have not been able to conclude on this point , and whether it is true or not remains an open question . both authors thank the referees for their carefull reading of the paper and their constructive remarks , which lead to a better introduction and justification of the model , and to a clearer paper . this work has been initiated during hai ha pham s phd studies in pau ( france ) , and has been supported by the conseil rgional daquitaine ( france ) . this work has also been supported for both authors by the french national research agency ( ammsi project , ref . anr 2011 bs01 - 021 ) . hao , c. su , and z .- z . reliability analysis for mechanical components subject to degradation process and random shock with wiener process . in _ 19th international conference on industrial engineering and engineering management _ , pages 531543 , 2013 . a. lehmann . joint modeling of degradation and failure time data . _ journal of statistical planning and inference _ , 1390 ( 5):0 1693 1706 , 2009 . special issue on degradation , damage , fatigue and accelerated life models in reliability testing . c. qian , s. nakamura , and t. nakagawa . cumulative damage model with two kinds of shocks and its application to the backup policy . _ journal of the operations research society of japan - keiei kagaku _ , * * 42**0 ( 4):0 501511 , 1999 .

the primordial cosmic microwave background ( cmb ) was generated when photons first decoupled from the baryonic fluid when the universe was only 400,000 years old . the vast majority of these photons travel unperturbed to the present day , and features of their angular power spectrum such as acoustic peaks and the damping tail @xcite record valuable information about cosmological parameters @xcite . baryons and dark matter evolve from small inhomogeneities at decoupling into increasingly complicated large - scale structure which can subtly perturb the observed pattern of cmb anisotropies . assuming that the primordial cmb is gaussian , non - gaussian correlations in the observed map can be used to reconstruct the intervening large - scale structure @xcite . in addition to the importance of learning about the large - scale structure itself , reconstruction of the weak - lensing potential generated by structure is essential to constraining tensor perturbations . weak lensing converts a fraction of the e - mode polarization generated by scalar perturbations at the last - scattering surface into b - mode polarization in the observed map . only by subtracting this b - mode polarization can one conclusively detect the primordial b - modes which serve as a model - independent signal of tensor perturbations @xcite . understanding lensing reconstruction requires a more detailed discussion of how weak lensing affects the cmb . weak gravitational lensing deflects the paths of cmb photons as they travel from the last - scattering surface to the observer . this deflection is accomplished by a projected lensing potential which is a weighted line - of - sight integral of the gravitational potential between the observer and the surface of last - scattering . at each point on the sky , lensing remaps the temperature and polarization to that of a nearby point at the last - scattering surface , the deflection angle being the gradient of the aforementioned projected lensing potential . assuming that this deflection angle is small , the temperature at any point can be expanded in a taylor series in the gradient of the lensing potential . in fourier space , this expansion appears as a series of convolutions of individual temperature and projected potential modes . the observed temperature - squared map in fourier space also appears as a convolution of individual fourier modes . subject to an overall normalization dependent on the scale of the fourier mode , these convolutions cancel in such a manner that each fourier mode of the temperature - squared map acts as an estimator for the _ same _ fourier mode of the projected lensing potential . lensing reconstruction as outlined above has been considered previously @xcite . in these works , two sources of noise were identified , and a filter of the temperature - squared map in fourier space was chosen to minimize the variance associated with lensing reconstruction subject to these noise sources . the first source is intrinsic signal variance ; the observed large - scale structure is one arbitrary member of an ensemble of realizations allowed by theory . the second source of noise , endemic to this method of lensing reconstruction , is a consequence of the nature of the primordial cmb . like the large - scale structure itself , the pattern of cmb anisotropies at the last - scattering surface is only one of many possible realizations allowed by theory . we do not know _ a priori _ which of these realizations nature has provided us , and this uncertainty hinders our ability to deconvolve the effects of lensing from true anisotropies at the last - scattering surface . even if the true pattern of anisotropies at the last - scattering surface was known , the finite amount of power in the cmb at small scales would still constrain lensing reconstruction . silk damping at the last - scattering surface suppresses cmb power at small scales , while the finite resolution of any real experiment would limit the detection of any signal that is present at small scales . lensing reconstruction fails below scales at which there is sufficient power , for the same reason that any remapping is indistinguishable given a uniform background . here , we consider a third source of noise neglected in previous studies . the filtered temperature - squared map is an unbiased estimator for the lensing potential in the approximation that a correlation between two given temperature modes is induced only by the single lensing mode whose wavevector is the sum of that of the two temperature modes . in actuality , any combination of two or more lensing modes whose wavevectors sum to this total induce correlations between the two temperature modes . there are many such combinations , but since these correlations add incoherently we do not expect a systematic bias . nonetheless , for estimators of each individual lensing mode we must use our knowledge of other lensing modes to subtract off this unwanted bias . this is an iterative process , and since our knowledge of the lensing map is imperfect it induces noise in lensing reconstruction . we calculate this additional variance for various estimators constructed from cmb temperature and polarization maps , and show how it compares to the dominant noise sources for an all - sky cmb experiment with a noise - equivalent temperature of 1 @xmath1k @xmath2 . since the lensing - potential power spectrum is a measure of the theoretical uncertainty with which we can predict the value of a given lensing mode , this noise associated with lensing reconstruction causes a systematic overestimation of the lensing - potential power spectrum . this systematic bias must be accounted for in order to compare observations with theoretical predictions . this paper is organized as follows . in [ s : lensing ] , we define the formalism we will use to explore the effects of weak lensing on the cmb . the taylor expansion of the lensed cmb map in gradients of the lensing potential is given in both real and fourier space , and the power spectra and trispectra of various components of the cmb temperature map are listed for later use . in [ s : estimators ] , we show that the fourier modes of the temperature - squared map when properly filtered can serve as estimators for the fourier modes of the displacement map with the same wave vector . using the power spectrum and trispectrum given in the preceding section , we calculate the variance associated with this estimator including a new component neglected in previous studies . this variance is evaluated numerically using the currently favored @xmath0cdm cosmological model with baryon density @xmath3 , matter density @xmath4 , cosmological constant density @xmath5 , hubble parameter @xmath6=0.65 , and power - spectrum amplitude @xmath7 . we then use the displacement estimator for individual fourier modes to construct an unbiased estimator for the lensing - potential power spectrum in [ s : power ] , and calculate the variance and covariance associated with this estimator . a few concluding remarks about the implications of our work for future studies are given in [ s : disc ] . the appendix contains useful formulae related to additional estimators of lensing based on polarization and a combination of temperature and polarization . we consider weak lensing under the flat - sky approximation following refs . @xcite . as discussed before @xcite , weak lensing deflects the path of cmb photons resulting in a remapping of the observed temperature pattern on the sky , @xmath8 \nonumber\\ & \approx & { \theta}({\hat{\bf n } } ) + \nabla_i { \phi}({\hat{\bf n } } ) \nabla^i { \theta}({\hat{\bf n } } ) + \frac{1}{2 } \nabla_i { \phi}({\hat{\bf n } } ) \nabla_j { \phi}({\hat{\bf n } } ) \nabla^{i}\nabla^{j } { \theta}({\hat{\bf n } } ) + \ldots\end{aligned}\ ] ] where @xmath9 is the unlensed primary component of the cmb in a direction @xmath10 at the last scattering surface . the observed , gravitationally - lensed temperature map @xmath11 in direction @xmath10 is that of the unlensed map in direction @xmath12 where @xmath13 represents the lensing deflection angle or displacement map . although a real cmb map will include secondary contributions such as the sz effect @xcite , we assume that such effects can be distinguished by their frequency dependence @xcite . they will not be further considered in this paper . a noise component denoted by @xmath14 due to finite experimental sensitivity must be included as well . thus the total observed cmb anisotropy will be @xmath15 . taking the fourier transform of the lensed map @xmath11 under the flat - sky approximation , we write @xmath16 where @xmath17 + \frac{1}{2 } { \int \frac{d^2 { \bf l''}}{(2\pi)^2}}{\phi}({{\mathbf{l^{\prime\prime } } } } ) \\ & & \quad \times { \phi}({{\mathbf{l}}}- { { \mathbf{l^{\prime}}}}- { { \mathbf{l^{\prime\prime } } } } ) \ , ( { { \mathbf{l^{\prime\prime}}}}\cdot { { \mathbf{l^{\prime } } } } ) \left [ ( { { \mathbf{l^{\prime\prime}}}}+ { { \mathbf{l^{\prime}}}}- { { \mathbf{l}}})\cdot { { \mathbf{l^{\prime}}}}\right ] + \ldots \ , . \nonumber\end{aligned}\ ] ] cmb correlations in fourier space can be described in terms of a power spectrum and trispectrum as defined in the usual manner , @xmath18 where the angle brackets denote ensemble averages over possible realizations of the primordial cmb , large - scale structure between the observer and the surface of last - scattering , and instrumental noise . the subscript @xmath19 denotes the connected part of the four - point function and the superscript @xmath20 denotes the temperature map being considered ( @xmath21 or @xmath22 ) . the lensing - potential power spectrum can be defined analogously , @xmath23 where here the angle brackets denote an average over all realizations of the large - scale structure . we make the assumption that primordial fluctuations at the last - scattering surface are gaussian . gaussian statistics are fully described by a power spectrum ; the gaussian four - point correlator , @xmath24 is zero . the instrumental noise @xmath22 is also assumed to be gaussian , as is the lensing potential @xmath25 . this second assumption is justified because the dominant contributions to the lensing potential come from intermediate redshifts @xmath26 at which linear theory holds . using these definitions , we can calculate the anticipated power spectrum and trispectrum of the observed cmb map . because the instrumental noise is uncorrelated with the signal , the power spectrum of the observed map is the sum of signal and noise power spectra , @xmath27 the power spectrum of the noise component is given by : @xmath28 where @xmath29 is the fraction of the sky surveyed , @xmath30 is the variance per unit area on the sky , and @xmath31 is the effective beamwidth of the instrument expressed in terms of its full - width half - maximum resolution @xmath32 . a cmb experiment that spends a time @xmath33 examining each of @xmath34 pixels with detectors of sensitivity @xmath35 will have a variance per unit area @xmath36 @xcite . the power spectrum of the lensed cmb can be determined by inserting eq . ( [ e : thetal ] ) into eq . ( [ e : specdef ] ) as discussed in @xcite , @xmath37 \ , c_l^{{\theta}{\theta } } + { \int \frac{d^2 { \bf l}_1}{(2\pi)^2 } } c_{| { { \bf l}}- { { \bf l}}_1|}^{{\theta}{\theta } } c^{\phi\phi}_{l_1 } [ ( { { \bf l}}- { { \bf l}}_1)\cdot { { \bf l}}_1]^2 \ , . \label{e : lenpow}\end{aligned}\ ] ] this result is given to linear order in the lensing - potential power spectrum @xmath38 . lensing neither creates nor destroys power in the cmb , but merely shifts the scales on which it occurs as seen by the fact that @xmath39 the observed cmb trispectrum can be calculated in a similar manner ; under our assumptions of gaussian instrumental noise and no secondary anisotropies the trispectrum of the lensed component @xmath40 is the sole contribution to the total observed trispectrum , @xmath41 [ ( { { { \bf l}_1}}+ { { { \bf l}_3 } } ) \cdot { { { \bf l}_4 } } ] + c^{{\phi\phi}}_{|{{\bf l}}_2+{{\bf l}}_3| } [ ( { { { \bf l}_2}}+{{{\bf l}_3 } } ) \cdot { { { \bf l}_3 } } ] [ ( { { { \bf l}_2}}+{{{\bf l}_3 } } ) \cdot { { { \bf l}_4 } } ] \big ] + \ , { \rm perm . } \ , . \nonumber \\ \label{e : trilens}\end{aligned}\ ] ] the term shown above is manifestly symmetric under the interchange @xmath42 , while the `` + perm . '' represents five additional terms identical in form but with the replacement of @xmath43 and @xmath44 with the other five combinations of pairs . the total trispectrum is symmetric under the interchange of any given pair as one would expect . having established a formalism within which to analyze weak lensing , we now consider the problem of reconstructing the lensing potential from an observed cmb temperature map . in this section , we examine lensing reconstruction following the approach of ref . @xcite , largely adopting their notation as well . the only important difference in notation is that we use @xmath40 to denote the lensed temperature field and @xmath45 for the unlensed field following ref . @xcite and most recent papers . ref . @xcite uses the opposite convention . for @xmath46 and to linear order in @xmath25 , @xmath47 where @xmath48 and @xmath49 . note that @xmath50 differs from the unmarked @xmath51 that first appeared in eq . ( [ e : specdef ] ) in that it denotes an ensemble average only over different gaussian realizations of the primordial cmb and instrument noise ; a fixed realization of the large - scale structure is assumed . for the purposes of estimating the large - scale structure actually realized in our observable universe , this is the appropriate average to take to ensure that our estimators are truly unbiased for a typical realization of the primordial cmb . when calculating the noise associated with lensing - potential estimators and again for power spectrum estimation in [ s : power ] , we will return to the full unmarked ensemble average . ( [ e : ea ] ) , an immediate consquence of eq . ( [ e : thetal ] ) , suggests that a temperature - squared map appropriately filtered in fourier space can serve as an estimator for the deflection field @xmath52 . hu and okamoto define five different estimators for the deflection field constructed from various combinations of the temperature and polarization ; we discuss the temperature - squared estimator in this section and relegate the analogous formulae for polarization estimators to the appendix . the minimum - variance temperature - squared estimator derived in ref . @xcite is @xmath53 where @xmath54 @xmath55^{-1 } \ , , \ ] ] and @xmath56 . substitution of eqs . ( [ e : ea ] ) , ( [ e : bias ] ) , ( [ e : filt ] ) , and ( [ e : norm ] ) into eq . ( [ e : est ] ) shows the desired result , @xmath57 namely that @xmath58 is indeed an unbiased estimator for the deflection field in fourier space . we now proceed to calculate the variance of this estimator . at first , we assume a complete knowledge of all lensing modes not examined by this estimator . in that case , we find @xmath59 where @xmath60 . evaluating the four - point function in the integrand of eq . ( [ e : estvar , cmb ] ) to second order in the lensing field , we obtain @xmath61 \nonumber \\ & & \quad \times \bigl [ { { \mathbf{l^{\prime}}}}\cdot ( { { { \bf l}_2}}- { { \mathbf{l^{\prime } } } } ) \bigr ] - \frac{1}{2 } { \int \frac{d^2 { \bf l'}}{(2\pi)^2}}{\phi}({{\mathbf{l^{\prime } } } } ) { \phi}(-{{\mathbf{l}}}- { { \mathbf{l^{\prime } } } } ) \bigl\ { c_{l_1}^{{\theta}{\theta } } ( { { { \bf l}_1}}\cdot { { \mathbf{l^{\prime } } } } ) \bigl [ { { { \bf l}_1}}\cdot ( { { \mathbf{l}}}+ { { \mathbf{l^{\prime } } } } ) \bigr ] + c_{l_2}^{{\theta}{\theta } } ( { { { \bf l}_2}}\cdot { { \mathbf{l^{\prime } } } } ) \bigl [ { { { \bf l}_2}}\cdot ( { { \mathbf{l}}}+ { { \mathbf{l^{\prime } } } } ) \bigr ] \bigr\ } \biggr ] \nonumber \\ & & \quad \times \biggl [ \bigl ( c_{l_{1}^{\prime}}^{{\theta}{\theta } } + c_{l_{1}^{\prime}}^{{\theta}{\theta}{{\rm n } } } \bigr ) ( 2 \pi)^2 \delta_{{\rm d}}({{\mathbf{l^{\prime } } } } ) + { \phi}({{\mathbf{l^{\prime } } } } ) { f_{\theta\theta}}({{{\bf l}_1 } } ' , { { { \bf l}_2 } } ' ) - { \int \frac{d^2 { \bf l'}}{(2\pi)^2}}c_{l^{\prime}}^{{\theta}{\theta } } { \phi}({{{\bf l}_1 } } ' - { { \mathbf{l^{\prime } } } } ) { \phi}({{{\bf l}_2 } } ' + { { \mathbf{l^{\prime } } } } ) \bigl [ { { \mathbf{l^{\prime}}}}\cdot ( { { { \bf l}_1 } } ' - { { \mathbf{l^{\prime } } } } ) \bigr ] \nonumber \\ & & \quad \times \bigl [ { { \mathbf{l^{\prime}}}}\cdot ( { { { \bf l}_2 } } ' + { { \mathbf{l^{\prime } } } } ) \bigr ] + \frac{1}{2 } { \int \frac{d^2 { \bf l'}}{(2\pi)^2}}{\phi}({{\mathbf{l^{\prime } } } } ) { \phi}({{\mathbf{l^{\prime}}}}- { { \mathbf{l^{\prime } } } } ) \bigl\ { c_{l_{1}^{\prime}}^{{\theta}{\theta } } ( { { { \bf l}_1 } } ' \cdot { { \mathbf{l^{\prime } } } } ) \bigl [ { { { \bf l}_1 } } ' \cdot ( { { \mathbf{l^{\prime}}}}- { { \mathbf{l^{\prime } } } } ) \bigr ] + c_{l_{2}^{\prime}}^{{\theta}{\theta } } ( { { { \bf l}_2 } } ' \cdot { { \mathbf{l^{\prime } } } } ) \bigl [ { { { \bf l}_2 } } ' \cdot ( { { \mathbf{l^{\prime}}}}- { { \mathbf{l^{\prime } } } } ) \bigr ] \bigr\ } \biggr ] { \rm + perm . } \nonumber \\\end{aligned}\ ] ] the terms given explicitly in eq . ( [ e:4pt , cmb ] ) correspond to the correlations between @xmath62 and @xmath63 and those between @xmath64 and @xmath65 . the `` + perm . '' stands for two additional terms , identical in form , arising from the pairings @xmath66 and @xmath67 . this expression indicates how uncertainty in the cmb at the last - scattering surface propagates into uncertainty in lensing reconstruction for a particular realization @xmath68 of the large - scale structure . to linear order in eq . ( [ e:4pt , cmb ] ) , correlations between the modes @xmath69 and @xmath64 are induced by those lensing modes whose wavevectors are the sums of any pair of wavevectors of these four modes . these lensing modes are precisely those forming the diagonals of the quadrilaterals depicted in fig . [ f:2quad ] . in practice , we do not know the large - scale structure between us and the last - scattering surface , so we assume a variance given by eq . ( [ e : lenpowdef ] ) with a model - dependent power spectrum @xmath70 . we must average eq . ( [ e : estvar , cmb ] ) over different realizations of the large - scale structure ( denoted by @xmath71 ) to obtain the total expected variance of our estimator , @xmath72 where @xmath73 the assumption that the lensing potential is gaussian imposes the constraint @xmath74 which closes the quadrilaterals of fig . [ f:2quad ] . the average of the four - point correlation function in eq . ( [ e : estvar ] ) can be calculated by further averaging eq . ( [ e:4pt , cmb ] ) over the large - scale structure . terms linear in the lensing field vanish when averaged over different realizations of the large - scale structure . quadratic terms in the lensing field arise as products either of two linear terms or of a zeroth and second - order term . averages over the product of two linear terms produce the connected part of the four - point correlation function , the trispectrum defined in eq . ( [ e : specdef ] ) . averages over the product of a zeroth and second - order term have no connected portion , but instead furnish an implicit dependence on @xmath70 in the total observed power spectrum of eq . ( [ e : temptot ] ) . the final result of averaging over the large - scale structure can be expressed in terms of the observed power spectrum and trispectrum , @xmath75 \ , , \nonumber \\\end{aligned}\ ] ] where the trispectrum can written in terms of @xmath76 as @xmath77 this form of the trispectrum is consistent with that of eq . ( [ e : trilens ] ) given directly in terms of power spectra . since @xmath78 , @xmath79 and the first term of eq . ( [ e:4pt ] ) vanishes . the remaining two terms containing pairs of delta functions , inserted into eq . ( [ e : estvar ] ) , yield the dominant contribution to the variance , @xmath80 \ , , \ ] ] where @xmath81 . notice that @xmath82 is zeroth order in the lensing potential @xmath25 ; it depends on the lensing potential power spectrum @xmath83 only implicitly though the total observed power spectrum @xmath84 . the ellipsis represents terms of higher order in @xmath83 that we now proceed to calculate . these terms arise from the trispectrum term of eq . ( [ e:4pt ] ) after @xmath85 is removed . substituting these results into eq . ( [ e : estvar ] ) , we find that to first order in @xmath83 , @xmath86 \ , , \ ] ] where @xmath87 is given by , @xmath88 the first - order contribution to the noise @xmath87 involves integrals over the lensing - potential power spectrum , and thus probes lensing modes with wave vectors different from that of the estimator @xmath58 . it can be interpreted physically as interference from these other modes in the determination of the mode @xmath89 being estimated . the filter @xmath90 was chosen to optimize the signal - to - noise ratio in the absence of the first - order contribution @xmath87 ; it is no longer an optimal filter once this additional noise is taken into account . as long as @xmath91 , the noise reduction that can be attained by re - optimizing our filter will not be significant . formulas analagous to those presented here relevant to the construction of estimators using polarization data are given in the appendix . the significance of @xmath87 for two different experiments is shown in fig . [ f : plrec ] using the currently favored @xmath0cdm cosmological model with baryon density @xmath3 , matter density @xmath4 , cosmological constant density @xmath5 , hubble parameter @xmath6=0.65 , and power - spectrum amplitude @xmath7 . the planck experiment is equivalent to a one - year , full - sky survey with temperature and polarization sensitivities of 12.42 and 26.02 @xmath92 respectively and resolution @xmath93 arcminutes as described in [ s : lensing ] . the reference experiment has the same resolution but superior sensitivities of 0.46 and 0.65 @xmath94 for temperature and polarization . these estimates of experimental parameters are identical to those given for the planck and reference experiments of ref . the @xmath95 and @xmath96 estimators have noise power spectra intermediate to those of the @xmath97 and @xmath98 estimators , while the @xmath99 estimator has substantially higher noise because the primordial cmb lacks true b - modes in the absence of inflationary gravitational waves . we see that for planck , with its comparatively inferior polarization sensitivity , the @xmath97 estimator will be best although it will be unable to detect individual fourier modes of the deflection field at the @xmath100 level . the reference experiment , and further experiments with similar sensitivity and even higher resolution , should be able to push @xmath100 detection of individual @xmath89 modes to @xmath101 by primarily relying on the @xmath98 estimator . in these cases the secondary noise @xmath102 is only smaller than the dominant noise @xmath103 by a factor of a few , whereas for higher sensitivity ( noisier ) experiments like planck it is smaller by at least an order - of - magnitude . this illustrates an interesting point , apparent from fig . [ f : plrec ] , that the zeroth - order noise @xmath82 declines dramatically with decreasing sensitivity until it becomes dominated by cosmic variance while the @xmath87 is largely unaffected by instrument sensitivity . the reasons for this trend are that instrument noise appears in @xmath81 through its contribution to the denominator of @xmath104 as shown by eqs . ( [ e : filt ] ) and ( [ e : norm ] ) . decreasing instrument noise raises the value of @xmath104 thereby lowering @xmath82 . by contrast instrument noise is reflected in @xmath87 through its effects on both @xmath105 and @xmath104 as shown in eq . ( [ e:1st ] ) . smaller instrument noise raises @xmath104 as before , driving @xmath87 up in this case , but this is compensated by a decrease in @xmath105 which appears as a prefactor outside the integrals . these two effects largely cancel each other out , rendering @xmath87 remarkably insensitive to instrument noise . although complete reconstruction of the deflection field @xmath89 can be an enormously powerful tool , such as for b - mode subtraction @xcite , for some purposes estimates of the lensing - potential power spectrum @xmath106 are sufficient . this power spectrum is a model - dependent prediction of theories of large - scale structure formation , and therefore estimates of the power spectrum from real data could be used to test these theories as well as the consistency of other determinations of cosmological parameters . furthermore , since estimates of all the modes @xmath89 with @xmath107 can be combined to estimate @xmath106 , @xmath100 detection of the power spectrum can be pushed to much higher @xmath108 than can that of individual modes . the deflection - field estimator @xmath58 derived in the preceding section can be used to construct an estimator for @xmath106 . our first guess for an appropriate lensing - potential power spectrum estimator is @xmath109 where @xmath110 is the area of the sky surveyed and @xmath111 is an annulus of radius @xmath108 and width @xmath112 . we ensemble average our estimator over different realizations of the cmb and large - scale structure using eq . ( [ e : prodave ] ) by bringing @xmath113 from the left to the right - hand side . this yields @xmath114 \ , .\ ] ] the definition of the dirac delta function , @xmath115 implies that @xmath116 . furthermore , in the limit that @xmath112 is small compared to the scales on which @xmath117 is varying , we can evaluate the integrand of eq . ( [ e : naive2 ] ) at its central value @xmath118 and extract it from the integral . the integral over the annulus @xmath111 cancels the factor @xmath119 in the denominator , reducing eq . ( [ e : naive2 ] ) to @xmath120 \ , .\ ] ] @xmath121 is indeed an estimator for the lensing - potential power spectrum @xmath106 , albeit a biaed one . note that the _ bias _ in the power - spectrum estimator @xmath121 is precisely the same as the _ variance _ shown in eq . ( [ e : prodave ] ) with which we were able to determine each individual lensing mode . this is no coincidence ; it reflects the fact that there are no grounds _ a priori _ on which to differentiate the variance with which we can reconstruct individual modes @xmath89 from the intrinsic variance @xmath122 of the underlying distribution from which they are drawn . to obtain an unbiased estimator to compare with theoretical predictions , we subtract off this unwanted reconstruction variance , @xmath123 \ , .\ ] ] since @xmath87 as defined in eq . ( [ e:1st ] ) itself depends on @xmath106 , this subtraction and evaluation must be performed iteratively until a self - consistent solution is obtained . the variance of our estimator @xmath124 can be calculated in the usual manner , @xmath125 evaluating this expression requires us to calculate @xmath126 \left [ { { \mathbf{d}}_{\theta\theta}}({{{\bf l}_2 } } ' ) \cdot { { \mathbf{d}}_{\theta\theta}}(-{{{\bf l}_2 } } ' ) \right ] \rangle \ , .\ ] ] since @xmath58 is a quadratic estimator in the temperature map , eq . ( [ e : dl2 ] ) includes the following integral over the eight - point correlation function in fourier space , @xmath127 \left [ { { \mathbf{d}}_{\theta\theta}}({{{\bf l}_2 } } ' ) \cdot { { \mathbf{d}}_{\theta\theta}}(-{{{\bf l}_2 } } ' ) \right ] \rangle = \frac{{a_{\theta\theta}}^2(l_{1}^{\prime } ) { a_{\theta\theta}}^2(l_{2}^{\prime})}{(l_{1}^{\prime})^2 ( l_{2}^{\prime})^2 } { \int \frac{d^2 { \bf k}_1}{(2\pi)^2 } } { \int \frac{d^2 { \bf k}_3}{(2\pi)^2 } } { \int \frac{d^2 { \bf k}_5}{(2\pi)^2 } } { \int \frac{d^2 { \bf k}_7}{(2\pi)^2 } } \nonumber \\ & & \quad \times \big\ { { f_{\theta\theta}}({{{\bf k}_1 } } , { { { \bf k}_2 } } ) { f_{\theta\theta}}({{{\bf k}_3 } } , { { { \bf k}_4 } } ) { f_{\theta\theta}}({{{\bf k}_5 } } , { { { \bf k}_6 } } ) { f_{\theta\theta}}({{{\bf k}_7 } } , { { { \bf k}_8 } } ) \langle { \theta}^{{\rm t}}({{{\bf k}_1 } } ) { \theta}^{{\rm t}}({{{\bf k}_2 } } ) { \theta}^{{\rm t}}({{{\bf k}_3 } } ) { \theta}^{{\rm t}}({{{\bf k}_4 } } ) { \theta}^{{\rm t}}({{{\bf k}_5 } } ) { \theta}^{{\rm t}}({{{\bf k}_6 } } ) { \theta}^{{\rm t}}({{{\bf k}_7 } } ) { \theta}^{{\rm t}}({{{\bf k}_8 } } ) \rangle \big\ } \ , , \nonumber \\\end{aligned}\ ] ] where @xmath128 , @xmath129 , @xmath130 , and @xmath131 . a fully general eight - point correlation function consists of a connected part , as well as terms proportional to the product of lower - order correlation functions . under the assumption that both the primordial cmb and the lensing potential are governed by gaussian statistics , all correlation functions higher than the four - point have vanishing connected parts @xcite . the temperature eight - point correlation function will therefore be composed of three groups of terms ; membership in a group being determined by whether the term contains zero , one , or two factors of the trispectrum . since the trispectrum given in eq . ( [ e : trilens ] ) is first order in the lensing - potential power spectrum @xmath106 , terms of these three groups are zeroth , first , and second order respectively in @xmath106 . combinatorics determines the number of terms in each group . there are : @xmath132 different ways of dividing @xmath133 into four pairs , and hence there will be 105 terms in the group containing no trispectra . similar calculations reveal that there are @xmath134 terms in the second group and @xmath135 terms in the third group . many terms in all three groups will vanish for the same reason that the first term vanished in the four - point correlation function of eq . ( [ e:4pt ] ) ; these terms are proportional to a dirac delta function evaluated at nonzero argument . consider now the first group of terms , those that are zeroth order in @xmath106 . the 60 nonvanishing terms in this group each contain four dirac delta functions ; they can be further segregated into the 12 terms that allow two of the integrals over @xmath136 appearing in eq . ( [ e:8pt ] ) to be immediately evaluated via dirac delta functions , and the 48 terms that allow evaluation of three @xmath136 integrals . the first 12 terms , inserted into eq . ( [ e:8pt ] ) and appropriately evaluated using the normalization of eq . ( [ e : norm ] ) , yield @xmath137 + \ldots \ , , \ ] ] while the remaining 48 terms give the final result to zeroth order in @xmath106 , @xmath138 \nonumber \\ & & \quad \quad \quad \quad \quad \quad + \frac{(2 \pi)^2}{a l^4 } \frac{2}{(2 \pi l \delta l)^2 } \int_{a_l } \frac{d^2 { { { \bf l}_1}}'}{(2 \pi)^2 } \ , \int_{a_l } \frac{d^2 { { { \bf l}_2}}'}{(2 \pi)^2 } \ , \frac{{a_{\theta\theta}}^2(l_{1}^{\prime } ) { a_{\theta\theta}}^2(l_{2}^{\prime})}{(l_{1}^{\prime})^2 ( l_{2}^{\prime})^2 } { \int \frac{d^2 { \bf k}_1}{(2\pi)^2 } } f_{{\theta}{\theta}}({{{\bf k}_1 } } , { { { \bf k}_2 } } ) p({{{\bf k}_1 } } , { { { \bf k}_2 } } , { { { \bf l}_1 } } ' , { { { \bf l}_2 } } ' ) \ , , \nonumber \\\end{aligned}\ ] ] where @xmath139 when @xmath140 is subtracted from @xmath141 in eq . ( [ e : sigpv ] ) , the first term in the square brackets of eq . ( [ e : zeropv ] ) will be eliminated . minimizing the variance associated with this estimator then consists of making an optimal choice of @xmath142 . the first noise term is proportional to @xmath143 . for a survey of area @xmath110 , @xmath144 is the specific area of an individual mode in @xmath145 space and @xmath146 is the area in @xmath145 space over which the power - spectrum estimator takes an average . this ratio is therefore the inverse of the number of individual @xmath58 modes whose inverse variances are added to determine the inverse variance of @xmath124 . it is obviously minimized by choosing @xmath147 . the second term , that involving @xmath148 , differs from the first noise term in that a dirac delta function has been used to evaluate an additional @xmath136 integral rather than an annulus integral . since the integrands are of the same order , we expect the second noise term to be suppressed relative to the first by a factor @xmath149 where @xmath150 is set by the resolution @xmath32 of the survey . under the conservative assumption @xmath151 , namely that we are probing scales well above our resolution , this term is assured to be small . we neglect such terms for the remainder of this paper . if we insert the portions of the eight - point correlation function that are first and second order in @xmath106 into eq . ( [ e:8pt ] ) and evaluate using eq . ( [ e:1st ] ) , we find @xmath152 \ , , \ ] ] and @xmath153 this result agrees with that given in ref . @xcite after subtracting our newly derived term @xmath87 . the term @xmath87 and corresponding terms for polarization - based estimators have two principal effects on power spectrum estimation . as shown in eq . ( [ e : finvar ] ) , they provide a fractional contribution to the variance of roughly @xmath154 when @xmath122 dominates the variance as in the reference experiment in the right - hand panel of fig . [ f : plrec ] . for the @xmath0cdm cosmological model considered here this represents an increase of @xmath155 in the variance of the @xmath98 estimator for @xmath156 . more importantly , @xmath87 acts as a bias for the naive estimator @xmath121 as shown by eq . ( [ e : naive3 ] ) . if this bias is not calculated and subtracted iteratively to form the unbiased estimator @xmath124 as in eq . ( [ e : pestsub ] ) , @xmath122 will be _ systematically _ overestimated by @xmath157 at low @xmath108 and by increasingly larger amounts at @xmath158 as the signal @xmath159 begins to plummet while @xmath87 remains comparatively flat . having evaluated the variance of our estimator @xmath124 , we consider whether this estimator has a substantial covariance @xmath160 the estimator @xmath121 as defined in eq . ( [ e : naive ] ) implies that : @xmath161 \left [ { { \mathbf{d}}_{\theta\theta}}({{{\bf l}_2 } } ' ) \cdot { { \mathbf{d}}_{\theta\theta}}(-{{{\bf l}_2 } } ' ) \right ] \rangle \ , , \ ] ] which can be evaluated using the same integral over the eight - point correlation function described in eq . ( [ e:8pt ] ) . whereas 60 of the 105 zeroth - order terms in @xmath106 coming from this equation were nonvanishing for the variance , for the covariance only 52 terms are nonzero provided that the widths @xmath112 and @xmath162 are chosen so that the annuli @xmath111 and @xmath163 do not overlap . this leads to a result analogous to eq . ( [ e : zeropv ] ) , @xmath164 note that the eight terms missing from the covariance when compared to the variance have altered the first term of eq . ( [ e : zerocv ] ) , and that it was precisely these terms that provided the dominant contribution to the variance of eq . ( [ e : finvar ] ) when @xmath142 was chosen appropriately . we therefore find that to zeroth order in @xmath106 , the covariance is given by @xmath165 where we have extracted the @xmath166 from the annular integrals since they are slowly varying over the widths @xmath112 and @xmath162 . for the same reasons that terms of this form were a subdominant contribution to the variance as discussed previously , we expect the covariance to be suppressed as well . if we define the ratio @xmath167 we can quantify this suppression . the triple integral of eq . ( [ e : fincov ] ) , appearing in the numerator of @xmath168 , involves integration over annuli with radii @xmath108 and @xmath169 and one integration over all fourier space . the triple integrals in the variances @xmath170 and @xmath171 appearing in the denominator of @xmath168 each consist of a single integration over an annulus of radius @xmath108 and @xmath169 respectively and two integrations over all fourier space . if we make the crude assumption that the integrand is constant , the ratio @xmath168 will simply be the ratio of these areas , @xmath172 the ratio @xmath168 is evaluated numerically for planck in fig . [ f : cov ] as a function of @xmath108 for various fixed values of @xmath169 . the estimators @xmath124 and @xmath173 were chosen such that @xmath174 , while integrals over fourier space were cut off at @xmath175 . substituting these values into eq . ( [ e : rcovest ] ) , we expect @xmath176 . this crude estimate is surprisingly close to the numerically obtained results of fig . [ f : cov ] ; in particular the slope of the curves is approximately 1/2 on this log - log plot . even at @xmath101 , @xmath177 suggesting that covariance in power - spectrum estimation can safely be neglected for planck . the estimate of eq . ( [ e : rcovest ] ) implicitly depends on the experimental resolution @xmath32 because the integrand appearing in expressions for the variance and covariance decreases rapidly for @xmath178 . for future experiments with better resolution than planck , @xmath179 will be higher implying by eq . ( [ e : rcovest ] ) that covariance will be even more negligible . weak gravitational lensing induces non - gaussian correlations between modes of the observed cmb temperature map as shown in eq . ( [ e : ea ] ) . these correlations , and assumptions about the gaussian nature of the primordial cmb , can be used to construct several temperature and polarization - based estimators of the fourier modes @xmath89 of the deflection field . this procedure was outlined in ref . @xcite , however in calculating the noise associated with this reconstruction , an assumption was made that the observed temperature map was gaussian . in the presence of lensing this assumption is invalid ; when calculating the variance of quadratic estimators all permutations of the observed trispectrum must be taken into account . one such permutation reflects the desired correlation making our estimator sensitive to @xmath89 , but the remaining two permutations induce additional variance proportional to the lensing - potential power spectrum @xmath106 . while subdominant , this variance will become increasing significant for future experiments as shown in the right - hand panel of fig . [ f : plrec ] . since the power spectrum @xmath106 is itself a measure of uncertainty in the deflection field , this additional variance in lensing reconstruction acts as a bias during power - spectrum estimation because there is no _ a priori _ way to distinguish it from the intrinsic variance of the underlying distribution . our calculation of the dependence of this variance on @xmath106 allows it in principle to be subtracted iteratively , which will prevent a systematic @xmath157 overestimate of @xmath106 at low @xmath108 . we close by considering several possible observational obstacles to the scheme for lensing reconstruction and power - spectrum estimation presented above . one hindrance is secondary contributions to the cmb such as the sz and isw effects . these effects increase the total temperature power spectrum appearing in the denominator of the optimum filter @xmath104 of eq . ( [ e : filt ] ) as would additional instrumental noise . they also correlate with the large - scale structure at low redshifts inducing further non - gaussian couplings and additional variance to lensing reconstruction . fortunately for our purposes the frequency dependence of the thermal sz effect differs from that of a blackbody . it can therefore be separated in principle from the lensed primordial cmb by an experiment with several frequency channels @xcite . the isw effect can not be removed in this manner , but is too small to significantly inhibit lensing reconstruction . polarization - dependent secondary effects are expected to appear at higher orders in the density contrast @xcite , and we therefore anticipate that they will not make a contribution at the levels considered here . a potentially more serious problem is that of galactic foregrounds , which though uncorrelated with the lensing signal may be substantial at certain frequencies . significant polarization has also been observed in some of these sources @xcite . we hope to understand and minimize the effects of galactic foregrounds in future work , and to pursue further refinements of lensing reconstruction . we thank wayne hu for useful discussions . this work was supported in part by nasa nag5 - 11985 and doe de - fg03 - 92-er40701 . kesden acknowledges the support of an nsf graduate fellowship and ac acknowledges support from the sherman fairchild foundation . here we provide the appropriate formulas for deriving the variance associated with polarization - based estimators of the deflection field @xmath89 . the cmb polarization can be decomposed into e and b - modes @xcite . these modes are mixed by weak lensing such that to linear order in @xmath68 , @xmath180 \phi({{\bf l}}-{{\bf l}}_1 ) \left [ ( { { \bf l}}-{{\bf l}}_1 ) \cdot { { \bf l}}_1 \right ] \ , , \nonumber \\ \tilde b({{\bf l } } ) & = & b({{\bf l } } ) - { \int \frac{d^2 { \bf l}_1}{(2\pi)^2 } } \left[e({{\bf l}}_1 ) \sin 2 ( \varphi_{{{\bf l}}_1 } -\varphi_{{\bf l } } ) + b({{\bf l}}_1 ) \cos 2 ( \varphi_{{{\bf l}}_1 } -\varphi_{{\bf l}})\right ] \phi({{\bf l}}-{{\bf l}}_1 ) \left [ ( { { \bf l}}-{{\bf l}}_1 ) \cdot { { \bf l}}_1 \right ] \ , .\end{aligned}\ ] ] we can exploit the sensitivity of the polarization modes to the lensing potential to construct lensing estimators from quadratic combinations of polarization modes . we generalize eq . ( [ e : ea ] ) to arbitrary combinations @xmath181 of @xmath45 , e , and b - modes as first derived in ref . @xcite , @xmath182 where @xmath183 \cos 2 ( \varphi_{{{\mathbf{l } } } } -\varphi_{{\mathbf{l^{\prime } } } } ) \ , , \nonumber \\ f_{eb}({{\mathbf{l } } } , { { \mathbf{l^{\prime } } } } ) & = & \left [ c_{l}^{ee } ( { { \mathbf{l}}}\cdot { { \mathbf{l } } } ) + c_{l^{\prime}}^{bb } ( { { \mathbf{l}}}\cdot { { \mathbf{l^{\prime } } } } ) \right ] \sin 2 ( \varphi_{{{\mathbf{l } } } } -\varphi_{{\mathbf{l^{\prime } } } } ) = f_{be}({{\mathbf{l^{\prime } } } } , { { \mathbf{l } } } ) \ , , \nonumber \\ f_{bb}({{\mathbf{l } } } , { { \mathbf{l^{\prime } } } } ) & = & \left [ c_{l}^{bb } ( { { \mathbf{l}}}\cdot { { \mathbf{l } } } ) + c_{l^{\prime}}^{bb } ( { { \mathbf{l}}}\cdot { { \mathbf{l^{\prime } } } } ) \right ] \cos 2 ( \varphi_{{{\mathbf{l } } } } -\varphi_{{\mathbf{l^{\prime } } } } ) \ , . \nonumber \\\end{aligned}\ ] ] in deriving these results we used parity considerations to demand @xmath184 . using these relations , we follow the approach of ref . @xcite to derive symmetric lensing estimators @xmath185 and @xmath186 { f_{xx'}}({{{\bf l}_1 } } , { { { \bf l}_2 } } ) \ , .\ ] ] we have explicitly symmetrized our estimators for @xmath187 to simplify the form of the optimal filter . the normalization bias of the estimators is removed by choosing @xmath188^{-1 } \ , , \ ] ] and @xmath189 { f_{xx'}}({{{\bf l}_1 } } , { { { \bf l}_2 } } ) \right]^{-1 } \ , .\ ] ] the minimum - variance filters @xmath190 for the various cases @xmath181 are given by @xmath191 using these optimal filters for the estimators defined in eqs . ( [ e : estxx ] ) and ( [ e : estxx ] ) , we can calculate the variances for these estimators in a fashion entirely analogoug to eq . ( [ e : estvar ] ) , @xmath192 @xmath193 \left [ x^{{\rm t}}({{{\bf l}_1 } } ' ) x'^{{\rm t}}({{{\bf l}_2 } } ' ) + x'^{{\rm t}}({{{\bf l}_1 } } ' ) x^{{\rm t}}({{{\bf l}_2 } } ' ) \right ] \rangle { f_{xx'}}({{{\bf l}_1 } } , { { { \bf l}_2 } } ) { f_{xx'}}({{{\bf l}_1 } } ' , { { { \bf l}_2 } } ' ) \nonumber \\ & & \quad \quad - ( 2 \pi)^2 \delta_{{\rm d}}({{\mathbf{l}}}- { { \mathbf{l^{\prime } } } } ) c_{l}^{dd } \ , . \nonumber \\\end{aligned}\ ] ] as for that of the temperature estimator , these variance will consist of zeroth - order terms in @xmath106 , @xmath194 and @xmath195 , and first - order terms , @xmath196 @xmath197 \nonumber \\ & & \quad \quad \quad + c_{|{{{\bf l}_1}}-{{{\bf l}_2}}'|}^{\phi\phi } \bigl [ f_{xx'}(-{{{\bf l}_1 } } , { { { \bf l}_2 } } ' ) f_{x'x}(-{{{\bf l}_2 } } , { { { \bf l}_1 } } ' ) + f_{xx}(-{{{\bf l}_1 } } , { { { \bf l}_2 } } ' ) f_{x'x'}(-{{{\bf l}_2 } } , { { { \bf l}_1 } } ' ) \nonumber \\ & & \quad \quad \quad \quad + f_{x'x'}(-{{{\bf l}_1 } } , { { { \bf l}_2 } } ' ) f_{xx}(-{{{\bf l}_2 } } , { { { \bf l}_1 } } ' ) + f_{x'x}(-{{{\bf l}_1 } } , { { { \bf l}_2 } } ' ) f_{xx'}(-{{{\bf l}_2 } } , { { { \bf l}_1 } } ' ) \bigl ] \big\ } \ , , \nonumber \\\end{aligned}\ ] ] the filters given in eq . ( [ e : filtxx ] ) are no longer optimal in the presence of this additional noise , but the difference between these filters and the optimal filters should be negligible provided that @xmath198 , @xmath199 . for the purposes of power - spectrum estimation , the terms @xmath200 and @xmath201 are not only an additional contribution to the variance , but are also a systematic bias if not subtracted iteratively following eq . ( [ e : naive3 ] ) . a final point to consider is that the six different estimators @xmath202 defined in this paper are not independent , as they are constructed from only three distinct maps . the covariance matrix for the six estimators will therefore not be diagonal , and this needs to be taken into account if the estimators are to be linearly combined to produce a single minimum - variance estimator . the off - diagonal elements of the covariance matrix can be evaluated in a straightforward manner involving pairs of double integrals similar to those of eqs . ( [ e : xxvar ] ) and ( [ e : xxvar ] ) . p. j. e. peebles and j. t. yu , astrophys . j. * 162 * , 815 ( 1970 ) ; r. a. sunyaev and ya . b. zeldovich , astrophys . space sci . * 7 * , 3 ( 1970 ) ; j. silk , astrophys . j. * 151 * , 459 ( 1968 ) ; w. hu and s. dodelson , ann . astrophys . * 40 * , 171 ( 2002 ) . g. jungman _ et al . _ , d * 54 * , 1332 ( 1995 ) ; j. r. bond , g. efstathiou , and m. tegmark , mon . not . r. astron . soc . * 291 * , l33 ( 1997 ) ; m. zaldarriaga and d. n. spergel , astrophys . j. * 488 * , 1 ( 1997 ) ; d. j. eisenstein , w. hu , and m. tegmark , astrophys . j. * 518 * , 2 ( 1999 ) . d. n. spergel and d. m. goldberg , phys . d * 59 * , 103001 ( 1999 ) ; d. m. goldberg and d. n. spergel , phys . rev . d * 59 * , 103002 ( 1999 ) ; a. cooray and w. hu , astrophys . j. * 534 * , 533 ( 2000 ) ; m. zaldarriaga and u. seljak , phys . rev . d * 59 * , 123507 ( 1999 ) ; h. v. peiris and d. n. spergel , astrophys . j. * 540 * , 605 ( 2000 ) . g. giardino _ et al . _ , astrophys . * 387 * , 82 ( 2002 ) ; f. r. bouchet , s. prunet , and s. k. sethi , mon . not . r. astron . 302 * , 663 ( 1999 ) ; s. prunet , s. k. sethi , and f. r. bouchet , mon . not . r. astron . 314 * , 348 ( 2000 ) .

diffractive scattering through strong interactions without any large momentum transfer has historically been described in terms of the exchange of a pomeron , a virtual hadron - like object with vacuum quantum numbers . the regge approach @xcite of the pre - qcd era provides a working phenomenology to describe such processes in a hadron basis since no parton structure is resolved . the idea @xcite to introduce a hard scale in a diffractive process opened the possibility to examine these processes at the level of quarks and gluons in the modern framework of qcd . the discovery of such _ diffractive hard scattering _ was made by the ua8 experiment @xcite by observing high-@xmath3 jets in single diffractive events in @xmath4 collisions at cern . many other hard processes in diffractive events have been observed later on , for a review see e.g.@xcite . of special importance are the measurements of diffractive deep inelastic scattering ( ddis ) at the electron - proton collider hera @xcite , where the well understood point - like electromagnetic interaction probes the parton structure in the diffractive reaction mechanism . this has resulted in theoretical descriptions of data based on essentially two different approaches , one being pomeron exchange in regge phenomenology and the other colour screening via soft gluon exchange in qcd . the pomeron approach starts with the initial proton state fluctuating in a soft non - perturbative process into a proton and a pomeron . the latter is assumed to have a partonic structure which is probed by the momentum transfer @xmath5 of the deep inelastic photon exchange to produce the hadronic final state , well separated in rapidity from the final state proton carrying most of the longitudinal momentum of the beam proton . hera data on ddis can then be described @xcite as a product of an effective pomeron flux factor from regge phenomenology and deep inelastic scattering on the pomeron having parametrised parton density functions ( pdf ) @xcite . alternatively , one may parametrise diffractive structure functions without an explicit pomeron flux but instead being conditionally dependent on the momentum of the final proton . however , this approach is not universal in the sense that such parametrisations do not reproduce diffractive hard scattering data in hadron - hadron collisions . for example , such pomeron pdfs overestimate substantially the cross - sections for diffractive hard scattering processes , such as production of jets or @xmath6 , at the tevatron @xcite . this has called for introducing a gap survival suppression factor @xmath7 , which can be given a qualitative theoretical motivation but which is difficult to calculate quantitatively . the colour exchange approach starts instead with the hard scattering process and then adds softer gluon exchanges to achieve the effective colour singlet exchange in diffractive processes . thus , the underlying hard process is assumed to be the same as in the corresponding non - diffractive process and its momenta naturally not affected by other exchanges at much lower momentum transfer scales . however , the formation of confining colour fields may well be affected by the softer gluon exchanges and thereby the hadronisation process such that a different distribution of the final state hadrons emerges . for example , when different colour singlet string - fields emerge separated in rapidity they will hadronise into two hadron systems separated by a rapidity gap with no hadrons . a simple , but phenomenologically rather successful model of this kind is the soft colour interaction ( sci ) model @xcite added to the lepto @xcite and pythia @xcite monte - carlo event generators . a large variety of diffractive data could then be reproduced with essentially the same value of a single new parameter , introduced to give the probability for exchange a soft colour octet gluon between any pair of partons . this colour exchange alter the formation of the colour string - fields and hence the application of the conventional lund string hadronisation model results in a different topology of the hadronic final state . the model gives an essentially correct description of diffractive dis @xmath8 scattering at hera @xcite as well as diffractive events at the tevatron having jets or quarkonia @xcite or gauge bosons @xcite . it has therefore been applied for predictions , e.g. of diffractive higgs production in double gap processes at lhc @xcite . the gap survival factor @xmath7 often used in other kinds of models , is not necessary here since the full event simulation accounts for such effects resulting in correct rates for the investigated diffractive processes in @xmath8 and @xmath9 and @xmath10 collisions . other models of a similar nature , with different forms of colour exchange , have been developed . the gal model @xcite for example considers soft colour exchanges between strings with a probability that favours minimization of the strings area in energy - momentum coordinates . a recent development of such colour string reconnection models makes a more elaborate account for @xmath11 colour statistics @xcite . a theoretical qcd basis for basic colour exchanges has been proposed in @xcite and , in a more elaborated form , in @xcite . the basic hard scattering process is treated by conventional perturbative qcd ( pqcd ) . its large momentum scale implies that it occurs on a small space - time scale compared to the bound state proton and is thus embedded in the proton . therefore , the emerging hard - scattered partons propagate through the proton s colour field and may interact with it . the amplitude for such multiple gluon exchanges is calculated in the eikonal approximation to all orders in perturbation theory resulting in an analytic expression for such a colour screening effect . the theoretical approach @xcite used here to resum multigluon exchange has similarities with the one in @xcite giving a similar eikonal factor @xmath12 in the amplitude , but there are also differences as will be discussed below . it is this amplitude that we here develop into a probabilistic model . we implement it for two different monte carlo event generators : lepto for general deep inelastic lepton - nucleon scattering including first order qcd matrix elements and parton showers based on conventional @xmath13 evolution from the dglap equations @xcite , and as well for cascade @xcite specialised on small-@xmath0 electron - proton scattering based on the off - shell @xmath14 first order matrix element and @xmath2 factorisation with ccfm evolution @xcite and unintegrated gluon density of the proton . the paper is organized as follows . in section [ sec : dynamic - rescattering ] we describe the basic process with the colour screening . section [ sec : ddis - via - dcs ] discusses the resulting cross - section for diffractive deep inelastic scattering and its monte carlo implementation . section [ sec : res ] shows our results in comparison to hera data . finally , section [ sec : conclusions ] presents our conclusions . in the virtue of the sci model , the skeleton of both inclusive and diffractive ( with rapidity gaps and/or leading proton ) dis process is provided by the same perturbative qcd diagram illustrated in fig . [ fig : ddis ] . a parton with longitudinal momentum fraction @xmath15 in the initial proton at the starting scale @xmath16 for pqcd is evolved to smaller momentum fractions , but higher transverse momenta and virtualities up to a hard scale @xmath17 . here , a virtual photon @xmath18 with momentum @xmath19 and virtuality @xmath20 resolves a quark at bjorken @xmath0 @xmath21 at small @xmath0 the process will dominantly be initiated by a gluon , which can radiate and splits into a @xmath22 pair . the total momentum fraction taken from the proton is @xmath23 , and the total mass @xmath24 of the parton system denoted as @xmath25 in fig . [ fig : ddis ] is @xmath26 for a @xmath27 pair ( without additional gluons ) having with quark transverse momentum @xmath2 and longitudinal momentum fraction @xmath28 , the corresponding invariant mass is @xmath29 with a subsequent rescattering of the @xmath30 dipole off the target colour field ( left panel ) . schematic diagram of the diffractive dis process @xmath31 accounting for final state rescattering by multiple gluon exchange at @xmath32 and perturbative parton shower off initial state parton which builds up the diffractive system @xmath25 ( right panel ) . the latter can be separated from the leading proton ( or small - mass system @xmath33 ) by a rapidity gap . the final state radiation is not shown as it does not affect the overall kinematics of the @xmath25 system.,title="fig:",scaledwidth=50.0% ] with a subsequent rescattering of the @xmath30 dipole off the target colour field ( left panel ) . schematic diagram of the diffractive dis process @xmath31 accounting for final state rescattering by multiple gluon exchange at @xmath32 and perturbative parton shower off initial state parton which builds up the diffractive system @xmath25 ( right panel ) . the latter can be separated from the leading proton ( or small - mass system @xmath33 ) by a rapidity gap . the final state radiation is not shown as it does not affect the overall kinematics of the @xmath25 system.,title="fig:",scaledwidth=35.0% ] at large @xmath34 , the parton distribution functions ( pdfs ) of the incident proton are dominated by valence quarks leaving practically no chance for the proton to survive such an interaction , and hence resulting in a non - diffractive event . at small @xmath34 , however , the pdfs are dominated by gluons , and the partonic system @xmath25 is created in photon - gluon fusion @xmath35 as depicted in fig . [ fig : ddis ] . in this case , the momentum exchange via multiple soft gluons with a small net fraction @xmath36 between the proton and the perturbative @xmath25 systems does not significantly change the momenta of partons emerging from the hard scattering . however , they do change the colour structure of the resulting @xmath25 and @xmath33 systems . the original sci model captures the main effects in many processes as ddis at hera and diffractive hard scattering at the tevatron . despite this success , it is not derived from a perturbative qcd amplitude . for the case of ddis , a derivation of the amplitude for the colour screened process has been done based on perturbative qcd in the large @xmath37 limit @xcite and provides a theoretical basis for the ddis process in terms of colour exchanges . it improves on the previous description by introducing a dependence on the kinematical details of the event which also leads to colour transparency . we start with the outline of the resummed colour screening amplitude and the derivation of probability for an event - by - event treatment in an event generator . consider the ddis amplitude in impact parameter representation in the target rest frame which corresponds to the colour dipole picture of the process . the lowest fock component of the virtual photon @xmath38 corresponds to its fluctuation to a @xmath30 dipole with transverse separation @xmath39 in the colour background field of the traget proton at impact distance @xmath40 from its centre . the prepared fock component then propagates through the field in the proton and softly interacts with it such that it can , in principle , change its colour but not kinematics ( the dipole size is frozen at the time scale of its propagation through the colour medium ) . we consider the forward limit where the total transverse momentum @xmath41 of gluon exchanges in the @xmath42-channel is small , @xmath43 . in this limit , as a straightforward consequence of the optical theorem in the limit of large @xmath44 c.m . energy , the ddis amplitude @xmath45 can be written with the ordinary gluon - initiated inclusive dis amplitude @xmath46 and the dynamical colour screening ( dcs ) amplitude @xmath47 as @xmath48 where @xmath49 is the relative quark transverse momentum in the @xmath30 dipole in the lowest order subprocess @xmath50 . the screening amplitude @xmath51 accounts for the soft gluon exchanges between the proton remnant @xmath33 and the rest of the final state commonly denoted as @xmath25 , with @xmath52 at the lowest order . these exchanges carry a small longitudinal fraction @xmath53 and the transverse momentum transfer @xmath54 is at a soft scale @xmath55 . @xmath51 is resummed to all orders in the large-@xmath56 limit where it acquires a simple eikonal form @xcite . @xmath57 here , @xmath58 is the colour factor for the single gluon exchange amplitude and @xmath59 is the effective coupling constant at the soft hadronic scale @xmath55 . the effective qcd coupling is not small in this case . several approaches dealing with the landau singularities at low momentum transfers were proposed in the literature , e.g. @xcite . in practice , we use the infrared - stable analytic perturbation theory ( apt ) approach @xcite . for our study we use similarly the inclusive amplitude @xmath60 in impact parameter space the fraction of the cross section with colour screening between the systems @xmath25 and @xmath33 is obtained from the ratio @xmath61 which defines the probability function for the overall colour singlet exchange . with @xmath62 , this leads to @xmath63 to apply this as the probability in a monte - carlo generated event at the parton level , we will associate @xmath64 and @xmath65 , and approximate by taking the average over the relative angle @xmath66 @xmath67 which is motivated by the fact that @xmath66 will be uniformly distributed in a sample of many collisions . for different values of @xmath68 ( upper to lower curve ) ] the resulting colour screening probability is shown in fig . [ fig : rescatter - prob - for - angles ] , for different choices of @xmath59 which enters as a normalisation factor . two important characteristic properties can here be observed . first , infrared safety with the probability levelling off at large @xmath69 , which resembles the saturation feature of the dipole scattering amplitude . second , vanishing colour screening probability for small dipoles @xmath70 , which is compatible with the colour transparency property . our basic theoretical approach @xcite for resumming multiple gluon exchanges has similarities with the theory developed in @xcite and resulting in an amplitude with an eikonal factor of the form @xmath12 in similarity with eq . ( [ eik - formula ] ) above . that approach can also be applied to both inclusive and diffractive dis starting from the same hard matrix element and employing factorization and resummation of soft @xmath42-channel gluons . despite these similarities , there are significant differences . in @xcite all exchanged gluons are treated on the same footing via a wilson line resumming them into a color singlet exchange . our approach separates one `` leading '' gluon , carrying the largest @xmath0-fraction , from the rest of the exchanged gluons which are much softer , carrying @xmath71 . these softer gluons are resummed to all orders and are required to be in a color octet state that matches the leading gluon to an overall color singlet exchange that provide the mechanism for rapidity gap formation . our leading gluon is treated via conventional @xmath2-factorisation in terms of the unintegrated gluon distribution in the proton , thus providing a model which is explicitly the same for diffractive and inclusive scattering without introducing any new kind of parton distribution functions , cf eq . ( [ sigmad ] ) below . this is in contrast with the formalism in @xcite which introduces diffractive parton distribution functions ( pdf ) containing non - perturbative dynamics and interpreted as the probability to find a parton with a momentum fraction @xmath72 in the proton , provided the proton emerges intact with momentum fraction @xmath73 . such diffractive pdfs can be analyzed theoretically @xcite which may introduce model parameters to be determined from data . naturally , one should study these , and other , theoretical approaches and via data find the optimal description of data to understand the rather complex diffractive processes in terms of basic theory and few free parameters . our approach has only two new parameters , which both have physical meanings that constrain their values to a rather narrow range . the following sections specify the monte carlo implementation of the model and show detailed numerical comparisons to data . the ddis cross section is then obtained using the inclusive cross section and standard inclusive parton densities together with the probability @xmath74 in eq . ( [ eq : prob - from - rb ] ) for dynamic colour screening resulting in @xmath75 where @xmath76 are the standard inclusive parton distributions . @xmath77 represents the differential distribution of the standard dis cross section in @xmath78 which is obtained from the parton evolution event - by - event in the monte carlo . since @xmath78 represents the transverse size of the @xmath22 together with the pqcd radiation and the amplitude for colour screening is dominated by a rescattering off large dipoles , we use the smallest @xmath2 difference within the partonic @xmath25-system and let @xmath79 . although this @xmath80 is typically related to the pqcd cutoff , the monte carlo simulation can give very small such relative @xmath2 due to random angular orientations of the momentum vectors . we therefore introduce a cut - off @xmath81 to avoid a spurious divergence and transverse sizes @xmath78 that are not perturbatively small . thus , we let @xmath82 . the impact parameter @xmath83 is related to the soft transverse momentum of the screening multiple gluon exchange , which is expected to be well below the factorisation scale for the pqcd processes . on the other hand @xmath84 is expected to be somewhat larger than the confining energy - momentum scale @xmath85{mev}$ ] in order for the screening process to occur fast enough that the proton state can stay quantum mechanically coherent into the final state . the colour screening probability therefore depends on the ratio @xmath69 given by @xmath86 where @xmath81 , as mentioned , regulates the divergence . the values of @xmath84 and @xmath81 constitute the two free parameters of the model and are to be determined from a comparison with experimental data . from the construction , we expect their values to be approximately between @xmath87 and the perturbative cutoff @xmath88 in the gluon pdf . using eq . ( [ eq : def - rb - ratio ] ) in eq . ( [ eq : prob - from - rb ] ) results in a probability @xmath89 for the effective colour screening that depends on the internal kinematics of the system @xmath25 . we calculate the diffractive reduced cross section @xmath90 within the same kinematic limits as applied by the experiment @xcite . in addition , we adopt two different notions of the diffractive cross section . the first definition @xmath91 is based on a forward remnant system with a mass @xmath92{gev}$ ] and proton quantum numbers . the other definition @xmath93 requires a large rapidity gap ( lrg ) of two units in pseudorapidity . this choice is potentially sensitive to the inner radiation structure of the system @xmath25 because the lrg is defined in terms of pseudorapidity . the emissions in the ccfm evolution @xcite are not strongly ordered in virtuality as they are by assumption in the dglap evolution @xcite . therefore , the parton in the hard interaction can no longer be approximated as on - shell and instead an off - shell matrix element is used for the hard interaction . likewise , the branching gluons in the ccfm evolution are described by an unintegrated gluon density function ( ugdf ) . in the ccfm evolution , one effectively resums the leading logarithms in the energy splitting @xmath94 and @xmath95 , as well as the leading logarithms in @xmath5 . experimental data on the ddis cross section @xmath96 covers a wide kinematic range where , in particular , one can have @xmath97 . because of such potentially very different hard scales , we expect corrections from large logarithms to become important in certain parts of the phase space . because of eq . ( [ mxdef ] ) we expect the ccfm evolution to be better suited for the ddis observables , at least , in the case of @xmath98 when the large leading logs @xmath99 are properly treated . because of the soft divergence in the first order qcd matrix element , the hard process @xmath100 will favour an uneven splitting of the energy between the quarks . specifically , if we define the fraction of energy taken by the quark as @xmath28 , then the matrix elements have soft divergences as @xmath94 and @xmath95 for @xmath101 or @xmath102 , respectively . one aspect of the soft divergence is that it favors a rather large ratio between @xmath0 and the energy fraction @xmath103 of the parton entering the matrix element . specifically , in the matrix element @xmath100 one can have a large ratio @xmath104 without any additional radiation from the quark propagator . such a large @xmath105 is not accounted for in the initial state parton evolution . in the case of the matrix elements @xmath106 and @xmath107 plus dglap evolution this phase space is in part taken into account , but the dglap evolution does not resum potentially large @xmath108 . another aspect of the uneven splitting is that one of the quarks will be very forward in pseudorapidity in the lab frame when @xmath28 is close to the divergence . in this case , the quark can have a large enough forward momentum @xmath109 to populate the gap region and the event therefore does not contribute to the diffractive cross section as defined in terms of a lrg in spite of having a leading proton . the divergent behaviour is unphysical and is usually avoided with a cutoff . still , an inclusion of higher order effects may be important for the cross section and also for the inner structure within the @xmath25 system . in particular , the energy of the @xmath22 system may be shared by additional gluons and therefore significantly reduce the rapidity range of the final @xmath25-system and therefore can have an influence on the lrg observable . in comparison with the h1 data @xcite . the model prediction uses dynamic colour screening with the parameters and and the cascade event generator with ccfm evolution . @xmath89 for the fitted parameters is shown in the upper - left corner . diffractive events in the model are defined as having a remnant system @xmath33 with proton quantum numbers and invariant mass @xmath92{gev}$ ] . rows for different values of @xmath110 are offset by a factor @xmath111 as indicated on the figure . ] in order to study the dynamic colour screening in more detail , we interface the model with different monte carlo event generators . in particular , we employ the program lepto @xcite which offers first order qed and first order qcd matrix elements combined with dglap @xcite parton showering and collinear pdfs . as a second program we use cascade @xcite which offers @xmath112 with @xmath2-factorised off - shell matrix elements and ccfm evolution @xcite which is intended to account for potentially large logarithms of incident momentum fractions of radiated partons . the photon - gluon fusion matrix element @xmath113 illustrated also in fig . [ fig : ddis ] is the dominant contribution to the diffractive dis cross section at small @xmath0 . lepto includes this process as a first order qcd matrix element , as well as via a combination of the qed hard process @xmath114 augmented by a @xmath115 dglap splitting . cascade provides this process as an off - shell first order qcd matrix element , but not as a first order qed matrix element with parton splitting . after generating events on parton level using matrix elements augmented with initial and final state parton showers , we apply the colour screening model before any special treatment of the remnant . in particular , we do not allow any cluster fragmentation of systems with a small invariant mass because the dynamic screening will potentially change the colour topology of the event before the scale of hadronization is reached and therefore change the possible outcomes of the fragmentation . this is understood as colour rescattering to happen on the scale between the perturbative cutoff at @xmath116{gev}$ ] and @xmath87 of hadronisation . the remnant system is in a monte carlo program treated by a non - perturbative model . for the case that the perturbative interaction resolves a gluon , which is the class of events which potentially leads to diffraction , the remnant is usually split into a @xmath117 pair with a certain sharing of momenta . this splitting typically introduces a relative transverse momentum representing the fermi motion in the bound state proton which is given by a gaussian distribution with a width @xmath118 . this relative @xmath2 affects the later hadronisation which introduces an uncertainty for the prediction of a forward proton spectrum . in this work , we are interested in diffraction defined by a forward small - mass system with proton quantum numbers or a large rapidity gap , which is insensitive to whether the hadronisation model maps the small - mass forward remnant state to a proton state or a resonance . all plots for the diffractive cross section in this paper show the reduced cross section @xmath119 which is related to the cross section via @xmath120 with @xmath121 given as @xmath122 with the negligible nucleon mass @xmath123 . , @xmath124 and for different values of @xmath5 . as the hard scale increases , there is more phase space for initial state radiation available and we see more gluon branchings . as the number of gluon branchings increases , a larger part of the full perturbative event and specifically its @xmath110-value is described using ccfm evolution which leads to a more accurate differential cross section . ] as in fig . [ fig : cascade_cutsnone_m01alphas70 ] but model results using the lepto event generator based on matrix elements for all hard processes to order @xmath125 and parton showers based on dglap @xmath1 evolution . the parameters for the dynamic colour screening model obtained from the fit against data @xcite are and . ] but with a constant probability of @xmath126 fitted from this data . we observe that especially at small @xmath110 the resulting diffractive cross section has a significantly different slope with respect to @xmath5 and the description of data is not as good as in the case of the dynamic rescattering model . ] [ fig : cascade_cutsnone_m01alphas70 ] shows the diffractive cross section @xmath127 obtained with dynamic colour screening and the process @xmath128 from the cascade event generator with ccfm evolution . events are selected according to the forward small - mass system prescription which requires a remnant system @xmath33 with proton quantum numbers and invariant mass @xmath92{gev}$ ] . the two parameters of the rescattering model are fitted and we obtain and . these values are physically reasonable in the sense that both are between @xmath87 and @xmath88 , and that the typical transverse momentum scale @xmath84 of the proton background is smaller than the minimal transverse momentum scale @xmath81 of the partonic @xmath25 system . we note that there is an overall good agreement with experimental data over a very wide region of the kinematical space . this agreement is remarkable because the model does not introduce specialised diffractive parton distributions , but uses standard proton ugdfs as input and introduces only two new physically motivated parameters . nevertheless , we note that one specific kinematic region is not very well described , namely , where @xmath110 and @xmath5 are both small and @xmath34 is large . the discrepancy develops for @xmath129 and increases for decreasing @xmath5 in the range of a few @xmath130 . a qualitative understanding of the problem can here be obtained from the principles of parton evolution via gluon radiation . at small @xmath110 , meaning large @xmath131 compared to @xmath5 , large logarithms of @xmath132 become important . on the other hand , at small @xmath5 scales , the event is mainly described by the matrix element , whereas the shower activity is low , which leads to a relative damping of the cross section with respect to the data . this lower radiative activity is illustrated in fig . [ fig : branchings-3em2 - 17em3 ] in terms of the number of branchings in the parton evolution for different @xmath5 at a representative value of @xmath124 . at very small @xmath133 there is a very long evolution path from the initial gluon with momentum fraction @xmath34 to a much smaller bjorken-@xmath0 at the quark - photon vertex . the cascade event generator has the advantage to resum both @xmath1 and @xmath134 contributions in its ccfm - based treatment of initial - state gluon radiation off the incoming gluon . however , gluon emissions from the quark propagator between the @xmath115 vertex and the quark - photon vertex ( _ cf . [ fig : ddis]b ) can also generate large @xmath134 contributions , in particular for the longest total evolution paths at very small @xmath110 . this radiation from the quark propagator can not be generated in cascade since the ccfm equation only includes gluon radiation from gluons and not from quarks . the simulation process starts by using the hard matrix element for @xmath135 , and then the initial gluon is evolved down to the starting scale of the proton state . thus , at very small @xmath110 in particular , there can be a substantial phase space available for gluon radiation off the quark propagator connecting to the virtual photon , which is not taken into account . gluon emission from the quark propagator can be taken into account by instead using the lepto monte carlo generator . it includes not only the hard matrix element for @xmath136 but also the other first order qcd process @xmath137 , where the gluon can be emitted from the incoming or outgoing quark , as well as the zeroth - order process @xmath114 . for both these processes , additional gluon radiation from the initial quark may occur through the initial - state parton showering , which in lepto is generated through conventional dglap evolution in @xmath1 , but without a resummation of @xmath134 contributions . [ fig : lepto_cutsnone ] shows the results of using lepto with the same dynamic colour screening model , with fitted parameter values and of expected magnitudes . the description of data is good in the inner region of the covered kinematic space , but with substantial discrepancies at small @xmath34 and very large @xmath110 . in the discussed problematic region , however , the agreement with data is better than that for cascade at large @xmath34 , small @xmath5 and small @xmath110 around 0.01 , although not at the very smallest @xmath138 . this indicates that , as discussed , emissions from the quark propagator is of some importance , but accounting for large @xmath134 contributions is also needed . no presently available event generator include all the mentioned effects that seems necessary in order to describe the diffractive cross - section over the entire kinematic region . the indication is , however , that combining all available theoretical formalisms in perturbative qcd , i.e. matrix elements and parton showers including @xmath1 and @xmath134 resummations , with the dynamical colour screening model presented here should provide a working description of the observed diffractive deep inelastic scattering process . constrained only by the requirement to be in the physically sensible range @xmath139 $ ] and to be fairly smooth . solid line : the probability from the fit of the colour screening model as used in fig . [ fig : cascade_cutsnone_m01alphas70 ] ( upper left corner ) . we note that both methods result in very similar functions for @xmath89 . ] it is interesting to compare the dynamic colour screening model with the results from having a fixed colour screening probability while keeping all other parameters equal . [ fig : cascade_cutsnone_const ] shows the diffractive cross section with @xmath126 , obtained by a fit to data . we note that the constant probability results in a significantly worse description of the data . the overall normalisation as well as the shape of @xmath96 with respect to @xmath5 is better described by the dynamic screening model . in the results from the dynamic model of fig . [ fig : cascade_cutsnone_m01alphas70 ] , we have fitted the two parameters @xmath81 and @xmath84 of the dynamic screening model to data . the parameters determine the overall normalisation and essentially the position of the slope where the screening probability @xmath89 falls off to zero . on the other hand , the general shape of this probability is given by the underlying model itself . it is interesting though to investigate what form of @xmath89 would result in a good fit without assuming an underlying model . to this end , we fit a mapping @xmath140 which is only constrained by the fact that it should lie in the physically sensible range @xmath141 $ ] and that it should be reasonably smooth . the result of such a fit is shown in fig . [ fig : prob - casc - m01-alphas70 ] . we observe that even though we did not place any particular constraint on the functional form , one obtains essentially the same result as in the dynamic screening model and therefore support to its qcd - basis . at the scales 1 , 4 and @xmath142{gev}$ ] for the solid , dashed and dotted lines , respectively , is shown in the left panel . the gluon density a1 for the same scales is shown in the right panel . the distribution starts out with a steeper slope at small scales and influences the observable as shown in fig . [ fig : cascade_gdf1013_cutsnone_m01alphas70 ] . ] at the scales 1 , 4 and @xmath142{gev}$ ] for the solid , dashed and dotted lines , respectively , is shown in the left panel . the gluon density a1 for the same scales is shown in the right panel . the distribution starts out with a steeper slope at small scales and influences the observable as shown in fig . [ fig : cascade_gdf1013_cutsnone_m01alphas70 ] . ] [ fig : gdf - plot-1013 ] but using a gluon density @xmath143 that increases stronger towards low-@xmath0 already at the starting scale ( a0 in @xcite ) . only the interesting subset of the kinematic plane is shown . by comparison with fig . [ fig : cascade_cutsnone_m01alphas70 ] it is seen that @xmath96 is at large @xmath110 and large @xmath34 sensitive to the shape of the gluon density at the starting scale . ] the result in fig . [ fig : cascade_cutsnone_m01alphas70 ] is obtained using the unintegrated gluon density @xmath144 illustrated in fig . [ fig : gdf - plot-1010 ] ( left ) . this density starts out flat at a low scale @xmath145 which can be compared with the @xmath146 behavior of the pomeron flux in regge - based models . the diffractive cross section in our model is sensitive to the slope especially in the kinematic region where @xmath110 is large . we can compare the main result in fig . [ fig : cascade_cutsnone_m01alphas70 ] with the result obtained by using a parton density which has a stronger increase towards small @xmath0 already at low scales , shown in fig . [ fig : gdf - plot-1010 ] ( right ) . the corresponding @xmath127 is shown in fig . [ fig : cascade_gdf1013_cutsnone_m01alphas70 ] . we note that especially the dependence of the cross section on @xmath5 is sensitive to the gluon density @xmath144 at small scales @xmath145 . by including diffractive data into the fit of a gluon density , this dependence could be used to further constrain the shape of the gluon distribution at low scales . but showing @xmath147 where the diffractive cross section is defined in terms of a large rapidity gap between @xmath148 and @xmath149 in pseudorapidity . ] range available in the experimental data is indicated by a horizontal thinner line for each row in @xmath110 . the fraction of events with a gluon in the lrg region is shown in the two rightmost plots . ] range available in the experimental data is indicated by a horizontal thinner line for each row in @xmath110 . the fraction of events with a gluon in the lrg region is shown in the two rightmost plots . ] [ fig : cascade_lrg40t60_m01alphas70 ] shows the diffractive cross section @xmath150 as defined by the presence of a large rapidity gap in the range @xmath151 in pseudorapidity . we note the close similarity to fig . [ fig : cascade_cutsnone_m01alphas70 ] . when requiring a minimum gap size @xmath152 in the event , the non - diffractive cross - section is exponentially suppressed by @xmath152 . while the presence of a lrg of size @xmath152 in the final state is a clear indication of diffraction , the exact location of the gap is sensitive to the description of the kinematics of the central system @xmath25 , and @xmath150 depends therefore additionally on the treatment of dynamics within the @xmath25-system . the dynamic colour screening model describes the probability of a colour screened interaction , but the additional soft exchanges do not change the kinematics of final state partons . we know on the other hand that our description in terms of matrix elements plus showering does not cover the full phase space . especially , the lack of additional gluon radiation from the quark propagator in the matrix element @xmath112 could be the reason why this @xmath22-system extends up to too high rapidity and can hadronize into the gap region . we therefore keep the gap size @xmath153 as in the data , but shift its starting value from 3.3 in data to 4 . this modification is irrelevant for our result at @xmath154 , but noticeable at @xmath155 where it causes at most a correction of @xmath156 at small @xmath110 and small @xmath5 , but becomes important at @xmath157 . we expect @xmath150 to be sensitive to higher order corrections especially close to the soft divergence @xmath94 where partons from the matrix element cause activity in the lrg region . this is illustrated in fig . [ fig : lrg - parton - veto ] where the fraction of events having a @xmath19 or @xmath158 with momentum vector into the lrg is shown . we note that the fraction of quarks in the lrg increases towards low @xmath5 which can be understood from the correlation between @xmath5 and the transverse momentum . on the other hand , the fraction of gluons in the lrg depends more strongly on @xmath34 and weaker on @xmath5 . this can be understood from the fact that gluons arise from the parton shower in contrast to the quarks which are defined by the matrix element . at the leading order of our computation , we note that the result is only mildly sensitive to the cuts employed to regulate the @xmath94 divergence . also , a comparison with massive matrix elements at leading order suggests that typical quark masses lead effectively to the same cuts on the energy sharing variable @xmath28 as used in our results . on the other hand , higher order corrections could significantly alter the internal event structure , especially in the region of the phase space where it is likely to have a large step @xmath28 between @xmath0 and @xmath103 . there , an additional final state gluon could modify the extension of the @xmath25 system in rapidity . an improved parton evolution based on ccfm with additional splittings @xmath159 and @xmath160 could therefore in principle improve the description further because the diffractive process at a low scale could be described with a qed matrix element and low @xmath0 resummed parton evolution . similarly , @xmath150 is also sensitive to the distribution of the transverse momenta and energy splitting in the parton shower . we have developed the probability for dynamic colour screening in dis in a way that can be used in monte carlo event generators and applied it with cascade and lepto . the resulting model predicts the diffractive dis cross section based on perturbative qcd matrix elements and standard inclusive parton densities in both collinear and @xmath2-factorisation approaches . this facilitates practical applications of previously obtained theoretical derivations of the amplitude for colour screening through semi - soft multiple gluon exchanges calculated in the eikonal approximation to all orders in perturbative qcd . the basic formalism gives a theoretical understanding why the soft colour interaction ( sci ) model has been phenomenologically successful , but goes beyond that model by leading to a colour screening probability that depends on the dynamics of the perturbative qcd parton dynamics . this dynamical screening probability exhibits a saturated behaviour at small transverse momenta of the emerging parton system as well as colour transparency at large transverse momentum . the monte carlo model has only two , physically motivated parameters . their values are obtained by fitting the hera diffractive cross section and found to be of the expected magnitude . the model successfully describes the data over a large kinematic range , significantly better than with a constant screening probability . interestingly , a fit of a free - form probability function results in the same shape as in our model , and hence gives support for our account of the basic qcd dynamics of relevance . it is noteworthy that we have not introduced any diffractive parton distribution functions to describe soft dynamics through unknown functions fitted to data . this explicitly demonstrates that diffractive and inclusive scattering can be described by the same basic qcd processes when explicit account is taken of color degrees of freedom in gluon exchanges at scales below normal cut - offs of @xmath161 gev of hard , perturbative qcd processes and @xmath162 of the hadronisation phase transition . in some kinematic regions there are two very different scales present , namely the invariant mass of the diffractive system @xmath131 and the photon virtuality @xmath5 . this calls for a resummation of large logarithms @xmath163 , or equivalently @xmath164 . to address this issue , we take the cross section in the @xmath2-factorisation approach from off - shell matrix elements and unintegrated gluon densities together with the ccfm evolution which provides a resummation of leading logarithms in @xmath165 . we show that this significantly improves the description of the diffractive hera data at @xmath166 corresponding to @xmath97 . nevertheless , there are residual deviations in the region where both @xmath110 and @xmath5 are at their lowest values . this may be attributed to the extreme region of very small @xmath110 when the quark propagator connected to the virtual photon have a significant phase space available for gluon radiation , which is not accounted for ; neither in the leading order matrix element for @xmath167 nor in the ccfm evolution that does not include the @xmath160 splitting . the dglap evolution does include this and also shows a slightly better result in this case , but is still not sufficient since here the effects of large @xmath168 is not included . to conclude and connect to the discussion in the introduction , our study has shown that the phenomenon of diffractive deep inelastic scattering can be described using a basic qcd - framework . the hard subprocess is treated in the same way as for non - diffractive events but a colour screening process occurs as a result of multiple gluon exchanges that are resummed to all orders . significant deviations from data occur in a special kinematic region , where potentially large logarithmic corrections are not yet fully included in available evolution equations for gluon radiation . still , the overall results show that gluonic colour screening in qcd is a viable approach to understand diffraction . j. r. forshaw and d. a. ross , cambridge lect . notes phys . * 9 * , 1 ( 1997 ) . g. ingelman and p. e. schlein , phys . b * 152 * , 256 ( 1985 ) . r. bonino _ et al . _ [ ua8 collaboration ] , phys . b * 211 * , 239 ( 1988 ) . g. ingelman , int . j. mod . a * 21 * , 1805 ( 2006 ) [ hep - ph/0512146 ] . s. chekanov _ et al . _ [ zeus collaboration ] , nucl . b * 816 * , 1 ( 2009 ) ; t. ahmed _ et al . _ , nucl . b * 429 * , 477 ( 1994 ) ; nucl . b * 435 * , 3 ( 1995 ) . f. d. aaron _ et al . _ [ h1 and zeus collaborations ] , eur . j. c * 72 * , 2175 ( 2012 ) [ arxiv:1207.4864 [ hep - ex ] ] . g. ingelman and k. prytz , z. phys . c * 58 * , 285 ( 1993 ) . t. affolder _ et al . _ [ cdf collaboration ] , phys . lett . * 84 * , 5043 ( 2000 ) . a. edin , g. ingelman and j. rathsman , phys . b * 366 * , 371 ( 1996 ) ; z. phys . c * 75 * , 57 ( 1997 ) . g. ingelman , a. edin and j. rathsman , comput . commun . * 101 * , 108 ( 1997 ) . t. sjstrand , s. mrenna and p. z. skands , jhep * 0605 * , 026 ( 2006 ) [ hep - ph/0603175 ] . a. edin , g. ingelman and j. rathsman , phys . d * 56 * , 7317 ( 1997 ) . r. enberg , g. ingelman and n. timneanu , phys . d * 64 * , 114015 ( 2001 ) . g. ingelman , r. pasechnik , j. rathsman and d. werder , phys . d * 87 * , no . 9 , 094017 ( 2013 ) [ arxiv:1210.5976 [ hep - ph ] ] . r. enberg , g. ingelman , a. kissavos and n. timneanu , phys . lett . * 89 * , 081801 ( 2002 ) . j. rathsman , phys . b * 452 * , 364 ( 1999 ) [ hep - ph/9812423 ] . s. j. brodsky , r. enberg , p. hoyer and g. ingelman , phys . d * 71 * , 074020 ( 2005 ) . r. pasechnik , r. enberg and g. ingelman , phys . d * 82 * , 054036 ( 2010 ) ; phys . b * 695 * , 189 ( 2011 ) . f. hautmann , z. kunszt and d. e. soper , phys . lett . * 81 * , 3333 ( 1998 ) ; nucl . b * 563 * , 153 ( 1999 ) . f. hautmann and d. e. soper , phys . d * 63 * , 011501 ( 2001 ) ; phys . d * 75 * , 074020 ( 2007 ) . v. n. gribov and l. n. lipatov , sov . j. nucl . * 15 * , 438 ( 1972 ) [ yad . fiz . * 15 * , 781 ( 1972 ) ] ; + g. altarelli and g. parisi , nucl . b * 126 * , 298 ( 1977 ) ; + y. l. dokshitzer , sov . jetp * 46 * , 641 ( 1977 ) [ zh . fiz . * 73 * , 1216 ( 1977 ) ] . h. jung _ et al . _ , eur . j. c * 70 * , 1237 ( 2010 ) . s. catani , f. fiorani and g. marchesini , nucl . b * 336 * , 18 ( 1990 ) ; + s. catani , f. fiorani and g. marchesini , phys . b * 234 * , 339 ( 1990 ) ; + s. catani , m. ciafaloni and f. hautmann , nucl . b * 366 * , 135 ( 1991 ) ; + g. marchesini , nucl . b * 445 * , 49 ( 1995 ) [ hep - ph/9412327 ] . b. andersson , g. gustafson , g. ingelman and t. sjstrand , phys . rept . * 97 * , 31 ( 1983 ) . s. j. brodsky , g. f. de teramond and a. deur , phys . d * 81 * , 096010 ( 2010 ) . d. v. shirkov and i. l. solovtsov , phys . * 79 * , 1209 ( 1997 ) . f. d. aaron _ et al . _ [ h1 collaboration ] , eur . j. c * 72 * ( 2012 ) 2074 [ arxiv:1203.4495 [ hep - ex ] ] .

x - ray observations of clusters of galaxies show that in the central regions of most clusters the cooling time of the intracluster gas is significantly less than a hubble time ( edge , stewart & fabian 1992 ; white , jones & forman 1997 ; peres 1998 ) . this cooling is thought to lead to a slow net inflow of material towards the cluster centre ; a process known as a cooling flow ( see fabian 1994 for a review ) . x - ray imaging data show that gas typically ` cools out ' throughout the central few tens to hundreds of kpc in clusters , with @xmath2 , where @xmath3 is the integrated mass deposition rate within radius @xmath4 ( thomas , fabian & nulsen 1987 ) . spatially resolved x - ray spectroscopy has confirmed the presence of distributed cooling gas in cooling flows , with a spatial distribution and luminosity in good agreement with the predictions from the imaging data and cooling - flow models ( allen & fabian 1997 ) . for some years , the primary uncertainty with the standard model for cooling flows was the fate of the cooled matter ( see fabian , nulsen & canizares 1991 ) . however , the discovery of large column densities of intrinsic x - ray absorbing material associated with cooling flows observed with the solid state spectrometer ( sss ) on the einstein observatory ( white 1991 ; johnstone 1992 ) opened one interesting possibility . follow - up spatially - resolved x - ray spectroscopy with the position sensitive proportional counter ( pspc ) on rosat ( allen 1993 ; irwin & sarazin 1995 ; allen & fabian 1997 ) confirmed the presence of intrinsic x - ray absorbing material in cooling - flow clusters and further showed this material to be distributed throughout , but centrally - concentrated within , the cooling flows . the x - ray data thus identify the intrinsic x - ray absorbing material as a plausible sink for the cooled gas deposited by the cooling flows . in this paper we examine the x - ray properties of the cooling flows in a sample of 30 of the most x - ray luminous ( @xmath5 ) clusters of galaxies known . using asca spectra and rosat high resolution imager ( hri ) data , we present independent determinations of the mass deposition rates in the cooling flows and measure the column densities of intrinsic x - ray absorbing material associated with these systems . we show that the cooling flow model provides a consistent description of the x - ray imaging and spectral data and that the observed masses of intrinsic x - ray absorbing material are in reasonable agreement with the masses expected to have been accumulated by the cooling flows over their lifetimes . we also discuss the evidence for mass deposition from the cooling flows in other wavebands . this is the final paper in a series that has examined the impact of cooling flows on mass measurements ( allen 1998 ) , the @xmath6 relation ( allen & fabian 1998a ) and metallicity measurements ( allen & fabian 1998b ) for x - ray luminous clusters . in this paper , we describe the data reduction and analysis procedures used in these works . a large number of other , previous studies have also analysed one or more of data sets included here ( see references in the papers listed above ) , although these studies have not , in general , examined the properties of the cooling flows in the clusters . exceptions are the deprojection analysis of peres ( 1998 ) , and the combined asca / rosat studies of allen ( 1996 ) , schindler ( 1997 ) , bhringer ( 1998 ) and rizza ( 1998 ) . the cosmological parameters @xmath7=50 , @xmath8 and @xmath9 are assumed throughout . our sample was identified from clusters contained in the rosat x - ray brightest abell - type cluster sample ( xbacs ; ebeling 1996 ) and rosat brightest cluster sample ( bcs ; ebeling 1998 ) with x - ray luminosities in the @xmath10 kev band exceeding @xmath11 . from this initial list , we selected for study those targets with asca x - ray spectra and rosat high resolution imager ( hri ) data available on the goddard space flight centre ( gsfc ) public archive as of 1997 july 15 . we supplemented this sample with a number of other , southern x - ray luminous clusters , known to exhibit strong gravitational lensing effects ( pks0745 - 191 , rxj1347.5 - 1145 , ms2137.3 - 2353 , ac114 ) and the distant , luminous cooling - flow clusters abell 1068 , 1704 and iras 09104 + 4109 . the primary goal of this project was to investigate the x - ray properties of cooling flows in x - ray luminous clusters and their effects on the integrated x - ray properties of their host systems . we have not included either the perseus ( abell 426 ) or coma ( abell 1656 ) clusters in our study , although these systems met our selection criteria , since both are too close ( @xmath12 and @xmath13 , respectively ) to allow their integrated cluster properties to be studied using the techniques used in this paper . detailed analyses of these clusters and other nearby cooling flows are discussed by allen ( 1999 ) . the clusters abell 370 , 115 and 1758 also met the selection criteria but were not included in our study since rosat hri images show them not be single , coherent structures but to consist of a number of merging subunits . the integrated x - ray spectra for such clusters will relate to the virial properties of the individual subclusters rather than the systems as a whole , and the assumptions of spherical symmetry and hydrostatic equilibrium required by the x - ray modeling will not apply . the x - ray images for the other clusters included in the sample do not exhibit any dramatic substructure that would suggest these assumptions to be invalid . our final sample consists of 30 clusters , spanning the redshift range @xmath14 , with a mean redshift of 0.21 . the asca ( tanaka , inoue & holt 1994 ) observations were made over a three - and - a - half year period between 1993 april and 1996 december . the asca x - ray telescope array ( xrt ) consists of four nested - foil telescopes , each focussed onto one of four detectors ; two x - ray ccd cameras , the solid - state imaging spectrometers ( sis ; s0 and s1 ) , and two gas scintillation imaging spectrometers ( gis ; g2 and g3 ) . the xrt provides a spatial resolution of @xmath15 arcmin half power diameter ( hpd ) in the energy range @xmath16 kev . the sis detectors provide good spectral resolution [ @xmath17 over a @xmath18 arcmin@xmath19 field of view . the gis detectors provide poorer energy resolution [ @xmath20 but cover a larger circular field of view of @xmath21 arcmin diameter . for our analysis we have used the screened event lists from the rev1 processing of the data sets available on the gsfc asca archive ( for a detailed description of the rev1 processing see the gsfc asca data reduction guide , published by gsfc . ) the asca data were reduced using the ftools software ( version 3.6 ) issued by gsfc , from within the xselect environment ( version 1.3 ) . further data - cleaning procedures as recommended in the asca data reduction guide , including appropriate grade selection , gain corrections and manual screening based on the individual instrument light curves , were followed . a summary of the asca observations , including the individual exposure times after all screening procedures were carried out , is provided in table 1 . spectra were extracted from all four asca detectors ( except in those few cases where the s1 data were lost due to saturation problems caused by flickering pixels in the ccds ) . the spectra were extracted from circular regions , centred on the peaks of the x - ray emission . for the sis data , the radii of the regions used were selected to minimize the number of chip boundaries crossed ( thereby minimizing the systematic uncertainties introduced by such crossings ) whilst covering as large a region of the clusters as possible . data from the regions between the chips were masked out and excluded . the final extraction radii for the sis data are summarized in table 2 . also included in that table are the chip modes used in the observations ( whether 1,2 or 4 chip mode ) and the number of chips from which the extracted data were drawn . for the gis data a constant extraction radius of 6 arcmin was used . [ we note that for abell 2142 , the 2 arcmin ( 3 arcmin ) radius region surrounding the x - ray bright seyfert-1 galaxy 1556 + 274 ( offset by @xmath22 arcmin from the x - ray centre of the cluster ) was masked out and excluded from the analysis of the sis ( gis ) data . ] for the gis observations , and sis observations of clusters in regions of low galactic column density ( @xmath23 ) , background subtraction was carried out using the ` blank sky ' observations of high galactic latitude fields compiled during the performance verification stage of the asca mission . for such data sets , the blank - sky observations provide a reasonable representation of the cosmic and instrumental backgrounds in the detectors . the background data were screened and grade selected in the same manner as the target observations and background spectra were extracted from the same regions of the detectors as the cluster spectra . for the sis observations of clusters in directions of higher galactic column density , background spectra were extracted from regions of the chips that were relatively free from foreground cluster emission . for the sis data , response matrices were generated using the ftools sisrmg software . where the spectra covered more than one chip , response matrices were created for each chip , which were then combined to form a counts - weighted mean matrix . for the gis analysis , the response matrices issued by gsfc on 1995 march 6 were used . the rosat hri observations were carried out between 1991 november and 1995 june . the hri provides a @xmath24 arcsec ( fwhm ) x - ray imaging facility covering a @xmath25 arcmin@xmath19 field of view ( david 1996 ) . reduction of the data was carried out with the starlink asterix package . x - ray images were created on a @xmath26 arcsec@xmath19 pixel scale , from which centres for the cluster x - ray emission were determined . where more than one observation of a source was made , a mosaic was constructed from the individual observations . for the cooling flow and intermediate clusters , the x - ray centres were identified from the peaks of the x - ray surface brightness distributions ( which are easily determined from the hri images ) . for the non - cooling flow clusters , the x - ray emission is not as sharply - peaked and for these systems we have identified the x - ray centres with the results from iterative determinations of the centroids within a circular aperture of fixed radius . for abell 665 , 2163 and ac114 a 1 arcminute radius aperture was used . for abell 2744 , 773 , 2218 and 2219 a 2 arcmin aperture was adopted , and for abell 520 a 3 arcmin aperture was used . for two of the clusters included in the sample , abell 665 and 1413 , rosat hri images were not available at the time of writing and for these clusters pspc imaging data have been used instead . a summary of the rosat observations and the x - ray centers for the clusters is given in table 3 . for the purposes of this paper , we have classified the clusters into subsamples of cooling - flow ( hereafter cf ) and non - cooling flow ( ncf ) systems . cfs are those clusters for which the upper ( 90 percentile ) limit to the central cooling time , as determined from the deprojection analysis of the rosat hri x - ray images ( section 4 ) , is less than @xmath27 yr . ( the ` central ' cooling time is the mean cooling time of the cluster gas in the innermost bin included in the deprojection analysis , which is of variable size . the use of a fixed physical size of 100 kpc for the central bin leads to similar results ; section 4 . ) using this simple classification scheme we identify 21 cfs and 9 ncfs in our sample . the mean redshift for the subsamples of both cf and ncf clusters is @xmath28 . the fraction of cfs in our sample is @xmath29 per cent , the same as that determined by peres ( 1998 ) from a study of a larger , flux - limited sample of clusters ( primarily at lower redshift ; @xmath30 ) using a similar classification scheme . the modeling of the x - ray spectra has been carried out using the xspec spectral fitting package ( version 9.0 ; arnaud 1996 ) . for the sis data , only counts in pulse height analyser ( pha ) channels corresponding to energies between 0.6 and 10.0 were included in the analysis ( the energy range over which the calibration of the sis instruments is best - understood ) . for the gis data only counts in the energy range @xmath31 were used . the spectra were grouped before fitting to ensure a minimum of 20 counts per pha channel , allowing @xmath32 statistics to be used . the spectra have been modeled using the plasma codes of kaastra & mewe ( 1993 ; incorporating the fe l calculations of liedahl , osterheld & goldstein 1995 , in xspec version 9.0 ) and the photoelectric absorption models of balucinska - church & mccammon ( 1992 ) . the data from all four asca detectors were analysed simultaneously ( except where the s1 data were lost due to chip saturation problems ) with the fit parameters linked to take the same values across the data sets . the exceptions to this were the emission measures of the ambient cluster gas in the four detectors which , due to the different extraction radii used and residual uncertainties in the flux calibration of the instruments , were allowed to fit independently . the spectra were examined with a series of spectral models . model a , consisted of an isothermal plasma in collisional equilibrium , at the optically - determined redshift for the cluster , and absorbed by the nominal galactic column density ( dickey & lockman 1990 ) . the free parameters in this model were the temperature ( @xmath33 ) and metallicity ( @xmath34 ) of the plasma and the emission measures in the four detectors . ( the metallicities are measured relative to the solar values of anders & grevesse ( 1989 ) , with the different elements assumed to be present in their solar ratios . ) the second model , model b , was the same as model a but with the absorbing column density @xmath35 also included as a free parameter in the fits . the third model , model c , was the same as model a but with an additional component explicitly accounting for the emission from the cooling flows . the material in the cooling flows is assumed to cool at constant pressure from the ambient cluster temperature , following the prescription of johnstone ( 1992 ) . the normalization of the cooling - flow component was parameterized in terms of a mass deposition rate , @xmath36 , which was a free parameter in the fits . the mass deposition rate is linked to take the same value in all four detectors , scaled by a normalization factor proportional to the total flux measured in that detector . the metallicity of the cooling gas was assumed to be equal to that of the ambient icm . the emission from the cooling flows was also assumed to be absorbed by an intrinsic column density , @xmath37 , of cold gas which was a further free parameter in the fits . the abundances of metals in the absorbing material were fixed to their solar values ( anders & grevesse 1989 ) although the effects of varying these values were examined . finally , a fourth spectral model , model d , was examined which was similar to model c but with the constant - pressure cooling flow replaced with a second isothermal emission component . the temperature and normalization of this second emission component were included as free parameters in the fits . ( as with model c , the normalizations were linked to the same value in all four detectors , scaled by the appropriate normalization factors . ) the second emission component was also assumed to be intrinsically absorbed by a column density of cold gas , @xmath37 , which was a free parameter in the fits . the metallicities of the two emission components were linked to take the same values . 1 shows the asca data and best - fitting spectral models for four of the clusters in the sample ; abell 1704 , 2029 , 2204 and 2219 . table 4 summarizes the fit results for all of the clusters using the four spectral models . for each cluster we list the best - fitting parameter values and 90 per cent ( @xmath38 ) confidence limits . the mass deposition rates ( @xmath39 ) , @xmath40 kev x - ray luminosities ( @xmath41 ) and bolometric luminosities ( @xmath42 ) are the values measured in the g3 detector . note that for the clusters at relatively low redshifts ( @xmath43 ) , we have corrected the @xmath41 and @xmath44 values for the emission arising at radii @xmath45 arcmin by scaling from the david ( 1993 ) results , which were based on ginga and einstein mpc observations . table 5 summarizes the goodness - of - fit measurements obtained with the different spectral models . the tabulated results are the probabilities of exceeding the @xmath32 values obtained , assuming in each case that the model correctly describes the spectral properties of the clusters . goodness - of - fit values @xmath46 may be regarded as indicating a formally unacceptable fit to the data . the results from the spectral analysis show that model a ( the isothermal model with the absorbing column density fixed at the nominal galactic value ) provides a reasonable description of most ( 21/30 ) of the clusters . however , model a fails significantly for the massive cf clusters abell 1068 , 1795 , 2029 and 2204 . the significance of the improvements to the fits obtained with spectral model b over model a have been evaluated using the f - test for the introduction of an additional free parameter ( bevington 1969 ) . these significances are listed in column 6 of table 5 . we see that the improvements obtained when including the column density as a free parameter in the fits with the isothermal models are significant at @xmath47 per cent confidence for 23 of the 30 clusters ( and would also be highly significant for abell 478 if the nominal dickey & lockman 1990 value for the galactic column density , rather than the x - ray value determined by allen & fabian 1997 were used in model a ) . model b provides a statistically acceptable fit for 27 of the 30 clusters . the results on the goodness - of - fit determined with spectral models c and d are summarized in columns 4 and 5 of table 5 . the only clusters that these models do not adequately describe are the two nearest , brightest systems ; abell 1795 and abell 2029 . spectral models c and d include an absorption component acting on the entire cluster emission which is normalized to the appropriate galactic column densities determined from hi studies ( dickey & lockman 1990 ) . we have examined the significance of the improvements to the fits obtained when allowing the galactic column density to also be a free parameter in the fits with these models . the fits were not significantly improved for any of the clusters except pks0745 - 191 , for which the galactic column density is most uncertain . ( the galactic column density to pks0745 - 191 varies between 3.2 and @xmath48 in the surrounding regions studied by dickey & lockman 1990 ) . we find @xmath49 with model c and @xmath50 with model d for pks0745 - 191 ( indicating improvements significant at @xmath51 per cent confidence ) when allowing the galactic column density to be free . the preferred value for the galactic column density for this cluster from the asca data is @xmath52 . for abell 1795 , we also find a marginal improvement to the fits with models c and d when allowing the galactic column density to be a free parameter ( @xmath53 with model c and @xmath54 for model d , which is significant at the 96.5 and 98 per cent level , respectively ) with a preferred value of @xmath55 . the agreement between the galactic column densities inferred from the 21 cm and x - ray data for most of the clusters with spectral models c and d provides further support for the validity of these models . the statistical improvements obtained by introducing an additional emission component into the fits with the single - temperature model ( model b ) are summarized in columns 7 and 8 of table 5 . we list the results obtained for both cases , where the extra emission component is modeled as a cooling flow ( for the cf systems only ) or a second isothermal emission component . ( essentially we use spectral models c and d , but also include the galactic column density as a free parameter it the fits so as to permit a direct comparison with the goodness of fit measured with spectral model b ) . the significance of the improvements to the fits obtained with the multiphase models with respect to model b have been quantified using an f - test for the introduction of 2 or 3 extra parameters ( for models c and d respectively ) . we find that the improvements with model d are significant at @xmath47 per cent confidence for 21 of the 30 clusters . the only clusters for which we find no significant improvement with the introduction of a second emission component are the ncf clusters abell 520 , 773 , 2218 and ac114 , and the more distant and/or less - luminous cf clusters abell 963 , ms1358.4 + 6245 , ms1455.0 + 2232 , ms2137.3 - 2353 and abell 2390 , which have amongst the lowest numbers of counts in their asca spectra . for the cf clusters ( the only clusters to which model c was applied ) the improvements obtained with the cooling - flow model over model b are significant at @xmath47 per cent confidence in 13 of 21 systems , with again those cf clusters with the lowest numbers of total counts in their asca spectra exhibiting the least - significant improvements . we thus find that most cf clusters , and a number of ncf systems ( in general those systems with the best signal - to - noise ratios in their asca spectra ) exhibit significant improvements to their fits with the use of multiphase , over single - phase , models . these improvements generally indicate the presence of emission from material cooler than the mean cluster temperatures ( see table 4 ) . for the cf clusters , this can be naturally understood as being due to emission associated with the cooling flows . for the ncf clusters , the improvements to the fits obtained with model d are likely to reflect the presence of merging subclusters with lower virial temperatures this is supported by optical , radio and x - ray imaging ( edge 1992 ; buote & tsai 1996 ; markevitch 1996 ; feretti , giovannini & bhringer 1997 ; rizza 1998 ) and gravitational lensing studies ( kneib 1995 ; smail 1995 , 1997 ; squires 1997 ; allen 1998 ) of the ncf clusters , which show that these systems typically exhibit complex morphologies and centroid shifts indicative of merger events . the results in table 4 show that spectral model d typically provides at least as good a fit to the asca spectra for the cf clusters as the cooling - flow model ( model c ) . this has sometimes been taken to indicate that the x - ray gas in these systems is distinctly two - phase , with the cooler phase being due to the dominant cluster galaxy ( makishima 1997 ; ikebe 1999 ) . however , when interpreting these results it is important to recall that the two - temperature model provides a more flexible fitting parameterization , with an extra degree of freedom , when applied to asca observations . simulated cooling - flow spectra constructed with spectral model c , including plausible levels of intrinsic absorption ( section 6 ) and observed at the spectral resolution and signal - to - noise levels typical of asca observations , are invariably well - described by two - temperature models . this is illustrated in fig . 2 where we show a simulated asca sis spectrum for a 7 kev cluster with a metallicity of 0.4@xmath56 , containing a cooling flow ( intrinsically absorbed by a column density of @xmath57 ) accounting for 30 per cent of the total @xmath40 kev flux . a count rate of 1ct s@xmath58 , an exposure time of 40ks and a galactic column density of @xmath0 have been assumed . overlaid , we show the best - fitting two - temperature model , which provides a good fit to the simulated data . thus , even in the case where the constant - pressure cooling flow model provides an exact description of the data ( _ i.e. _ in the simulations ) , the two - temperature model provides a similarly good fit . in a real cluster , where spectral model c will undoubtedly over - simplify the true situation , the more flexible two - temperature model ( model d ) is likely to provide a better match to the observations . thus , our finding that spectral model d often provides a better fit to the observed cluster spectra than model c , only shows that the constant pressure cooling flow model over - simplifies the true spectra of the cooling flows in the clusters ( see also section 6.4 ) . we note that such tendencies would be enhanced by the presence of additional cool components in the spectra , for example due to temperature gradients at large radii ( markevitch 1998 ) . finally , we note that the very high luminosities associated with the cool emission components in clusters like rxj1347.5 - 1145 ( with @xmath59 a few @xmath11 ) are difficult to explain with models in which the cool emission is due to the interstellar medium of the central cluster galaxies . the analysis of the hri imaging data has been carried out using an extensively updated version of the deprojection code of fabian ( 1981 ) . azimuthally - averaged x - ray surface brightness profiles were determined for each cluster from the hri images . these profiles were background - subtracted , corrected for telescope vignetting and re - binned to provide sufficient counts in each annulus to allow the analysis to be extended to radii of at least 500 kpc . ( bin sizes of 8 - 24 arcsec were used . ) with the x - ray surface brightness profiles as the primary input , and under assumptions of spherical symmetry and hydrostatic equilibrium , the deprojection technique can be used to study the basic properties of the intracluster gas ( temperature , density , pressure , cooling rate ) as a function of radius . the deprojection code uses a monte - carlo method to determine the statistical uncertainties on the results and incorporates the appropriate hri spectral response matrix issued by gsfc . the cluster metallicities were fixed at the values determined from the spectral analysis ( table 4 ) . the absorbing column densities were fixed at the appropriate galactic values ( table 1 ) . the deprojection code requires the total mass profiles for the clusters ( which define the pressure profiles ) to be specified . we have iteratively determined the mass profiles that result in deprojected temperature profiles ( which approximate the mass - weighted temperature profiles in the clusters ) that are isothermal within the regions probed by the hri data ( the central @xmath60 mpc ) and which are consistent with the spectrally - determined temperatures ( section 3 ) . the assumption of approximately isothermal mass - weighted temperature profiles in the central regions of the clusters is supported by the following evidence : firstly , asca observations of nearby cooling flows show that in the central regions of these systems the gas is multiphase , but that the bulk of the x - ray gas there has a temperature close to the cluster mean ( fabian 1994b ; fukazawa 1994 ; matsumoto 1996 ; ikebe 1999 ) . secondly , combined x - ray and gravitational lensing studies of cf clusters ( allen 1998 ) show that approximately isothermal mass - weighted temperature profiles in the cores of such clusters lead to excellent agreement between their x - ray and gravitational - lensing masses . thirdly , the use of approximately constant mass - weighted temperature profiles implies a more plausible range of initial density inhomogeneities in the clusters than would be the case if the temperature profiles decreased within their cores ( thomas , fabian & nulsen 1987 ) . finally , the use of approximately isothermal mass - weighted temperature profiles in the deprojection analyses leads to independent determinations of the mass deposition profiles in the cooling flows , from the x - ray spectra and imaging data , in excellent agreement with each other ( allen & fabian 1997 ) . we note that the assumption of a constant mass - weighted deprojected temperature profile is consistent with measurements of a decreasing emission - weighted temperatures in the cores of many cf clusters ( waxman & miralda - escud 1995 ) . the mass profiles for the clusters were parameterized as isothermal spheres ( equation 4 - 125 of binney & tremaine 1987 ) with adjustable core radii ( @xmath61 ) and velocity dispersions ( @xmath62 ) . the core radii were adjusted until the temperature profiles determined from the deprojection code became isothermal . the velocity dispersions were then adjusted until the temperatures determined from the deprojection code came into agreement with the spectrally - determined values . errors on the velocity dispersions are the range of values that result in isothermal deprojected temperature profiles that are consistent , at the 90 per cent confidence limit , with the spectrally - determined temperature results . estimates of the thermal gas pressure in the outermost radial bins used in the analysis are also required by the deprojection code and were determined iteratively . ( the uncertainties on the outer pressure estimates do not significantly affect the results presented here . ) although the deprojection method of fabian ( 1981 ) is essentially a single - phase technique , we note that it produces results in good agreement with more detailed multiphase treatments ( thomas , fabian & nulsen 1987 ) and , due to its simple applicability at large radii in clusters , is better - suited to the present project . the mass distributions determined from the deprojection analysis are summarized in columns 3 and 4 of table 6 . these simple parameterizations permit direct comparisons with independent mass constraints from gravitational lensing and dynamical studies ( allen 1998 ) . we note that for a few of the clusters ( primarily the brightest cfs with the best data ) the single - component mass models can not adequately satisfy the requirement for approximately isothermal deprojected temperatures profiles and for these systems a significantly better match was obtained by introducing a second ` linear ' mass component , truncated at a specified outer radius . the clusters requiring the additional mass components were abell 586 ( @xmath63 kpc@xmath58 within the central 20 kpc ) , abell 2029 ( @xmath64 kpc@xmath58 within the central 20 kpc ) , abell 2219 ( @xmath65 kpc@xmath58 within the central 30 kpc ) , abell 478 ( @xmath66 kpc@xmath58 within the central 20 kpc and an adjustment of the cluster parameters to @xmath67 and @xmath68 kpc ) and abell 1795 ( @xmath65 kpc@xmath58 within the central 20 kpc , and an adjustment of the cluster parameters to @xmath69 and @xmath70 kpc ) . the core radii of the mass distributions determined from the deprojection analysis are similar to the values measured from simple ` @xmath71model ' fits to the x - ray surface brightness profiles ( cavaliere & fusco - femiano 1976 , 1978 ) . it is well known that the @xmath72-model does not provide a good match to the x - ray surface brightness profiles of the brightest cf clusters when their central regions are included in the analysis ( jones & forman 1984 ) . however , where @xmath72-models are used to estimate the mass core radii in cf clusters , the central regions of these clusters should not simply be excised from the fits . the central surface brightness peaks in cf clusters trace the presence of sharp central density rises in these systems and are not only the result of excess emission due to cooling gas . ( it is the high central densities that lead to the high cooling rates . ) although fully accounting for the effects of cooling on the central density measurements is difficult with current data , simply excising the inner regions of the clusters can lead to overestimates of the mass core radii and corresponding underestimates of the central gas and total mass densities . the basic results on the cooling flows from the deprojection analysis are summarized in columns @xmath73 of table 6 . for those clusters in common , the results obtained are generally in good agreement with values reported from previous works ( edge 1992 , white 1997 ; peres 1998 ) . the spectral and image deprojection analyses discussed in sections 3 and 4 provide essentially independent estimates of the mass deposition rates in the clusters . a comparison of the results obtained from these analyses therefore provides a test of the validity of the cooling flow model . the deprojection method describes the x - ray emission from a cluster as arising from a series of concentric spherical shells . the luminosity in a particular shell , @xmath74 , may be written as the sum of four components ( arnaud 1988 ) . @xmath75 , \label{swa_eq1}\ ] ] where @xmath76 is the mass deposited in shell @xmath74 , @xmath77 is the enthalpy of the gas in shell @xmath74 , and @xmath78 is the gravitational energy released in crossing that shell . @xmath79 is the mass flow rate through shell @xmath74 , and @xmath80 the change of enthalpy of the gas as it moves through that shell . the first term in equation [ swa_eq1 ] thus accounts for the enthalpy of the gas deposited in shell @xmath74 . the second term is the gravitational work done on the gas deposited in shell @xmath74 . the third and fourth terms respectively account for the gravitational work done on material flowing through shell @xmath74 to interior radii , and the enthalpy released by that material as it passes through the shell . in any particular shell , the densest material in the cooling flow is assumed to cool out and be deposited . since the cooling time of this material will be short compared to the flow time , the cooling can be assumed to take place at a fixed radius . thus , the luminosity contributed by the first term in equation 1 should have a spectrum that can be approximated by gas cooling at constant pressure from the ambient cluster temperature the same spectrum as the cooling component incorporated into the spectral analysis with model c ( section 3 ) . for the bulk of the material continuing to flow inwards towards the cluster centre , the cooling via x - ray emission is assumed to be offset by the gravitational work done on the gas as it moves inwards . the emission accounted for in the second and third terms of equation 1 should therefore have a spectrum that can be approximated by an isothermal plasma at the appropriate ambient temperature for the cluster the spectrum of the isothermal emission component in model c. since the mass - weighted temperature profiles in the clusters are assumed to remain approximately isothermal with radius , the luminosity contributed by the fourth term of equation [ swa_eq1 ] should be negligible . the mass deposition rates , @xmath81 , listed in table 6 are the mass flow rates ( @xmath82 ) determined from the deprojection analysis , at the point where the mean cooling time of the cluster gas first exceeds the hubble time ( @xmath83 yr ) . thus , if the cooling flow model is correct , and the cooling flows have been undisturbed for a significant fraction of a hubble time , the mass deposition rates determined from the deprojection analysis should be similar to those measured independently from the spectral data . 3 shows the mass deposition rates determined from the spectral analysis ( @xmath39 ) versus the image deprojection results ( @xmath81 ) , for those clusters for which both quantities have been measured . we see that the results exhibit an approximately linear correlation ; a fit to the data with a power - law model of the form @xmath84 , using the akritas & bershady ( 1996 ) bisector modification of the ordinary least - squares statistic , gives a best - fitting power - law slope of @xmath85 ( where the error is the standard deviation determined by bootstrap re - sampling ) , although the spectrally - determined mass deposition rates typically exceed the deprojection results by a factor of @xmath86 . however , the spectral analysis presented in section 3 also requires that the cooling gas is intrinsically absorbed by equivalent hydrogen column densities of , typically , a few @xmath1 ( model c ) . the mass deposition rates determined from the deprojection analysis must therefore also be corrected for the effects of this absorbing material . the corrections for the effects of intrinsic absorption on the deprojection results have been carried out by re - running the deprojection analysis with the absorbing column densities set to the total values determined with spectral model c. we assume that the absorption is due to cold gas with solar metallicity . if , however , the intrinsic absorption were instead due to dust , a possibility examined in more detail in section 7 , then the required correction factors would likely be reduced by a few tens of per cent . ( the introduction of a simple oik absorption edge at @xmath87kev , such as might be associated with oxygen - rich , silicate dust grains , generally provides as good a description of the intrinsic absorption detected in the @xmath88kev asca spectra as a cold , gaseous absorber . ) since the intrinsic column densities measured with model c are redshifted quantities , we set the total column densities used in the revised deprojection analysis , which assumes zero redshift for the absorber , to be @xmath89 , where @xmath90 is the galactic column density ) . the absorption - corrected mass deposition rates ( @xmath91 ) so determined are summarized in table 7 . 4 compares the spectrally - determined mass deposition rates with the absorption - corrected deprojected values . we see that once the presence of intrinsic absorption has been accounted for in a consistent manner in the imaging and spectral analyses , the results on the mass deposition rates in the clusters show good agreement . table 8 lists the fractions of the total @xmath40 kev x - ray fluxes from the cf clusters associated with cooling gas , as determined with spectral model c. the results range from @xmath92 per cent for systems like abell 1689 and 2142 to @xmath93 per cent for zwicky 3146 , abell 1068 , rxj1347 - 1145 and abell 2204 . such a range of measurements is consistent with the results of peres ( 1998 ) , from a deprojection analysis of a flux - limited sample of the brightest clusters known . the fraction of the x - ray flux from a cluster due to it s cooling flow can be expected to increase with the age of the flow . it is therefore likely that the cooling flows in abell 1689 and 2142 are younger than those in abell 2204 and zwicky 3146 . we note that the x - ray data are not , in general , of sufficient quality to place firm constraints on the ages of the cooling flows in these clusters ( for a discussion of cooling flow age measurements see allen 1999 ) . it is interesting that the clusters with the largest cooling - flow flux fractions ( with the exception of abell 1413 and 2261 ) are also amongst the most optically line - luminous cooling flows known ( crawford 1999 ) . the production of powerful optical emission lines and associated uv / blue continuum emission is known to be due , at least in part , to the formation of massive stars at the centres of cooling flows ( johnstone , fabian & nulsen 1987 ; allen 1995 ; cardiel 1995 , 1998 ; mcnamara 1996 ; voit & donahue 1997 ) . the results presented here are consistent with the idea that such star formation is typical of undisturbed cooling flows . it is also interesting to note that the clusters with the largest cooling flow flux fractions and most optically line - luminous emission line nebulae also tend to have the shortest cooling times averaged over their central 100 kpc regions , with @xmath94gyr ( see also peres 1998 ) . the equivalent column densities of intrinsic x - ray absorbing material inferred from the spectral analysis are summarized in table 9 . all three of the spectral models incorporating a free - fitting absorption component ( models b , c , d ) reveal the presence of excess x - ray absorption over and above the nominal galactic values along the lines of sight to the clusters ( dickey & lockman 1990 ) , although the measured column densities are sensitive to the spectral model used . using the simple isothermal model with free - fitting absorption ( model b ) , we determine a mean excess column density for the whole sample of 30 clusters of @xmath95 . when we examine the cf and ncf clusters separately , we find that both subsamples exhibit intrinsic absorption , at a similar level on average . for the cf clusters , the mean excess column density measured with spectral model b is @xmath96 . for the ncf systems , the value is @xmath97 . the application of a students t - test ( accounting for the possibility of unequal variances in the two distributions ; press 1992 ) shows the mean excess column densities for the cf and ncf clusters , determined with spectral model b , to differ at only the 45 per cent confidence level . the application of a kolmogorov - smirnov test shows the two subsamples to be drawn from different parent populations at only the @xmath92 per cent confidence level . thus , our analysis with spectral model b suggests that intrinsic absorption is not confined to cf clusters . the results on the column densities are quite different , however , when the more - sophisticated multiphase spectral models are used . as discussed in section 3 , the multiphase models generally provide a better description of the x - ray properties of the cf clusters ( and some of the brighter ncf systems ) . using spectral model c , our preferred model for the cf clusters in that it provides a consistent description of the spectral and imaging x - ray data for these systems ( section 5.1 ) , we determine a mean intrinsic column density acting on the cooling - flow components of @xmath98 . using the two - temperature spectral model ( model d ) , we determine a mean intrinsic column density acting on the cooler emission components in the cf systems of @xmath99 . thus , the intrinsic column densities for the cf clusters measured with the multiphase models ( models c and d ) are similar and approximately an order of magnitude larger than the values inferred using the single - phase model b. these results demonstrate the need for adopting an appropriate spectral when attempting to measure the column densities of absorbing material in cf clusters . we also note that where spectral model d provided a significant improvement to the fits to the ncf clusters , with respect to model b , the measured column densities acting on the cooler emission components were also significantly larger than the values determined using the single - phase models . we have also measured the intrinsic column densities in the cf clusters using one further spectral model , referred to in table 9 as model c. model c is identical to model c except that it assumes that the excess absorption acts on the entire cluster spectrum , rather than just the cooling gas , and that the absorbing material lies at zero redshift . model c has been used in a number of previous studies and is included here for comparison purposes . the mean excess column density for the cf clusters determined with spectral model c is @xmath100 . allen & fabian ( 1997 ) present results from an x - ray colour deprojection study of 18 clusters observed with the rosat pspc . these authors determine intrinsic column densities across the central 30 arcsec ( radius ) regions of abell 478 , 1795 and 2029 , using spectral model c , of @xmath101 , @xmath102 and @xmath103 , respectively . the results for abell 478 and abell 1795 are in excellent agreement with those presented here , although the allen & fabian ( 1997 ) value for abell 2029 is @xmath22 times smaller than our result . abell 2029 is unusual in that it hosts a strong cooling flow without associated optical line emission . this may indicate that the central regions of the cluster have been disrupted ( perhaps by a minor merger event or the strong central radio source in this system ) , complicating the distributions of cooling and absorbing material ( _ cf . _ section 5.2 ) . four of the cf clusters studied here were also examined by white ( 1991 ) , using einstein observatory sss data . these authors measured intrinsic column densities for abell 478 , 1795 , 2029 and 2142 ( in a 3 arcmin radius circular aperture , using spectral model c ) of @xmath104 , @xmath105 , @xmath106 and @xmath107 , respectively . ( the sss result for abell 478 has been corrected to account for the different value of galactic absorption assumed in that study ) . the intrinsic column densities determined by white ( 1991 ) are @xmath108 times larger than the values measured from the asca data . allen & fabian ( 1997 ) compared the results from their x - ray colour deprojection study of pspc data with the white ( 1991 ) sss analysis and concluded that if the intrinsic absorption were due to cold gas , then agreement between the measured column densities could only be obtained if the covering fraction of the x - ray absorbing material were , in general , @xmath109 . we have therefore examined the constraints on the covering fraction , @xmath110 , that can be obtained from the asca data using spectral model c. the results are also listed in table 9 . in all cases we determine a best - fit covering fraction for the intrinsic x - ray absorbing material of unity . for many clusters , covering fractions of @xmath111 per cent can be firmly ruled out . ( the physical significance of these results are further explored in section 6.4 ) . the results on the covering fractions , and the reasonable agreement of the asca and pspc results for abell 478 and 1795 , suggest that the white ( 1991 ) results may have systematically over - estimated the intrinsic column densities in clusters by a factor of a few . before considering in more detail the possible origin and nature of the absorbing material , it is pertinent to consider whether the intrinsic absorption , inferred to be present using a variety of spectral models , could in fact be an artifact due to an unconsidered emission process in the clusters . 5 shows the sis0 spectra for abell 1068 and 1795 , which exhibit two of the clearest absorption signatures . the data between @xmath112 and @xmath113 kev have been fitted with a single - temperature emission model with galactic absorption ( _ i.e. _ spectral model a ) . the residuals to these fits , over the full @xmath88 kev band , are shown in the lower panels . we see that the residuals , which can be naturally explained by the introduction of cool emission components with associated intrinsic absorption ( using spectral models c and d ) , take the form of a large excess of counts between @xmath114 kev . we therefore include the caveat that if some extra emission process which can account for such residuals in the x - ray spectra were identified , then the requirement for intrinsic absorption from the x - ray data would be greatly diminished . finally , we note that the fits with spectral model a over the restricted @xmath115 energy range lead to determinations of the cluster temperatures for abell 1068 and 1795 of @xmath116 and @xmath117 kev , respectively , in excellent agreement with the results obtained for the full data sets with spectral model c. the results presented in section 6.1 show that the cooling flows in our sample are typically intrinsically absorbed by equivalent hydrogen column densities of a few @xmath118 ( taking spectral model c as our preferred model ) . assuming , in the first case , that this absorption is due to cold gas with solar metallicity , the mass of absorbing gas , @xmath119 , within a radius , @xmath120 , may be approximated as @xmath121 where @xmath120 is in units of kpc and @xmath37 in @xmath1 . for metallicities of @xmath122 in the absorbing gas ( the mean emission - weighted value for the x - ray emitting gas in the cf clusters ; allen & fabian 1998b ) the implied masses are a factor @xmath123 larger . for zero metallicity these masses are a factor @xmath15 times larger than the values given by equation 2 . ( the scaling factors for different metallicities in the absorbing material have determined from spectral simulations using the xspec code . ) we have adopted @xmath120 as the radius where the cooling time of the cluster gas first exceeds @xmath124 yr in the clusters ( a plausible age for the cooling flows in most of the cf systems ) . these radii , and the masses of absorbing gas implied by equation 2 ( using the column densities determined with spectral model c ) for both solar metallicity and @xmath125 , are listed in columns @xmath126 of table 10 . the observed masses of absorbing material may be compared to the masses expected to have been accumulated by the cooling flows @xmath127 within the same radii over the past @xmath124 yr . assuming that from time @xmath128 to @xmath129 yr , the integrated mass deposition rate within radius @xmath120 increases approximately linearly with time , the total mass accumulated within radius @xmath120 after @xmath124 yr is approximately @xmath130 at times @xmath131 yr , the mass deposition rate within @xmath120 should remain approximately constant . if instead we assume @xmath132 to be constant during the first 5 gyr , the accumulated mass will be twice the value indicated by equation 3 . ( exactly how the mass deposition rate within @xmath120 grows with time during the first 5 gyr is unclear and will depend upon the evolution of the cluster and the detailed properties of the cluster gas . ) the mass deposition rates within radii @xmath120 ( corrected for the effects of absorption due to cold gas , as discussed in section 5.1 ) and the estimated masses of cooled material accumulated by the cooling flows within these radii over a 5 gyr period , are summarized in table 10 . 6 compares the accumulated masses with the masses of absorbing material determined from the spectral data . the agreement between the results is reasonable and is maximized for a metallicity in the absorbing gas of @xmath133 ( depending on the growth of the cooling flow over the first 5gyr ) . this supports the idea that the observed x - ray absorption may be due to material accumulated by the cooling flows . we note that if the x - ray absorption in the cf clusters were due to dust rather than cold gas , a possibility examined in more detail in section 7 , then the agreement between the observed and predicted masses of absorbing material shown in fig . 6 might still be expected to hold if the dust were contained in the material deposited by the cooling flows . in this case , the mass of the absorber calculated with equation 2 would be the mass of cooled gas associated with the dust when it was deposited from the cooling flow . for a galactic dust to gas ratio , the mass in dust would be @xmath134 times smaller than the associated gas mass . the luminosities absorbed at x - ray wavelengths must eventually be reprocessed in other wavebands . if the absorbing material is dusty ( as is likely to be the case in the central regions of the clusters voit & donahue 1995 ; fabian , johnstone & daines 1994 ; allen 1995 ) then the bulk of this reprocessed emission is likely to emerge in the infrared band ( dwek , rephaeli & mather 1990 ) . table 11 summarizes the reprocessed luminosities ( _ i.e. _ the bolometric luminosities absorbed within the clusters ) measured with spectral model c. the reprocessed luminosities range from @xmath135 for abell 1795 to @xmath136 for rxj1347.5 - 1145 , with a mean value of @xmath137 . allen ( 1999 ) report detections of spatially extended @xmath138 m emission coincident with the x - ray centroids in the nearby centaurus cluster and abell 2199 . the infrared fluxes from those clusters were also shown to be in good agreement with the predicted values due to reprocessed x - ray emission from their cooling flows . we have also used the ipac scanpi software and archival iras data to measure the 60 and @xmath138 m fluxes within a four arcmin ( radius ) aperture centred on the x - ray centroids for the clusters in the present sample . ( the medians of the co - added scanpi results were used . ) the results are listed in table 11 . the error bars associated with the measured fluxes are the root - mean - square deviations in the residuals , external to the source extraction regions , after baseline subtraction . where no detection was made , an upper limit equal to three times the r.m.s . deviation in the residuals is listed . where a detection was made , we also list the in - scan separations ( in arcmin ) between the peak of the @xmath138 m emission ( or the @xmath139 m emission in the case of abell 1835 ) and the x - ray centre . two of the clusters in our sample , iras 09104 + 4109 and abell 1068 have known infrared point sources coincident with their x - ray centroids ( kleinmann 1988 ; moshir 1989 ) . the presence of these sources is confirmed by our analysis . the 60 and @xmath138 m data for abell 1704 and @xmath139 m data for abell 1835 also suggest the presence of unresolved sources , coincident with the x - ray centroids for the clusters . abell 478 and zwicky 3146 have @xmath138 m emission originating from close to their x - ray centroids , which appears spatially extended , although the infrared flux from abell 478 is probably contaminated by galactic cirrus . the data for abell 2142 , 2204 , 2261 and 2390 also provide significant detections within the four arcmin ( radius ) source apertures , although the peaks of the infrared emission from these systems are spatially offset from their x - ray centroids , suggesting that the detected flux is likely to originate , at least in part , from some other source . several of the clusters included in our study ( abell 586 , 963 , 1413 , 1795 , 2029 , 2142 and pks0745 - 191 ) have previously been studied at infrared wavelengths , using iras data , by wise ( 1993 ) and/or cox , bregman & schombert ( 1995 ) . these authors also report no clear detections of infrared emission from these sources . edge ( 1999 ) report measurements of @xmath139 m emission from abell 1835 and 60 and @xmath138 m emission from abell 2390 , using a similar analysis to that presented here . these authors also present detections of @xmath140 m emission from these clusters which , for abell 1835 , they suggest is likely to be due to dust heated by the vigorous star formation observed in the cluster core or an obscured active galactic nucleus . following helou ( 1988 ) and wise ( 1993 ) , we can estimate the total infrared luminosities implied by the observed iras fluxes and flux limits using the relation @xmath141 where @xmath142 and @xmath143 are the 60 and @xmath138 m iras fluxes in units of jy . this relation assumes a dust temperature of @xmath144k ( which is consistent with the observations ; allen 1999 ; edge 1999 ) and an emissivity index , @xmath145 , in the range @xmath146 , where the emissivity is proportional to the frequency , @xmath147 . where measurements , rather than upper limits , were obtained , we associate a systematic uncertainty of @xmath148 per cent with the estimated @xmath149 m luminosities , which is combined in quadrature with the random errors . the @xmath149 m luminosities calculated from this relation are summarized in table 11 . in general , the measurements and upper limits to the observed @xmath149 m luminosities exceed the predicted luminosities due to reprocessed x - ray radiation from the cooling flows . where measurements ( rather than upper limits ) are made , this may suggest the presence of an additional source of infrared flux , such as star formation or agn associated with the central cluster galaxies , or some other contaminating ( possibly galactic ) source . at some level , the massive starbursts at the centres of systems like abell 1068 , 1835 , 2204 2390 and zwicky 3146 ( allen 1995 ; crawford 1999 ; edge 1999 ) must contribute to the detected infrared flux . as discussed in section 6.1 , the measured intrinsic equivalent hydrogen column densities of absorbing material inferred from the asca spectra exhibit significant systematic variation , depending upon whether the single - phase or multiphase emission models are used . we have adopted the results obtained with the cooling - flow model ( model c ) as our preferred values , since this model provides a consistent description of the spectral and imaging x - ray data for the clusters . however , a number of systematic uncertainties affecting the absorption results remain with the cooling flow model . spectral model c undoubtedly over - simplifies the situation in a real cooling flow . in particular , the model assumes that the absorbing material lies in a uniform screen in front of the cooling flows . if the absorber is intrinsically associated with the clusters , then the absorbing material must , to some extent , be distributed throughout the cooling flows . allen & fabian ( 1997 ; see also wise & sarazin 1999 ) discuss two different , limiting geometries for the absorbing material ; partial covering and multilayer absorption . the partial covering model , with a covering factor @xmath150 , can be expected to apply or where there are @xmath151 individual absorbing clouds along each line of sight and/or where the absorbing material is strongly clumped on large scales . however , as discussed in section 6.1 , covering fractions significantly less than unity are found to provide a poor description of the asca spectra , suggesting that the absorbing material is likely to be more smoothly distributed throughout the cooling flows . the multilayer absorption model discussed by allen & fabian ( 1997 ) is applicable where the absorbing material is made up of a large number of similarly sized absorbing clouds , each with a column density much less than the total intrinsic column density ( @xmath37 ) , homogeneously distributed throughout the x - ray emitting medium . for an intrinsic emission spectrum @xmath152 , the observed spectrum , @xmath153 , emerging from the emitting / absorbing region may be written ( where @xmath154 is the absorption cross - section ) as @xmath155 we have simulated cooling flow spectra with multilayer absorption , with total column densities of between @xmath1 and @xmath156 , and fitted these with simple models which assume that the absorbing material acts as a uniform screen in front of the cooling flow ( as with model c ) . for the simulations we adopt an emission spectrum appropriate for a constant pressure cooling flow , with an upper temperature of 7 kev and a metallicity of @xmath157 . we use a response matrix appropriate for the asca sis detectors . the comparison between the fitted ( uniform screen ) column densities and the true total ( multilayer ) values are shown in fig . we see that for multilayer column densities of @xmath158 , the fitted column densities underestimate the true values by factors of @xmath159 . ( these results are not sensitive to any reasonable choice of galactic column density . ) if the multilayer model approximates the true distribution of x - ray emitting and absorbing material in cooling flows , then the results obtained with spectral model c would imply true equivalent column - densities of intrinsic absorbing gas in the cluster cores of , typically , a few @xmath160 . importantly , we also note that for signal - to - noise ratios appropriate for asca observations of bright clusters ( with count rates of @xmath161ct s@xmath58 in the sis detectors and exposure times of @xmath162 ks ) , the simple uniform - screen absorption model provides a good fit to the simulated spectra with multilayer absorption , for total column densities @xmath163 . ( for larger total column densities , complex residuals are detected in the spectra at energies @xmath164kev . ) the large intrinsic column densities implied by the multilayer models would have strong implications for the nature of the absorbing matter . in particular , column densities of @xmath165 of gaseous material with solar metallicity would be very difficult to reconcile with current hi and co limits , even if this material were very cold and highly molecular . ( we also note that in the case of a multilayer distribution , the implied masses of absorbing gas would often exceed the predicted accumulated masses due to the cooling flows and imply , at least in part , some other origin for the absorbing medium . ) in the following section , we explore these issues in more detail and summarize the current observational constraints on the physical nature of the absorbing matter . the results presented in this paper provide a consistent picture for the x - ray properties of cooling flows in x - ray luminous clusters of galaxies . the asca spectra and rosat images provide independent determinations of the mass cooling rates in the clusters in good agreement with each other , when a consistent method of analysis is employed ( section 5.1 ) . the spectral data require the presence of large column densities of intrinsic x - ray absorbing material associated with the cooling gas ( section 6 ) . the implied masses of absorbing material are in reasonable agreement with the masses expected to have been accumulated by the cooling flows over their lifetimes ( section 6.2 ) , identifying this material as a plausible sink for the bulk of the cooled gas from the flows . despite the evidence for intrinsic x - ray absorption associated with cooling flows , however , searches for this material in other wavebands have , to date , proved largely unsuccessful ( at least outside of the central @xmath166 kpc ) . we here briefly review these observations and comment on their implications for the physical state of the x - ray absorbing matter . more detailed discussions of a number of these issues are given by ferland , fabian & johnstone ( 1994 ) , daines , fabian & thomas ( 1994 ) , fabian ( 1994a ) , odea ( 1994a ) , voit & donahue ( 1995 ) , odea & baum ( 1996 ) , arnaud & mushotzky ( 1988 ) and henkel & wiklind ( 1998 ) . as mentioned in section 5.2 , the central ( @xmath166 kpc ) regions of cf clusters are commonly observed to be sites of ongoing star formation with powerful , associated optical line emission ( johnstone 1987 ; heckman 1989 ; mcnamara & oconnell 1989 ; crawford & fabian 1992 ; allen 1995 ; cardiel 1995 , 1998 ; mcnamara 1996 ; voit & donahue 1997 ; crawford 1999 ) . nearly all cf clusters with short central cooling times ( @xmath167gyr ) exhibit such phenomena ( peres 1998 ) , whereas systems with longer central cooling times ( including all ncf clusters ) generally do not . the observed levels of star formation imply the presence of at least moderate masses of cooled gas in the cores of cf clusters , sufficient in a few cases to account for the bulk of the mass deposited by the cooling flows in those regions ( allen 1995 ) and at least some causal connection with the cooling flows . at radio wavelengths , extensive searches have been carried out for 21 cm line - emission associated with warm atomic hydrogen in the cores of cf clusters ( valentijn & giovanelli 1982 ; odea & baum 1996 ) . the negative results obtained ( with the exception of m87 ; jaffe 1992 ) typically constrain the column densities of optically thin material to be less than a few @xmath168 and , for an assumed spin temperature , @xmath169k , the number of clouds along the line of sight to be @xmath170 ( odea & baum 1996 ) . searches have also been made for the 21 cm absorption signature of cold atomic hydrogen ( mcnamara , bregman & oconnell 1990 ; jaffe 1992 ; dwarakanath , van gorkom & owen 1994 ; odea , gallimore & baum 1995 ; odea & baum 1996 ; johnstone 1998 ) . such material has been detected across the central few kpc of a few cf clusters , including abell 426 ( crane , van der hulst & haschick 1982 ; sijbring 1989 ; jaffe 1990 ) 2a0335 + 096 ( mcnamara 1990 ) , hydra a ( dwarakanath , van gorkom & owen 1995 ) , mkw3s ( mcnamara 1990 ) and abell 2597 ( odea , baum & gallimore 1994 ) with typical column densities of a few @xmath171 . however , for most clusters only negative results have been obtained , with upper limits to the column densities of 10k gas ( assuming a covering fraction of unity and a velocity width of @xmath172 kms@xmath58 ; dwarakanath 1994 , odea & baum 1996 ) of @xmath173 . laor ( 1997 ) presents tight limits on the column density of atomic hydrogen towards the core of abell 426 ( @xmath174 ) from ly@xmath175 observations made with the hubble space telescope ( see also johnstone & fabian 1995 ) . together , the results on hi emission and absorption , and ly@xmath175 absorption , imply that the bulk of the material responsible for the observed x - ray absorption can not be in the form of atomic hydrogen . an issue of crucial importance in interpreting the 21 cm hi results and the results from co studies ( see below ) is the expected temperature of x - ray absorber , if in gaseous form , which has proved controversial . johnstone , fabian & taylor ( 1998 ; updating earlier calculations by ferland , fabian & johnstone 1994 and fabian 1994a ) suggest that cooled gas clouds , in equilibrium with the cooling - flow environment , are likely to be very cold . johnstone ( 1998 ) suggest that dust - free clouds form an extended outer envelope , with a column density of @xmath176 and a temperature of 13 - 17k , beyond which the gas drops to the microwave background temperature . for galactic dust / gas ratios , however , the results are significantly modified , such that the cloud temperature drops to the microwave background temperature after a column density of only @xmath177 . at such low temperatures , dusty gas is likely to be highly molecular . however , these predictions differ from those of odea ( 1994a ) and voit & donahue ( 1995 ) who suggest a minimum temperature for such clouds of @xmath178k , even in the presence of dust . braine ( 1995 ) also suggest a minimum temperature of @xmath179k for the absorbing gas . possible reasons for the discrepancies between these results ( with contrasting views ) are discussed by voit & donahue ( 1995 ) and ferland ( in preparation ) . extensive searches have been carried out for co associated with molecular gas in cooling flows . to date co emission has only been detected in abell 426 ( lazareff 1989 ; mirabel , sanders d.b . & kazs i. 1989 ) . braine ( 1995 ) also present limits on co absorption in the central regions of this cluster ) . the upper limits to the column density of molecular hydrogen in the inner regions of other clusters ( for an assumed kinetic temperature of @xmath180k and a covering fraction of unity ) are typically @xmath181 a few @xmath0 ( odea 1994a , antonucci & barvainis 1994 , mcnamara & jaffe 1994 ; braine & dupraz 1995 ; odea & baum 1996 ) which is significantly below the x - ray inferred column densities . the co results also constrain the number of clouds along the line of sight ( for @xmath180k ) to be @xmath182 ( odea 1994a ; odea & baum 1996 ) . the co emission limits appear to exclude the possibility of large column densities ( _ i.e. _ values consistent with the x - ray measurements ) of molecular gas , with a kinetic temperature @xmath183k . however , if the bulk of the x - ray gas cools to the microwave background temperature , as suggested by johnstone ( 1998 ) , the co limits may be consistent with the x - ray data . it should be noted that the determinations of molecular hydrogen masses from co observations are affected by a number of additional uncertainties . firstly , the co / h@xmath184 ratios in cluster cooling flows may differ from galactic values . in particular , where the temperature and metallicity of the absorbing gas is low , the use of a standard galactic co / h@xmath184 ratio may underestimate the mass of molecular gas ( maloney & black 1988 ; madden 1997 ) . it is unclear whether x - ray absorbing gas accumulated by cooling flows could have substantially sub - solar metallicity . allen & fabian ( 1998b ) determine a mean emission - weighted metallicity for the x - ray emitting gas in cf clusters of @xmath185 . these authors also show that cf clusters generally contain metallicity gradients , so that the metallicity of the x - ray gas within the cooling radii of cf systems will , on average , exceed this value . thus , it seems unlikely that the material currently being deposited from the cooling flows will have substantially sub - solar metallicities . however , if the x - ray absorbing material is formed from the matter deposited by the cooling flows over their lifetimes , the metallicity of this material could be somewhat lower . within an inhomogeneous cooling flow ( nulsen 1986 , 1998 ; thomas , fabian & nulsen 1987 ) the denser gas at any particular radius will tend to have a smaller volume filling factor . if the metals are evenly distributed throughout the volume of the cooling flow at any particular radius ( whether this will occur is unclear ) , the least dense gas will contain most of the metals ( reisenegger , miralda - escud & waxman 1996 ) . since the densest material in cooling flows will have the shortest cooling time and be deposited first , the material accumulated by the cooling flows over their histories could have a significantly lower metallicity than the material being deposited today . conversely , if the metals in cooling flows were concentrated in the densest material , the metallicity of the x - ray absorbing material could exceed that of the x - ray emitting gas . fabian ( 1994a ) , have suggested that the masses of molecular gas inferred from co observations could be significantly underestimated if most of the co in the absorbing material has frozen onto dust grains . such freezing should occur rapidly ( on timescales @xmath186 yr ) if even small amounts of dust are initially present in the absorbing gas ( voit & donahue 1995 ) . daines ( 1994 ) and fabian ( 1994a ) discuss how dust is likely to form even in clouds that are initially free of dust , on timescales @xmath187 yr ( although see also voit & donahue 1995 ) . optical , uv and sub - mm studies of the cores of cooling flows show that dust lanes ( sparks , macchetto & golombek 1989 ; mcnamara & oconnell 1992 ) and intrinsic reddening ( hu 1992 ; allen 1995 ; edge 1999 ) are common . elston & maloney ( 1994 ) and jaffe & bremer ( 1997 ) report detections of h@xmath184(1 - 0)s(1 ) emission from warm ( @xmath188k ) molecular hydrogen from k - band spectroscopy of the inner few kpc of a number of cf systems ( including abell 478 and and pks0745 - 191 ) . no detections of such emission have been made in ncf systems . the h@xmath184s emission is likely to arise from the heated skin of cold molecular clouds which could also be responsible for the optical emission - line phenomena discussed above . a final possibility , first discussed in detail by voit & donahue ( 1995 ; see also arnaud & mushotzky 1998 ) , is that the x - ray absorption could be due to dust , with little or no associated gas . the k - absorption edge of oxygen is a primary source of x - ray absorption and its signature is fairly insensitive to the physical state of the oxygen , so long as it is not highly ionized . the introduction of a simple oik absorption edge at @xmath189 kev in the asca analysis ( such as might be associated with oxygen - rich , silicate dust grains ) typically provides at least as good a model for the intrinsic absorption as a cold gaseous absorber . arnaud & mushotzky ( 1998 ) have shown that @xmath190 kev broad band x - ray telescope data for the perseus cluster require excess absorption which is significantly better - explained by a simple oxygen absorption edge ( which they associate with dust ) than by a cold , gaseous absorption model . verification of this important result will be possible in the near future using observations made with the chandra observatory and xmm . optical and uv spectroscopy of the central emission - line nebulosities in cooling flows indicate the presence of significant amounts of dust ( hu 1992 ; allen 1995 ; crawford 1999 ) . allen ( 1995 ) also show that the column densities of absorbing gas inferred from the optical reddening studies are in reasonable agreement with the x - ray values , for galactic dust / gas ratios . the infrared observations reported here and elsewhere , together with the sub - mm results of edge ( 1999 ) , require the presence of significant dust masses in the core regions of at least a few nearby and/or exceptionally massive cooling flows ( for which the best data exist ) and are consistent with dust being a common feature of cf clusters . limits on background counts of galaxies and quasars behind clusters are consistent with extinctions ranging from @xmath191 ( boyle , fong & shanks 1988 ; romani & maoz 1992 ) , although measurements of galaxy colours in nearby clusters constrain the reddening on cluster - wide scales to be @xmath192 mag ( ferguson 1993 ) . voit & donahue ( 1995 ) argue that the levels of dust required to account for the observed x - ray absorption are unlikely to lie entirely within the cores of the clusters , and should be distributed more widely throughout the cluster gas . rosat results on the spatial distribution of the x - ray absorbing material ( allen & fabian 1997 ) show it to be centrally concentrated within the cooling radii of clusters and possibly be confined within cooling flows . however , the x - ray data are also consistent with a scenario in which the absorbing material is distributed in a manner similar to the x - ray emitting gas . this is illustrated in fig . 8(a ) , where we show the equivalent column densities of hydrogen ions in the x - ray emitting gas through two simulated clusters ; one cf and one ncf system . the gas distributions have been modeled as isothermal spheres , with core radii of 50 and 300 kpc , respectively ( and have been normalized to provide total x - ray gas masses within 3mpc of the cluster centres of @xmath193 ) . we see that the distribution of x - ray emitting gas is much more centrally concentrated in the cf system , particularly within the central @xmath134 kpc . 8(b ) shows the corresponding x - ray absorbing column densities as a function of radius ( as would be determined from observations with asca - like ccd detectors in the @xmath88 kev band ) in the case where the intracluster gas has an equivalent metallicity in dust grains of @xmath194 solar . for simplicity , we assume that the dust has an absorption spectrum in the asca band similar to that of metal - rich cold gas ( morrison & mccammon 1983 ; qualitatively similar results are obtained for an oxygen edge at 0.54 kev ) . we also assume that the absorber follows a multilayer distribution within the x - ray emitting medium , but is modeled as a uniform absorbing screen in front of the emitting regions . the results imply mean emission - weighted column densities ( where we assume the x - ray emissivity to be proportional to square of the gas density ) of @xmath195 for the cf cluster , and a few @xmath196 for the ncf system . ( the absolute results on the excess column densities will depend upon the chemical composition of the dust ) . we thus see that if the x - ray absorption is due to dust , with a spatial distribution that follows the x - ray gas , then we can expect absorption signatures to be present in both cf and ncf clusters , but to be stronger in cf systems ( assuming in each case that an appropriate spectral model is used in the analysis ) . we would also expect the absorption to be concentrated towards the cores of cf clusters , in agreement with the observations . we note that if the x - ray absorption were entirely due to a pre - existing distribution of dust grains distributed throughout the clusters , then the agreement between the observed masses of absorbing matter ( calculated assuming that the absorption is due to cold gas confined within the cooling flows ) and the masses predicted to have been accumulated by the cooling flows over their lifetimes ( fig . 6 ) must be regarded as coincidence . however , it remains possible that both material accumulated by the cooling flows and a more extended distribution of large dust grains ( presumably due to supernovae enrichment at early epochs ) could contribute to the x - ray absorption . the lifetime of grains of radius @xmath197 m to sputtering in hot gas of density @xmath145 is @xmath198 ( draine & salpeter 1979 ) . provided that individual grains exceed 10@xmath199 m in radius , they should survive for a hubble time or longer throughout ncf clusters and beyond the cooling radius in clusters with cooling flows . within cooling flows , the gas density rises inward so the grains are increasingly sputtered , releasing the metals into the gas phase which thus becomes increasingly metal rich towards the cluster centre . this provides one possible explanation for the abundance gradients inferred to be present in most cf clusters ( allen & fabian 1998b ; see also irwin & bregman 1999 ) . however , @xmath200 m grains are optically thick to soft x - rays and so some distribution of grain sizes extending to smaller values is required if dust is responsible for the observed x - ray absorption ( laor & draine 1993 ) . we note that the sputtering of small ( @xmath201 m ) grains will significantly modify the reddening law in clusters with respect to the standard galactic relation , reducing the optical / uv reddening ( laor & draine 1993 ) . in conclusion , the available data from other wavebands provide some support for the large column densities of intrinsic absorbing material inferred to be present from the x - ray data , at least in the innermost regions of cooling flows . optical and sub - mm observations of star formation in the cores of cf clusters and k - band observations of emission lines from molecular hydrogen suggest that significant masses of dusty , molecular gas are present in the central few tens of kpc of many cf clusters . the constraints from 21 cm observations suggest that the bulk of the material accumulated by cooling flows can not remain in a long - lived reservoir of atomic hydrogen . substantial masses of molecular gas may be distributed throughout cooling flows and as yet have avoided detection , but this material must be very cold ( @xmath202k ) and dusty , a possibility that remains controversial . the very large x - ray column densities ( @xmath165 ) required if the absorbing material has a multilayer distribution are probably incompatible with a gaseous absorber . it remains possible that dust grains , either present in the cooled material deposited by the cooling flows or distributed throughout the clusters in a manner similar to the x - ray gas , may be responsible for much of the observed x - ray absorption . the main conclusions that may be drawn from this paper may be summarized as follows : 0.2 cm \(i ) we have demonstrated the need for multiphase models to consistently explain the spectral and imaging x - ray data for the cf clusters included in our study . the mass deposition rates from the cooling flows , independently inferred from multiphase analyses of the asca spectra and deprojection analyses of the rosat hri images , exhibit good agreement , especially once the effects of intrinsic x - ray absorption have been accounted for in a consistent manner . the mass deposition rates from the largest cooling flows exceed 1000 , identifying these as some of the most massive cooling flows known . \(ii ) we have confirmed the presence of intrinsic x - ray absorption in the cluster spectra using a variety of spectral models . the measured equivalent hydrogen column densities are sensitive to the spectral models used in the analysis , ranging from a few @xmath0 for a simple isothermal emission model to a few @xmath1 using our preferred cooling - flow models , assuming in each case that the absorber lies in a uniform foreground screen . both the cf and ncf systems exhibit excess absorption at a similar level ( on average ) when analysed with the same , simple isothermal model . \(iii ) the masses of x - ray absorbing material inferred to be present in the cf clusters ( assuming the absorption to be due to cold gas with a covering fraction of unity ) are in reasonable agreement with the masses expected to have been accumulated by the cooling flows over their lifetimes . \(iv ) the asca spectra constrain the covering fraction of the absorbing material acting on the cooling flows to be close to ( or exceed ) unity . if the x - ray absorption is due to many small , similarly - sized clouds along each line of sight , intermixed with the x - ray emitting gas , then the column densities inferred from the spectral analysis may significantly underestimate the true column densities of absorbing material in the cooling flows . \(vii ) we have summarized the constraints on the physical properties of the x - ray absorbing material from observations in other wavebands . substantial 60 and 100@xmath199 m fluxes , sufficient to account for the bolometric luminosities absorbed and reprocessed within the clusters , are detected from several of the largest cooling flows . optical observations of star formation in the cores of cooling flows also suggests that significant masses of molecular gas are present in the central few tens of kpc of many cf systems . constraints from 21 cm observations suggest that the bulk of the material accumulated within the cooling radii can not remain in a long - lived reservoir of atomic hydrogen . substantial masses of molecular gas may be distributed throughout cooling flows and as yet have avoided detection , although this material must be very cold ( @xmath202k ) , a possibility that remains controversial . the very large x - ray column densities ( @xmath165 ) required if the absorbing material has a multilayer distribution are probably incompatible with a gaseous absorber . it remains possible that dust grains , either present in the material accumulated by the cooling flows or distributed more widely in a manner similar to the x - ray gas , may be responsible for much of the x - ray absorption . i thank andy fabian , roderick johnstone and dave white for many helpful discussions and harald ebeling , alastair edge and carolin crawford for their continuing efforts with the bcs project . i acknowledge the support of the royal society . allen s.w . , 1995 , mnras , 276 , 947 allen s.w . , 1998 , mnras , 296 , 392 allen s.w . , fabian a.c . , 1997 , mnras , 286 , 583 allen s.w . , fabian a.c . , 1998a , mnras , 297 , l57 allen s.w . , fabian a.c . , 1998b , mnras , 297 , l63 allen s.w . , fabian a.c . , johnstone r.m . , white d.a . , daines s.j . , edge a.c . , stewart g.c . , 1993 , mnras , 262 , 901 allen s.w . , fabian a.c . , edge a.c . , bhringer h. , white d.a . , 1995 , mnras , 275 , 741 allen s.w . , fabian a.c . , edge a.c . , bautz m.w . , furuzawa a. , tawara y. , 1996 , mnras , 283 , 263 allen s.w . , fabian a.c . , johnstone r.m . , arnaud k.a . , nulsen p.e.j . , 1999 , mnras , submitted akritas m.g . , bershady m.a . , 1996 , apj , 470 , 706 anders e. , grevesse n. , 1989 , geochemica et cosmochimica acta 53 , 197 antonucci r. , barvainis r. , 1994 , aj , 107 , 448 arnaud , k.a . , 1996 , in astronomical data analysis software and systems v , eds . jacoby g. and barnes j. , asp conf . series volume 101 , p17 arnaud k.a . , 1988 , in fabian a.c . , ed , cooling flows in clusters and galaxies . kluwer , dordrecht , p. 31 arnaud k.a . , mushotzky r.f . , 1998 , apj , 501 , 119 balucinska - church m. , mccammon d. , 1992 , apj , 400 , 699 bevington p.r . , 1969 , data reduction and error analysis for the physical sciences . mcgraw - hill , new york binney j. , tremaine s. , 1987 , galactic dynamics , princeton univ . press , princeton bhringer h. , tanaka y. , mushotzky r.f . , ikebe y. , hattori m. , 1998 , a&a , 334 , 789 boyle b.j . , fong r. , shanks t. , 1988 , mnras , 231 , 897 braine j. , dupraz c. , 1994 , a&a , 283 , 407 braine j. , wyrowski f. , radford s.j.e . , henkel c. , lesch h. , 1995 , a&a , 293 , 315 buote d.a . , tsai j.c . , 1996 , apj , 458 , 27 cardiel n. , gorgas j. , aragon - salamanca a. , 1995 , mnras , 277 , 502 cardiel n. , gorgas j. , aragon - salamanca a. , 1998 , mnras , 298 , 977 cavaliere a. , fusco - femiano r. , 1976 , a&a , 49 , 137 cavaliere a. , fusco - femiano r. , 1978 , a&a , 49 , 137 cox c.v . , bregman j.n . , schombert j.m . , 1995 , apjs , 99 , 405 crane p , van der hulst j.m . , haschick a. , 1982 , iaus , 97 , 307 crawford c.s . , fabian a.c . , 1992 , mnras , 259 , 265 crawford c.s . , allen s.w . , ebeling h. , edge a.c . , fabian a.c . , 1999 , mnras , 306 , 857 daines s.j . , fabian a.c . , thomas , p.a . , 1994 , mnras , 268 , 1060 david l.p . , slyz a. , jones c. , forman w. , vrtilek s.d . , arnaud k.a . , 1993 , apj , 412 , 479 david l.p . , harnden f.r . , kearns k.e . , zombeck m.v . , 1996 , the rosat hri calibration report , @xmath203 dickey j.m . , lockman f.j . , 1990 , ara&a , 28 , 215 draine b.t . , salpeter e.e . , 1979 , apj , 231 , 77 dwarakanath k.s . , van gorkom j.h . , owen f.n . , 1994 , apj , 432 , 469 dwarakanath k.s . , owen f.n . , van gorkom j.h . , 1995 , apj , 442 , l1 dwek e. , rephaeli y. , mather j.c . , 1990 , apj , 350 , 104 ebeling h. , voges w. , bhringer h. , edge a.c . , huchra j.p . , briel u.g . , 1996 , mnras , 281 , 799 ebeling h. , edge a.c . , bhringer h. , allen s.w . , crawford c.s . , fabian a.c . , voges w. , huchra j.p . , 1998 , mnras , 301 , 881 edge a.c . , stewart g.c . , fabian a.c . , 1992 , mnras , 258 , 177 edge a.c . , ivison rr.j . , smail i. , blain a.w . , kneib j .- . , 1999 , mnras , 306 , 599 elston r. , maloney p. , 1994 , in infrared astronomy with arrays , assl , 190 , kluwer , dordrecht , p. 169 fabian a.c . , 1994 , a&ar , 32 , 277 fabian a.c . , nulsen p.e.j . , canizares c.r . , 1991 , a&ar , 2 , 191 fabian a.c . , johnstone r.m . , daines s.j . , 1994a , mnras , 271 , 737 fabian a.c . , hu e.m . , cowie l.l . , grindlay j. , 1981 , apj , 248 , 47 fabian a.c . , arnaud k.a . , bautz m.w . , tawara y. , 1994b , apj , 436 , l63 ferguson h.c . , 1993 , mnras , 263 , 343 feretti l. , giovannini g. , bhringer h. , 1997 , newa , 2 , 501 ferland g.j . , fabian a.c . , johnstone r.m . , 1994 , mnras , 266 , 399 fukazawa y. 1994 , pasj , 46 , l55 heckman t.m . , baum s.a . , van breugel w.j.m . , mccarthy p. , 1989 , apj , 338 , 48 helou g. , khan i.r . , malek l. , boehmer l. , 1988 , apjs , 68 , 151 henkel c. , wiklind t. , 1997 , ssrv , 81 , 1 hu e.m . , 1992 , apj , 391 , 608 ikebe y. , makishima k. , fukazawa y. , tamura t. , xu h. , ohashi t. , matsushita k. , 1999 , apj , in press irwin j.a . , bregman j.n . , 1999 , apj , submitted irwin j.a . , sarazin c.l . , 1995 , apj , 355 , 497 jaffe w. , 1990 , a&a , 240 , 254 jaffe w. , 1992 , in fabian a.c . , ed . , clusters and superclusters of galaxies , kluwer , dordrecht , p. 109 jaffe w. , bremer m.n . , 1997 , mnras , 284l , 1 johnstone r.m . , fabian a.c . , 1995 , mnras , 273 , 625 johnstone r.m . , fabian a.c . , nulsen p.e.j . , 1987 , mnras , 224 , 75 johnstone r.m . , fabian a.c . , taylor g.b . , 1998 , mnras , 298 , 854 johnstone r.m . , fabian a.c . , edge a.c . , thomas p.a . , 1992 , mnras , 255 , 431 jones c. , forman w. , 1984 , apj , 276 , 38 kaastra j.s . , mewe r. , 1993 , legacy , 3 , heasarc , nasa kleinmann s.g . , hamilton d. , keel w.c . , wynn - williams c.g . , eales s.a . , becklin e.e . , kuntz k.d . , 1988 , apj , 328 , 161 kneib j .- p . , mellier y. , pell r. , miralda - escud j. , le borgne j .- f . , bhringer h. , picat j .- , 1995 , a&a , 303,27 laor a. , 1997 , apj , 483 , l103 laor a. , draine b.t . , 1993 , apj , 402 , 441 lazareff b. , castets a. , kim d - w . , jura m. , 1989 , apj , 336 , l13 liedahl d.a . , osterheld a.l . , goldstein w.h . , 1995 , apj , 438 , l115 madden s.c . , poglitsch a. , geis n. , stacey g.j . , townes c.h . , 1997 , 483 , 200 maloney p. , black j.h . , 1988 , apj , 325 , 389 makishima k. , 1997 , in makino f. , mitsuda k. , eds . , x - ray imaging and spectroscopy of cosmic hot plasmas , universal academy press , tokyo , p. 137 markevitch m. , 1996 , apj , 465 , l1 markevitch m. , forman w.r . , sarazin c.l . , vikhlinin a. , 1998 , apj , 503 , 77 matsumoto h. , koyama k. , awaki h. , tomida h. , tsuru t. , mushotzky r. , hatsukade i. , 1996 , pasj , 48 , 201 mcnamara b.r . , jaffe w. , 1994 , a&a , 281 , 673 mcnamara b.r . , oconnell r.w . , 1989 , aj , 98 , 2018 mcnamara b.r . , oconnell r.w . , 1992 , apj , 393 , 589 mcnamara b.r . , bregman j.n . , oconnell , 1990 , apj , 360 , 20 mcnamara b.r . , jannuzi b.t . , elston r. , sarazin c.l . , wise m. , 1996 , apj , 469 , 66 mirabel i.f . , sanders d.b . , kazs i. , 1989 , apj , 340 , l9 morrison r. , mccammon d. , 1983 , apj , 270 , 119 moshir m. , 1989 . explanatory supplement to the iras faint source survey . ipac preprint 44 nulsen p.e.j . , 1986 , mnras , 221 , 377 nulsen p.e.j . , 1998 , mnras , 297 , 1109 odea c.p . , baum s.a . , 1996 , in cold gas at high redshift , eds bremer m.n . , van der werf p.p . , rttgering h.j.a . , carilli c.l . , kluwer , dordrecht , p.199 odea c.p . , baum s.a . , gallimore j.f . , 1994b , apj , 436 , 669 odea c.p . , gallimore j.f . , baum s.a . , 1995 , aj , 109 , 26 odea c.p . , baum s.a . , maloney p.r . , tacconi l.j . , sparks w.b . , 1994a , apj , 422 , 467 peres c.b . , fabian a.c . , edge a.c . , allen s.w . , johnstone r.m . , white d.a . , 1998 , mnras , 298 , 416 press w.h . , teukolsky s.a . , vetterling w.t . , flannery b.p . , 1992 , numerical recipes , cambridge university press reisenegger a. , miralda - escud j. , waxman e. , 1996 , apj , 457 , 11l rizza e. , burns j.o . , ledlow m.j . , owen f.n . , voges w. , bliton m. , 1998 , mnras , 301 , 328 romani r.w . , maoz d. , 1992 , apj , 386 , 36 schindler s. , hattori m. , neumann d.m . , bhringer h. , 1997 , a&a , 317 , 646 sijbring d. , de bruyn a.g . , jaffe w.j . , sancisi r. , in meurs e.j.a , fosbury r.a.e . , extranuclear activity in galaxies , european southern observatory , 1989 , garching bei munchen , p107 smail i. , hogg d.w . , blandford r. , cohen j.g . , edge a.c . , djorgovski s.g . , 1995 , mnras , 277 , 1 smail i. , ellis r.e . , dressler a. , couch w.j . , oemler a. , sharples r.m . , butcher h. , 1997 , apj , 479 , 70 sparks w.b . , macchetto f. , golombek d. , 1989 , apj , 345 , 153 squires g. , neumann d.m . , kaiser n. , arnaud m. , babul a. , bhringer h. , fahlman g. , woods d. , 1997 , apj , 482 , 648 tanaka y. , inoue h. , holt s.s . , 1994 , pasj , 46 , l37 thomas p.a . , fabian a.c . , nulsen p.e.j , 1987 , mnras , 228 , 973 valentijn e.a . , giovanelli r. , 1982 , a&a , 114 , 208 voit g.m . , donahue m. , 1995 , apj , 452 , 164 voit g.m . , donahue m. , 1997 , apj , 486 , 242 waxman e. , miralda - escud j. , 1995 , apj , 451 , 451 white d.a . , fabian a.c . , johnstone r.m . , mushotzky r.f . , arnaud k.a . , 1991 , mnras , 252 , 72 white d.a . , jones c. , forman w. , 1997 , mnras , 292 , 419 wise m.w . , sarazin c.l . , 1999 , apj , submitted wise m.w . , oconnell r.w . , bregman j.n . , roberts m.s . , 1993 , apj , 405 , 94

modelling of elliptical galaxies has a long history , extending back to a time when it was taken for granted that elliptical galaxies were axisymmetric and that all of their mass could be accounted for in stars . we now know that dark matter is a commmon component of galaxies , both at large ( ashman 1992 ) and small ( kormendy & richstone 1995 ) radii , and that triaxial configurations are possible ones for stellar systems ( schwarzschild 1979 ) . one goal of modern dynamical studies is accordingly to verify the presence of dark matter ; another is to detect departures from axisymmetry . because triaxial models are difficult to construct , one often begins by looking for simpler models that are consistent with the data . this `` model - building '' approach is popular since it requires a minimum of thought about the information content of the data : one simply builds a model and checks whether it reproduces the observations . but one can take a more sophisticated approach , and ask whether the data imply anything definite about the observed stellar system . for instance , one might attempt to falsify the axisymmetric hypothesis in a model - independent way . this `` inverse problem '' approach is more difficult but also more rewarding , since it leads to more secure conclusions about the dynamical state of the galaxy . beginning with binney , davies & illingworth s ( 1990 ) pioneering study , a standard scheme has been developed for modelling elliptical galaxies . one assumes that the observed galaxy is axisymmetric and that the distribution of mass is known for instance , mass might be assumed to follow light . the gravitational potential is then derived from this mass distribution using poisson s equation ; the unique two - integral distribution function @xmath1 that reproduces the stellar density in this potential ( or more exactly , the even part of @xmath2 ) can also be found ( lynden - bell 1962 ; hunter 1975 ) . the odd part of @xmath2 , which determines the rotational velocity field , is usually represented via some ad hoc parametrization . one then projects this derived @xmath2 back into observable space to find the predicted kinematical variables and compares them with the observations . if there is agreement , one can claim to have found an axisymmetric model that is consistent with the data . this technique has been used to reproduce the kinematical data for a few , well - observed elliptical galaxies ( van der marel et al . 1994 ; dehnen 1995 ; qian et al . but it is difficult to state precisely what has been learned from studies like these . are the observed galaxies actually characterized by a constant @xmath3 , axisymmetry , and a two - integral distribution function , as assumed ? or might there exist models with spatially - varying @xmath3 s or three - integral @xmath2 s that reproduce the data equally well ? and if no model consistent with the data can be found , which of the various assumptions made in the model - building has been violated ? a useful comparison can be made here with spherical systems ( 2.1 ) . if one assumes that the gravitational potential @xmath4 of a spherical galaxy is known , then the observed velocity dispersion profile can be used to infer the unique dependence of anisotropy on radius ( binney & mamon 1982 ) . alternatively , one can assume that the distribution of stellar velocities is everywhere isotropic , in which case the velocity dispersion profile implies a unique form for @xmath4 ( merritt & gebhardt 1994 ) . assuming _ both _ isotropy _ and _ a constant @xmath3 would clearly be an over - determination of the problem , since the two assumptions together imply a unique velocity dispersion profile , and this predicted profile would almost certainly be different from the observed profile indeed , the kinematical data would not have been used at all in the construction of the model . but this is precisely what is commonly done in modelling axisymmetric galaxies : one assumes both isotropy in the meridional plane ( i.e. @xmath5 ) as well as an ad hoc form for the potential , and these two assumptions together imply a unique dependence of the mean square velocity on position over the image of the galaxy . it is unlikely that a model constructed in this way would reproduce the two - dimensional velocity dispersion data ; and if it did not , there would be no clear indication of which assumption had been violated . it is clearly desirable to think more carefully about the information content of kinematical data in axisymmetric stellar systems . unlike the spherical case , the functions to be derived depend on two spatial variables , but the kinematical data are likewise two - dimensional , depending on both projected radius and position angle . many such data sets now exist , for stars in globular clusters ( meylan et al . 1995 ) and in galaxies ( bacon et al . 1994 ) ; emission - line objects around the galactic bulge ( te linkel hekkert et al . 1991 ) and around other galaxies ( hui et al . 1995 ) ; and galaxies in clusters ( colless & dunn 1996 ) . these data are most often in the form of discrete velocities , and in the best - studied systems , the measured velocities number in the hundreds or even thousands . here it is shown ( 2.2 ) that the availability of two - dimensional kinematical data allows one to approach the dynamical study of axisymmetric stellar systems as an inverse problem rather than as a model - building problem . from the single _ ansatz _ @xmath5 , one can infer the unique gravitational potential @xmath6 that is consistent with the observed first and second velocity moments of @xmath2 . the rotational velocity field can likewise be obtained in a model - independent way . by using the kinematical data in the construction of the model , rather than assuming that mass follows light , one thus arrives at the unique pair of functions @xmath7 that are consistent with the two - integral hypothesis . if this unique model can be falsified for instance , by using line - of - sight velocity distributions , proper motions , or some other data then the two - integral , axisymmetric hypothesis will have been convincingly ruled out . in the model - building approach , on the other hand , one can not discard the two - integral assumption until one has explored an effectively infinite number of possible forms for the potential ( 2.3 ) . the numerical techniques that are required for converting the velocity data into a map of the potential are rather more subtle than the ones that have been applied up to now in the study of axisymmetric galaxies . these techniques are accordingly described in some detail ( 3.1 ) before applying them to pseudo - data generated from a family of oblate models ( 3.2 , 3.3 ) . although this approach yields the unique potential consistent with the two - integral assumption for @xmath2 , the true potential might be different if @xmath2 depends on a third integral . however it would be premature to investigate three - integral models until one had first used the algorithms described here , or equivalent ones , to rule out a two - integral model . having done so , one could then search for the pair of functions @xmath8 , @xmath9 that are most consistent with the data . this is a hard problem , and even in the spherical geometry there is no published algorithm that can extract @xmath4 and @xmath10 in a completely model - independent way from the data ; the most sophisticated such algorithms are still based on parametrized forms for the potential ( saha & merritt 1993 ; merritt 1993b ) . the focus here is on situations where the velocities are measured discretely , as would be the case for stars in a globular cluster . most large , kinematical data sets are of this form . it is also assumed throughout that the observer lies in the equatorial plane of the observed system . the reason for this unpleasant assumption is that deprojection of the luminosity density becomes underdetermined if the inclination angle is less than @xmath11 , even if this angle is assumed known ( rybicki 1986 ; gerhard & binney 1996 ) . one can therefore probably not hope to uniquely infer the dynamical state of an axisymmetric galaxy that is not viewed edge - on . this fact is routinely ignored in the model - building studies but must be faced squarely if one wishes to solve the inverse problem . also presented here ( 3.4 ) is the first regularized algorithm capable of deriving @xmath1 for an axisymmetric system from its density and potential ( the lynden - bell hunter problem ) given imperfect or incomplete information about those functions . our starting point for the axisymmetric inverse problem is the simpler spherical problem . a spherical stellar system has a distribution function that depends on the orbital energy @xmath12 and angular momentum @xmath13 ( both defined here per unit mass ) . spherical symmetry in velocity space is assumed as well ; then @xmath14 $ ] , and the radial and azimuthal velocity dispersions are everywhere equal , @xmath15 . having measured a set of discrete positions and velocities , one can compute smooth approximations to the surface density profile @xmath16 and the line - of - sight velocity dispersion profile @xmath17 . the function @xmath18 that defines the intrinsic velocity dispersion is then related to the observed velocity dispersions via @xmath19 which when inverted gives @xmath20 here @xmath21 is the spatial density of the kinematical sample , obtained by deprojecting the surface density @xmath16 : @xmath22 the mass within @xmath23 and the potential follow from jeans s equation in spherical symmetry : @xmath24 and the mass density is @xmath25 finally , the unique isotropic distribution function describing the kinematical sample is given by eddington s formula : @xmath26 thus both @xmath2 and @xmath27 follow uniquely from the observed surface density and velocity dispersion profiles , under the sole assumption of spherical symmetry in velocity space . does not imply @xmath28 ; one can construct `` isotropic '' stellar systems with distribution functions that depend on @xmath29 as well . ] here we are treating the stars as if they were ions in an x - ray emitting gas ; the quantity @xmath30 plays the role of @xmath31 in the gas , and no assumption has been made that mass follows light . in spite of its limitations , this approach has the following things to recommend it . one begins from a statement of the problem that is fully determined : there is a unique pair of functions @xmath32 that reproduce any set of observed profiles @xmath33 . in practice one does not measure @xmath16 or @xmath17 precisely , but nonparametric algorithms can be defined which yield estimates @xmath34 that are suitably close to the true functions ( merritt & gebhardt 1994 ) . the dependence of @xmath27 on @xmath23 is determined from the velocity data , as it should be , rather than being derived from the luminosity density under the always - questionable assumption that mass follows light . the validity of the single assumption that @xmath2 is spherically symmetry in velocity space can be checked . one simply projects the derived model to find the predicted , joint distribution of positions and line - of - sight velocities : @xmath35dv'^2.\ ] ] if this function is consistent with the observed line - of - sight velocity distribution at every radius , then the model is fully consistent with the data ( assuming that no other relevant data exist ) ; if not , isotropy can be ruled out , independent of any preconceptions about the radial variation of m / l . often the kinematical data will be insufficient in quality or quantity to produce a statistically significant discrepancy with the computed @xmath36 , because anisotropy generates only slight changes in the form of the line profiles and very high quality data are required to detect these changes ( gerhard 1993 ; merritt 1993b ) . nevertheless this approach makes manifest a point that is often misunderstood : one can not , in principle , make inferences about the degree of velocity anisotropy in a spherical system from the velocity dispersion data alone , since one can always construct the potential in such a way as to reproduce any observed @xmath17 . \4 . for some stellar systems e.g. globular clusters the assumption of isotropy in velocity space can be physically motivated . our goal is to identify a similar inverse problem for axisymmetric stellar systems . we wish to show that complete knowledge of the line - of - sight velocity moments @xmath37 and @xmath38 on the plane of the sky , and of the surface density @xmath39 of the kinematical population , is equivalent to knowledge of the two - integral distribution function @xmath1 describing that population and of the potential @xmath6 in which it moves independent of any assumptions about the relative distribution of mass and light . as discussed above , it is assumed throughout that the observer lies in the equatorial plane of the observed system . the jeans equations that relate the potential @xmath6 of an axisymmetric system to gradients in the velocity dispersions are @xmath40 here @xmath41 are the usual cylindrical coordinates ; all quantities are assumed independent of @xmath42 . now set @xmath5 . the velocity dispersions are then isotropic in the meridional plane , @xmath43 , and the jeans equations reduce to @xmath44 define cartesian coordinates @xmath45 on the plane of the sky , with @xmath46 parallel to the symmetry axis and @xmath47 along the line of sight . we assume that @xmath48 has been derived from @xmath49 via the usual inversion , unique for an edge - on galaxy ( rybicki 1986 ) . @xmath6 is still unknown . but if we multiply the first jeans equation by @xmath50 and the second by @xmath51 and equate , we find a relation between @xmath52 , @xmath53 and @xmath54 that is independent of @xmath27 : @xmath55 this relation may be seen as specifying a unique @xmath52 given @xmath54 and @xmath53 , i.e. @xmath56 dz + \sigma^2(\varpi , z ) . \label{magic2}\ ] ] now consider the rotational velocity field . the mean line - of - sight velocity @xmath57 is related to the internal streaming velocity @xmath58 via the projection integral @xmath59 ( fillmore 1986 , eq . equation ( [ sigmavy ] ) has inversion @xmath60 { dx\over\sqrt { x^2-\varpi^2}}. \label{nuvphi}\ ] ] note that @xmath53 is completely determined by @xmath61 independent of the two - integral assumption all that is required here is the assumption of axisymmetry . the possibility of computing the rotational velocity field via direct inversion of the data has been pointed out ( though never implemented ) by a number of authors including meylan & mayor ( 1986 ) and merrifield ( 1991 ) . the mean square line - of - sight velocity is @xmath62{\varpi d\varpi\over\sqrt{\varpi^2-x^2 } } \label{disp}\ ] ] ( fillmore 1986 , eq . the integrand on the right hand side contains the _ two _ unknown functions @xmath63 and @xmath64 , and at first sight does not admit of a unique inversion . but we have a relation between @xmath63 and @xmath65 , equation ( [ magic2 ] ) . we can therefore replace @xmath65 in the integrand by a linear expression that depends only on @xmath63 and on @xmath66 . although no analytic expression for the inverse relation could be found by the author , it will be shown below that @xmath67 is numerically well - determined by this equation . having derived @xmath67 , @xmath68 then follows from equation ( [ magic2 ] ) , and @xmath69 . it has thus been shown that @xmath58 , @xmath70 and @xmath71 are determined by the two - dimensional velocity moment data in an edge - on , two - integral system without any assumptions aside from isotropy in the meridional plane . it follows that the potential @xmath6 and the mass density @xmath72 are also known : the former from equations ( 10 ) or ( 11 ) , the latter from poisson s equation : @xmath73.\ ] ] so derived , @xmath74 will of course not necessarily be proportional to @xmath66 . finally , there is a unique distribution function @xmath1 describing the kinematical population that reproduces @xmath66 and @xmath53 in the derived potential . as usual , we divide @xmath2 into odd and even parts with respect to @xmath75 : @xmath76.\ ] ] the number density is related to @xmath77 via @xmath78 and the rotational velocity field is related to @xmath79 via @xmath80 these equations imply unique functions @xmath77 and @xmath79 given @xmath66 , @xmath53 and @xmath27 ( lynden - bell 1962 ; hunter 1975 ; dejonghe 1986 ) , although the inverse relations are complex . once again , since mass is not assumed proportional to light , the population described by this @xmath2 will not necessarily have the same spatial distribution as the matter that determines the potential . practical algorithms for carrying out the inversions , suitable for real ( i.e. noisy and incomplete ) data , are described below . just as in the spherical case , the validity of the assumed relation between @xmath81 and @xmath52 can be verified by computing the full distribution of line - of - sight velocities @xmath82 from the model and comparing with the observed velocity distributions ; or by carrying out the same procedure with the predicted and observed proper motions ; etc . if these predicted and observed distributions are significantly different , one can conclude that the assumption of isotropy in the meridional plane is not valid ( or that the system is not being viewed edge - on , or is not axisymmetric ) and proceed to investigate more general models . note that just as in the spherical case one can not infer the presence of velocity anisotropy ( @xmath83 ) based on inspection of the velocity dispersions alone , since the potential can always be constructed so as to reproduce the observed dispersions without a third integral . information about the dependence of @xmath2 on a third integral is contained only in the higher - order moments of the line - of - sight velocity distribution ( or the proper motions ) ; and conversely , there is no need to use that extra information unless one intends to investigate three - integral models the two - integral problem is fully constrained by the the first and second velocity moments of @xmath2 . it is interesting to compare this approach to the one pioneered by binney , davies & illingworth ( 1990 ) ; their technique has recently formed the basis of a number of data - based studies by van der marel and collaborators ( van der marel , binney & davies 1990 ; van der marel 1991 ; van der marel et al . similar approaches to the axisymmetric inverse problem have been worked out by merrifield ( 1991 ) , dejonghe ( 1993 ) , emsellem , monnet & bacon ( 1994 ) , dehnen ( 1995 ) , kuijken ( 1995 ) , and qian et al . all of these authors assume at the outset that mass follows light ( or that the mass distribution has some other ad hoc form , often including a central point mass ) , and derive the two - dimensional potential from the assumed mass distribution using poisson s equation . the depth of the potential , or equivalently the mass - to - light ratio , is obtained from the virial theorem . the observed velocities are not otherwise used in the construction of the model , except insofar as they constrain the form of the odd part of @xmath2 , which is typically represented via some simple parametrization . the advantage of this method is its simplicity : once the potential is specified , the even part of the distribution function @xmath84 or any of its moments can be obtained ( lynden - bell 1962 ; hunter 1975 ; dejonghe 1986 ) . but this simplicity is achieved at the cost of having to assume the forms of @xmath6 and @xmath79 at the outset . it was shown above that the potential follows uniquely from the observed @xmath85 and @xmath86 in a two - integral , edge - on system one is not free to assume arbitrary functional forms for @xmath27 . furthermore the odd part of @xmath2 can be derived in a model - independent way from the observed rotational velocity field . it follows that the approach taken by these authors is overdetermined . models constructed in an assumed potential are unlikely to reproduce a two - dimensional set of velocity dispersions , since the information contained in those data was not used . and if the model should fail to reproduce the kinematical data , the failure could be due to the breakdown of either of two assumptions : that mass follows light , or that @xmath2 depends on only two integrals . in the method described here , on the other hand , the potential is constructed to have the unique form that is consistent with the two - integral assumption for @xmath2 and with its lowest observable moments . a failure of the model to reproduce any additional kinematical data could only mean that @xmath87 . the model - building technique has nevertheless been successfully applied to a few galaxies , notably m32 ( van der marel et al . 1994 ; dehnen 1995 ; qian et al . 1995 ) . but this apparent success should be interpreted with caution . it is possible that these galaxies satisfy all of the assumptions made in the model building , i.e. axisymmetry , @xmath5 , and @xmath88 constant . but it is also possible that the published models would fail to reproduce a full , two - dimensional set of velocities if such data were available . future observations of these galaxies should reveal which of these two possibilities is correct . of course , binney , davies & illingworth ( 1990 ) were concerned with situations where the kinematical data are confined to just a few cuts across the image of the galaxy . given such limited data , their approach seems entirely justified . at the same time , our discussion makes clear that when two - dimensional data _ are _ available , as is increasingly often the case one can and should infer the kinematics and potential directly from the velocities . merrifield ( 1991 ) proposed a test for the sufficiency of two - integral models when describing edge - on , axisymmetric galaxies . his test is based on the assumption that mass follows light , and uses as a discriminant the measured velocity dispersions . since one can always construct the potential so as to reproduce the velocity dispersion data in a two - integral system , as shown here , conclusions based on merrifield s test are contingent on the assumption that mass follows light . the formal solutions given above to the axisymmetric inverse problem are of little use when dealing with real data , since the inversion equations are ill - conditioned in the sense understood by statisticians ( miller 1974 ; osullivan 1986 ) . this means that errors or incompleteness in the data will be amplified when going from data space to model space unless some objective smoothness condition is placed on the solution . the degree to which errors are amplified depends on the number of effective differentiations that separate the model from the data . for instance , the distribution function @xmath1 is essentially a second derivative of the number density ( eq . [ fela ] ) which is itself a half - order derivative of the surface brightness ( eq . [ nur ] ) hence the determination of @xmath2 is strongly ill - conditioned even though formally unique . some authors have ignored this element of the problem and carried out direct deprojections without imposing smoothness constraints . the results tend to be spectacularly noisy ( e.g. figure 2 of kuijken 1995 ) . another approach , equally ill - advised , is to replace the data by ad hoc fitting functions for which the inversions can be carried out exactly ( e.g. qian et al . unless these smooth functions are generated from the data using a nonparametric algorithm , they will contain a bias that is likewise amplified by the inversion . solutions obtained in this way may look appealing in a @xmath89 sense but are likely to be far from the true solutions . richardson - lucy iteration ( e.g. dehnen 1995 ) is genuinely nonparametric , going in discrete steps from an initial guess to the ( ill - defined ) solution that would have been obtained via direct inversion of the data . however the success of this technique depends on the accuracy of the initial guess . a bad guess will require many iterations before the projected solution approximates the data , at which point the solution is likely to be unacceptably noisy . one typically halts the iteration before this occurs but the solution so obtained will accordingly be biased in the direction of the initial guess by some unknown amount . furthermore , solutions obtained in this way tend to achieve their optimal degree of smoothness in one part of solution space while remaining undersmoothed in other parts . our goal is an algorithm which generates solutions that are characterized by acceptably small values of _ both _ the noise and the bias . modern nonparametric methods ( mller 1980 ; hrdle 1990 ; wahba 1990 ; scott 1992 ; green & silverman 1994 ) are designed with precisely this goal in mind . these methods make no assumptions about the global form of the solutions ; they deal with the ill - conditioning by imposing smoothness via an effectively local constraint on the level of fluctuations in the solution , an approach called `` regularization . '' they thus avoid the pitfalls of both richardson - lucy iteration ( no regularization ) and model fitting ( parametric bias ) . our focus is on situations where the velocities are measured discretely , as would be the case for stars in a globular cluster , planetary nebulae around a galaxy , galaxies in a cluster , etc . modifying the algorithms to deal with spatially - continuous data is straightforward . we assume that the number density profile @xmath48 of the kinematical sample is known . nonparametric algorithms for estimating @xmath66 in the spherical geometry have been described by merritt ( 1993a ) in the case that the surface brightness is measured directly , and by merritt & tremblay ( 1994 ) in the case that the surface brightness must first be inferred from discrete positions . these algorithms are easily generalized to the axisymmetric case . we also continue to assume , as discussed above , that the observer lies in the equatorial plane of the observed system , since otherwise the inverse problem is guaranteed not to have a unique solution ( gerhard & binney 1996 ) . the inverse problems to be solved here belong to a class of problems that have been well studied by statisticians in the last few years ( e.g. wahba 1990 ) . let the data be @xmath90 and the model @xmath91 . the data are related to the model via @xmath92 where the @xmath93 are known linear functionals in our case , projections of @xmath91 , and the @xmath94 represent measurement errors or scatter from some other source . for instance , in the determination of the rotational velocity field @xmath53 , we would have @xmath95 and @xmath96 the observed rotation at point @xmath97 . if the line - of - sight rotational velocity were measured directly , then @xmath98 and the @xmath94 would be measurement errors . if instead discrete velocities were measured , then @xmath99 , the observed velocity of the @xmath100th star , and @xmath101 , the deviation ( due both to measurement errors and to the intrinsic velocity dispersion ) of the @xmath100th star s measured velocity from the true mean line - of - sight velocity at point @xmath97 . we seek a smooth function @xmath102 such that @xmath103 is suitably close to the data . one possible approach is to construct a continuous approximation to the function represented by the data e.g. @xmath57 using a kernel or spline smoother , and then to operate mathematically on this function via the inverse operator @xmath104 ( if it exists ) to produce the estimate @xmath105 . this approach is consistent in the sense defined by statisticians as long as the functions that are fit to the data are asymptotically unbiased , i.e. nonparametric ( wahba 1990 , p. 19 ) , and such an approach has been used with success by merritt & gebhardt ( 1994 ) to solve the dynamical inverse problem in spherical geometry . such an approach will not be followed here because one of the inverse problems that we wish to solve the recovery of @xmath1 from @xmath66 , @xmath53 and @xmath27 does not have a simple inverse operator . in addition , one sometimes wishes to place physical constraints ( e.g. positivity ) on the solutions and this is extremely difficult to do if the function approximations are carried out in data space rather than solution space . instead we define our solutions implicitly , as the functions @xmath102 that minimize functionals of the form @xmath106 the first term in equation ( [ opt ] ) measures the fidelity of the model to the data . minimizing this term alone would lead to an unacceptably noisy solution due to the ill - conditioning of the inverse operator . the second term measures the degree to which the solution is unsmooth . one standard choice for the `` penalty function '' @xmath107 is @xmath108 ( here and below , a two - dimensional @xmath109 is assumed . ) the function @xmath102 that minimizes equation ( [ opt ] ) with penalty function ( [ penaltya ] ) is the so - called `` thin plate spline '' solution ( duchon 1977 ; meinguet 1979 ) . remarkably , this solution can be found in essentially analytic form for many simple operators @xmath110 ( wahba & wendelberger 1980 ) . the qualitative nature of the solution to equations ( [ opt ] ) and ( [ penaltya ] ) depends on the choices of @xmath111 and @xmath112 . the penalty function acts as a low - pass filter , with @xmath111 controlling the half - power point of the filter and @xmath112 the steepness of the roll - off ( craven & wahba 1979 ) . typically one chooses @xmath113 ; the smoothness of the solution is then controlled by varying @xmath111 . too large a value for @xmath111 will oversmooth the solution , i.e. reduce its curvature ; too small a value will yield an overly noisy solution . the larger the data set and the smaller the measurement errors , the smaller the value of @xmath111 that can be profitably used and the more accurate the solution . thus the bias the average deviation of the solution from the true function falls to zero as the sample size @xmath114 is increased . in parametric algorithms , by contrast , the bias remains finite even as @xmath114 tends to infinity , since the adopted parametric form is guaranteed to be incorrect at some level and this incorrectness does not diminish as @xmath114 is increased . it is for this reason that parametric methods are unsuited to data - based inverse problems . data are typically incomplete or cover only a limited region . the use of a penalty function guarantees that the solution will tend toward a predictable form in parts of solution space that are not strongly constrained by the data . for instance , the thin - plate spline solution with @xmath113 yields a solution that is linear in @xmath115 and @xmath116 wherever the data do not force a different functional form , since any linear function is assigned zero penalty by expression ( [ penaltya ] ) . this very useful feature of solutions found via penalty functions means that there is never any need to extend or augment the data in ad hoc ways . nonparametric techniques based on direct smoothing of the data , e.g. via kernels , do not share this nice property . many variations on this basic scheme exist . if the data errors are not normally distributed , one can choose the first term in equation ( [ opt ] ) to be some more robust measure of the fidelity of the model to the data . such a modification would be appropriate in the study of galaxy cluster dynamics where interlopers are common . another variation is to modify the form of the penalty function . the simplest such variation is to add a factor @xmath117 with @xmath118 some positive constant ; changing @xmath118 is equivalent to changing the units in which @xmath116 is measured . one can also include in the penalty function a weighting function @xmath119 that is large where the solution @xmath91 is expected to be small , thus encouraging the fluctuations in the solution to remain a fixed fraction of @xmath91 in amplitude everywhere . finally , one can replace the differential operator @xmath120 by some more complicated operator , which defines as `` perfectly smooth '' some different class of functions ( e.g. silverman 1982 ) . for many choices of @xmath107 , including the thin plate penalty function defined above , the quantity to be minimized is quadratic in @xmath91 . this means that a standard quadratic programming routine ( rustagi 1994 ) can be used to find the solution on some grid in model space . the use of a quadratic programming algorithm also allows the easy imposition of linear constraints , such as positivity , on the solution if desired . there is a huge literature concerning the choice of the smoothing parameter @xmath111 . while `` objective '' schemes exist , a standard practice is to slowly decrease @xmath111 until the solution begins to exhibit small - scale fluctuations , then stop . consider first the recovery of the rotational velocity field in the meridional plane @xmath58 . we seek the function @xmath121 that minimizes @xmath122 d\varpi dz , \label{opta}\ ] ] with @xmath123 the measured velocities and @xmath124 we choose @xmath53 as our function to be optimized rather than the physically more natural choice @xmath125 , since @xmath53 is expected to be a more nearly linear function of the coordinates and hence better suited to a second - derivative penalty function . we begin by defining a cartesian grid in model space , @xmath126 , @xmath127 , and @xmath128 , @xmath129 . the variables to be solved for are the @xmath130 values @xmath131 of the rotational velocity on the grid . the derivatives and integrals on the right hand side of ( [ opta ] ) can easily be represented as linear combinations of the @xmath131 using standard expressions from numerical analysis ; for details from a very similar problem , see merritt & tremblay ( 1994 ) . placing the @xmath131 into a linear vector @xmath132 , one can then write @xmath133 where the quantities @xmath134 , @xmath135 and @xmath136 are straightforward to compute but messy and will not be given here . the resulting expression is quadratic in the unknowns , and the constraints @xmath137 are linear . the values @xmath138 that minimize @xmath89 can accordingly be found via quadratic programming . figure 1 shows an estimate of @xmath139 derived using 5000 discrete velocities drawn randomly from an oblate model belonging to lynden - bell s ( 1962 ) family . the adopted model is the flattest one consistent with non - negative @xmath2 , having @xmath140 in lynden - bell s notation and @xmath141 in the notation of dejonghe ( 1986 ) . ( these models have non - constant ellipticity , becoming spherical at large radii for all values of @xmath142 . ) the odd part of @xmath2 , which specifies the degree of ordered motion , was set to the function derived in the appendix of lynden - bell s paper when generating the monte - carlo data . the resultant rotational velocity field has @xmath143 everywhere , an `` oblate isotropic rotator . '' note that even this extreme model from lynden - bell s sequence is only mildly flattened , except very near the center , and the rotation is correspondingly small ; the peak rotation velocity is only about one half of the central , one - dimensional velocity dispersion . in spite of the small signal , however , the inversion algorithm recovers the rotational velocity field reasonably well near the center ; the main defect is a too - shallow rise of the rotation velocity near @xmath144 , an inevitable result of the smoothing . this example suggests that data sets consisting of a few thousand velocities are none too large for recovering the rotational velocity field of an axisymmetric system in a model - independent way . we now wish to make estimates of @xmath71 and @xmath70 . we search for the functions @xmath145 and @xmath146 that minimize @xmath147 ^ 2 \nonumber \\ & + & \lambda\int\int\left[\left({\partial^2 \sigma^2\over \partial\varpi^2 } \right)^2 + 2\left({\partial^2 \sigma^2 \over \partial\varpi\partial z } \right)^2 + \left({\partial^2 \sigma^2 \over \partial z^2}\right)^2\right ] d\varpi dz \nonumber \\ & + & \lambda\int\int\left[\left({\partial^2\overline{v^2_{\phi}}\over \partial \varpi^2}\right)^2 + 2\left({\partial^2 \overline{v^2_{\phi}}\over\partial\varpi \partial z}\right)^2 + \left({\partial^2 \overline{v^2_{\phi } } \over \partial z^2 } \right)^2\right ] d\varpi dz \label{optb}\end{aligned}\ ] ] subject to the constraints of equation ( [ magic ] ) : @xmath148 as well as the usual positivity constraints . the operator @xmath93 is now defined as @xmath149 { d\varpi\over\sqrt{\varpi^2-x_i^2}}. \label{projb}\ ] ] if the velocities contain measurement errors , these should be subtracted in quadrature from the @xmath123 . we define a double grid in @xmath150 , with the first @xmath130 points containing the @xmath63 values and the second @xmath130 points the @xmath65 values . the constraints ( [ magic3 ] ) are approximated via finite - difference derivatives , which have the effect of linking together the values on the two grids . finally , one obtains @xmath52 via @xmath151 , with @xmath152 having been recovered in the previous step . before applying the algorithm to pseudo - data , one would like some assurance that the information contained in the observed mean - square velocity field @xmath86 is sufficient , in principle , to uniquely constrain both @xmath71 and @xmath70 . ideally this hypothesis would be demonstrated by finding analytical expressions for the inverse relations . such expressions could not be found ( which may simply reflect limitations in the author s mathematical abilities ) , and instead the case will here be made numerically . figure 2 shows how the integrated square error in the solutions @xmath153 , @xmath154 , @xmath155 depends on the degree to which the constraints ( [ magic3 ] ) are imposed . that is , these constraints were replaced by @xmath156 with the parameter @xmath157 a constant at all points on the solution grid determining how stringently the contraints are imposed . the algorithm was given as input the exact values , on a grid , of the mean square velocity @xmath86 of the rotating oblate model of figure 1 . solutions were then obtained as a function of @xmath158 . figure 2 demonstrates that the error in the solutions tends to zero along with @xmath158 , which suggests at least for this particular model that the constraints ( [ magic3 ] ) contain just enough information to yield the correct inversion . this experiment was repeated for a number of other models from the lynden - bell sequence of models . for each model examined , the solution with @xmath159 was found to be extremely close to the true solution . these experiments hardly constitute a rigorous proof that the inversion is uniquely defined for every possible two - integral oblate model , however . figures 3a and 3b illustrate the recovery of @xmath71 and @xmath70 given the same 5000 velocities drawn from the rotating oblate model of figure 1 . given the uniqueness , just demonstrated , of the inversion for this model , the discrepancies apparent in the numerical solutions of figure 3 are primarily due to the finiteness of the data set rather than to any limitations of the algorithm . different , 5000-velocity monte - carlo realizations of the same model were found to yield solutions that were different in their details but that had roughly the same average error . the gravitational potential is obtained by applying either of equations ( 10 ) or ( 11 ) , in finite - difference form , to the derived @xmath54 and @xmath52 . the result is shown in figures 3c and 3d . the derived potential is remarkably similar to the correct one except for slightly reduced gradients , again an unavoidable result of the smoothing . we now want to find the functions @xmath84 and @xmath160 that generate the estimated number density @xmath161 and rotational velocity field @xmath162 in the reconstructed potential @xmath163 . this is a simpler version of a problem solved in merritt ( 1993b ) , and we follow a similar approach here . a rectangular grid is defined in @xmath164-space , and the unknowns @xmath165 and @xmath166 are assumed constant over each cell ( figure 4 ) . the configuration - space density at any point @xmath167 is the sum of the contributions @xmath168 , for each cell such that @xmath169}$ ] . ( cells that lie partly below and partly above this curve contribute by an amount proportional to the area that lies below the curve . ) similarly , each cell contributes @xmath170 to the estimate of @xmath171 . the quantity to be minimized is @xmath172 de dl_z , \label{optc}\end{aligned}\ ] ] in the case of @xmath77 , and @xmath173 de dl_z , \label{optd}\end{aligned}\ ] ] in the case of @xmath79 . the integrals are understood to extend over the region in the @xmath164 plane such that @xmath174 , where @xmath175 is the maximum value of @xmath75 attainable at energy @xmath12 ( fig . 4 ) . the smoothing parameters @xmath176 and @xmath177 are allowed to depend on the energy since @xmath2 falls off much more rapidly than linearly with @xmath12 , and a penalty function with constant @xmath111 would not strongly penalize a solution that was smooth at large @xmath178 but noisy for @xmath179 . furthermore , a different degree of smoothing is permitted in the @xmath12 and @xmath75 directions since @xmath2 is expected to be much more homogeneous in @xmath75 than in @xmath12 . figure 5 illustrates the recovery of @xmath77 and @xmath79 using the approximations to @xmath6 and @xmath53 obtained in the previous sections . the smoothing parameters were chosen to vary as @xmath180 , with @xmath181 . once again , the only obvious systematic error takes the form of a bias due to the smoothing . the derivation of @xmath1 given @xmath48 and @xmath182 is a classical problem in stellar dynamics . lynden - bell ( 1962 ) , hunter ( 1975 ) and dejonghe ( 1986 ) demonstrated the uniqueness of the solution , and recently hunter & qian ( 1993 ) have presented an elegant algorithm for carrying out the inversion given analytic expressions for @xmath66 and @xmath27 . hunter & qian s algorithm is not regularized and hence not suitable for use with real data . it has nevertheless been applied in this way , after first replacing the data with analytic models ( qian et al . this approach is dangerous at best since any slight error in the choice of fitting functions will be amplified enormously by the inversion operator . solutions so obtained for @xmath2 may be smooth and physically appealing , but they are unlikely to be close to the true solutions . from a numerical point of view , the only non - trivial aspect of solving the lynden - bell hunter problem is the regularization ; the inversion itself is essentially just a matrix operation once the integral operator has been represented via some discrete approximation . much effort has been directed recently toward refining the discretization ( kuijken 1995 ) or the inversion ( hunter & qian 1993 ) , with curiously little attention paid to the more crucial question of regularization . given measurements of the line - of - sight rotational velocity and the velocity dispersion over the two - dimensional image of an edge - on , axisymmetric galaxy , one can derive the unique functions @xmath6 and @xmath1 that reproduce those data , without the necessity of assuming that mass follows light . the validity of such a model can then be tested using other data , such as the full line - of - sight velocity distributions or the proper motions . failure to reproduce the data using this method would imply that @xmath2 depends on a third integral of motion ( or that one of the geometric assumptions , axisymmetry or zero inclination , was violated ) . this approach is a generalization of the one pioneered by binney , davies and illingworth ( 1990 ) which assumes that mass follows light and which derives both @xmath27 and @xmath2 from the luminosity distribution alone , without using the kinematical data except in the normalization of @xmath27 . we have presented numerical algorithms , suitable for use with noisy and incomplete data , for carrying out the required inversions and shown that they provide smooth and accurate estimates of @xmath2 and @xmath27 using simulated data sets derived from oblate models . the second - derivative penalty functions adopted here are almost certainly non - optimal and it would be worthwhile to investigate more general forms . in addition , it is possible that the solutions to the roughness - penalized inverse problems defined here can be found analytically , as in wahba & wendelberger ( 1980 ) . \3 . as shown most recently by gerhard & binney ( 1996 ) , the axisymmetric inverse problem becomes under - determined when the stellar system is viewed from a direction that is not parallel to the symmetry plane . additional information must therefore be added if one is to solve the inverse problem for an arbitrary orientation . ( one would also like to _ infer _ that orientation from the data . ) there is at present no clear understanding of what form that extra information should take . one can define an alternate inverse problem based on the _ ansatz _ that mass follows light , or more generally that the form of the potential @xmath6 is known a priori ; for instance , the potential might be derived from x - ray data , from the statistics of gravitational lensing , etc . this is of course the usual assumption that is made when modelling axisymmetric galaxies ; however the goal would be to learn something unique about the degree of velocity anisotropy @xmath184 from the velocity moment data , as in the analogous spherical problem ( binney & mamon 1982 ) . unfortunately one can show that the assumption of a known potential is _ not _ sufficient in itself to uniquely constrain the anisotropy , even for an edge - on system ; one must make additional assumptions , e.g. that the velocity ellipsoid is everywhere aligned with the coordinate axes . nevertheless , more work along these lines would help to elucidate the more difficult inverse problem for a three - integral @xmath2 . this work was supported by nsf grant ast 90 - 16515 and nasa grant nag 5 - 2803 . t. fridman kindly assisted in the plotting of the figures . i thank w. dehnen and o. gerhard for lively discussions on the topic of axisymmetric modelling , and for pointing out some mathematical errors in an early version of the paper . i also thank h. dejonghe for many useful insights on problems of this sort . ashman , k. 1992 , pasp , 104 , 1109 bacon , r. , emsellem , e. , monnet , g. & nieto , j. l. 1994 , aap 281 , 691 binney , j. , davies , r. & illingworth , g. 1990 , apj , 361 , 78 binney , j. & mamon , g. a. 1982 , mnras , 200 , 361 colless , m. & dunn , a. m. 1996 , apj , 458 , 435 craven , p. & wahba , g. 1979 , numer . 31 , 377 dehnen , w. 1995 , mnras , 274 , 919 dejonghe , h. 1986 , phys . 133 , 217 dejonghe , h. 1993 , in galactic bulges , iau symp . h. dejonghe & h. j. habing ( kluwer : dordrecht ) , 73 de zeeuw , p. t. 1985 , mnras , 216 , 273 duchon , j. 1977 , in constructive theory of functions of several variables ( springer : berlin ) , 85 emsellem , e. , monnet , g. , & bacon , r. 1994 , aap , 285 , 723 fillmore , j. a. 1986 , aj , 91 , 1096 gerhard , o. 1993 , mnras , 265 , 213 gerhard , o. & binney , j. 1996 , preprint green , p. j. & silverman , b. w. 1994 , nonparametric regression and generalized models ( chapman & hall : london ) hrdle , w. 1990 , applied nonparametric regression ( cambridge university press : cambridge ) hui , x. , ford , h. , freeman , k. c. & dopita , m. 1995 , apj 449 , 592 hunter , c. 1975 , aj , 80 , 783 hunter , c. & qian , e. 1993 , mnras , 262 , 401 kormendy , j. & richstone , d. 1995 , ann . , 33 , 581 kuijken , k. 1995 , apj , 446 , 194 kuzmin , g. g. & kutuzov , s. a. 1962 , bull . abastumani ap . 27 , 82 lynden - bell , d. 1962 , mnras , 123 , 447 meinguet , j. 1979 , j. appl . math . phys . 30 , 292 merrifield , m. r. 1991 , aj , 102 , 1335 merritt , d. 1993a , in structure , dynamics and chemical evolution of elliptical galaxies , ed . i. j. danziger , w. w. zeilinger & k. kjr ( eso : garching ) , 275 merritt , d. 1993b , apj , 413 , 79 merritt , d. & gebhardt , k. 1994 , in clusters of galaxies , proceedings of the xxixth rencontre de moriond , ed . f. durret , a. mazure & j. tran thanh van ( editions frontiere : singapore ) , p. 11 . merritt , d. & saha , p. 1993 , apj , 409 , 75 merritt , d. & tremblay , b. 1994 , aj , 108 , 514 meylan , g. & mayor , m. 1986 , aa 166 , 122 meylan , g. , mayor , m. , duquennoy , a. & dubath , p. 1995 , aa 303 , 761 miller , g. f. 1974 , in numerical solution of integral equations , ed . l. m. delves & j. walsh ( clarendon press : oxford ) , 175 . mller , h .- 1980 , nonparametric regression of longitudinal data ( springer : berlin ) osullivan , f. 1986 , stat . sci . 1 , 502 qian , e. e. , de zeeuw , p. t. , van der marel , r. p. & hunter , c. 1995 , mnras , 274 , 602 rustagi , j. s. 1994 , optimization techniques in statistics ( academic press : boston ) rybicki , g. b. 1986 , in structure and dynamics of elliptical galaxies , iua symp . p. t. de zeeuw ( kluwer : dordrecht ) , 397 scott , d. w. 1992 , multivariate density estimation , new york : john wiley & sons schwarzschild , m. 1979 , apj , 232 , 236 silverman , b. 1982 , ann . stat . , 10 , 795 te linkel hekkert , p. , caswell , j. l. , habing , h. j. , haynes , r. f. & norris , r. p. 1991 , aap suppl . 90 , 327 van der marel , r. p. 1991 , mnras , 253 , 710 van der marel , r. p. , binney , j. & davies , r. 1990 , mnras , 245 , 582 van der marel , r. p. , evans , n. w. , rix , h .- w . , white , s. d. m. & de zeeuw , t. 1994 , mnras , 271 , 99 wahba , g. 1990 , spline models for observational data ( siam : philadelphia ) wahba , g. & wendelberger , j. 1980 , monthly weather review , 108 , 1122 recovery of the rotational velocity field @xmath58 given 5000 velocities derived from an oblate , lynden - bell ( 1962 ) model with maximum flattening . heavy contours : estimated @xmath53 ; thin contours : exact @xmath53 . recovery of the velocity dispersions and potential from the same data used fig . heavy contours : estimated values ; thin contours : exact values . ( a ) @xmath54 ; ( b ) @xmath52 ; ( c ) @xmath27 ; ( d ) @xmath185 . grid - based scheme for recovery of @xmath84 . ( a ) solution grid . upper curve is the maximum @xmath75 as a function of @xmath12 ; lower curve is the characteristic curve for a given @xmath167 . ( b ) projection of a single @xmath164 grid cell into @xmath167-space .

the recent discovery of strongly coupled quark - gluon plasma at rhic @xcite has spurred a significant theoretical effort towards understanding the properties of matter described by strongly coupled quantum field theories at finite temperature . the transport properties of such theories are much more sensitive to the strength of the coupling than stationary thermodynamic properties . in particular , the near perfect fluidity in such theories is viewed as a tell - tale sign of the strong coupling . the transport properties of a theory are closely related to the spectral functions of the operators such as stress - energy tensor . in particular , viscosity can be determined from the low frequency limit of the spectral function by the well - known kubo formula . the first - principles calculation of the spectral functions using lattice monte carlo techniques is a challenging task @xcite , especially in the low - frequency regime , and the prior knowledge of the properties of spectral functions is essential . therefore , constraints on the spectral functions in the form of the sum rules for the integral of the spectral function has been the subject of recent attention@xcite . the discovery of the ads / cft holographic correspondence@xcite has opened new possibilities to study thermodynamics and transport in strongly coupled theories ( for reviews see @xcite ) . although not yet directly applicable to qcd , at least until we know its holographic dual , these methods allow one to study generic properties of the strongly coupled plasmas using model theories , such as @xmath0 susy ym , or holographic models which incorporate qcd - like features such as confinement . in particular , the shear channel sum rule derived in @xcite has been verified in the @xmath1 susy ym using ads / cft correspondence . in this paper , we concentrate on the sum rule for the spectral function in the bulk channel , corresponding to uniform dilatation or isotropic expansion . the bulk channel sum - rule is trivial in a conformal theory such as @xmath1 susy ym . however , nontrivial sum rules in qcd do exists and have been a subject of recent studies @xcite . unlike the shear channel , the bulk channel correlation function ( and associated bulk viscosity ) is sensitive not only to the strength of the coupling , but also to the amount of the scaling violation . therefore we consider a holographic model with the simplest mechanism of conformality violation , which is similar to qcd . we shall assume that this model describes field theories where scale invariance is broken by the presence of a scalar operator @xmath2 with dimension @xmath3 in the action . the scale anomaly , @xmath4 , is proportional to this operator , and in qcd this role is played by the gluon condensate operator @xmath5 . thermodynamics and transport in such theories have been first studied in ref . in particular , it has been shown that the speed of sound @xmath6 approaches the conformal value @xmath7 universally from _ below _ @xcite . the bulk sum rule has been also tested recently in ref . @xcite in a theory whose holographic dual is a dilaton gravity with the chamblin - reall dilaton potential @xcite , which has the virtue of being analytically tractable . however , the mechanism of the scaling violation in such a putative field - theory ( if the field - theory dual exists ) would differ from that in qcd . in particular , the speed of sound does not approach its conformal value in the high - temperature limit . in the general class of theories we consider , we shall find that the temperature dependent part of the bulk spectral function @xmath8_t}}$ ] obeys the following sum rule @xmath9_t}}}{\omega } d\omega.\ ] ] in the limit when the operator @xmath10 becomes marginal , @xmath11 , the sum rule coincides with the sum rule in qcd , derived in @xcite . we shall also find that in this limit , which we shall refer to as marginal , the sum rule in eq . ( [ eq : sum - rule - dplus ] ) exhibits the same puzzle as discussed in refs.@xcite in the context of qcd : the l.h.s . of the sum rule is of order @xmath12 , while @xmath13 , where @xmath14 is the qcd coupling . here , in the marginal limit , the l.h.s . of eq.(1 ) is of order @xmath15 while @xmath16 , where @xmath17 we shall demonstrate that a delicate cancellation indeed occurs between the high frequency tail of the integral , @xmath18 , and the intermediate region of @xmath19 , of the kind needed to occur in qcd . keeping only the leading @xmath20 terms in @xmath21_t}}$ ] , we find that the integral in eq . ( [ eq : sum - rule - dplus ] ) converges . however , the integral over the contribution of the subleading @xmath22 terms in @xmath21_t}}$ ] has support over an interval of @xmath23 which stretches to infinity as @xmath24 in the marginal limit @xmath25 . we evaluate the resulting additional @xmath26 contribution from this long tail to the r.h.s . ( [ eq : sum - rule - dplus ] ) analytically and show that it cancels the @xmath20 contribution from the @xmath19 region . this paper is organized as follows . in the next section , we present the definitions of quantities involved in the sum rule and its derivation . in section [ sec : sum - rule ] , we derive the sum rule , eq . ( [ eq : sum - rule - dplus ] ) , in the general class of theories we consider . section [ sec : holographic - model ] introduces the holographic description of such theories . section [ sec : two - point - functions ] explains our method of calculating the spectral function and the related green s functions . in section [ sec : large - freq - asympt ] , we make analytical calculation of the large @xmath23 asymptotics of @xmath27 , which we use in section [ sec : spec ] to subtract from the numerically determined @xmath27 to obtain @xmath8_t}}$ ] . the sum rule is verified numerically in section [ sec : sum ] for a sample of values of @xmath28 . we demonstrate the cancellation required to satisfy the sum rule in the marginal limit in section [ sec : sum ] . cross - check of the results with the existing analytical result for bulk viscosity ( ref.@xcite ) is made in section [ sec : viscosity ] . we conclude in section [ sec : conc ] . appendix [ sec : appa ] contains the relevant results from @xcite , used throughout the paper . we shall consider only the response to homogeneous ( @xmath29 ) perturbations and define the spectral function for the trace of the stress - energy tensor @xmath30 , as usual , by @xmath31\rangle.\end{aligned}\ ] ] the definition of the retarded green s function @xmath32 is subject to the usual ambiguity due to the contribution of the product of the operators at the same point at @xmath33 . however , these contact terms do not have imaginary parts and do not contribute to @xmath34 . it is convenient , as in @xcite , to define green s functions of @xmath30 by considering the response of the system to the perturbation of the background metric @xmath35 around the flat minkowski metric @xmath36 . euclidean time correlators can be obtained similarly from variations of the euclidean partition function @xmath37 $ ] : @xmath38=\frac12 \int\ ! d^4x\ , \langle t^{\mu\nu}(x)\rangle \delta g_{\mu\nu}(x ) + \frac18\int\!\!\int\ ! d^4x\ , d^4y\ , \langle t^{\mu\nu}(x)t^{\lambda\rho}(y)\rangle\ , \delta g_{\mu\nu}(x)\delta g_{\lambda\rho}(y ) + { \cal o}(g^3).\ ] ] the correlation functions of the trace @xmath39 can be also defined via variations of the partition function with respect to the metric variations of a special form ( dilatations ) : @xmath40 @xmath41=-\int\ ! d^4x\ , \langle\theta(x)\rangle \omega(x ) + \frac12\int\!\!\int\ ! d^4x\ , d^4y\ , \langle\theta(x)\theta(y)\rangle\ , \omega(x)\omega(y ) + { \cal o}(\omega^3).\ ] ] the two definitions of the two - point function of the trace differ by a contact term : @xmath42 the corresponding retarded correlators can be defined via linear response to perturbation @xmath43 : @xmath44 the definition ( [ eq : gr - theta ] ) is convenient because the value of its fourier transform ( for @xmath29 ) @xmath45 at vanishing frequency @xmath23 follows from the conservation of entropy in the ideal hydrodynamics @xcite : @xmath46 where @xmath47 is the entropy density . the sum rule derived in @xcite applies to the zero temperature subtracted green s function @xmath48_t } } \equiv g_r - g_r^{(t=0)}\ ] ] and its imaginary part on the real axis @xmath21_t}}=-{\rm im}\,{{\left[g_r\right]_t}}$ ] : @xmath49_t } } -{{\left[g_r(0)\right]_t } } = \frac2\pi \int_0^\infty \ ! d\omega\ , \frac{{{\left[\rho\left(\omega\right)\right]_t}}}{\omega}.\ ] ] using eq . ( [ eq : theta - theta-0-ds ] ) and @xmath50_t } } = 3p-\epsilon$ ] , one can write @xmath51_t } } = - \left ( 3 s \frac{\partial}{\partial s } -4\right)(\epsilon-3p)\ ] ] where @xmath52_t}}$ ] and @xmath53_t}}$ ] are equilibrium thermal energy and pressure at given entropy density @xmath47 . in qcd , due to the asymptotic freedom , @xmath54_t}}=0 $ ] and one obtains the sum rule found in @xcite : @xmath55_t}}}{\omega}\ ] ] in this paper , we consider a generic conformal field theory perturbed by an operator @xmath10 of dimension @xmath56 sourced by the field @xmath57 . we place this theory on a nontrivial gravitational background given by metric @xmath58 . the change of the partition function of the theory under dilatations ( [ eq : g - omega ] ) is equivalent to the rescaling of the only dimensionful external field @xmath57 , the source of the operator @xmath10 : @xmath59 = z[g(0),e^{-{\delta_-}\omega}c],\ ] ] where @xmath60 . thus @xmath61 thus @xmath62 and @xmath63 to evaluate the fourier transform of the correlation function ( [ eq : theta - theta - o ] ) at large ( imaginary ) frequency @xmath23 , we can use the operator product expansion ( ope ) for @xmath64 . the leading contribution comes from the unit operator , and behaves as @xmath65 , but cancels in @xmath66_t}}$ ] because it is independent of temperature . the contribution of an operator of dimension @xmath67 comes with the wilson coefficient which behaves as @xmath68 . assuming that there are no operators ( with vacuum quantum numbers ) of dimension equal to or lower than @xmath69 , we conclude that the contribution of the first term in eq . ( [ eq : theta - theta - o ] ) vanishes in the @xmath70 limit . the assumption implies , in particular , that @xmath3 , i.e. , @xmath71 . the contribution of the last term in eq . ( [ eq : theta - theta - o ] ) then gives the value for @xmath72 : @xmath73_t } } = -{{\left[\langle\theta\theta\rangle(i\infty)\right]_t } } = \delta_-^2\ , c\ { { \left[\langle{{o}}\rangle\right]_t } } = \delta_-\,(\epsilon-3p),\ ] ] where we used the fact that analytical continuation of a retarded correlation function to matsubara frequencies on the imaginary axis equals negative of the euclidean correlator . we shall verify eq . ( [ eq : gr - infty ] ) explicitly and analytically in the holographic model . combining eqs . ( [ eq : delta - t - gr-0 ] ) , ( [ eq : gr - infty ] ) and ( [ eq : sum - rule ] ) we obtain eq . ( [ eq : sum - rule - dplus ] ) . as discussed above , we will consider a four - dimensional ( 4d ) conformal theory in which the conformal symmetry is broken by the operator @xmath10 with scaling dimension @xmath56 . the holographically dual description of such a theory must therefore include a five - dimensional ( 5d ) scalar field with mass @xmath74 . as a model for the qcd thermodynamics , such a theory was first considered in @xcite . it models in the most straightforward way the breaking of the scaling invariance in qcd . the minimum set of fields required to describe correlators of the stress energy tensor and of the operator @xmath10 include the 5d metric @xmath75 ( @xmath76 and the scalar field ( dilaton ) @xmath77 . the minimal action is thus given by @xmath78 , \end{split}\ ] ] where @xmath79 is the ricci scalar , @xmath80 is the determinant of the metric , @xmath81 is the determinant of the induced metric on the uv boundary @xmath82 at @xmath83 , @xmath84 is the extrinsic curvature on @xmath85 , and @xmath86 is the 5d einstein gravitational constant . the value of @xmath86 is inversely proportional to the number of the degrees of freedom in the dual four - dimensional theory , e.g. , @xmath87 in a gauge theory with large number of colors @xmath88 . the smallness of @xmath86 ( i.e. , the largeness of the number of colors ) controls the semiclassical approximation which we use . the last term in the action is the gibbons - hawking term , which removes the boundary terms arising upon integration by parts of the terms in @xmath79 linear in second derivatives of the metric @xcite . the potential for the dilaton , @xmath89 , is the function which was tuned in @xcite to best `` mimic '' the qcd equation of state . here , similarly to @xcite , we shall concentrate on the results which are universal in the class of models described by eq . ( [ eq : s5 ] ) with any ( sensible ) potential . for example , ref . @xcite found that the speed of sound approaches the conformal value @xmath7 as @xmath90 universally from below in such theories . here we shall also make use of the large @xmath91 limit to the extent that it makes only the curvature of the potential @xmath92 matter . the negative cosmological constant provided by choosing @xmath93 ensures necessary asymptotics of the metric near the boundary @xmath94 while the dilaton field asymptotics is fixed by @xmath95 : @xmath96 . the singular behavior at @xmath83 can be regulated by setting the boundary conditions at @xmath97 and taking the uv regulator @xmath98 to 0 after necessary renormalization . the rules of the holographic correspondence require us to extremize the action @xmath99 w.r.t . the metric and the dilaton field subject to the boundary conditions on the uv boundary @xmath100 : @xmath101 the holographic duality then implies that the extremal value of the euclidean 5d action as a functional of the boundary values @xmath102 and @xmath103 is the same functional as @xmath104 $ ] in the 4d theory . in order to calculate the green s function @xmath105 we use eq . ( [ eq : theta ] ) and the relationship ( [ eq : z - g - tmunu ] ) between the derivatives of @xmath106 and the correlators of @xmath30 . the one - point functions @xmath107 are determined by the solutions of the equations of motion with boundary conditions homogeneous in the 4d coordinates @xmath108 . we follow ref . @xcite to determine the solution to the equations of motion homogeneous in 4d . in the gauge chosen in ref . @xcite , the metric has the form @xmath109 where @xmath110 and @xmath111 are functions only of the extra dimensional coordinate @xmath112 , and @xmath110 has a simple zero at some value of @xmath113 . the functions , @xmath114 and @xmath115 , and the background dilaton field , @xmath77 , are all determined by extremizing the action . the details can be found in @xcite and for completeness presented in appendix [ sec : appa ] . the extremum of the action with boundary conditions ( [ eq : b - c ] ) is a one - parameter family of solutions to the equations of motion . a convenient choice of the parameter is the enthalpy density , @xmath116 , which arises as an integration constant ( see eq . ( [ eq : ein2b ] ) ) . the temperature is determined , as usual , by considering the periodicity in the euclidean time necessary to avoid conical singularity at @xmath113 : @xmath117 and is related to @xmath116 via equation of state ( see e.g. eq . ( [ eq : w - t ] ) ) . the calculation of two - point correlation functions requires us to consider non - homogeneous solutions to the equations of motion . we only need to consider infinitesimal variations around the homogeneous solution ( [ eq : metric ] ) and expand the action @xmath99 to quadratic order in these variations . we parameterize these inhomogeneous solutions as @xmath118 where summation over @xmath119 is not implied . we exercise the freedom of the gauge choice to set @xmath120 , i.e. , we set variations of @xmath77 around homogeneous solution to 0 ( we can do this as long as we are not interested in calculating correlation functions of @xmath10 , i.e. , as long as @xmath57 remains homogeneous ) . we find that this gauge choice provides nontrivial simplification of the equations of motion , allowing us to reduce them to a second order equation for metric variation @xmath121 , instead of the third . we then consider solution homogeneous in the spatial coordinates @xmath122 , and depending only on @xmath123 and @xmath112 , since we are interested in @xmath29 variations of @xmath58 . for such solutions we can use the remaining gauge freedom to set @xmath124 , but not @xmath125 . we can also use @xmath126 symmetry to simplify our analysis by separating the spatial part of metric perturbations into trace and traceless parts : @xmath127 at quadratic order , @xmath121 mixes with components @xmath128 and @xmath125 , but decouples from the traceless part @xmath129 as well as from the off - diagonal components @xmath130 . therefore , we shall focus only on the metric perturbations @xmath121 , @xmath128 and @xmath125 . we can express the coupling of the metric perturbation to @xmath30 as @xmath131 where @xmath132 is the trace and @xmath133 is the traceless part of the _ stress _ tensor @xmath134 . the two - point function @xmath105 in eq . ( [ eq : theta ] ) , since @xmath135 , can be expressed in terms of correlation functions of @xmath136 and @xmath137 using @xmath138 the corresponding retarded correlators can be found using holographic correspondence and the recipe @xcite : @xmath139 since we need the fourier transform of @xmath105 it would be convenient to express @xmath99 directly in terms of the fourier modes of @xmath140 defined as @xmath141 the metric perturbations @xmath121 , @xmath125 and @xmath128 , defined by eqs . ( [ eq : g - h ] ) and ( [ eq : h - h - ht ] ) , satisfy a set of equations derived from the linearized einstein s equations : @xmath142 with the primes denoting derivatives with respect to @xmath112 . in order for eqs . ( [ eq : t0000])([eq : tssss ] ) to give the _ retarded _ correlation functions , the solutions to eqs . ( [ eq : eom])([eq : eom - hzz ] ) must satisfy the in - falling wave conditions at the horizon @xmath113 . in order to calculate the two - point functions , we expand the holographic action in eq . ( [ eq : s5 ] ) to quadratic order in metric variations around the homogeneous background . this expansion can be written in a compact form ; the bulk part of the action is given by , @xmath143 where the dagger refers to the transposed complex conjugate ( @xmath144 ) , and @xmath145 integrating by parts and using equation of motion , we reduce the action at the extremum to the boundary term : @xmath146 the action also receives a contribution from the gibbons - hawking boundary term : @xmath147 combining all of this , the action at the extremum can be expressed in terms of the boundary values of @xmath128 , @xmath121 and @xmath148 : @xmath149 from this one can immediately read off the expressions for the fourier transforms of the stress - energy two - point functions : @xmath150 where in the last equality in eqs . ( [ eq:0000 ] ) and ( [ eq:00xx ] ) we used expressions for one - point functions ref . as expected , the correlators involving @xmath136 are frequency independent and satisfy energy - momentum conservation ward identies ref . the above expressions contain contributions divergent in the limit @xmath151 . they can be removed by subtracting a polynomial of @xmath152 sufficient to cancel the divergences . these terms do not affect @xmath27 . putting this all together , the finite ( as @xmath153 ) part of the bulk green s function is found to be , @xmath154 where the polynomial @xmath155 combines the uv divergent , temperature independent contact terms . the bulk spectral function is then given by @xmath156 in order to calculate the bulk retarded green s function and the bulk spectral function , one needs to solve the equation of motion for the field @xmath157 , eq . ( [ eq : eom ] ) . in general , this can not be done analytically . however , in the large @xmath23 limit , the solution can be found similarly to the born approximation in quantum mechanics . we perform a louiville transformation to bring eq . ( [ eq : eom ] ) to the schrdinger form . in terms of the new coordinate @xmath158 , given by @xmath159 and the new function @xmath160 , given by @xmath161 eq . ( [ eq : eom ] ) takes the form @xmath162 with the schrdinger potential , as an implicit function of @xmath158 , is given by @xmath163 the in - falling boundary condition on @xmath164 at the horizon @xmath113 transforms into the outgoing wave condition on @xmath160 at @xmath165 . the bulk spectral function can be expressed in terms of the new function @xmath166 as @xmath167 or somewhat more intuitively ( using eq . ( [ eq : ein1 ] ) ) as @xmath168 } { |\psi(x)|^2 } \right|_{z=\varepsilon}\ ] ] which has the quantum mechanical interpretation of the probability flux for the wave function @xmath166 ( normalized as @xmath169 at @xmath97 ) . the positivity of the spectral function can be seen then as a consequence of the conservation of the flux and the outgoing wave boundary condition at @xmath165 . at large @xmath23 , we would like to treat the potential as a perturbation and calculate the `` wave function '' @xmath166 using the born approximation . there is one difficulty , however : the schrdinger potential @xmath170 diverges at @xmath33 . we shall separate the leading divergence explicitly @xmath171 and treat the remaining part of the potential , @xmath172 , as a perturbation . we can calculate @xmath160 iteratively , using the green s function method . the solution to eq . ( [ eq : phi2 ] ) with @xmath173 , satisfying the outgoing wave boundary condition at @xmath165 , is given by , up to unimportant normalization , @xmath174 where @xmath175 , @xmath176 is the hankel function the first kind . substituting @xmath177 into ( [ eq : htophi ] ) and then into ( [ eq : tt ] ) we find , using eq . ( [ eq : b - small - z ] ) for @xmath178 , @xmath179 where the dots stand for corrections to the leading order which would come from iterations of @xmath180 . the first term on the r.h.s . is the leading term in the @xmath181 limit . it is temperature independent and has correct @xmath23-scaling to be identified with the leading contribution of the unit operator to the ope of @xmath182 in eq . ( [ eq : theta - theta - o ] ) . this term has nontrivial , but temperature independent , imaginary part and it is subtracted when we calculate @xmath66_t}}$ ] . the second term on the r.h.s . of eq . ( [ eq : grlw ] ) gives @xmath183_t } } = - \frac{c\ , { { \left[d\right]_t}}}{2\kappa^2}\ , \delta_-^2\left(\delta_+-\delta_-\right ) = \delta_- ( \epsilon - 3p).\ ] ] where we used eq . ( [ eq : e - p ] ) . this agrees with eq . ( [ eq : gr - infty ] ) as expected . the leading correction , @xmath184 , to @xmath177 can be found , similarly to the born approximation , by using green s function @xmath185 : @xmath186 the green s function satisfies equation @xmath187 with boundary condition @xmath188 at @xmath189 and outgoing wave condition at @xmath165 . the solution is given by @xmath190 where @xmath191 substituting into eq . ( [ eq : psi1-gpsi0 ] ) we find @xmath192.\ ] ] to zeroth order in @xmath180 , @xmath193 and eq . ( [ eq : rho - psi ] ) gives : @xmath194 to next order in @xmath180 , @xmath195 and @xmath196 , where @xmath197 in the limit @xmath181 , the integral in eq . ( [ eq : rhoint ] ) is dominated by the region of small @xmath198 . expanding @xmath180 in powers of @xmath158 , we generate a @xmath199 series ( possibly asymptotic ) for the integral . to enable further analytic calculation , we shall limit it to leading terms in the expansion in powers of @xmath200 . as we also discuss in appendix [ sec : appa ] , this corresponds to the regime of high temperatures , i.e. , @xmath201 , since @xmath57 is the only dimensionful parameter in the theory . in addition to enabling analytic calculations , this limit has the virtue of yielding results which do not depend on the form of the dilaton potential beyond its curvature at the minimum ( related to @xmath28 ) . in this regime , to the order we work , we can set @xmath57 to zero in @xmath202 . the leading term in the taylor expansion of @xmath202 is given by @xmath203 where @xmath204 , as defined in appendix [ sec : appa ] , and @xmath205 combining eq . ( [ eq : rho0 ] ) and eq . ( [ eq : rho1-rho0 ] ) we see that @xmath206 as @xmath181 . we note that , strictly speaking , not all @xmath207 corrections to this result are negligible , since some of them grow with @xmath23 . however , as one can check , those corrections do not depend on @xmath116 ( i.e. , on @xmath91 ) and are part of the @xmath208 corrections to the wilson coefficient of the unit operator in the ope of @xmath182 . these terms would be also subtracted if @xmath8_t}}$ ] was calculated to the next order in @xmath200 . in order to calculate the bulk spectral function , one needs to first solve the equation of motion for the functions @xmath110 , @xmath111 and @xmath209 and then use those results to solve eq . ( [ eq : eom ] ) for @xmath164 . this can not be done analytically for an arbitrary potential @xmath89 . however , the first step of this calculation can be done analytically in the limit where the temperature is large compared with the conformal breaking scale , @xmath210 . another advantage of this limit is that the results are universal in the sense that the only property of the potential @xmath89 which matters is the curvature at the minimum , which determines the dimension of the operator @xmath56 . equation ( [ eq : eom ] ) would still need to be solved numerically , however . we use a shooting procedure starting from the ir boundary , i.e. , the horizon @xmath113 . the equation has a regular singular point at @xmath113 . a frbenius expansion around this point can be constructed iteratively . among the two linearly independent solutions the in - falling wave is chosen . the expansion is used to calculate @xmath121 and @xmath148 at a point near the ir boundary . these values are then used as the initial conditions to integrate the differential equation numerically all the way to the uv boundary @xmath97 . the spectral function can then be determined from the numerical results using eq . ( [ eq : bulk ] ) . as a representative example , in fig . [ fig : rhoa ] , we plot the spectral function at order @xmath200 for @xmath25 . as expected , it is non - negative and it diverges for large @xmath23 . the subtracted spectral function @xmath8_t}}$ ] to order @xmath207 is obtained by subtracting @xmath211 determined analytically in eq . ( [ eq : rho0 ] ) . figures [ fig : rhob ] and [ fig : drho ] show @xmath8_t}}/\omega$ ] plotted at different values of @xmath28 , while the dashed line corresponds to the analytic expressions for the large @xmath23 behavior given by @xmath212 in eq . ( [ eq : rho1-rho0 ] ) . it is interesting to compare these plots with the lattice calculations of the spectral function in fig . 5 of ref . @xcite . in both cases one finds the sign oscillation of the spectral function in the region @xmath213 , which is absent in the weak coupling ( boltzmann equation ) result @xcite . for @xmath28 equaling 0.2 , 0.5 , 0.7 , and 1.0 . the dashed curve in each plot is from the analytic expression for the spectral function in the large @xmath23 limit in eq . ( [ eq : rho1-rho0 ] ) . the dotted line is drawn at the value of the bulk viscosity calculated from eq . ( [ eq : bulk - visc ] ) in units of @xmath214 . this value should be compared with the intercept of the plots at @xmath215 . , title="fig : " ] for @xmath28 equaling 0.2 , 0.5 , 0.7 , and 1.0 . the dashed curve in each plot is from the analytic expression for the spectral function in the large @xmath23 limit in eq . ( [ eq : rho1-rho0 ] ) . the dotted line is drawn at the value of the bulk viscosity calculated from eq . ( [ eq : bulk - visc ] ) in units of @xmath214 . this value should be compared with the intercept of the plots at @xmath215 . , title="fig : " ] for @xmath28 equaling 0.2 , 0.5 , 0.7 , and 1.0 . the dashed curve in each plot is from the analytic expression for the spectral function in the large @xmath23 limit in eq . ( [ eq : rho1-rho0 ] ) . the dotted line is drawn at the value of the bulk viscosity calculated from eq . ( [ eq : bulk - visc ] ) in units of @xmath214 . this value should be compared with the intercept of the plots at @xmath215 . , title="fig : " ] for @xmath28 equaling 0.2 , 0.5 , 0.7 , and 1.0 . the dashed curve in each plot is from the analytic expression for the spectral function in the large @xmath23 limit in eq . ( [ eq : rho1-rho0 ] ) . the dotted line is drawn at the value of the bulk viscosity calculated from eq . ( [ eq : bulk - visc ] ) in units of @xmath214 . this value should be compared with the intercept of the plots at @xmath215 . , title="fig : " ] in the high - temperature limit @xmath210 , the left - hand side ( thermodynamic side ) of the sum rule eq . ( [ eq : sum - rule - dplus ] ) can be analytically calculated in the holographic model . the trace anomaly is given by eq . ( [ eq : tanom ] ) , while the derivative with respect to entropy can be expressed via derivative with respect to enthalpy @xmath116 . using the expression for @xmath216 in the high @xmath91 limit , eq . ( [ eq : dc ] ) , the derivative can be performed , and the left - hand side can be expressed as , @xmath217 the right - hand side can be numerically calculated from @xmath21_t}}$ ] . for efficiency , we make use of the analytical result for the asymptotics of @xmath21_t}}$ ] eq . ( [ eq : rho1-rho0 ] ) . we calculate numerically only the contribution to the integral up to some relatively large frequency @xmath218 . the contribution from @xmath218 to @xmath219 is calculated analytically using asymptotics @xmath21_t}}\approx\rho_1 $ ] from eq . ( [ eq : rho1-rho0 ] ) . we chose @xmath220 . table [ tab : sum ] shows that the sum rule holds to at least within a fraction of a percent . .verification of the sum rule in eq . ( [ eq : sum - rule - dplus ] ) in the holographic model for a sample of values of @xmath28 in the high - temperature limit @xmath201 . the columns lhs and the rhs are the values of the left- and the right- hand side of eq . ( [ eq : sum - rule - dplus ] ) expressed in units of @xmath221 , which removes the dependence on @xmath222 , @xmath57 , @xmath91 and the leading ( @xmath15 ) dependence on @xmath28 . the lhs is calculated analytically using eq . ( [ eq : lhs ] ) , while the rhs is a numerical calculation of the integral as described in text . the last column shows the percent discrepancy between the two sides . [ cols="^,^,^,^",options="header " , ] as @xmath223 the left - hand side of the sum rule given by eq . ( [ eq : lhs ] ) scales as @xmath15 ( @xmath224 according to eq . ( [ eq : dc ] ) ) . this is also apparent in table [ tab : sum ] . however , the spectral function @xmath21_t}}$ ] on the right - hand side scales as @xmath225 . this comes from factor @xmath178 in eq . ( [ eq : bulk ] ) which scales as @xmath225 according to eq . ( [ eq : b - small - z ] ) . therefore , in order to determine the @xmath226 part of @xmath27 we can simply set @xmath227 in eq . ( [ eq : eom ] ) which then becomes ) , @xmath228 , which means , to @xmath207 we are working , we can drop terms like @xmath178 , but not terms like @xmath229 . ] @xmath230 where @xmath114 is given by eq . ( [ eq : fz ] ) in the high - temperature limit . this is the well known equation of motion of a massless scalar field on a pure ads background . however , the analytic solution to this equation with required boundary conditions has not yet been found ( see @xcite for discussion ) . nevertheless , the solution can be obtained numerically as discussed in sec . [ sec : sum ] . the resulting spectral function is displayed in fig . [ fig : rho ] . the @xmath226 part of @xmath8_t}}$ ] shown in fig . [ fig : rhob ] oscillates around 0 , but the contribution from it to the spectral integral in the sum rule is non - vanishing . we calculate the integral numerically and find @xmath231_t}}}{\omega } \,d \omega = 2 c^2 \delta_-^2 \left(\frac{w}{4}\right ) \left(.60000(1)\right ) + { { \cal o}}(\delta_-^3)\,,\ ] ] where @xmath116 is the enthalpy . the integral of the @xmath225 part of @xmath8_t}}$ ] is converging very fast ( apparently exponentially ) at @xmath232 and we picked relatively large @xmath233 for this numerical calculation . the number in parenthesis is @xmath234 to at least four significant digits . how then is the sum rule satisfied if the left - hand side is only @xmath235 ? we shall find the missing contribution in the tail of the function @xmath8_t}}$ ] . indeed , taking @xmath218 to @xmath219 requires extra care . because @xmath226 part of @xmath8_t}}$ ] decreases very fast as @xmath181 , for sufficiently large @xmath23 the dominant part in @xmath8_t}}$ ] is of order @xmath15 . as we see analytically in eqs . ( [ eq : rho1-rho0 ] ) , ( [ eq : rho0 ] ) , @xmath21_t}}\to\rho_1\sim\delta_-^3\omega^{-2\delta_-}$ ] . at any finite @xmath28 the negative power - law tail effectively cuts off the integral at large @xmath236 . however , the contribution of this long tail to the integral grows with @xmath237 as @xmath238_t}}/\omega\sim \delta_-^3\log(\omega_{\rm tail}/\omega_{\rm max})\sim\delta_-^2 $ ] , i.e. , the length of the tail compensates for the extra power of @xmath28 . let us calculate this @xmath226 contribution from the tail . using eqs . ( [ eq : rho1-rho0 ] ) , ( [ eq : rho0 ] ) we can write for large @xmath23 and small @xmath28 : @xmath239_t}}= \rho_1(\omega)+\ldots = -\frac{6}{5 } \pi c^2 \delta_-^3 \omega^{-2\delta_- } \left(\frac{w}{4}\right)+\ldots.\ ] ] thus @xmath240_t } } = - 2 c^2 \delta_-^2 \left ( \frac{w}{4}\right ) \left(\frac{3}{5}\right)+{{\cal o}}(\delta_-^3).\ ] ] we see that the contribution of the high - frequency tail exactly cancels the contribution from the region of @xmath19 in eq . ( [ eq : smalld ] ) . therefore the rhs of the sum rule is proportional to @xmath15 just as the lhs . remarkably , this cancellation mechanism is very similar to the one which was found in qcd by caron - huot in ref.@xcite . in the case of qcd the left - hand ( thermodynamic ) side of the sum rule is of order @xmath12 , while the spectral function on the right - hand side is @xmath241 , with logarithmically long high - frequency tail of order @xmath242 . we would like also to comment that the @xmath225 contribution to the bulk spectral function @xmath21_t}}$ ] in the high - temperature regime is the same , up to a constant , as the spectral function in the @xmath243 susy ym theory . in fact , the spectral integral in eq . ( [ eq : smalld ] ) has been also performed numerically in ref . @xcite , using a different method and with the same result . unlike the case we consider , however , for the shear channel sum rule considered in ref . @xcite that integral saturated the sum rule . also worth noting is that the shear and the bulk channel spectral functions are proportional to each other in the chamblin - real dilaton model examined in @xcite and @xcite . in such theories the mechanism of the saturation of the bulk sum rule is the same as that of the shear sum rule and no high / low frequency cancellations are needed in either case . as another cross - check of our results , we can compare them to the analytic calculation of the bulk viscosity in the high temperature limit of ref . @xcite , which finds @xmath244 this formula was derived by matching gradient expansion of the stress - energy tensor to the gradient expansion of the background metric . ideally , one should also be able to derive eq . ( [ eq : bulk - visc ] ) by solving eq . ( [ eq : eom ] ) at small @xmath23 and using kubo formula @xmath245 indeed , this can be done analytically for @xmath246 . in this case , eq . ( [ eq : eom ] ) reduces to eq . ( [ eq : eom - small - d ] ) . for small @xmath23 , this equation can be easily integrated . normalizing as @xmath247 and using in - falling boundary condition at the horizon we find it is valid close enough to the horizon to allow matching it to the in - falling wave solution . ] @xmath248 this gives for the spectral function at small @xmath23 , @xmath249 and for the bulk viscosity @xmath250 this agrees with the @xmath251 limit of eq . ( [ eq : bulk - visc ] ) . unfortunately , a similar approach at finite @xmath28 appears intractable , since an analytic solution to eq . ( [ eq : eom ] ) , even at small @xmath23 , for arbitrary @xmath28 is not known . however , the bulk viscosity can be calculated by applying the kubo formula eq . ( [ eq : zeta - rho ] ) to our numerical results for the spectral function . we have verified the agreement of eq . ( [ eq : bulk - visc ] ) with these numerical calculations and illustrated it in fig . [ fig : drho ] . we studied the spectral function corresponding to time - dependent bulk deformation ( uniform expansion ) in a class of field theories where conformality is broken `` softly '' in the sense that at high temperature the equation of state approaches conformal limit @xmath252 , similar to qcd . the breaking is due to a scalar operator of conformal dimension @xmath3 , which is an analogue of the operator of gluon condensate in qcd . we find that in such theories the bulk spectral function satisfies the sum rule given by eq . ( [ eq : sum - rule - dplus ] ) . the sum rule in eq . ( [ eq : sum - rule - dplus ] ) is similar to the sum rule in an asymptotically free theory such as qcd derived by romatschke and son in ref . @xcite . in the marginal limit @xmath11 , the two sum rules are identical . we used this similarity to address an interesting puzzle noted in ref . @xcite . in order to satisfy the sum rule , a delicate cancellation must occur between the regions of @xmath19 and of @xmath253 in the spectral integral . ref.@xcite showed that this cancellation can be indeed seen in the high - temperature limit of qcd , using weak coupling calculation . our study of the holographic model shows that such a cancellation is very generic to the whole class of strongly coupled theories with softly broken conformal symmetry . to make the connection of our model with qcd more tangible , we can observe that in qcd the action contains operator @xmath5 and thus identify the operator @xmath10 with @xmath5 , up to a numerical coefficient . this coefficient can be determined by matching correlation functions ( e.g. , the bulk spectral function ) of the holographic model to qcd , but we shall not need it here . the anomalous dimension of this operator in qcd is given by @xmath254 . in qcd this anomalous dimension is a function of scale , vanishing logarithmically , i.e. , very slowly , with increasing energy - momentum scale . in the class of qcd - like theories we consider , the corresponding quantity is @xmath255 , which is a constant . , have been discussed in , e.g. , refs.@xcite . ] this scaling dimension taken at @xmath232 , determines the value of @xmath54_t}}$ ] in the sum rule ( [ eq : sum - rule ] ) , according to eq . ( [ eq : gr - infty ] ) , and it is responsible for the difference of the general sum rule ( [ eq : sum - rule - dplus ] ) from qcd . the holographic theories we consider can serve only as a qualitative or semi - quantitative guide to the qcd phenomena , since qcd becomes a weakly coupled theory at sufficiently high energy - momentum scale due to asymptotic freedom . but this guide might be useful by offering a view complementary to the weak coupling extrapolation out of the domain of asymptotically high energies . the experiments at rhic provide a powerful argument that the domain of interest in heavy - ion collisions is a strongly coupled domain . it will be very interesting to see to what extent this remains true at lhc energies . also notable in this regard are lattice calculations of the spectral function . the striking feature of the lattice results is the sign oscillation of the spectral function in the @xmath256 region . this oscillation is absent in weakly coupled calculations @xcite , but appears to be a generic feature in the spectral functions obtained in holographic models , as our results in figs . [ fig : rhob ] and [ fig : drho ] illustrate . this qualitative difference appears to be another manifestation of the now familiar fact that real - time response ( in particular , hydrodynamics ) is more sensitive to the coupling strength than , e.g. , equation of state . this work is supported by the doe grant no . de - fg0201er41195 . this appendix summarizes , for completeness , the relevant results of ref . @xcite . the background geometry corresponding to homogeneous boundary conditions on @xmath58 and @xmath77 can be found by solving einstein s equations for the metric in eq . ( [ eq : metric ] ) which take the form : @xmath257 where a dot denotes a @xmath258 derivative , _ e.g. _ , @xmath259 . the holographic correspondence provides the boundary conditions at the uv boundary @xmath97 given by eqs . ( [ eq : b - c ] ) . minkowski metric at the boundary requires @xmath260 equation ( [ eq : ein2 ] ) can be integrated once to give @xmath261 the integration constant @xmath262 must be positive if the metric is to possess a horizon @xmath263 at some value of @xmath264 . since @xmath262 determines the position of the horizon , it is related to temperature , and we find that it is proportional to the enthalpy , eq . ( [ eq : w - bar ] ) . a boundary condition on @xmath115 is not needed because the value of @xmath115 is determined by eq . ( [ eq : ein3 ] ) , which is algebraic in @xmath115 . the role of the second boundary condition for eq . ( [ eq : ein4 ] ) is played by the requirement that @xmath77 is finite at the horizon , @xmath113 , which is a regular singular point of the second order differential equation ( [ eq : ein4 ] ) . near the @xmath100 boundary , @xmath265 , @xmath266 , @xmath267 and @xmath268 . equation ( [ eq : ein4 ] ) for @xmath77 can be linearized and the asymptotic behavior of @xmath77 near the boundary can be determined easily : @xmath269 where the curvature of the potential @xmath270 determines the indices @xmath271 . the coefficient of the first term is related to @xmath57 by eq . ( [ eq : b - c ] ) . the coefficient @xmath216 of the second linearly independent solution should be determined by the finiteness condition at the horizon and is a function of @xmath262 ( i.e. , temperature ) and @xmath57 . by calculating the derivative of the 5d action with respect to @xmath57 and matching it , by holographic correspondence , to the expectation value @xmath272 , one finds ( see also @xcite ) @xmath273 where `` '' denote uv divergent but temperature independent terms . one can thus see that the integration constant @xmath216 is related to the uv finite ( and temperature dependent ) part of @xmath274 . by considering homogeneous variations of the boundary condition on the metric , the one - point functions of the stress - energy tensor can be calculated using eq . ( [ eq : z - g - tmunu ] ) as in ref . @xcite . @xmath277 the thermal energy and pressure , @xmath278_t}},\qquad p = \langle t^{11}\rangle - \langle t^{11}\rangle_{t=0 } \equiv{{\left[\langle t^{11}\rangle\right]_t}},\end{aligned}\ ] ] are finite at @xmath279 and equal to zero at @xmath280 . the enthalpy , @xmath281 is related to the constant @xmath262 : @xmath282 after solving eq . ( [ eq : ein3 ] ) for @xmath115 at @xmath97 with @xmath77 given by eq . ( [ eq : asy ] ) , we find that the energy and pressure can be expressed as @xmath283_t}}}{8\kappa^2}\ , \delta_-(\delta_+-\delta_-),\qquad p = w - \epsilon . \ ] ] therefore the temperature dependence of the expectation value of the trace anomaly is given by , @xmath284_t } } = 3p-\epsilon = \frac{c\,{{\left[d\right]_t}}\,\delta_-}{2 \kappa^2}\,(\delta_+-\delta_- ) = -\delta_-\ , c\ , { { \left[\langle { { o}}\rangle\right]_t}}.\ ] ] in accordance with the anomaly equation ( [ eq : theta - o ] ) . the equations of motion can be simplified and analytically solved if one considers the high temperature limit , or the limit where @xmath210 . since the enthalpy is related to the temperature , @xmath285 this limit can also be expressed as @xmath286 . one can begin by observing that at large @xmath262 the function @xmath114 varies very rapidly according to eq . ( [ eq : ein2b ] ) . this means one can neglect variation of the function @xmath115 between the boundary @xmath97 and the horizon @xmath113 , since @xmath264 becomes small ( as @xmath287 ) . since on the boundary @xmath288 ( up to terms of order @xmath289 , negligible here , according to eq . ( [ eq : ein3 ] ) ) , we find from eq . ( [ eq : ein2b ] ) @xmath290 another consequence is that @xmath77 , which is small at @xmath97 , remains small up to @xmath264 ( @xmath291 ) , and the linearized approximation to eq . ( [ eq : ein4 ] ) is valid not only near the boundary , but all the way to the horizon . with @xmath288 and @xmath114 from eq . ( [ eq : fz ] ) we obtain @xmath292 equation ( [ eq : dphi ] ) can be solved analytically @xmath293 where the coefficients follow the notations of eq . ( [ eq : asy ] ) ( up to terms @xmath294 , here negligible ) . both linearly independent solutions are logarithmically divergent at the horizon @xmath113 , where @xmath295 . the condition @xmath296 requires us to select the linear combination in which these divergences cancel . this fixes @xmath216 in terms of @xmath57 : @xmath297 where the function @xmath298 is given by @xmath299 this solution for @xmath209 can be used in eq . ( [ eq : ein1 ] ) to calculate @xmath300 , @xmath301 from this expression , the function @xmath111 can be calculated to order @xmath200 , which is the leading order in the high - temperature limit . by iteratively solving the equations of motion , the higher order corrections to @xmath115 as well as to @xmath114 and @xmath77 can be determined if necessary . h. b. meyer , pos * lattice2008 * , 017 ( 2008 ) [ arxiv:0809.5202 [ hep - lat ] ] . n. iqbal and h. b. meyer , jhep * 0911 * , 029 ( 2009 ) [ arxiv:0909.0582 [ hep - lat ] ] . h. b. meyer , jhep * 1004 * , 099 ( 2010 ) [ arxiv:1002.3343 [ hep - lat ] ] . d. kharzeev and k. tuchin , jhep * 0809 * , 093 ( 2008 ) [ arxiv:0705.4280 [ hep - ph ] ] . f. karsch , d. kharzeev and k. tuchin , phys . b * 663 * , 217 ( 2008 ) [ arxiv:0711.0914 [ hep - ph ] ] . p. romatschke , d. t. son , phys . * d80 * , 065021 ( 2009 ) . [ arxiv:0903.3946 [ hep - ph ] ] . j. m. maldacena , adv . theor . * 2 * , 231 ( 1998 ) [ int . j. theor . phys . * 38 * , 1113 ( 1999 ) ] ; s. s. gubser , i. r. klebanov and a. m. polyakov , phys . lett . b * 428 * , 105 ( 1998 ) ; e. witten , adv . * 2 * , 253 ( 1998 ) . o. aharony , s. s. gubser , j. m. maldacena , h. ooguri and y. oz , phys . rept . * 323 * , 183 ( 2000 ) [ arxiv : hep - th/9905111 ] . d. t. son , a. o. starinets , ann . rev . nucl . part . * 57 * , 95 - 118 ( 2007 ) . [ arxiv:0704.0240 [ hep - th ] ] .

nonthermal emission is commonly observed among the lowest - mass stars , traced by optical line ( e.g. , ca ii , h i sequence ) , x - ray , uv and radio emission . this emission can be both persistent ( quiescent ) and eruptive , with short - duration flares from m dwarfs occurring at a rate of roughly 3% @xcite . the incidence and strength of quiescent magnetic activity as traced by the h@xmath6 line reaches @xmath380% and @xmath7 @xmath8 @xmath04 , respectively , among nearby late - type m dwarfs @xcite , but both metrics decline precipitously for the cooler l and t dwarfs @xcite and sources far from the galactic plane @xcite . similar declines are seen in x - ray emission , but surprisingly not at radio frequencies . assuming that this nonthermal emission arises from magnetic interaction , the decline with vertical scale height among m dwarfs is likely an age effect , as angular momentum loss over time results in weakened magnetic dynamos . however , spin - down timescales exceed 5 gyr beyond spectral type m5 @xcite , and many late - type dwarfs are found to be rapid rotators ( p @xmath8 2 - 10 hr ; @xcite ) exhibiting kilogauss magnetic fields at their photospheres @xcite . hence , the decline in magnetic emission with spectral type must arise from a different effect . the favored cause is the decoupling of cooler , increasingly neutral photospheres from magnetic structures , which reduces magnetic stresses and the frequency of magnetic reconnection above the ( sub)stellar surface ( e.g. , ) . deeper reconnection events may continue to power strong flaring bursts observed in a handful of weakly active or inactive late - m and l dwarfs ( e.g. , @xcite ) . contrary to these trends , a very rare set of low - temperature `` hyperactive '' dwarfs exhibit unusually prodigious and persistent nonthermal emission . one of the first examples of these to be identified was 2mass j13153094@xmath02649513 ( hereafter 2mass j1315@xmath02649 ; @xcite ) , a high proper motion l5e dwarf which has exhibited sustained but variable h@xmath6 emission on no fewer than seven epochs spanning nearly a decade . with @xmath7 @xmath8 @xmath04 , 2mass j1315@xmath02649 is as active as a mid - type m dwarf , but is 12 orders of magnitude brighter in h@xmath6 than comparably - classified l dwarfs @xcite . in addition , h@xmath9 and na i d lines have also been observed in emission . since the photosphere of 2mass j1315@xmath02649 is cool and likely to be highly neutral , the origin of its unexpected emission remains a mystery . neither the kinematics nor spectral characteristics of 2mass j1315@xmath02649 indicate youth , and the absence of mid - infrared excess argues against accretion from a protoplanetary or debris disk @xcite . other hyperactive dwarfs , such as the l1e dwarf 2mass j10224821 + 5825453 ( @xmath7 @xmath8 @xmath03.5 to @xmath02.7 ; @xcite ) and the t6.5e dwarf 2mass j1237392 + 652615 ( @xmath7 @xmath8 @xmath04.6 to @xmath04.2 ; @xcite ; hereafter 2mass j1237 + 6526 ) also lack evidence of disk accretion and do not appear to be particularly young . alternative mechanisms , such as acoustic heating , unusually strong magnetic fields , and roche lobe overflow from a substellar companion have been proposed , but none of these scenarios have been validated . in this article , we report new observations of 2mass j1315@xmath02649 that reveal both continued nonthermal optical line emission in several neutral atomic species , and the presence of a t dwarf companion at a projected separation of 7 au . in section 2 we describe our combined - light optical and near - infrared spectroscopic observations , the latter of which yields preliminary evidence for a brown dwarf companion . in section 3 we describe adaptive optics ( ao ) imaging and spectroscopic observations that confirm the presence of the companion and allow determination of its separation and classification . in section 4 we analyze the observed and inferred physical properties of the components , the latter based on comparison to evolutionary and atmospheric models . we also perform a coevality test to examine the reliability of these models . in section 5 we discuss how the properties of 2mass j1315@xmath02649 argue for a magnetic origin to its persistent emission , powered by a strong magnetic field retained by a relatively old and massive cool dwarf . we summarize our results in section 6 . moderate - resolution optical spectra of 2mass j1315@xmath02649 were obtained on 2011 march 26 ( ut ) with the magellan echellette ( mage ; @xcite ) , mounted on the 6.5 m landon clay telescope at las campanas observatory . conditions during the observations were clear with 0@xmath106 seeing . two exposures totaling 3000 s were obtained at an average airmass of 1.003 using the 0@xmath107 slit aligned with the parallactic angle ; this setup provided 320010050 spectroscopy at a resolution @xmath11 @xmath8 4000 . we also observed the spectrophotometric flux standard eg 274 @xcite on the same night for flux calibration . thar lamps were obtained after each source observation for wavelength calibration , and internal quartz and dome flat field lamps were obtained during the night for pixel response calibration . data were reduced using the mase reduction pipeline @xcite , following standard procedures for order tracing , flat field correction , wavelength calibration ( including heliocentric correction ) , optimal source extraction , order stitching , and flux calibration . the red portion of the reduced spectrum is shown in figure [ fig_optspec ] , flux - calibrated to an apparent @xmath12-band magnitude of 20.16@xmath10.14 as estimated from two micron all sky survey ( 2mass ) photometry and a mean @xmath13 = 4.97@xmath10.13 color for l5 dwarfs @xcite . we confirm the characteristic mid - l dwarf features identified in previous studies , including strong feh and crh bands ; weak tio absorption ( relative to late - m and early - l dwarfs ) ; line absorption from na i , rb i and cs i ; and the strongly pressure - broadened 7700 k i doublet . the overall spectral shape is well - matched to the l5 dwarf 2mass j15074769@xmath01627386 @xcite , consistent with previously reported classifications @xcite . we confirm the absence of 6710 li i absorption as reported by @xcite to an equivalent width ( ew ) limit of 0.5 . this is well below measured ews for equivalently - classified l dwarfs ( e.g. , @xcite ) . we also see no evidence of any peculiar spectral features associated with low surface gravities , such as enhanced vo absorption or weakened alkali lines @xcite . line center measurements of the alkali lines indicate a heliocentric radial velocity of @xmath07@xmath19 km s@xmath14 . lccc + li i ( 6710 ) & @xmath150.5 & & + rb i ( 7802 ) & 4.9@xmath10.8 & & + rb i ( 7950 ) & 5.4@xmath10.6 & & + na i ( 8186 ) & 1.2@xmath10.3 & & + na i ( 8197 ) & 2.5@xmath10.3 & & + cs i ( 8523 ) & 4.2@xmath10.2 & & + cs i ( 8946 ) & 2.9@xmath10.7 & & + + h@xmath16 ( 4342 ) & & 4.0@xmath11.8 & -5.12@xmath10.20 + h@xmath9 ( 4863 ) & & 13.1@xmath11.5 & -4.61@xmath10.08 + h@xmath6 ( 6565 ) & @xmath058@xmath14 & 34.9@xmath11.2 & -4.18@xmath10.06 + na i ( 5892 ) & & 1.6@xmath10.2 & -5.50@xmath10.08 + na i ( 5898 ) & & 2.1@xmath10.3 & -5.40@xmath10.08 + k i ( 7667 ) & & 0.6@xmath10.1 & 5.92@xmath10.10 + k i ( 7701 ) & & 0.9@xmath10.2 & 5.79@xmath10.13 + the most striking feature in the optical spectrum of 2mass j1315@xmath02649 is its pronounced h@xmath6 emission . we measure an equivalent width of @xmath058@xmath14 in the mage data , in the middle of prior measurements that span @xmath024 to @xmath0160 . we also detect h@xmath16 ( 4342 ) and h@xmath9 ( 4863 ) in emission , and confirm the presence of na i emission as reported in . in addition , we detect weak emission in the cores of the 7700 k i doublets , but no emission from the 5877 he i d3 ( ew @xmath17 18 ) or 85008660 ca ii triplet lines ( ew @xmath17 0.3 ) . line profiles of detected emission features ( figure [ fig_emission ] ) show no appreciable broadening at the 75 km s@xmath14 velocity resolution of mage , nor do we detect any significant velocity shift between emission and absorption lines ( @xmath18@xmath19 = 4@xmath116 km s@xmath14 ) . as most of these lines are superimposed on an undetected continuum , we report in table [ tab_emission ] line fluxes and relative line - to - bolometric luminosities , that latter computed using the bolometric correction / spectral type relations of @xcite . our measurement of @xmath7 = @xmath04.18@xmath10.06 is similar to the first detection made by @xcite , and is again 12 orders of magnitude greater than measured quiescent or flaring fluxes for equivalently classified l dwarfs . the balmer decrement @xmath20 = 2.7@xmath10.3 is roughly comparable to the mean values for non - flaring m dwarfs @xcite . h@xmath6 and h@xmath9 emission contribute 61% and 23% of the total measured line flux of @xmath21 @xmath8 @xmath04 . low resolution near - infrared spectra of 2mass j1315@xmath02649 were obtained on 2009 june 30 ( ut ) with the 3 m nasa infrared telescope facility spex spectrograph @xcite . observing conditions were clear and dry with 0@xmath107 seeing at @xmath22-band . we used the prism - dispersed mode of spex with a 0@xmath105 slit ( aligned to the parallactic angle ) to obtain a continuous 0.72.5 @xmath23 spectrum with resolution @xmath11 @xmath24 . a total of 8 exposures of 120 s each were obtained in two abba dither pairs , nodding along the slit , at an average airmass of 1.50 . we also observed the a0 v star hd 125438 ( @xmath25 = 7.10 ) for flux calibration and telluric absorption correction , as well as internal flat field and argon arc lamps for pixel response and wavelength calibration . data were reduced with the idl spextool package , version 3.4 @xcite , using standard settings ; see @xcite for details . the reduced spectrum of 2mass j1315@xmath02649 is shown in figure [ fig_nirspec ] , compared to the l5 dwarf 2mass j21373742 + 0808463 ( hereafter 2mass j2137 + 0808 ; @xcite ) . its near - infrared spectral morphology is consistent with its l5 optical type , with strong h@xmath4o and co absorption bands , feh absorption at 1.0 and 1.6 @xmath23 , and ( unresolved ) alkali line absorption in the 1.11.3 @xmath23 region . the overall spectral shape is again consistent with a normal l5 field dwarf , with no evidence of an unusual surface gravity , metallicity or cloud content ( e.g. , @xcite ) . there is , however , a subtle `` notch '' feature present at 1.62 @xmath23 that differs from the spectrum of 2mass j2137 + 0808 ( see inset box of figure [ fig_nirspec ] ) . this feature has previously been noted in the combined - light spectra of l dwarf plus t dwarf binaries , arising from the overlap of feh absorption in the primary and ch@xmath26 absorption in the secondary ( e.g. @xcite ) . to characterize this feature , we performed a spectral template fitting analysis similar to that described in @xcite , using 295 l2t8 spectral templates from the spex prism spectral libraries . fluxes of these templates were scaled to the @xmath27/spectral type relation of @xcite , and all spectra were interpolated onto a common wavelength scale . restricting potential secondaries to have t spectral types , a total of 17889 binary templates were constructed and compared to the spectrum of 2mass j1315@xmath02649 over the wavelength ranges 0.951.35 @xmath23 , 1.451.80 @xmath23 and 2.002.35 @xmath23 using the @xmath28 statistic . + figure [ fig_binfit ] shows the four best - fitting binary templates from these comparisons , a combination of 2mass j2137 + 0808 and either early- or late - type t dwarfs . the addition of a t - type companion fills in the `` missing '' flux at 1.58 @xmath23 , creating the distinct notch feature at 1.62 @xmath23 ; and at 1.27 @xmath23 , producing a somewhat sharper @xmath29-band flux peak . importantly , all of the binary templates shown in figure [ fig_binfit ] provide statistically superior fits to the spectrum of 2mass j1315@xmath02649 as compared to 2mass j2137 + 0808 alone , based on the f - test statistic ( eqns . 26 in @xcite ) . however , we can not precisely constrain the properties of the secondary from this analysis ; the average spectral type of all of the template fits weighted by the f - test probability distribution function ( f - pdf ) is t3@xmath14 . to more accurately characterize this putative companion , high - resolution , near - infrared images of 2mass j1315@xmath02649 were obtained with the 10 m keck ii laser guide star adaptive optics system ( lgsao ; @xcite ) and facility nirc2 near - infrared camera . observations were conducted on two runs , 2010 march 24 and 2010 may 13 ( ut ) , both with clear skies and fair seeing ( @xmath171@xmath30 and 0@xmath105 , respectively ) . we used the narrow camera with image scale 9.963@xmath10.011 mas pixel@xmath14 @xcite covering a 10@xmath102@xmath3110@xmath102 field of view . images were obtained through the @xmath29 , @xmath22 , and @xmath32 filters , using a three - point dither pattern that avoided the noisy , lower left quadrant of the focal plane array . exposure times ranged from 30 s with 8 coadds to 120 s with 2 coadds per pointing position , with total integrations of 360 s to 720 s in a given filter . the sodium lgs provided the wave front reference source for ao correction , while tip - tilt aberrations and quasi - static changes were measured by monitoring the @xmath33 = 12.8 field star usno - b1.0 0631 - 0348160 @xcite located @xmath34 = 38@xmath106 from 2mass j1315@xmath02649 . images were reduced using custom idl scripts , as described in @xcite . @xmath311@xmath30 nirc2 @xmath29 , @xmath22 and @xmath32 images of the 2mass j1315@xmath02649ab system . images are linearly scaled to optimize visibility of the faint secondary ( southeast of the primary ) , and are oriented with north up and east to the left . [ fig_nirc2],title="fig:",scaledwidth=30.0% ] @xmath311@xmath30 nirc2 @xmath29 , @xmath22 and @xmath32 images of the 2mass j1315@xmath02649ab system . images are linearly scaled to optimize visibility of the faint secondary ( southeast of the primary ) , and are oriented with north up and east to the left . [ fig_nirc2],title="fig:",scaledwidth=30.0% ] @xmath311@xmath30 nirc2 @xmath29 , @xmath22 and @xmath32 images of the 2mass j1315@xmath02649ab system . images are linearly scaled to optimize visibility of the faint secondary ( southeast of the primary ) , and are oriented with north up and east to the left . [ fig_nirc2],title="fig:",scaledwidth=30.0% ] figure [ fig_nirc2 ] displays the reduced nirc2 images from our may observations . a faint source is clearly present southeast of 2mass j1315@xmath02649 , and was visible during both imaging epochs . relative astrometry for each epoch was measured on the individual frames using a centroiding algorithm , and these values were then averaged and multiplied by the pixel scale ( uncertainties include the standard deviations of the position measurements and 0.1% pixel scale uncertainty ) . relative photometry was performed on the coadded mosaics through aperture photometry . as the point spread function ( psf ) of the brighter component contributes significant flux ( @xmath825% ) to the brightness of the fainter object , photometry for the latter was extracted from a primary psf - subtracted image , constructed by rotating the frame 180@xmath35 about the centroid of the primary and differencing . systematic errors in the primary subtraction were estimated by offsetting the rotation axis over a 5@xmath315-pixel grid centered on the original centroid , subtracting , and measuring aperture photometry on the secondary . the final photometric values for each band and epoch were taken as the means and standard deviations of these 25 measurements . lc + @xmath36 ( @xmath30 ) & 185@xmath16 + @xmath37 ( @xmath30 ) & @xmath0277@xmath13 + @xmath34 ( @xmath30 ) & 333@xmath17 + @xmath38 ( @xmath39 ) & 143.1@xmath11.1 + @xmath40 ( mag ) & 3.12@xmath10.05 + @xmath41 ( mag ) & 4.29@xmath10.14 + @xmath2 ( mag ) & 4.91@xmath10.18 + + @xmath36 ( @xmath30 ) & 187@xmath12 + @xmath37 ( @xmath30 ) & @xmath0286@xmath17 + @xmath34 ( @xmath30 ) & 342@xmath17 + @xmath38 ( @xmath39 ) & 143.9@xmath10.5 + @xmath40 ( mag ) & 3.00@xmath10.03 + @xmath41 ( mag ) & 4.59@xmath10.07 + @xmath2 ( mag ) & 5.17@xmath10.12 + + @xmath36 ( @xmath30 ) & 180@xmath15 + @xmath37 ( @xmath30 ) & @xmath0286@xmath16 + @xmath34 ( @xmath30 ) & 338@xmath14 + @xmath38 ( @xmath39 ) & 147.8@xmath11.2 + results are listed in table [ tab_astrometry ] . separations in right ascension and declination are consistent between both epochs , and ( in conjunction with the osiris observations described below ) yield a mean separation of 336@xmath16 mas at position angle of @xmath42 , measured from primary to secondary . the mean relative magnitudes are also statistically consistent between epochs ( to within 2@xmath43 ) , and indicate that the companion is both considerably fainter ( @xmath40 = 3.03@xmath10.03 , @xmath2 = 5.09@xmath10.10 ) and bluer in the near - infrared . to our knowledge , this is the second most extreme near - infrared flux ratio measured for an ultracool dwarf binary to date2649 is exceeded at @xmath29-band only by the young tw hydrae binary 2massw j1207334 - 393254ab ( @xmath40 = 7.0@xmath10.2 , @xmath44 = 4.98@xmath10.14 ; ) ; and at @xmath45-band only by the old , widely - separated binary sdss j141624.08 + 134826.7ab ( @xmath40 = 4.31@xmath10.02 , @xmath44 = 6.85@xmath10.17 ; ) . ] . we examine the physical association of the companion in section 4.3 ; hereafter , we refer to the two sources as 2mass j1315@xmath02649a and b. resolved @xmath22-band spectroscopy of 2mass j1315@xmath02649ab was obtained using the keck ii oh - suppressing infrared integral field spectrograph ( osiris ; @xcite ) and lgsao in mostly clear conditions on 2010 may 19 ( ut ) . we used the 35 mas - scale camera and h@xmath46 filter , providing 1.471.80 @xmath23 spectroscopy at an average resolution of 3800 and dispersion of 2.1 pixel@xmath14 over a 0@xmath1056@xmath312@xmath1024 field of view . the instrument rotator was set at a position angle of @xmath045@xmath35 to accommodate both components in the rectangular field of view . six exposures of 600 s each were obtained at an average airmass of 1.52 using a linear dither pattern with steps of 0@xmath104 along the long axis , and tip - tilt correction for lgsao operation was again provided by usno - b1.0 0631 - 0348160 . these observations were followed by a 600 s exposure of a blank sky frame . we also obtained three dithered 20 s exposures of the a0 v star hd 107120 ( @xmath25 = 9.90 ) in natural guide star ( ngs ) ao mode at an airmass of 1.56 for telluric absorption correction and flux calibration . data were reduced with the osiris data reduction pipeline @xcite , version 2.3 . we first subtracted the 600 s sky frame from each of the 2mass j1315@xmath02649 images , and a median - combined dark frame from the calibrator images . we then used the pipeline to adjust bias levels , remove detector artifacts and cosmic rays , extract and wavelength - calibrate the position - dependent spectra ( using the most current rectification files as of february 2011 ) , assemble 3d data cubes and correct for dispersion . spectra for the 2mass j1315@xmath02649 primary and hd 107120 were extracted directly from the data cube by aperture photometry in each image plane , using a 3-pixel ( 105 mas ) aperture and 10 - 20 pixel ( 350700 mas ) sky annulus . for the faint companion , light contamination from the primary was a concern , so we first performed a partial subtraction of the primary s radial brightness profile . we sampled the profile over two position angle ranges 2040@xmath35 away from the separation axis , generating a mean profile as a function of wavelength and separation . we then subtracted this profile @xmath125@xmath35 about the separation axis . figure [ fig_osiris_image ] displays mosaics of both the original and subtracted images , illustrating the reduced background achieved around the companion . we measured aperture photometry for this component in each of the subtracted image planes , using a more restricted 1.5 pixel ( 52 mas ) aperture and 35 pixel ( 105 - 175 mas ) sky annulus . the individual spectra for all three sources were scaled and combined using the _ xcombspec _ routine in spextool @xcite . flux calibration and telluric correction of the 2mass j1315@xmath02649ab spectra were performed using the _ xtellcor_general _ routine in spextool , assuming a 20 nm gaussian kernel for the a0v h i lines @xcite . figure [ fig_osiris_spec ] displays the reduced spectra of the two components of 2mass j1315@xmath02649 . the spectrum of 2mass j1315@xmath02649a has exceptionally high signal - to - noise ( s / n @xmath8 200 ) , and is similar to that of the combined - light spex spectrum , with strong h@xmath4o absorption wings shortward of 1.55 @xmath23 and longward of 1.7 @xmath23 , and weak feh absorption in the 1.571.64 @xmath23 region , all indicative of a mid - type l dwarf . the notch feature , however , is no longer present . the spectrum of 2mass j1315@xmath02649b ( s / n @xmath8 25 at 1.6 @xmath23 ) is unambiguously that of a late - type t dwarf , with strong ch@xmath26 absorption at 1.6 @xmath23 . component spectral types were determined by comparing the resolved osiris spectra to the spex prism spectral libraries templates , restricting the template sample to optically - classified l dwarfs and near - infrared - classified t dwarfs . following a @xmath28-fitting procedure similar to that described above over the 1.51.8 @xmath23 region , we identified the l5 2mass j21373742 + 0808463 @xcite and the t7 2mass j07271824 + 1710012 @xcite as the best - fitting templates to 2mass j1315@xmath02649a and b , respectively ( figure [ fig_osiris_spec ] ) . an f - test weighted average of all templates indicates mean classifications of l3.5@xmath12.5 and t7@xmath10.6 for the components . the large uncertainty for the former is largely due to the broad diversity of near - infrared spectra exhibited by l dwarfs at a given optical spectral type , arising from variations in surface gravity , metallicity and cloud properties @xcite . we therefore adopt the optical l5 classification for this component and the near - infrared t7 classification for the secondary . component brightnesses on the mko system were determined by converting the combined - light 2mass @xmath47 and relative nirc2 @xmath47 magnitudes to mko @xmath48 . these conversions were computed directly from the spex prism spectra of 2mass j1315@xmath02649ab ( 2mass @xmath49 mko ) and the best - fit templates in figure [ fig_osiris_spec ] ( nirc2 @xmath49 mko ) using the appropriate filter profiles and a kurucz model spectrum of vega ( see @xcite ) . the resulting combined light and component magnitudes are listed in tables [ tab_system ] and [ tab_components ] , respectively . both components appear to have relatively normal near - infrared colors for their spectral types @xcite . lcl spectral type & l5 & 1,2 + est . distance ( pc ) & 19@xmath13 & 1 + mko @xmath29 ( mag ) & 15.07@xmath10.05 & 1,3 + mko @xmath22 ( mag ) & 14.12@xmath10.04 & 1,3 + mko @xmath45 ( mag ) & 13.45@xmath10.04 & 1,3 + mko @xmath50 ( mag ) & 1.63@xmath10.07 & 1,3 + @xmath51 ( mas yr@xmath14 ) & @xmath0682@xmath113 & 4 + @xmath52 ( mas yr@xmath14 ) & @xmath0282@xmath114 & 4 + @xmath53 ( km s@xmath14 ) & 65@xmath110 & 1,4 + @xmath19 ( km s@xmath14 ) & @xmath07@xmath19 & 1 + u ( km s@xmath14 ) & @xmath038@xmath18 & 1 + v ( km s@xmath14 ) & @xmath041@xmath19 & 1 + w ( km s@xmath14 ) & @xmath013@xmath15 & 1 + @xmath34 ( mas ) & 336@xmath13 & 1 + @xmath34 ( au ) & 6.6@xmath10.9 & 1 + @xmath38 ( @xmath35 ) & 336@xmath13 & 1 + @xmath40 ( mag ) & 3.03@xmath10.03 & 1 + @xmath41 ( mag ) & 4.53@xmath10.06 & 1 + @xmath2 ( mag ) & 5.09@xmath10.10 & 1 + age ( gyr ) & @xmath30.81.0 & 1 + est . orbit period ( yr ) & 4560 ( 1595 ) & 1 + lccc + nir spt & l3.5@xmath12.5 & t7@xmath10.6 & + mko @xmath29 ( mag ) & 15.14@xmath10.05 & 18.20@xmath10.06 & + mko @xmath22 ( mag ) & 14.14@xmath10.03 & 18.66@xmath10.07 & + mko @xmath45 ( mag ) & 13.45@xmath10.04 & 18.79@xmath10.11 & + mko @xmath50 ( mag ) & 1.69@xmath10.06 & @xmath00.59@xmath10.12 & + distance ( pc ) & 18@xmath14 & 36@xmath19 & 17@xmath19 + @xmath54 ( dex ) & -4.19@xmath10.16 & @xmath05.86@xmath10.16 & + + ( k ) & 1760@xmath170 & 790@xmath170 & + @xmath55 ( cm s@xmath56 ) & @xmath35.2 & 5.0@xmath10.5 & + + mass ( m@xmath57 ) & 60@xmath16 & 16@xmath13 & 0.26@xmath10.03 + t@xmath58 ( k ) & 1720@xmath1150 & 630@xmath160 & + @xmath55 ( cm s@xmath56 ) & 5.26@xmath10.04 & 4.57@xmath10.09 & + radius ( r@xmath57 ) & 0.90@xmath10.01 & 1.01@xmath10.02 & + + mass ( m@xmath57 ) & 74@xmath12 & 28@xmath14 & 0.38@xmath10.05 + t@xmath58 ( k ) & 1770@xmath1140 & 670@xmath170 & + @xmath55 ( cm s@xmath56 ) & 5.41@xmath10.01 & 4.93@xmath10.08 & + radius ( r@xmath57 ) & 0.84@xmath10.02 & 0.91@xmath10.02 & + + mass ( m@xmath57 ) & 76@xmath12 & 37@xmath15 & 0.48@xmath10.06 + t@xmath58 ( k ) & 1790@xmath1140 & 690@xmath170 & + @xmath55 ( cm s@xmath56 ) & 5.44@xmath10.02 & 5.10@xmath10.08 & + radius ( r@xmath57 ) & 0.83@xmath10.02 & 0.85@xmath10.02 & + + mass ( m@xmath57 ) & 77@xmath11 & 48@xmath15 & 0.62@xmath10.06 + t@xmath58 ( k ) & 1790@xmath1140 & 720@xmath170 & + @xmath55 ( cm s@xmath56 ) & 5.44@xmath10.02 & 5.28@xmath10.07 & + radius ( r@xmath57 ) & 0.83@xmath10.02 & 0.79@xmath10.02 & + component distances were computed using the mko @xmath48 absolute magnitude / spectral type relations of @xcite ; we considered both the `` bright '' and `` faint '' relations . propagating uncertainties in component spectral types and photometry , and scatter in the relations , through monte carlo simulation yields consistent distances for each component in all three bands and both relations , with mean values of 18@xmath14 pc and 35@xmath19 pc for 2mass j1315@xmath02649a and b , respectively . these distances are formally consistent with each other , differing at the 1.8@xmath43 level ; the latter has a larger uncertainty due to the larger photometric error for this component ( table [ tab_components ] ) . the error - weighted mean distance of 20@xmath13 pc matches the 21.7 pc estimate of @xcite . at this distance , the projected separation of the components is 6.6@xmath10.9 au , right at the peak of the separation distribution of resolved very low - mass field binaries @xcite . component luminosities were computed from the individual mko @xmath48 magnitudes using bolometric correction ( bc ) relations as a function of spectral type , as quantified in @xcite . apparent bolometric magnitudes were converted to absolute bolometric magnitudes by adopting a common distance of 20@xmath13 pc , and luminosities calculated assuming @xmath59 = 4.74 . luminosities computed in each of the @xmath48 bands were again mutually consistent , resulting in @xmath54 = @xmath04.19@xmath10.16 and @xmath05.86@xmath10.16 for 2mass j1315@xmath02649a and b , respectively . the luminosity for 2mass j1315@xmath02649a is similar to other l4.5l5.5 field dwarfs as compiled by @xcite and @xcite , while 2mass j1315@xmath02649b is somewhat underluminous for its spectral type . the similar distances and relative small projected separation of 2mass j1315@xmath02649a and b indicates that these sources are cospatial ; we also find that their space motions are aligned . the three epochs of nirc2 and osiris relative astrometry are consistent with each other in both right ascension and declination , and despite the short period between the observations the high proper motion of 2mass j1315@xmath02649 ( @xmath51 = @xmath0682@xmath113 mas yr@xmath14 , @xmath52 = @xmath0282@xmath114 mas yr@xmath14 ; @xcite ) allows us to rule out either component as a ( non - moving ) background star at the 9@xmath43 level . furthermore , the magnitudes and position angles of the component proper motions are identical to within @xmath147 mas yr@xmath14 ( 4.4 km s@xmath14 at 20 pc ) and @xmath14@xmath35 . the radial motions of the two components are also equivalent . cross - correlation of the osiris spectra with zero - velocity spectral model templates from ( * ? ? ? * see section 4.4 ) yield identical velocities to within the uncertainties ( @xmath18@xmath19 = 0@xmath19 km s@xmath14 ) . with a projected tangential velocity of @xmath53 = 70@xmath118 km s@xmath14 , and adopting the radial velocity from the combined - light optical spectrum above , we find @xmath60 space velocities in the local standard of rest = 10 km s@xmath14 , @xmath25 = 5.25 km s@xmath14 and @xmath61 = 7.17 km s@xmath14 in the lsr @xcite , where the directions of @xmath60 follow a right - handed coordinate system . ] ( lsr ) of @xmath62 = @xmath038@xmath18 km s@xmath14 , @xmath25 = @xmath041@xmath19 km s@xmath14 and @xmath61 = @xmath013@xmath15 km s@xmath14 . these values lie outside the 1@xmath43 velocity spheroid for the `` cold '' population of nearby l dwarfs reported by @xcite , but are within the velocity spheroid of local thick disk stars ( e.g. , ) . both the kinematics and spectral properties of 2mass j1315@xmath02649 are therefore consistent with an older dwarf system in the galactic disk population . assuming that 2mass j1315@xmath02649ab comprises a coeval system , insight into the physical properties of its components can be obtained by joint comparison to atmospheric and evolutionary models . the spectral and kinematics analysis above suggests that 2mass j1315@xmath02649ab is an older system ; a more quantitative constraint comes from the absence of li i absorption in the combined light optical spectrum , which is dominated by the l5 primary . the absence of this line sets a minimum mass of @xmath630.06 m@xmath64 for this component @xcite . combined with its luminosity , the evolutionary models of @xcite , and @xcite indicate a minimum age for the primary ranging from 0.8 to 1.0 gyr . this implies a minimum secondary mass of 0.013 m@xmath64 , around the deuterium - burning minimum mass limit . + figure [ fig_evol ] displays the regions in t@xmath58/@xmath55 space occupied by the components as constrained by their luminosities , the minimum mass of 2mass j1315@xmath02649a , and the evolutionary models listed above . table [ tab_components ] details specific physical properties for select ages based on the evolutionary models of . with no empirical upper limit on the age of this system , the mass of 2mass j1315@xmath02649a could be above the hydrogen - burning mass limit ( m @xmath3 0.072 m@xmath64 for @xmath65 @xmath3 2 gyr ) , while the secondary must be substellar at any age . the estimated mass ratio of the system , @xmath66 m@xmath4/m@xmath5 , is one of the smallest inferred for a very low - mass binary : @xmath630.3 ( @xmath630.6 ) for an age of 1 gyr ( 10 gyr ) . this makes 2mass j1315@xmath02649ab a unique system , given that @xmath6390% of resolved brown dwarf binaries identified to date have @xmath67 @xcite . indeed , if low - mass binaries prefer to be in higher mass ratio systems , these values further support the hypothesis that 2mass j1315@xmath02649 is quite old . t@xmath58 constraints on the components15701930 k for the primary and 570790 k for the secondary ( 1@xmath43 ranges)are again consistent with comparably classified field dwarfs @xcite , with 2mass j1315@xmath02649b being somewhat on the cool side for its spectral type . the lower mass limit for 2mass j1315@xmath02649a tightly constrains its surface gravity to @xmath55 = 5.225.46 cm s@xmath56 , while 2mass j1315@xmath02649b has a broader constraint of 4.465.35 cm s@xmath56 . an independent assessment of the component atmospheric parameters was made by fitting the osiris spectra to the bt - settl models of @xcite . these models are based on the phoenix code @xcite , and reflect an update to the original settl models of @xcite with a microturbulence velocity field determined from 2d hydrodynamic models and updated solar abundances from . we followed the same fitting procedure described in ( * ? ? ? * see also @xcite and @xcite ) , using a set of solar - metallicity models sampling t@xmath58 = 6002500 k ( 100 k steps ) and @xmath55 = 4.05.5 cm s@xmath56 ( 0.5 cm s@xmath56 steps ) . model surface fluxes ( in @xmath68 units ) and the osiris spectra were smoothed to a common resolution of @xmath11 = 3500 using a gaussian kernel , and interpolated onto a common wavelength grid . the data were then scaled to the appropriate @xmath22-band apparent magnitude . data and models were compared over the 1.51.75 @xmath23 region using a @xmath28 statistic , with the degrees of freedom equal to the number of resolution elements sampled . an optimal scaling factor @xmath69 was computed for each fit to minimize @xmath28 , where @xmath33 is the radius of the brown dwarf and @xmath70 its distance from the sun @xcite . we further constrained our model selection by requiring that the model - inferred distance be within 2@xmath43 of the estimated spectrophotometric distance of the system ( 1326 pc ) . means and uncertainties in the atmospheric parameters were determined using the f - pdf as a weighting factor , as above ; we also propagated sampling uncertainties of 50 k and 0.25 dex for t@xmath58 and @xmath55 , respectively . note that these uncertainties quantify _ experimental _ errors ; they do not account for _ systematic _ errors that arise from the fidelity of the model fits , as discussed below . figure [ fig_osirisfit ] displays the best - fitting models for the osiris spectra . for wavelengths shortward of 1.55 @xmath23 , the models provide reasonably good fits to the forest of h@xmath4o lines present in both component spectra . however , at longer wavelengths we see deviations in the primary arising from missing feh opacity and a premature downturn in fluxes longward of 1.65 @xmath23 . the latter is symptomatic of overly blue spectral energy distributions across the near - infrared range exhibited by the models at these temperatures . there are also deviations in model fits to the secondary near the 1.58 @xmath23 peak and within the 1.61.75 @xmath23 ch@xmath26 absorption system . note that , quantitatively , the fit to the secondary is better ( @xmath71 = 1.52 ) than that to the primary ( @xmath71 = 38.4 ) , but this mainly stems from the high s / n data for the latter ; neither fit reproduces the observed spectrum with great fidelity . nevertheless , visual inspection confirms that these are the best fits among the model sample , and the inferred mean parameters t@xmath58 = 1760@xmath170 k and @xmath55 @xmath3 5.2 cm s@xmath56 for the primary and t@xmath58 = 790@xmath170 k and @xmath55 = 5.2@xmath10.4 cm s@xmath56 for the secondary are roughly in line with estimates from the evolutionary model parameters above . with independent determinations of luminosity and t@xmath58 for both components , we can now examine whether the evolutionary and atmospheric models are consistent with each other assuming the system is coeval ; this is the so - called coevality test ( e.g. , @xcite ) . prior studies of brown dwarf binaries have produced mixed results with respect to this test , with a few low - temperature systems showing evidence of @xmath6350100 k offsets ( both high and low ) between evolutionary and atmospheric models @xcite . however , spectroscopic t@xmath58s have generally been taken from estimates of comparably classified field dwarfs , rather than spectroscopic fits to the binary components themselves . such temperature estimation by proxy could result in systematic biases . the single exception is the t1+t6 binary @xmath72 indi bc , for which resolved optical and near - infrared spectroscopy have enabled determinations of individual component luminosities and t@xmath58s , and comparison of these components on the hr diagram indicate that they are coeval with each other and with their stellar primary @xcite . + for 2mass j1315@xmath02649ab , we find good agreement between atmospheric and evolutionary models for the primary but not for the secondary . figure [ fig_evol ] compares the t@xmath58 and @xmath55 constraints from our spectral models fits to those from the evolutionary model comparisons . the atmospheric t@xmath58 and @xmath55 constraints for the primary overlap reasonably well for all four evolutionary model sets , although there is less agreement for the cloudy models of @xcite . moreover , these values are consistent with the @xmath630.8 - 1.0 gyr minimum age of the system based on the absence of li i in the optical spectrum . for the secondary , overlap in t@xmath58 and @xmath55 regions is not as good , with essentially no overlap for the @xcite cloudy models . this discrepancy can also be seen in figure [ fig_hr ] , which compares the hr diagrams for all four evolutionary models to the luminosities and spectral model fit t@xmath58s for the 2mass j1315@xmath02649 components . both sit at the lower envelope of the evolutionary tracks , consistent with older ages ; however , 2mass j1315@xmath02649b falls off the @xcite cloudy tracks entirely . the sense of the deviations between the model comparisons of 2mass j1315@xmath02649b is that the spectral model fit t@xmath58s are systematically higher than the evolutionary model t@xmath58s . as we do not have any other independent constraints on the system ( e.g. , age , metallicity or component masses ) , we can not _ a priori _ determine whether this mismatch is specifically attributable to errors in the spectral or evolutionary models . however , based on the fits shown in figure [ fig_osirisfit ] , we suspect the former given the poor match between the bt - settl models and spectral data around the 1.6 @xmath23 ch@xmath26 band . this feature causes persistent problems in spectral model ftis due to incomplete ch@xmath26 opacities at t dwarf temperatures @xcite . a decrease of just 100200 k in the derived secondary t@xmath58 would bring both components in precise alignment with evolutionary models in both t@xmath58/@xmath55 and hr diagrams , a shift previously suggested in prior low - mass binary analyses ( although not necessarily in the same direction ; @xcite ) . it is also notable that spectral model fits for 2mass j1315@xmath02649a are even worse than those for 2mass j1315@xmath02649b . it may be that the alignment of evolutionary and spectral model parameters for this component is merely an example of `` chance agreement '' . in any case , without more accurate fitting of l and t dwarf spectral data , such coevality tests are fundamentally inconclusive about the underlying accuracies of model - derived parameters . we emphasize that the discrepancies noted here are only at the 1@xmath43 level , and should be verified through more precise constraints on the component luminosities and t@xmath58s . these can best be accomplished through a parallax distance measurement of the system , as the distance uncertainty dominates the luminosity uncertainty . in addition , resolved spectroscopy spanning the near - infrared ( and possibly optical ) range would provide a more robust test of the atmosphere models , and allow us to test different sets of models . unfortunately , mass measurements for this widely - separated system are probably not feasible in the near future . a probability analysis of the possible orbits for this system.[foot_periodsim ] ] predicts likely periods of 4560 yr ( 15 - 95 yr at 90% confidence ) . this rules out a `` fast '' astrometric orbit determination , and the relative orbital radial velocities ( @xmath730.5 km s@xmath14 ) are comparable to current systematic uncertainties for isolated late - type dwarfs ( e.g. , @xcite ) . despite these challenges , the 2mass j1315@xmath02649 system is an important benchmark for empirical tests of atmospheric and evolutionary models given its proximity , well - separated components , component types and relatively old age . while the presence of a faint , substellar companion provides useful constraints on the physical properties of the 2mass j1315@xmath02649 system , it also suggests that a binary interaction could be responsible for the unusual strength and persistence of its nonthermal emission . in a study of the comparably active t6.5e dwarf 2mass j1237 + 6526 , @xcite proposed that accretion of material from a binary companion via roche lobe overflow could be a mechanism for sustained emission , as the inverted mass / radius relationship for brown dwarfs allows for sustained mass loss for @xmath74 , precisely the mass limit we find for 2mass j1315@xmath02649ab . however , the maximum separation for roche lobe overflow is only a few jupiter radii , well below the @xmath815,000 r@xmath57 projected separation for this pair . moreover , the ballistic velocity of material impacting the surface of the primary , @xmath75 km s@xmath14 @xmath8 500 km s@xmath14 ( where m@xmath76 and r@xmath76 are the mass and radius of the primary in jupiter units ) would have been readily detectable from line broadening over a broad range of viewing geometries . combined with the absence of infrared excess associated with a circumstellar disk , we rule out accretion from the observed binary companion as the source of line emission . the most likely explanation is that 2mass j1315@xmath02649a possesses an unusually active chromosphere for its spectral type , perhaps reflecting an unusually strong magnetic field . @xcite postulated an inverse relationship between age and activity among late - type m and l dwarfs , finding that stellar - mass ( and hence older ) l dwarfs were more likely to exhibit h@xmath6 emission than younger l dwarfs in their sample ( note that @xcite find marginal evidence for the opposite trend ) . @xcite and infer a similar mass - dependence in the strength and persistence of magnetic fields on brown dwarfs based on the total energy flux available for field generation . in their model , objects above the hydrogen burning mass limit retain kilogauss fields for up to 10 gyr . if strong fields correlate with strong chromospheres in l dwarfs ( as they do for late - type m dwarfs ; @xcite ) , then the presence of nonthermal emission may align with kinematic , spectroscopic and binary mass ratio evidence that 2mass j1315@xmath02649 is an old system , and 2mass j1315@xmath02649a a relatively massive brown dwarf or low - mass star with a strong magnetic field . a direct measurement of this component s magnetic field could be obtained through zeeman line broadening measurements in the 0.99 @xmath23 feh wing - ford band . however , the high spectral resolution required for such a measurement ( @xmath11 @xmath77 30,000 ) makes it a challenge for this faint system ; indeed , no zeeman broadening measurements have been reported for an l dwarf to date . an independent issue is how such a field could generate a persistent chromosphere in the presence of a highly neutral photosphere . variability in the strength of h@xmath6 emission from 2mass j1315@xmath02649 over the past decade may indicate microflaring as a viable source of heating , and a possible indicator of vigorous turbulent field generation @xcite . alternately , @xcite have hypothesized that dust grain ionization and electron avalanche ( i.e. , lightning ) could locally increase the photospheric ionization fraction and magnetic field coupling in dusty l dwarfs , although early models appear to favor this mechanism in younger , lower - mass brown dwarfs and exoplanets . a connection between magnetic activity and `` cloudiness '' has yet to be explored . for completeness , we note that because of its relatively wide separation , magnetic interaction between 2mass j1315@xmath02649a and b is an unlikely source of emission in this system . magnetospheric interactions over scales of up to @xmath6350 r@xmath78 have been implicated in outbursts from solar - mass t tauri binaries . however , the estimated @xmath6330 r@xmath78 size of a typical l dwarf magnetic field @xcite / r@xmath79 @xmath80 b@xmath81 . ] is still several orders of magnitude smaller than the projected separation of the 2mass j1315@xmath02649 binary . in addition , the possibility that the observed nonthermal emission arises from the secondary , rather than the primary , is unlikely . this scenario would imply a relative emission luminosity of @xmath7 @xmath8 @xmath02.5 for the secondary , an amplitude seen only during exceptionally large flare bursts from m and l dwarfs ( e.g. , @xcite ) . while other interactions could be postulated ( e.g. , magnetic wind interactions , kozai - like perturbations of an unseen third component ) , these scenarios are far more complex than the simple hypothesis of enhanced chromospheric emission powered by a strong magnetic field around a relatively old and massive cool dwarf . in other words , while the presence of a low - mass t dwarf secondary in this system is certainly intriguing , it appears to play no role in the observed nonthermal emission . we have identified a t7 brown dwarf companion to the unusually active l5e dwarf 2mass j1315@xmath02649 , a source that continues to exhibit strong h i and alkali line emission despite its late spectral type . resolved imaging and spectroscopy confirm the pair to be co - moving and co - spatial , and evolutionary models indicate an unusually low mass ratio as compared to other low - mass field binaries . the spectral and kinematic properties of both components confirm prior indications that this system is relatively old ( @xmath65 @xmath3 0.81.0 gyr ) , and likely a member of the old galactic disk population . joint comparison to atmospheric and evolutionary models ( a coevality test ) also supports an older age for this system , but reveals discrepancies in the case of the secondary ; we suspect these are due to continued shortcomings in the modeling cool brown dwarf spectra . the age and separation of the system , coupled with the narrow emission lines and absence of mid - infrared excess , rule out accretion from or magnetic interaction with the t dwarf secondary as the emission source . rather , we attribute it to an unusually strong magnetic field as predicted by energy flux scaling arguments for relatively old and massive low - mass dwarfs . a direct test of this hypothesis can be achieved through zeeman line broadening measurements , although the source of chromospheric heating remains a separate issue . whether the larger sample of hyperactive cool dwarfs are also old , relatively massive , and possess strong magnetic fields remains to be determined , but they hint at a remarkable inversion of the standard age - activity relationship for low - mass stars . the authors would like to thank paul sears and john rayner at irtf , al conrad , heather hersley and marc kassis at keck , and sergio vera and david osip at las campanas for their assistance in the observations reported here . thanks also go to shelley wright for her help with the osiris data reduction , and our anonymous referee for her / his careful and timely critique . this publication makes use of data from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center , and funded by the national aeronautics and space administration and the national science foundation . 2mass data were obtained from the nasa / ipac infrared science archive , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . this research has made use of the simbad database , operated at cds , strasbourg , france ; the very - low - mass binaries archive housed at http://www.vlmbinaries.org and maintained by nick siegler , chris gelino , and adam burgasser ; and the m , l , and t dwarf compendium housed at dwarfarchives.org and maintained by chris gelino , davy kirkpatrick , and adam burgasser . the authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of mauna kea has always had within the indigenous hawaiian community . we are most fortunate to have the opportunity to conduct observations from this mountain . , b. , leggett , s. k. , lucas , p. w. , pinfield , d. j. , smart , r. l. , day - jones , a. c. , jones , h. r. a. , murray , d. , nickson , e. , tamura , m. , zhang , z. , lodieu , n. , tinney , c. g. , & zapatero osorio , m. r. 2010 , , 404 , 1952 , j. d. , looper , d. l. , burgasser , a. j. , schurr , s. d. , cutri , r. m. , cushing , m. c. , cruz , k. l. , sweet , a. c. , knapp , g. r. , barman , t. s. , bochanski , j. j. , roellig , t. l. , mclean , i. s. , mcgovern , m. r. , & rice , e. l. 2010 , , 190 , 100 , a. , gasaway , t. , song , i. , iserlohe , c. , weiss , j. , larkin , j. e. , barczys , m. , & lafreniere , d. 2004 , in presented at the society of photo - optical instrumentation engineers ( spie ) conference , vol . 5492 , society of photo - optical instrumentation engineers ( spie ) conference series , ed . a. f. m. moorwood & m. iye , 14031410 , j. , barczys , m. , krabbe , a. , adkins , s. , aliado , t. , amico , p. , brims , g. , campbell , r. , canfield , j. , gasaway , t. , honey , a. , iserlohe , c. , johnson , c. , kress , e. , lafreniere , d. , lyke , j. , magnone , k. , magnone , n. , mcelwain , m. , moon , j. , quirrenbach , a. , skulason , g. , song , i. , spencer , m. , weiss , j. , & wright , s. 2006 , in presented at the society of photo - optical instrumentation engineers ( spie ) conference , vol . 6269 , society of photo - optical instrumentation engineers ( spie ) conference series , j. l. , burles , s. , thompson , i. b. , shectman , s. a. , bigelow , b. c. , burley , g. , birk , c. , estrada , j. , jones , p. , smith , m. , kowal , v. , castillo , j. , storts , r. , & ortiz , g. 2008 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 7014 , society of photo - optical instrumentation engineers ( spie ) conference series , d. g. , levine , s. e. , canzian , b. , ables , h. d. , bird , a. r. , dahn , c. c. , guetter , h. h. , harris , h. c. , henden , a. a. , leggett , s. k. , levison , h. f. , luginbuhl , c. b. , martini , j. , monet , a. k. b. , munn , j. a. , pier , j. r. , rhodes , a. r. , riepe , b. , sell , s. , stone , r. c. , vrba , f. j. , walker , r. l. , westerhout , g. , brucato , r. j. , reid , i. n. , schoening , w. , hartley , m. , read , m. a. , & tritton , s. b. 2003 , , 125 , 984 , f. j. , henden , a. a. , luginbuhl , c. b. , guetter , h. h. , munn , j. a. , canzian , b. , burgasser , a. j. , kirkpatrick , j. d. , fan , x. , geballe , t. r. , golimowski , d. a. , knapp , g. r. , leggett , s. k. , schneider , d. p. , & brinkmann , j. 2004 , , 127 , 2948 , a. a. , morgan , d. p. , bochanski , j. j. , andersen , j. m. , bell , k. j. , kowalski , a. f. , davenport , j. r. a. , hawley , s. l. , schmidt , s. j. , bernat , d. , hilton , e. j. , muirhead , p. , covey , k. r. , rojas - ayala , b. , schlawin , e. , gooding , m. , schluns , k. , dhital , s. , pineda , j. s. , & jones , d. o. 2011 , , 141 , 97

levitation melting ( elm ) was invented in the twenties of the last century @xcite whereas its usage started only at the beginning of the fifties @xcite when high - frequency power generators became available . the basic principle of elm is simple : a conducting sample , usually metallic , weighting from several tens of grams up to several kilograms is placed in a coil fed by an ac current with a typical frequency ranging from about @xmath2 up to several @xmath3 . the ac magnetic field induces eddy currents in the sample which , in turn , give rise to two effects . on one hand , eddy currents interact with those in the coil giving rise to a lorentz force that repels the sample from the coil . on the other hand , the ohmic dissipation due to the induced currents provides heating of the sample . in such a way , the sample can be levitated and also melted , provided that the coil configuration and the current in it are properly adjusted . elm is particularly useful for melting reactive metals with high melting points , for example such as _ ti _ , _ zr _ , _ v _ , _ ta _ , _ mo _ , which often react with the crucible material and so get polluted by it @xcite . elm avoids contamination of the melt and allows one to carry out solidification from deeply under - cooled states which is of interest for certain material science applications @xcite . besides , elm is a well - known method for measurements of material properties of liquid metals used on ground as well as in space @xcite . the balance of gravity and electromagnetic forces is necessary but not sufficient for a successful levitation . in addition , the sample has to be stable at least to perturbations of sufficiently small amplitude . first , it means that the reaction force due to the displacement of the sample from its equilibrium position has to act against that displacement . otherwise , the equilibrium will be statically unstable and the solid sample will fall out of the coil or touch it when it is slightly perturbed . similarly , a molten sample may leak out of the coil in the result of surface folding and deformation @xcite . the static stability alone may also be insufficient for a successful levitation because sometimes the sample exhibits overstability . namely , there is a restoring force that makes the sample to execute oscillations with increasing amplitude so that the sample eventually hits the coil or leaves it @xcite . similarly , sometimes the sample is observed to spin - up @xcite . a purely electromagnetic theory trying to account for such dynamic instabilities has been proposed in our previous work and applied to some simple configurations of magnetic fields @xcite . in this study , we apply our previous theory to analyze both the static and dynamic stability of a spherical sample in more realistic axisymmetric magnetic systems made of a set of coaxial circular current loops . although the basic configuration is axisymmetric , the perturbation fields are , in general , three - dimensional that renders the problem mathematically more complicated . to calculate 3d fields for a spherical sample we use similar analytic techniques as in refs . however , our approach differs from the previous ones by an original use of the gauge transformation for the 3d vector potential in order to satisfy boundary conditions for the induced current . this allows us to carry out the analysis only in terms of the vector potential without considering the scalar electrostatic potential . therefore our approach is considerably simpler compared to @xcite . the paper is organized as follows . the stability of small amplitude oscillations of arbitrary direction is considered in sec . 2 beginning with the governing equations and analytic solutions for the vector potential of a circular current loop in spherical harmonics . in this section we also derive solutions for the axisymmetric base state and the perturbation field due to a small displacement . section 3 presents governing equations and analytic solution for the spin - up instability . numerical results for both instabilities in several simple inductors are discussed in section 4 and summarized in section 5 . consider a sphere with radius @xmath4 and conductivity @xmath5 moving at velocity @xmath6 in a magnetic field @xmath7 alternating with circular frequency @xmath8 . the induced electric field follows from the first maxwell equation as @xmath9 where @xmath10 is the scalar potential of the electric field and @xmath11 is the vector potential of the magnetic field . the density of the electric current induced in a moving medium is given by ohm s law @xmath12 the electric field induced by a translational solid body motion with @xmath6 being a spatially invariant vector can be represented as @xmath13 assuming that the frequency of the alternating magnetic field is sufficiently low to neglect the displacement current , the second maxwell equation leads to the following advection - diffusion equation @xmath14 where the gauge invariance of @xmath11 has been employed to define the scalar potential as @xmath15 boundary conditions at the surface follow from the continuity of the magnetic field @xmath16_{s}=\left[\partial_{n}\vec{a}_{0}\right]_{s}=0 \label{eq : bcnd-1}\ ] ] where @xmath17_{s}$ ] denotes a jump of the quantity @xmath18 across the boundary @xmath19 ; @xmath20 is the derivate normal to the boundary . sketch to the formulation of the problem . ] the magnetic field of an inductor consisting of a set of coaxial circular loops represents a superposition of the fields of the separate loops . therefore , we first consider a single loop lying parallel to the _ @xmath21 _ plane and centered with respect to the _ z_-axis of a cartesian coordinate system with a sphere at its origin , see fig . [ fig : sketch ] . approximation of a toroidal inductor of small cross - section by a circular loop is considered in more detail in appendix 1 . the loop is supplied with an ac current with amplitude @xmath22 alternating harmonically with a circular frequency @xmath8 as @xmath23 . henceforth we use the magnetic diffusion time @xmath24 and the radius of the sphere @xmath4 as time and length scales while the current density and the vector potential are scaled by @xmath25 and @xmath26 , respectively . in the following , we assume all quantities to be dimensionless with the same notation as for the dimensional counterparts used so far . further we consider the vector potential generated by the loop @xmath27 with the axisymmetric amplitude @xmath28 where @xmath29 is the dimensionless ac frequency which will be the main parameter throughout this paper . the amplitude of the current density of such a loop can be represented by using the dirac @xmath30 function : @xmath31 where @xmath32 and @xmath33 are the meridional angle and the spherical radius of the loop with respect to the center of the sphere as shown in fig . [ fig : sketch ] . further , it is advantageous to introduce a complex unit vector @xmath34 that allows us to represent the azimuthal unit vector as @xmath35 where @xmath36 and @xmath37 are @xmath38 and @xmath39 unit vectors , respectively , and the asterisk denotes the complex conjugate . then the current density can be expanded in spherical harmonics as @xmath40 where @xmath41 are spherical harmonics as defined in @xcite . the expansion coefficients are @xmath42 using the generating function of the spherical harmonics @xcite the vector potential of a circular loop is obtained as @xmath43 where @xmath44 is the cylindrical radius of the loop ; @xmath45 ; @xmath46 . in order to facilitate the following algebra , it is advantageous to introduce the functions @xmath47 and @xmath48 , which are the outer and the inner solutions of the laplace equation associated to the spherical harmonic @xmath41 . basic properties of those functions are given in appendix 2 . for the region occupied by the sphere ( @xmath49 ) the solution above takes the form @xmath50 where @xmath51 for @xmath52 is assumed . for a sphere at rest the vector potential is sought as @xmath53,$ ] where @xmath54 is complex . for the interior of the sphere , ( [ eq : vecpot ] ) takes the form @xmath55 while for the exterior we have @xmath56 whereas ( [ eq : gauge ] ) takes the form of the coulomb gauge @xmath57 . for the interior , the solution is @xmath58 where @xmath59 are unknown coefficients to be determined from the boundary conditions ; @xmath60 is the spherical bessel function of index @xmath0 @xcite . for the exterior , the solution is represented as a superposition of external and induced fields @xmath61 where the latter is sought as @xmath62 from the conditions ( [ eq : bcnd-1 ] ) applied at the surface of the sphere at @xmath63 we find @xmath64 some useful properties and an efficient algorithm for the calculation of @xmath65 and its derivative are given in appendix 3 . the time - averaged total force on the sphere is @xmath66dv,\ ] ] where the integral is taken over the volume of the sphere @xmath67 . taking into account that @xmath68 and @xmath69 where @xmath70 is a purely real quantity , we obtain @xmath71\times\vec{b}_{0}^{e}dv.$ ] according to the momentum conservation law the force exerted by a magnetic field on a body is opposite to that exerted by the body on the source of the magnetic field @xmath72dv$ ] where the integral is taken over the space outside the sphere @xmath73 the last integral can be taken straightforwardly because the external current ( [ eq : j0-sph ] ) is defined in terms of a dirac @xmath30 function . taking into account that the induced magnetic field outside the sphere is @xmath74 where @xmath75 , we find @xmath76 f_{l-1}f_{l}\left(\frac{l^{2}-1}{4l^{2}-1}\right)^{1/2 } \label{eq : f0}\ ] ] where @xmath77 involves the normalised legendre function @xmath78 related to the spherical harmonics as @xmath79 @xcite . in case of @xmath80 current loops we have @xmath81 where @xmath82 is the dimensionless current carried by the _ n_-th loop ; @xmath83 , @xmath84 , and @xmath85 are defined analogously as for the single loop . note that the solution ( [ eq : f0 ] ) alone could be obtained in a simpler way as in ref . our following analysis , however , is aimed at non - axisymmetric solutions which require a more complicated algebra . consider a small displacement @xmath86 of the center of the sphere from the position @xmath87 to @xmath88 in the frame of reference related to the sphere , this corresponds to a perturbation of the external magnetic field @xmath89 where @xmath90 @xmath91 and @xmath92 now the external field represents a superposition of the axisymmetric base field @xmath93 and a small but in general non - axisymmetric perturbation @xmath94 . the perturbation of the inner field @xmath95 is governed by the same eqs . ( [ eq : a0-ex ] , [ eq : a0-in ] ) as the base field with the difference that now the external field is given by @xmath94 . consequently , the perturbation of the inner field and the corresponding induced field outside the sphere can be obtained straightforwardly by replacing the coefficients @xmath96 by @xmath97 in the corresponding solutions ( [ eq : a0-insol ] , [ eq : a0-exsol ] ) for the base field . although the solution of the perturbed field obtained in such a way satisfies ( [ eq : a0-ex ] , [ eq : a0-in ] ) and the boundary conditions , it turns out that the induced field outside does not satisfy the coulomb gauge unless the problem is axisymmetric , _ i.e. _ , the offset is along the symmetry axis . thus , although the laplace equation is satisfied we formally have @xmath98 which implies that the current is leaking outside the sphere . in addition note that the continuity of the tangential components of induction following from the continuity of the vector potential and its normal derivative at the surface of the sphere ensures continuity of the radial , _ i.e. _ normal , component of the current because @xmath99 . thus , the vector potential satisfying the laplace equation and the coloumb gauge outside the sphere and being continuous at the surface ensures that the current is closed in the sphere . as shown above , the induced vector potential unambiguously follows from the imposed one . thus , the induced vector potential can be modified only by changing the imposed one . this can be done by employing the gauge invariance which allows us to add to the external vector potential @xmath94 a gradient of some gauge potential defined as @xmath100 in order to fulfill the coulomb gauge for the external vector potential , @xmath101 has to be a harmonic function which is reflected in the expression above . so , we obtain a set of free coefficients @xmath102 which can be chosen to satisfy the coulomb gauge for the induced field . this gauge transformation results in the replacement of the original coefficients @xmath97 by @xmath103 then , similarly to the base field , the perturbation of the induced field is obtained as @xmath104 from the coulomb gauge we find @xmath105 the corresponding induction is @xmath106.\end{aligned}\ ] ] according to our previous theory @xcite , small - amplitude oscillations of the sphere with @xmath107 give rise to a perturbation of the force on the sphere @xmath108 = -\vec{k}\cdot\vec{x}+\vec{\g}\cdot\frac{d\vec{x}}{dt}.\ ] ] where @xmath109 and @xmath110 are the effective electromagnetic stiffness and damping tensors , respectively . further we refer to the elements of @xmath109 and @xmath111 as the spring constants and growth rates , respectively . note that due to the axial symmetry both @xmath109 and @xmath111 are purely diagonal matrices with zero azimuthal components . at first , the axial symmetry implies that a purely axial displacement gives rise to a similar reaction force . second , if the displacement is purely radial the reaction force , on the one hand , has to change to the opposite together with the displacement because of linearity but , on the other hand , it has to rotate around the symmetry axis together with the latter . thus , a purely radial displacement causes a purely radial reaction force and , therefore , both @xmath109 and @xmath111 are purely diagonal . both diagonal elements of @xmath109 have to be positive for the static stability whereas the corresponding growth rates have to be negative for dynamic stability . note that the growth rates can become positive and , thus , destabilizing when the dimensionless ac frequency exceeds a certain threshold depending on the configuration of the field @xcite . both the spring constants and the growth rates can be calculated for the given direction of the displacement specified by @xmath112 by using the corresponding base and perturbation fields @xcite : @xmath113\times\vec{b}_{0}^{e } + \re\left[\vec{j}_{0}\right]\times\vec{b}_{1}^{e}\right]dv\\ \vec{\g}\cdot\vec{\epsilon}_{x } & = & \frac{1}{2}\partial_{\bar{\omega } } \int_{v}\im\left[\vec{j}_{1}\right]\times\vec{b}_{0}^{e}dv.\end{aligned}\ ] ] similarly to the integral force , it is advantageous to change the region of integration from the sphere to the current loop @xmath114+\vec{j}_{1}^{e}\times \re\left[\vec{b}_{0}^{i}\right]\right]dv,\\ \vec{\g}\cdot\vec{\epsilon}_{x } & = & -\frac{1}{2}\partial_{\bar{\omega } } \int_{\bar{v}}\vec{j}_{0}^{e}\times\im\left[\vec{b}_{1}^{i}\right]dv.\end{aligned}\ ] ] after some algebra we find @xmath115\frac{f_{l}^{2}}{2l+1}(l^{2}-1)\\ & & { \times}\left[\frac{1}{2}\vec{e}_{z}\times\vec{\epsilon}_{x}\times \vec{e}_{z}(1-\frac{1}{l})+\vec{e}_{z}(\vec{\epsilon}_{x}\cdot\vec{e}_{z})\right ] . \nonumber \label{eq : gamma}\end{aligned}\ ] ] taking into account that @xmath116 because @xmath117 where @xmath118 is defined via dirac @xmath30 functions we eventually obtain @xmath119\left\{\frac{f_{l}^{2}}{2l+1}(l^{2}-1)\right.\\ & & \left . { \times}\left[\frac{1}{2}\vec{e}_{z}\times\vec{\epsilon}_{x}\times \vec{e}_{z}(1-\frac{1}{l})+\vec{e}_{z}(\vec{\epsilon}_{x}\cdot \vec{e}_{z})\right]\right.\\ & & \left . { -}\:\frac{f_{l-1}f_{l+1}}{2l-1}\left ( \frac{l(l+2)(l^{2}-1)}{(2l-1)(2l+3)}\right)^{1/2}\right.\\ & & \left . { \times}\left[\frac{1}{2}\vec{e}_{z}\times\vec{\epsilon}_{x}\times\vec{e}_{z } -\vec{e}_{z}(\vec{\epsilon}_{x}\cdot\vec{e}_{z})\right]\right\}.\end{aligned}\ ] ] now , we can use this series to calculate both the stiffness and the damping coefficients for a given coil defined by the coefficients @xmath120 according to ( [ eq : fl ] ) . we consider a solid sphere , as in section 2 , rotating with angular velocity @xmath121 in an alternating magnetic field . the density of the induced current is given by ohm s law for a moving medium @xmath122 where @xmath123 is the velocity of a solid - body rotation . the electric field induced by such a rotation may be represented as @xmath124 then from the second maxwell equation with neglected displacement current , that corresponds to the quasistationary approximation assumed throughout this study , we obtain @xmath125 note that this equation governs the vector potential in the laboratory frame of reference where the body rotates while the source of the magnetic field is at rest . further we proceed to the frame of reference rotating together with the sphere . in a rotating frame of reference , where the sphere is at rest while the source of the magnetic field rotates with velocity @xmath126 , the equation above takes the form as for a body at rest @xmath127 where @xmath128 is a radius vector , which is time - dependent in the rotating frame of reference with @xmath129 . in a rotating frame of reference , the body rotation appears as an additional time - dependence of the external magnetic field modulating the time - dependence of the applied magnetic field . the time - averaged total torque on the sphere is @xmath130\times\vec{b}_{0}^{e}dv.\ ] ] according to the conservation of the angular momentum the torque exerted by the magnetic field on a body is opposite to that exerted by the body on the source of the magnetic field @xmath131dv,\ ] ] where the integral is taken over the volume @xmath132 outside the sphere . the last integral can be taken straightforwardly because @xmath133 is defined via dirac @xmath30 functions . because @xmath134 the integral above may be represented as @xmath135dv\ ] ] further we consider a sphere rotating about an axis perpendicular to the symmetry axis of an axisymmetric external magnetic field of a circular loop . the coordinate system related to the loop with the _ @xmath136 _ axis as symmetry axis is chosen so that the rotation takes place about the @xmath38-axis , as shown in fig . [ fig : sketch ] . it is convenient to analyze this rotation in a coordinate system turned about the @xmath39-axis by 90 degrees so that the _ @xmath137_-axis of the transformed coordinate system coincides with the original @xmath38-axis which is the axis of rotation and , respectively , the _ @xmath138_-axis is directed opposite to the _ @xmath136 _ axis of the original coordinate system . then the basis vectors of both coordinate systems are related as @xmath139 @xmath140 and @xmath141 respectively , the complex unity vector is @xmath142 where @xmath143 to represent the vector potential of a circular loop in the turned coordinate system we use the addition theorem for legendre polynomials @xcite written as @xmath144 together with the relation @xmath145 which yields @xmath146 where @xmath147.\end{aligned}\ ] ] similarly , the current density of the loop can be rewritten as @xmath148 in this case , we define a gauge potential @xmath149 where @xmath150 @xmath102 are unknown coefficients to be determined subsequently from the coulomb gauge for the vector potential induced outside the sphere . addition of @xmath151 to @xmath152 results in the replacement of the coefficients @xmath153 by @xmath154 now we can simply proceed to the frame of reference rotating together with the sphere by redefining the azimuthal angle as @xmath155 where @xmath156 is the azimuthal angle in the rotating frame of reference and @xmath157 is the dimensionless angular velocity of rotation . respectively , the complex unit vector and the spherical harmonics change in a rotating frame of reference as @xmath158 and @xmath159 in a rotating frame of reference , the vector potential takes the form @xmath160 further we consider the current alternating harmonically with the dimensionless frequency @xmath161 as @xmath162 and the vector potential alternating in the laboratory frame of reference , respectively , as @xmath163.\ ] ] in the laboratory frame of reference , the complex amplitude of the vector potential induced outside the sphere is obtained as @xmath164 where @xmath165 and @xmath166 . from the coulomb gauge @xmath167 we find @xmath168&{}={}&\\ { } = { } \left [ \frac{(\vec{e}_{\eta}'^{*}\cdot\vec{i}_{l}^{m-1})}{\sqrt{2}\left(n_{l}^{m-1}\right)^{2 } } -\frac{(\vec{e}_{\eta}'\cdot\vec{i}_{l}^{m+1})}{\sqrt{2}\left(n_{l}^{m+1}\right)^{2 } } -\frac{(\vec{e}_{z}'\cdot\vec{i}_{l}^{m})}{\left(n_{l}^{m}\right)^{2 } } \right ] & & \end{aligned}\ ] ] for the field induced outside the sphere we obtain @xmath169.\end{aligned}\ ] ] the gauge potential @xmath170 vanishes in the expression above and so it does not affect the total torque which is @xmath171\\ & & { \times}\left[(l - m+1)(l - m+2)\bar{p}_{l+1}^{m-1}(0)^{2}\right.\\ & & \left . { -}\:(l+m+1)(l+m+2)\bar{p}_{l+1}^{m+1}(0)^{2}\right].\end{aligned}\ ] ] other components of the torque , which are not parallel to the angular velocity of rotation , vanish because of @xmath172 . for slow rotations with @xmath173 we have @xmath174 that results in @xmath175 , where @xmath176f_{l}^{2}}{(2l+1)^{2}(l+1)}\sum_{m=1}^{l}m\\ & & { \times}\left[(l - m+1)(l - m+2)\bar{p}_{l+1}^{m-1}(0)^{2}\right.\nonumber\\ & & \left . { -}\:(l+m+1)(l+m+2)\bar{p}_{l+1}^{m+1}(0)^{2}\right].\nonumber\end{aligned}\ ] ] for the rest state of the sphere to be stable to rotational perturbations , the torque caused by these perturbations has to damp them that corresponds to @xmath177 in the opposite case , the rest state is unstable . the state is marginally stable when @xmath178 which is the equation defining the critical frequency @xmath179 for the spin - up instability . in the following we present some numerical results for an inductor consisting of two circular loops of equal radii placed coaxially at the distance @xmath180 from the mid - plane . the current having the same amplitude may flow either in the same or in opposite directions in both loops . subsequently , both these cases are referred to as symmetric and antisymmetric , respectively . we start with a simple symmetric case when both loops coincide ( @xmath181 ) making up a single loop with doubled current amplitude . the forces along with axial and radial spring constants versus the distance of the sphere from such a loop of dimensionless radius @xmath182 are shown in fig . [ fig : force ] for various dimensionless frequencies . as seen in fig . [ fig : force](_a _ ) , at large enough distances the force , which is repulsing , raises with decreasing distance between the loop and the sphere . at some distance , the force attains a maximum and tends to zero as the sphere approaches the center of the loop . such a reducing force implies that the position of the sphere is statically unstable to axial perturbations . the position is statically stable to axial perturbations when the axial spring constants shown in fig . [ fig : force](_b _ ) are positive that requires the sphere to be placed further away from the loop than the point of maximum force which is at @xmath183 . a position being axially stable may become unstable radially when the distance from the loop becomes too large ( see fig . [ fig : force]_b _ ) . thus the range of statically stable positions for a single loop is rather limited ( @xmath184 ) and does not change significantly with the frequency of the magnetic field . further let us turn briefly to the static stability of a sphere in an arrangement of two coaxial current loops of the same radius placed symmetrically at distance @xmath180 from the center of the sphere . for the sake of brevity we restrict the static stability analysis to the perfect - conductor approximation @xmath185 axial and radial spring constants are shown in fig . [ fig : finfz0 ] versus the distance @xmath180 at various radii of the loops for both symmetric and antisymmetric arrangements . in the symmetric arrangement , which is usually used for heating , the position of the sphere is statically unstable in radial direction when the loops are too much separated , and it becomes unstable in the axial direction when the loops are moved too tightly together . overlapping of the stability ranges for radial and axial perturbations depends on the radius of the loops . for @xmath182 the range of axial stability ends at @xmath186 where the range of radial stability begins . thus , there is no overlapping of the stability ranges in this case that implies a static instability regardless of the distance between the loops . a range of positions statically stable to both axial and radial perturbations is possible only for sufficiently small radii of the loops . for instance at @xmath187 this range is approximately @xmath188 in contrast to the symmetric arrangement , an antisymmetric one , which is usually used for positioning of the sample , ensures a statically stable state to radial perturbations regardless of the distance between the loops ( see fig . [ fig : finfz0]_b _ on right ) . the position of the sphere is stable to axial perturbations for any radius of the loop provided the loops are separated by a distance larger than the radius of the sphere ( @xmath189 ) . there might be an axial instability for smaller separations of the loops at @xmath190 ( see fig . [ fig : finfz0]_a _ on right ) . ( 0,0)(a ) ( 0,0)(b ) ( 0,0)(a ) ( 0,0)(b ) ( 0,0)(c ) forces together with axial and radial spring constants versus the distance of the sphere from the mid - plane of both symmetric and antisymmetric arrangements of two loops are shown in fig . [ fig : finfr2 ] for various distances between the loops in the perfect - conductor approximation ( @xmath191 ) . it is seen that the position of the sphere in the middle - point ( @xmath192 ) of a symmetric arrangement ( fig . [ fig : finfr2 ] on left ) is unstable either in radial or axial directions for almost all distances between the coils . however , there is one distance between the coils , @xmath186 , at which both spring constants cross zero almost simultaneously . thus , this distance is an optimal one for the static stability of the sphere in the middle - point . note that the distance @xmath193 provides the most uniform magnetic field in the vicinity of the middle - point of the symmetric arrangement . this is because the next to the leading order contribution for the magnetic field of symmetric arrangement vanishes at the aspect ratio @xmath194 , _ i.e_. , @xmath195 in ( [ eq : a0e ] ) . similarly , the distance @xmath196 providing the most uniform gradient of the magnetic field in the vicinity of the middle - point of an antisymmetric arrangement is expected to ensure the widest range of statically stable positions for loops of large enough radius . ( 0,0)(a ) ( 0,0)(b ) ( 0,0)(c ) ( 0,0)(d ) the dynamic stability of small - amplitude radial and axial oscillations is characterized by the corresponding growth rates @xmath197 and @xmath198 defined by ( [ eq : gamma ] ) . the threshold of linear stability is given by the frequency above which the corresponding growth rate becomes positive . at the threshold frequency , the amplitude of oscillations turns from decaying to growing in time . a characteristic feature of this instability is that the growth rate , becoming positive above the threshold frequency , reaches a maximum at a certain frequency and tends to zero at higher frequencies . the frequency at which the growth rate attains a maximum is referred to as the most dangerous one . threshold frequencies along with the most dangerous one and the corresponding maximal growth rates are plotted in fig . [ fig : oscz0 ] versus the half - distance between the loops at various radii of the loops for both axial and radial oscillations of a sphere about the middle - point of both symmetric and antisymmetric arrangements of two loops . the most dangerous frequency is always above the threshold one . as seen in figs . [ fig : oscz0](_a _ ) and [ fig : oscz0](_b _ ) , at large enough distances between the loops , the threshold and the most dangerous frequencies tend to @xmath199 and @xmath200 in the symmetric arrangement , and to @xmath201 and @xmath202 in the antisymmetric one , respectively . note that these critical frequencies for the oscillatory instability in spatially quadratic and linear magnetic fields coincide with those for the spin - up instability in linear and uniform magnetic fields , respectively , which will be considered later . for each radius of a symmetric arrangement , there is a certain optimal distance between the loops that maximizes the critical frequency . for large enough radii the optimal distance between the loops is approximately equal to the radius , @xmath203 , which is optimal also for the static stability as considered above . with decreasing of the loop radius , the optimal distance becomes slightly larger ( see figs . [ fig : oscz0]_a _ and [ fig : oscz0]_b _ on left ) . in the antisymmetric case , the critical frequency changes weakly on reducing of the distance @xmath180 until the loops approach a distance @xmath204 to the mid - plane ( see figs . [ fig : oscz0]_a _ and [ fig : oscz0]_b _ on right ) . further approaching of the loops results in the increase of the critical frequency which is the larger , the smaller the radius . this raise is caused by the nonuniformity of the magnetic field in the vicinity of the loop . although the critical frequency for axial oscillations is , in general , slightly lower than that for the radial oscillations , there is no significant difference between both . the maximal growth rate for the symmetric case changes weakly with reducing distance between the loops until the distance becomes comparable to the radius of the sphere ( @xmath205 ) ( see figs . [ fig : oscz0]_c _ and [ fig : oscz0]_d _ on left ) . closer approaching of the loops causes a significant increase of the growth rate for both axial and radial oscillations . again , the smaller the radius of the loops , the larger the increase of the growth rate . in the antisymmetric case , the growth rate reaches a maximum as the loops approach to the mid - plane at a distance comparable to the radius of the sphere ( @xmath206 ) ( see figs . [ fig : oscz0]_c _ and [ fig : oscz0]_d _ on right ) . spin - up instability occurs when the frequency of the alternating magnetic field exceeds the threshold @xmath179 defined above . the spin - up rate being proportional to @xmath207 becomes positive as the frequency raises over the threshold and attains a maximum at some higher frequency @xmath208 subsequently referred to as the most dangerous one . these critical frequencies together with the corresponding maximal spin - up rates for a sphere at the middle - point between symmetric and antisymmetric arrangements of two loops are plotted in fig . [ fig : rotz0 ] versus the half - distance between the loops of various radii . as seen in fig . [ fig : rotz0](_a _ ) , at large enough distances between the loops the threshold and the most dangerous frequencies tend to @xmath201 and @xmath202 in the symmetric arrangement , and to @xmath199 and @xmath200 in the antisymmetric one , which correspond to the spatially uniform and linear fields , respectively . in the symmetric case , the critical frequencies have a minimum at the distance between the loops equal to their radius : @xmath203 ( see fig . [ fig : rotz0]_a _ on left ) . note that this is converse to the critical frequencies for the oscillatory instability considered previously which have a maximum at this distance . this minimum is due to the maximal uniformity of the field in the vicinity of the middle - point achieved at this distance between the loops . critical frequencies increase with the non - uniformity of the field which becomes particularly significant when the loops are approached closer than the radius of the sphere ( @xmath209 ) . in this case , the smaller the radius of the loops , the higher the critical frequencies . the spin - up rate raises as the loops are approached and attains a maximum at @xmath209 . for the loops with large enough radius the maximum of the spin - up rate is attained when both loops merge together forming a single one ( see fig . [ fig : rotz0]_b _ on left ) . in the antisymmetric case , the critical frequencies begin to raise significantly when the loops approach to the mid - plane at a distance of the radius of the sphere : @xmath186 ( see fig . [ fig : rotz0]_a _ on right ) . as seen in fig . [ fig : rotz0]_b _ on right , the spin - up rate attains a maximum approximately at the same distance . ( 0,0)(a ) ( 0,0)(b ) this work presents an analytic stability analysis of a solid sphere levitated electromagnetically in an axisymmetric magnetic field induced by a set of coaxial circular loops . stability is considered with respect to both oscillations of small amplitude and arbitrary direction and rotations perpendicular to the axis of symmetry of the field . both oscillations and rotations are found to become growing in time when the dimensionless frequency of the field exceeds a certain critical threshold @xmath179 depending on the configuration of the field for each kind of instability . the growth rates of both instabilities become positive as the frequency is raised over the corresponding threshold , attain a maximum at some higher frequency and tend to zero at high frequencies . the frequency @xmath208 , at which the maximum is attained , is referred to as the most dangerous one for the corresponding instability . critical frequencies and the corresponding maximal growth rates are calculated for arrangements of two loops of the same radius carrying the same total current which may flow either in the same or in opposite directions . these arrangements , corresponding to the usual heating and positioning fields , are referred to as the symmetric and antisymmetric ones . it is found that critical frequencies for both oscillatory and rotational instabilities increase with the nonuniformity of the magnetic field . the lowest dimensionless critical frequencies are @xmath201 and @xmath202 which are the same for both the spin - up instability in a uniform magnetic field and the oscillatory instability in a spatially linear field . this coincidence represents a particular case of a more general rule : the critical frequencies for the spin - up instability in a spherically harmonic field of degree @xmath0 coincide with the critical frequencies for oscillatory instability in a spherically harmonic field of degree @xmath1 . thus , the critical frequencies @xmath199 and @xmath200 for the spin - up instability in a linear field coincide with the critical frequencies for the oscillatory instability in a spatially quadratic magnetic field which is the case at the mid - point of a symmetric arrangement of two loops except the case that the distance between the loops is equal to the radius of the loops . in this particular case , the quadratic term vanishes and the instability is dominated by the fourth - order term which has critical frequencies @xmath210 and @xmath211 . thus , the distance @xmath203 is optimal in order to avoid the oscillatory instability of the sphere at the mid - point of the symmetric arrangement . this distance , which ensures the most uniform magnetic field at the middle - point , yields a minimum of the critical frequency for the spin - up instability . note that there is no similar optimal distance for the oscillatory instability in antisymmetric arrangement where the critical frequency can be increased only by the nonuniformity of the magnetic field when the loops are approached close to the surface of the sphere . hence , the stabilization of a sphere by an optimal design of the inductor is rather limited and active means of stabilization , like an additional steady magnetic field , may be necessary . note that the maximal dimensionless growth rates for both oscillatory and spin - up instabilities are small , typically @xmath212 . from the physical point of view we have introduced the growth rate as a viscous - type friction coefficient at the velocity . thus the physical dimension of @xmath213 is @xmath214 and , consequently , its scale is @xmath215 because we have used the magnetic diffusion time @xmath216 and the radius of the sample @xmath4 as time and length scales , respectively . in order to obtain the actual friction force we have to take the product of velocity , friction coefficient and the magnitude of the characteristic total electromagnetic force @xmath217 which usually is comparable to the gravity of the sample @xmath218 , where @xmath219 is the mass of the sample and @xmath220 the free fall acceleration . by comparing the negative effective friction force to the inertia we obtain an estimate of the characteristic growth time of the instability @xmath221 which for a characteristic size @xmath222 and a conductivity @xmath223 of the sample leads to @xmath224 . thus , although the magnetic field has a dynamically destabilizing effect , the development of the instability is expected to be very slow and , thus , hardly observable . in conclusion , note that the frequency of the magnetic field is the only parameter determining the threshold of dynamic instabilities as long as no external damping , for instance due to a surrounding gas , is taken into account . in the opposite case , the threshold of instability would depend not only on the frequency but also on the amplitude of the current . in conclusion note that the dynamic instabilities resulting from the effect of motion of a conducting body in an ac magnetic field can be interpreted from an alternative physical point of view which can explain the weakness of this instability and also suggests other possible instability mechanisms . in the given ac magnetic field , the electromagnetic force on the spherical body at rest depends only on its position . if the position changes , it takes some time for the electromagnetic force to relax to its time - averaged stationary value for the given position . this delay is caused by the finite time of magnetic diffusion in the body which , taking place over the skin depth , is thus comparable to the ac oscillation period @xmath225 the electromagnetic force at instant @xmath226 on the body in motion may be represented as @xmath227 assuming a sufficiently high ac frequency , we can expand this representation in a power series of small @xmath228 yielding @xmath229 thus , the delay of the magnetic field results in a force component proportional to the velocity which is obviously analogous to a viscous friction force . since for a statically stable position the effective electromagnetic reaction force @xmath230 is directed against the displacement @xmath86 , the effective friction force @xmath231 is directed with the motion . consequently , the delay of the magnetic field results in an effective electromagnetic friction force with a negative and , thus , destabilizing friction coefficient . it is important to note that the friction coefficient is small because it involves the delay time which is short and comparable to the ac period when the relaxation of the magnetic field is solely due to the electromagnetic diffusion over the skin depth . the same arguments may be extended to the coupling of the magnetic field and the temperature distribution in the body caused by its temperature - dependent electrical conductivity . in this case , the delay of the magnetic field would be dominated by the thermal relaxation time which is much longer than the electromagnetic diffusion time and , thus , expected to cause stronger dynamic instabilities than the pure electromagnetic mechanism considered here . in addition , the decrease of the electrical conductivity for increasing temperature , which is caused by the joule heating of the ac magnetic field , and the related reduction of the non - dimensional frequency @xmath161 might be important in levitation experiments . consider a toroidal inductor of arbitrary cross - section carrying an axisymmetric and purely azimuthal current . the magnetic field of such an inductor may be represented as a superposition of the fields of separate circular loops constituting the inductor , and the corresponding vector potential may be written as the integral over the inductor cross - section @xmath19 : @xmath232 where @xmath233 is the nondimensionalized vector potential of a single circular loop defined by the radius vector @xmath234 with @xmath235 and @xmath137 being the radius and axial position of the loop in cylindrical coordinates . subsequently assume the skin effect to be negligible as if the inductor would be fed by a direct current driven by the gradient of an electrostatic potential @xmath236 an axisymmetric and purely azimuthal current distribution with density @xmath237 implies @xmath10 to depend solely on the azimuthal angle @xmath238 . consequently , @xmath239 where @xmath240 is a constant that can be related to the total current in the inductor @xmath241 and its effective radius @xmath242 as @xmath243 further we assume the inductor cross - section to be located about some position @xmath244 which will be specified later , and approximate the field distribution by the following multipole type expansion : @xmath245\frac{ds'd\phi'}{\left|\vec{r}-\vec{r}_{0}\right| } , \label{eq : multexp}\end{aligned}\ ] ] where the operator @xmath246 acts on @xmath247 now , if we define @xmath52 as the position of the mass center of the cross - section @xmath248 the second dipole - like term cancels in the expression above which , thus , reduces to @xmath249 the last expression defines the vector potential of a circular loop located at @xmath52 and carrying the effective current @xmath250 for example , in the case of an inductor represented by a torus with major and minor radii @xmath33 and @xmath251 the effective radius can easily be found from ( [ eq : effr ] ) as @xmath252 while the mass center of the cross - section coincides with its geometrical center . note that the solution ( [ eq : a_approx ] ) is accurate up to the quadrapole - like term neglected in ( [ eq : multexp ] ) the magnitude of which relative to the remaining term may be estimated as @xmath253 where @xmath254 is the characteristic size of the cross - section and @xmath4 is the characteristic distance to the cross - section center . obviously , the neglected term becomes significant only at distances comparable to @xmath255 similarly to ref . @xcite we also use a complex unity vector @xmath258 where @xmath36 and @xmath37 are the corresponding cartesian unit vectors , and employ the outer and inner solutions of the scalar laplace equation , @xmath259 and @xmath260 associated with the spherical harmonic @xmath261 @xcite . these variables provide a simple algebra for calculation of the gradient operator @xmath262 used throughout this study : @xmath263 where the asterisk denotes the complex conjugate and @xmath75 . the corresponding relations for the complex conjugate functions follow straightforwardly from the above relations and the property of spherical harmonics : @xmath264 another relation used for the calculation of the torque in sec . 3 is : @xmath265.\end{aligned}\ ] ] note that our definition of @xmath266 is slightly different from that used in @xcite leading to a bit more complicated algebra but simpler resulting expressions . the coefficients @xmath268 and @xmath269 in ( [ eq : f0],[eq : gamma ] , [ eq : dm ] ) can efficiently be calculated using continued fractions @xcite . for this purpose we rewrite @xmath270 where @xmath271 and @xmath272 is the spherical bessel function of index @xmath273 further , applying a recurrence relation for the spherical bessel function of index @xmath0 + 1 we obtain the continued fraction : @xmath274 allowing us to calculate @xmath275 provided that @xmath276 is known . but asymptotic properties of bessel functions suggest that for large index @xmath0 @xmath277 . thus for any @xmath136 we can choose sufficiently large @xmath278 and then truncate the fraction by approximating @xmath279 by the previous expression that allows us to calculate back the necessary @xmath280 note that the forward recurrence for @xmath275 is not practically applicable because it is numerically unstable similarly to its counterpart for bessel functions . the other necessary quantity @xmath281 contained in ( [ eq : gamma],[eq : dm ] ) , can be calculated in a similar way by expressing it as @xmath282 . for @xmath283 we have @xmath284 which follows from the corresponding asymptotics of the bessel functions @xcite . i. egry , a. diefenbach , w. dreier and j. piller , `` containerless processing in space - thermophysical property measurements using electromagnetic levitation , '' int . j. thermophys . 2 , pp . 569578 , 2001 .

the life cycles of dwarf galaxies with present - day stellar masses @xmath13 illustrate many challenges to our current understanding of galaxy formation and evolution . while we see that starbursts do not play an important role in the global star - formation at the present epoch @xcite , it is likely that the star formation histories of dwarf galaxies are complex and varied @xcite and that their typical star - formation rates were higher in the past ( e.g. * ? ? ? that star formation in low - mass galaxies may be very burst - like is predicted by hydrodynamical simulations ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in general , star formation appears to be regulated by stellar feedback in the form of supernovae and winds that heat and deplete the central reservoirs of cold gas required for continued star formation . in simulations of lower mass systems , feedback is predicted to eject gas out of the galaxy and into the halo , resulting in an episodic star formation history across the entire galaxy @xcite . repeated bursts have been cited as the driving force behind intense feedback mechanisms that can change the dynamical profile of the systems by driving baryons out of the center of the halo on short time scales , which also displaces the dark matter from the center and creates a cored dark matter profile @xcite , potentially addressing one of the principal challenges to the standard @xmath14cdm cosmology . until recently , however , the progenitors of contemporary dwarf galaxies at high - redshift could not be studied directly . `` archaeological '' studies of resolved stellar populations in local dwarfs ( e.g. , * ? ? ? * ; * ? ? ? * ) have confirmed that their star - formation histories are indeed not smooth , and have also found that a large fraction of their stars formed at early epochs ( @xmath15 ) . for such old stellar populations ( @xmath16 gyr ) , the age resolution achieved by the archaeological approach is insufficient to distinguish star formation events with durations of @xmath17 and @xmath18 years , leaving the actual `` burstiness '' of the star formation history unconstrained . look - back studies , directly observing the star formation activity in very low - mass systems at @xmath15 , were impractical until recently . but , near - infrared spectroscopy and deep imaging from the wide - field camera 3 ( wfc3 ) on the _ hubble space telescope _ ( hst ) has provided a new opportunity . an abundant population of galaxies at @xmath19 with extremely high equivalent widths ( ews ) was found by @xcite using data from the cosmic assembly near - infrared deep legacy survey ( candels , * ? ? ? * ; * ? ? ? they find a large number of objects with unusual @xmath20 and @xmath21 colors which imply the presence of a bright emission line in the @xmath22-band , likely to be [ o iii ] after considering other photometric constraints . slitless grism spectroscopy confirms the redshifts and emission - line interpretation of four of these objects . these are the high - ew tail of the emission line galaxies found in @xcite via slitless grism spectroscopy at the hst , a tail which is also probed by @xcite . other objects with similarly low masses and metallicities at these redshifts have been discovered in @xcite ( 2012b ; @xmath23,rest}$]=1499 ) and @xcite ( 2013 ; @xmath23,rest}$]=1200 ) assisted by strong gravitational lensing , in @xcite ( 2014 ; @xmath23,rest}$]=154 ) , in @xcite ( 2010 ; @xmath23,rest}$]=285 ) , and in @xcite ( 2013 ; discussed further below ) . all of these objects are emission - line dominated systems with low metallicities and high equivalent widths , the so - called `` extreme emission line galaxies '' ( eelgs ) . these systems are likely the high - ew tail of the distribution of high - redshift dwarf galaxies , and resemble the class of blue compact dwarf galaxies ( bcds , * ? ? ? * ) observed locally in several ways : low masses , high sfrs relative to their mass , and strong emission lines . however , the eelgs are indeed `` extreme , '' with ssfr values an order of magnitude higher than the bcds , similar to the strong [ o iii ] emitters ( `` green peas '' ) discovered photometrically in the sdss by @xcite , as well as spectroscopically by @xcite and @xcite . as suggested by those authors , the star - formation mode exhibited in the `` green peas '' is likely a relic of a mode that was much more prevalent in the earlier universe : their comoving number density is one to two orders of magnitude lower than the value of @xmath24 for eelgs at @xmath19 @xcite . the implication of strong emission lines and an extremely faint continuum are that these systems have low masses and are undergoing an intense burst of star formation . equivalent widths @xmath25 100 @xmath26 and stellar masses of @xmath27 imply specific star formation rates ( ssfr ) in excess of 10 gyr@xmath28 , which is more than an order of magnitude higher than star - forming systems of equivalent masses at lower redshifts @xcite . these low stellar and dynamical masses are confirmed in @xcite , who also rule - out significant contributions to the total mass from older stellar populations for objects with restframe [ o iii]@xmath29 equivalent widths @xmath30 @xmath26 and intimate that the bursts are intense and have low metallicities . the implied star formation rates and masses have only been reproduced recently in hydrodynamical simulations ( e.g. in * ? ? ? although these previous observational studies have placed constraints on various quantities , many uncertainties remain . in the case of @xcite and @xcite , all of the observed [ o iii ] , h@xmath31 , and h@xmath7 emission is attributed to star formation and not agn . in @xcite , upper - limits are placed on the black hole masses from the uv - continuum slopes , but their starbursting nature is merely plausible given the lack of knowledge about low - metallicity agn @xcite . low metallicies are simply inferred from the consistency of the sed fits using low - z ( 0.2 @xmath8 ) templates with the observed photometry . these objects represent a field of growing importance , both for studies of low - mass dwarf galaxies and for the highest - z galaxies observed . depending on the strength , duration , and initial mass of these bursts , the descendants could display a wide range of masses , from present - day @xmath32 dwarf galaxies to potentially milky way - like systems , particularly if merging is important . in any case , these galaxies are also significant contaminants in searches for higher - z dropout sources @xcite , as in the case of _ udfj-30546284 _ as mentioned in @xcite and @xcite , great care must be taken in the interpretation of high - z candidates , as eelgs can potentially reproduce their observed colors ( although the implied ews in the latter case exceeded those that can be produced from star formation alone ) . some issues still remain in our understanding of such systems , such as stringent limits to the low masses and metallicities , including those with ews @xmath33 500 , and the starburst origin of their strong emission lines . here we combine both high- and low - resolution near - ir spectroscopy with broadband photometry . with the low - resolution spectra , we select candidates for follow - up high - resolution spectroscopy and obtain emission line fluxes . the high - resolution spectra constrain various emission line ratios , some of which are useful diagnostics of agn activity , as well as line widths , which are themselves a probe of the dynamical masses of the systems . this provides strong evidence for their low masses and low metallicities , as well as confirming their starbursting nature . sophisticated modeling of the broadband seds constrains the stellar masses and ages , as well as providing information on the dust content and metallicities . together , this tells us about the strength and duration of the star - forming event . the remainder of this paper is organized as follows . in @xmath34 2 we describe the near - infrared spectroscopy and multi - band photometry used in the initial candidate selection process and the subsequent follow - up observations . @xmath34 3 presents the results of the spectroscopic study , including emission - line widths , physical sizes , and masses . in @xmath34 4 we confirm their low metallicities and rule out agn as a significant source of contamination . @xmath34 5 presents the implications of this work for the gas content of these systems , and @xmath34 6 summarizes our findings and puts the results into the overall context of the formation history of dwarf galaxies . we adopt a flat @xmath14cdm cosmology with @xmath35 and h@xmath36 km s@xmath28 mpc@xmath28 and a @xcite imf throughout . in order to search for and investigate these `` starbursting dwarf galaxies , '' we take a multi - faceted approach . our preliminary search utilizes data from the 3d - hst survey @xcite , a near - infrared spectroscopic treasury program utilizing the wfc3 . this program provides wfc3/ir primary and advanced camera for surveys ( acs ) parallel imaging and grism spectroscopy over approximately three - quarters ( 625 square arcminutes ) of the candels fields . the main source of spectroscopic data comes from the wfc3 g141 grism , with an effective wavelength coverage of 1.1 to 1.65 @xmath37 m . @xmath38 and squares represent equivalent widths from h@xmath7 . the equivalent widths were determined from photometry and grism spectroscopy , see section [ sec : sel ] . for objects in which both [ o iii ] and h@xmath7 are visible in the grism spectrum , both equivalent widths are plotted . open symbols show objects without luci1 line detections , due to intrinsic faintness or skyline contamination . our emission - line detection rate is close to 100% for objects with ew @xmath25 300 .,scaledwidth=47.5% ] the grism data allow us to select and confirm strong line emitters spectroscopically . photometric cuts , such as the @xmath39 cut of @xcite and a similar @xmath40 selection ( excess in @xmath41 compared to a blue continuum , all from candels ) , are used to preselect objects with strong features in their seds . the g141 grism data are reduced according to @xcite for the ukidss ultra - deep survey ( uds ) , goods - south ( goods - s ) , and cosmos fields , and are then used to confirm bright lines with little or no associated continuum . while we find numerous examples of these objects at @xmath15 ( maseda et al . in prep . ) , we focus here on objects where the emission lines do not fall in the wavelength range between the @xmath22- , @xmath41- , and @xmath42-bands , enhancing the chances for detectability from the ground . the low - resolution wfc3 grism data ( @xmath43 ) provide redshift information such that targets can be selected with [ o iii ] in the redshift range 1.15 @xmath44 2.40 and h@xmath7 in the redshift range 0.64 @xmath44 1.59 to @xmath45z / z @xmath12 0.005 . however , our photometric preselection relies on flux excesses such that we mostly see objects at 1.3 @xmath46 1.8 and 2.1 @xmath46 2.3 . note that we do not resolve the continuum in our ground - based observations ( discussed presently ) , so all ew values are calculated from the grism spectra directly . in total , we select 31 objects for ground - based spectroscopic observations using this method . an additional five candidates are taken from the sample of @xcite and one from @xcite . their equivalent widths as a function of redshift are shown in figure [ fig : ewz ] . non - detections are due to either intrinsic faintness in the lines ( we are sensitive to line flux in our ground - based spectra , not ew ) or contamination from oh - skylines . we observed our grism - selected sample with the luci1 multi - object spectrograph @xcite on the 8.4 m _ large binocular telescope _ ( lbt ) . we use luci1 in mos mode , splitting our 31 candidates between four masks , during april 2012 ( two masks in the cosmos field ) , october 2012 ( one mask in the uds field ) , and march 2013 ( an additional mask in the cosmos field ) . approximately two hours are spent observing in each of the @xmath22- and @xmath41-bands for the first two cosmos masks ( a and b ) using 1@xmath47 slits , two hours in each of the @xmath41- and @xmath42-bands for the uds mask ( c ) with 0.6@xmath47 slits , and three total hours on the final cosmos mask ( d ) in the @xmath22- and @xmath41-bands using 0.6@xmath47 slits . all data are taken using the high - resolution @xmath48 grating ( @xmath49 , @xmath50 , @xmath51 ) . the exposures are dithered by 3@xmath47 and are of varying durations , depending on the band : @xmath22-band data using 600s exposures , @xmath41-band data with 300s exposures , and @xmath42-band data with 120s exposures . the shorter integrations in the @xmath41- and @xmath42-bands lead to lower signal - to - noise ( s / n ) due to additional readnoise , but are necessary so as not to saturate the detector in the regions with the brightest oh sky lines . seeing was generally good ( between 0.5@xmath47 and 1@xmath47 in the optical ) during the cosmos observations , with good transparency . for the uds observations , seeing was generally better than 1@xmath47 . table [ tab : obs ] details the observations of individual objects , including ids ( from the candels catalogs , discussed in section [ sec : sed ] ) , coordinates , and the main line detections . lcccc + goods - s-7892&53.17194 & -27.75915& ... &[o iii ] , h@xmath7 + goods - s-43693&53.07129&-27.70580& ... &h@xmath7 + goods - s-43928&53.05158&-27.70476& ... &[o iii ] + uds-6195&34.42648&-5.25577& ... &[o iii ] , h@xmath7 + uds-6377&34.42857&-5.25532& ... &[o iii ] , h@xmath7 + uds-7665&34.39076 & -5.25080 & c&[o iii ] + uds-10138&34.42336 & -5.24226&c&[o iii ] + uds-12435&34.41087 & -5.23481&c&h@xmath7 + uds-12539&34.47389&-5.23423& ... &[o iii ] , h@xmath7 + uds-12920&34.39870 & -5.23320&c & ... + uds-15319&34.40516 & -5.22493&c & ... + uds-19167&34.43140 & -5.21212&c&[o iii ] + uds-24154&34.39137 & -5.19531&c&[o iii ] , h@xmath7 + cosmos-8509&150.09837 & 2.26596 & b & ... + cosmos-8700&150.09740 & 2.26848 & b & ... + cosmos-10599&150.09535 & 2.28725 & b&[o iii ] + cosmos-11530&150.11931 & 2.29688 & b & ... + cosmos-12102&150.09728 & 2.30252 & b&[o iii ] , h@xmath7 + cosmos-13184&150.12424 & 2.31367 & b&[o iii ] + cosmos-14249&150.11011 & 2.32459 & b & ... + cosmos-14435&150.16232&2.32602&d & ... + cosmos-15091&150.15955&2.33330&d&[o iii ] + cosmos-16152&150.18762 & 2.34469&a & ... + cosmos-16286&150.17699&2.34539&d&[o iii ] + cosmos-16566&150.17067&2.34830&d&[o iii ] , h@xmath7 + cosmos-17118&150.15114&2.35410&d&[o iii ] + cosmos-17162&150.15134 & 2.35482&a & ... + cosmos-17295&150.18318&2.35537&d & ... + cosmos-17539&150.12814 & 2.35810&a & ... + cosmos-17839&150.15677 & 2.36080&a&[o iii ] + cosmos-18299&150.17098 & 2.36536&a & ... + cosmos-18358&150.16719 & 2.36689&a&[o iii ] + cosmos-18582&150.13281 & 2.36878&a & ... + cosmos-18777&150.18628 & 2.37054&a & ... + cosmos-19049&150.13886 & 2.37340&a&[o iii ] , h@xmath7 + cosmos-19077&150.18309 & 2.37295&a&[o iii ] + cosmos-20589&150.18056&2.38822&d & ... we first mask regions of the spectra that are affected by persistence due to the acquisition and alignment exposures . this effect is reduced with each readout , so we only mask the regions if they are 2@xmath52 higher than the background level . we then create flat - field images from lamp - illuminated exposures and remove cosmic rays using a median - stacking technique . the most important cosmetic step is the removal of bad pixels , which are identified in the lamp - illuminated exposures , and the hot pixels in the spectra , which are identified in dark exposures . additionally , for an as - yet unknown reason , the first exposure of every series has small `` halos '' around the hot pixels . as such , we remove slightly larger regions around these hot pixels in the first exposure of every series . our wavelength calibration is done using the oh sky lines , with a code based on ` xidl ` routines ] . we also use ` xidl ` for the final sky subtraction , which uses a spline - fitting algorithm to measure and remove the lines . to maximize s / n , we do not use frame - frame subtraction and instead measure the sky from the individual frames , with the exception being some objects with particularly bad skyline contamination , where frame - frame subtraction better removes the skylines but adds noise to the spectrum . dithering the exposures by 3@xmath47 ensures a decreased dependence on the pixel - to - pixel variations in the detector . since our objects of interest have virtually no visible continuum , one - dimensional spectral extraction must be done carefully . we must visually search for the lines in the stacked reduced spectra . since we know the wavelengths of the brightest lines from the grism data , this exercise is straightforward . we isolate the line region according to a signal - to - noise cut , and then collapse that region in the wavelength direction to create a slice containing the spatial line profile . this profile is fit with a standard gaussian function . the width of the distribution , @xmath52 , and the center , @xmath37 , are then used in the full spectral extraction : a gaussian function with these same @xmath52 and @xmath37 values is fit for each pixel row in the spatial direction tracing a constant distance from the edge of the ( curved ) slit with the amplitude as the only free parameter , reflecting the electron counts at that particular wavelength . all sets of observations were reduced and analyzed separately and the resolution is incorporated into the calculation of the intrinsic line widths to remove systematics in the velocity dispersion measurements . flux calibration is done by comparing the integrated counts from the luci1 spectrum to the integrated line flux from the 3d - hst grism spectrum when possible . for an additional six sources , we obtained near - ir and visible spectra using the x - shooter spectrograph @xcite on the 8.2 m _ eso very large telescope _ ( vlt ) . we observed five of the objects initially found in @xcite : three with previous grism - spectroscopic confirmation in @xcite and weiner et al . ( in prep . ) , and the two remaining candidates with the largest photometrically inferred line fluxes . a sixth candidate was selected from the @xcite sample . observations were done in long - slit mode from august to december 2012 with 40 minute integrations using 1@xmath47/0.9@xmath47/0.9@xmath47 ( uvb / vis / nir ) slits and the 100k/1pt / hg/1@xmath532 readout mode . see table [ tab : obs ] for the targets and observing dates . the proximity of objects _ uds-6377 _ and _ uds-6195 _ allowed for them to be observed in the same slit . although the x - shooter spectrograph also observes in the uv - blue , four of our six objects were not observed during dark time , rendering the data unusable . the near - ir region of x - shooter spans the combined @xmath54 region from 1024@xmath32048 nm , while the visible region spans 559.5@xmath31024 nm . reduction of the x - shooter data is performed using version 2.0.0 of the eso xshooter pipeline , which provides merged , 2d near - ir and visible spectra . extraction is performed in a similar manner to the luci1 data . in order to confirm our hypothesis that these systems represent starbursting dwarf galaxies , we must confirm their low stellar masses , low metallicities , and high star formation rates . stellar masses can be constrained through sed fits to broadband photometry , and metallicities can be constrained by observing emission - line ratios , such as [ o iii]/h@xmath31 . any star formation rate is contingent on the nature of the emission lines , since agn can also produce very high excitations . lccccccccc goods - s-33131 & 23.66@xmath550.08 & 0.68@xmath550.62 & ... & ... & 7.36@xmath552.99 & 693@xmath5547 & 1.687 & 48.7@xmath554.3 & 52.3@xmath555.7 + goods - s-43693 & 24.36@xmath550.12 & 0.35@xmath550.06 & ... & ... & 16.9@xmath550.88 & 861@xmath5566 & 1.738 & 54.4@xmath554.5 & ... + goods - s-43928 & 24.59@xmath550.15 & 1.9@xmath550.48 & 4.5@xmath551.2 & 199 & 3.7@xmath551.6 & 176 & 1.472 & ... & 31.4@xmath558.2 + uds-6195 & 24.26@xmath550.13 & 1.4@xmath550.42 & ... & ... & ... & 701@xmath5595 & 1.687 & 69.9@xmath554.9 & 54.7@xmath556.1 + uds-6377 & 24.53@xmath550.17 & 0.67@xmath550.04 & ... & ... & ... & 731@xmath5586 & 1.664 & 54.5@xmath554.5 & 48.2@xmath555.9 + uds-7665 & 25.40@xmath550.14 & 0.51@xmath550.08 & ... & ... & 13.7@xmath552.76 & 803@xmath55162 & 2.298 & ... & 57.8@xmath559.7 + uds-10138 & 23.77@xmath550.03 & 0.75@xmath550.08 & ... & ... & 10.9@xmath550.57 & 322@xmath5517 & 2.151 & ... & 80.9@xmath5510.0 + uds-12435 & 23.42@xmath550.03 & 1.0@xmath550.06 & ... & ... & 12.2@xmath550.94 & 263@xmath5531 & 1.611 & 65.2@xmath5511.3 & ... + uds-12539 & 23.39@xmath550.06 & 1.3@xmath550.07 & ... & ... & 35.5@xmath552.73 & 713@xmath5542 & 1.621 & 81.3@xmath554.3 & 71.1@xmath555.7 + uds-19167 & 23.99@xmath550.04 & 1.1@xmath550.20 & ... & ... & 21.5@xmath552.82 & 723@xmath5595 & 2.185 & ... & 54.2@xmath559.4 + uds-24154 & 23.78@xmath550.04 & 1.8@xmath550.16 & ... & ... & 21.9@xmath553.09 & 503@xmath5534 & 2.297 & 72.5@xmath5513.1 & 61.0@xmath5510.8 + cosmos-10599 & 24.47@xmath550.26 & 0.67@xmath550.16 & ... & ... & 11.8@xmath550.89 & 714@xmath5585 & 2.220 & ... & 30.9@xmath559.0 + cosmos-12102 & 22.82@xmath550.06 & 1.6@xmath550.08 & 19.9@xmath551.0 & 360@xmath5518 & 49.4@xmath552.28 & 630@xmath5529 & 1.463 & 230.8@xmath5514.7 & 241.3@xmath5512.7 + cosmos-13184 & 23.91@xmath550.16 & 0.53@xmath550.10 & ... & ... & 11.5@xmath552.94 & 598@xmath55189 & 2.199 & ... & 40.3@xmath558.9 + cosmos-15091 & 25.46@xmath550.13 & 0.75@xmath550.11 & ... & ... & 5.99@xmath553.61 & 628@xmath55152 & 1.583 & ... & 38.2@xmath5510.0 + cosmos-16286 & 24.64@xmath550.11 & 1.1@xmath550.13 & 0.39@xmath553.32 & 41@xmath55345 & 8.76@xmath553.46 & 888@xmath55351 & 1.444 & ... & 46.7@xmath5514.4 + cosmos-16566 & 24.60@xmath550.09 & 1.4@xmath550.09 & 4.22@xmath553.26 & 560@xmath55432 & 7.28@xmath553.58 & 388@xmath55191 & 1.437 & 25.5@xmath5514.0 & 32.8@xmath558.4 + cosmos-17118 & 24.16@xmath550.13 & 4.8@xmath550.33 & ... & ... & 12.3@xmath554.22 & 493@xmath55169 & 1.656 & ... & 46.5@xmath558.8 + cosmos-17839 & 24.36@xmath550.24 & 0.99@xmath550.06 & 3.27@xmath553.20 & 325@xmath55230 & 16.3@xmath553.58 & 1126@xmath55247 & 1.412 & ... & 43.3@xmath558.9 + cosmos-18358 & 22.84@xmath550.04 & 1.6@xmath550.02 & ... & ... & 30.2@xmath550.97 & 316@xmath5526 & 1.645 & ... & 55.9@xmath559.0 + cosmos-19049 & 22.93@xmath550.05 & 2.3@xmath550.06 & 16.4@xmath551.2 & 368@xmath5528 & 20.9@xmath552.20 & 330@xmath5535 & 1.370 & 81.9@xmath5550.2 & 122.0@xmath5511.0 + cosmos-19077 & 23.69@xmath550.12 & 1.6@xmath550.05 & ... & ... & 20.5@xmath550.77 & 536@xmath5520 & 1.649 & ... & 47.7@xmath559.5 in our near - ir spectra , the most prominent lines seen are [ o iii]@xmath95007 and h@xmath7 , along with their associated complexes ( [ o iii]@xmath94959+h@xmath31 , and [ n ii]@xmath566548,6584 , respectively ) . with the exception of _ cosmos-12102 _ , the emission lines can be well - fit by a gaussian function , as described in @xcite . _ cosmos-12102 _ has the broadest emission lines of the sample . in addition to their broadness , they display some degree of asymmetry and are thus not well - fit by gaussian functions . the skewness could be caused by several processes , such as the presence of strong outflows . such interpretations are beyond the scope of this paper and demand additional observations . the best - fitting redshifts , velocity dispersions , and line ratios are given in table [ tab : lines ] . as described in @xcite , velocity dispersions of the strong emission lines can be used to estimate the dynamical masses , assuming the line width comes entirely from gravitational motion in a virialized system such that @xmath57 here , @xmath52 is the observed line width from our nir spectrum and @xmath58 is the effective radius of the galaxy from the public candels catalog released in @xcite . typical objects are 1.0@xmath550.1 kpc in both the @xmath59- and @xmath60-bands . these dynamical masses are listed in table [ tab : dp ] , ranging from @xmath61 to @xmath62 , with a median mass of @xmath63 . the uncertainty in the dynamical mass estimate comes primarily from the systematic uncertainty in the proportionality constant of 3 , which relates the intrinsic velocity @xmath64 to the observed velocity dispersion @xmath52 , which we assume to be the same factor of 33% as @xcite , since in most cases our observed line widths and physical sizes are well - constrained . further details can be found in @xcite . @xcite , in a study of `` green peas , '' observe multiple star - forming regions and gas flows which is seen as asymmetries and broad , low - intensity wings . while we do not see such clear evidence for outflows via asymmetric line profiles ( _ cosmos-12102 _ excepted ) or broad wings , we can not currently rule - out contributions of non - gravitational motions to the observed line widths since we do not resolve the continuum and can only observe the bright , central line regions . any such contributions would mean that the intrinsic dispersion is lower and that our dynamical mass estimates are upper - limits . [ sec : sed ] multi - band photometry is obtained from 3d - hst @xcite , covering 0.3 - 24 @xmath37 m for the goods - s ( 23 bands ) , uds ( 17 bands ) , and cosmos ( 31 bands ) fields . visual inspection of the spitzer - irac frames show contamination from nearby objects in _ uds-7665 _ , _ cosmos-13184 _ , and _ cosmos-15091 _ , so their seds do not include the contaminated points . for the same reason , we do not include any the data from the 5.8 and 8@xmath65 m irac channels for this sample , even though it is available as part of the publicly - released catalogs . only three objects have detections at 24 @xmath37 m with spitzer - mips ( discussed in section [ sec : starburstagn ] ) , and an upper limit of 10 @xmath37jy is adopted for the remainder of the sample . we fit the broadband spectral energy distributions of our galaxies in the same manner as @xcite using a custom version of the ` magphys ` code @xcite , which computes the emission by stellar populations and the attenuation by dust in a two - component ism , and includes nebular line emission computed self - consistently using the @xcite model ( @xcite @xcite ; c. pacifici et al in prep ) . the broad - band fluxes computed with this model include the contamination by emission lines , so they can be directly and robustly compared with the observed fluxes that we know are likely emission - line contaminated for these galaxies ( at these and higher redshifts , it has been shown that an improper treatment of nebular contamination to broadband magnitudes results in an overestimate of stellar masses and hence an underestimate in ssfrs , e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? ` magphys ` compares the input photometry to an extensive library of sed templates spanning a wide range in parameters such as star formation history , metallicity , age , and dust optical depth using a bayesian method . as such , all results quoted are the medians of the posterior probability distributions for each parameter , with uncertainties corresponding to the 16@xmath66 and 84@xmath66 percentiles for the distribution . in cases where the output probabilities are not well constrained , typically due to the ( potentially sytematic ) uncertainties in the photometry , we adopt an uncertainty of 0.3 dex in the relevant parameter : a formal error of 0 is indicative of the models not fitting the data well . 0.3 dex is the typical uncertainty in determining stellar masses from fits to broadband photometry @xcite . an example probability distribution for some of the various parameters is shown in the appendix , figure [ fig : pdf ] . results of the sed fitting are given in table [ tab : dp ] , showing high ssfrs , low stellar masses , low metallicities , young ages , and low dust extinction . since the nir sizes represent the restframe optical and hence the stellar continuum at these redshifts , we are probing the same physical region in both mass estimates . the median @xmath67 ( @xmath68-band optical depth seen by young stars in the birth clouds ) is 0.2 , consistent with the very blue observed seds . even with the lack of infrared data to directly probe the dust content of these systems , we can place a limit on the dust mass based on the total dust attenuation and luminosity inferred from the sed fits and a prior on the dust temperature as in @xcite . resulting limits are @xmath69 and hence negligible compared to the stellar masses . as mentioned before , the critical piece of additional information that we include to perform these fits is the line fluxes . we see a median sfr of @xmath129 @xmath70 which , combined with the low stellar masses , justifies our emission line criteria for the selection of starbursts . by separating the emission lines from the stellar continuum light , we are better able to trace the gas - phase metallicities . this results in metallicity estimates consistent with direct probes of the oxygen abundance using emission - line ratios ( see section [ sec : metals ] ) . in addition , it allows for better estimates of the extinction in the hii regions , which produce the aforementioned @xmath67 values . our model library of stellar population seds contains a broad range of complex star formation histories ( sfhs ) , including bursts on top of extended sfhs with a variety of evolutionary trends ( rising , falling , and constant ) . despite these efforts , which far exceed the still common use of exclusively exponentially - declining sfhs , systematic uncertainties remain . in particular for galaxies with significant star formation activity in the past @xmath71 myr , as is the case here , red supergiants with individual luminosities of @xmath72 can easily outshine more massive populations of stars with any age @xmath73 myr , especially in the near - infrared . prior knowledge of the sfh is needed to address this issue , producing a degree of circularity in the problem of stellar mass determinations . keeping this in mind , we proceed and note where necessary that for galaxies with estimated ages @xmath74 myr the mass ( and age ) estimates must be lower limits , as seen in tails to higher masses and mass - weighted ages , e.g. figure [ fig : pdf ] . extracted near - ir spectra and sed fitting results for _ goods - s-33131 _ are shown in figure [ fig : caspec ] ; all remaining objects in our sample are shown in figures 2a , 2b , and [ fig : seds ] . telluric corrections are applied as needed . and 2b for h@xmath7 ; the seds for the remaining objects can be seen in the appendix , figure [ fig : seds ] . sed fits are performed as described in section [ sec : sed ] , with red points denoting the measured photometry ( open points are upper limits ) , the blue curve denoting the non - attenuated sed , and the black curve denoting the observed sed including dust attenuation . the direct f140w image is shown on the left and the dispersed g141 grism image is shown to the right , with important spectral lines labeled and contamination subtracted as described in momcheva et al . ( in prep . ) . the x - shooter spectrum is smoothed by 3 pixels and is flux - calibrated according to the grism line flux . the shaded gray area represents the + /@xmath3 1@xmath52 flux uncertainties and the red curve shows the best- fitting model of the emission lines . ] lcccccc goods - s-33131 & 9.05@xmath550.30&8.91@xmath75&7.78@xmath76&0.321@xmath77&0.892@xmath78&0.242@xmath79 + goods - s-43693 & 8.86@xmath550.31&8.28@xmath80&7.50@xmath81&0.147@xmath82&0.557@xmath83&0.137@xmath84 + goods - s-43928 & 9.12@xmath550.38&8.83@xmath85&8.39@xmath86&0.169@xmath87&0.262@xmath88&0.327@xmath89 + uds-6195 & 9.47@xmath550.33&7.60@xmath90&6.85@xmath91&0.299@xmath92&0.447@xmath93&0.027@xmath94 + uds-6377 & 9.04@xmath550.31&8.32@xmath95&6.96@xmath96&0.081@xmath97&1.08@xmath98&1.12@xmath99 + uds-7665 & 9.07@xmath550.33 & 8.52@xmath100 & 7.04@xmath100 & 0.155@xmath100 & 1.21@xmath100 & 0.862@xmath100 + uds-10138 & 9.53@xmath550.31&8.43@xmath101&7.10@xmath102&0.177@xmath103 & 1.05@xmath104&0.112@xmath105 + uds-12435 & 9.47@xmath550.33&9.42@xmath106&8.60@xmath107&0.189@xmath108 & 0.677@xmath109&0.082@xmath110 + uds-12539 & 9.66@xmath550.30&8.67@xmath100&7.05@xmath100&0.581@xmath100&1.29@xmath100&0.667@xmath100 + uds-19167 & 9.35@xmath550.34 & 8.96@xmath100 & 8.58@xmath100 & 0.085@xmath100 & 1.02@xmath100 & 0.192@xmath100 + uds-24154 & 9.67@xmath550.33&9.10@xmath111&7.49@xmath112&0.531@xmath113 & 1.38@xmath114&0.462@xmath115 + cosmos-10599 & 8.65@xmath550.40&8.72@xmath116 & 7.49@xmath117 & 0.531@xmath118 & 1.00@xmath119&0.462@xmath120 + cosmos-12102 & 10.8@xmath550.30&9.57@xmath100 & 8.40@xmath100 & 0.155@xmath100 & 1.63@xmath100&1.77@xmath100 + cosmos-13184 & 8.78@xmath550.36&9.00@xmath121 & 7.28@xmath122 & 0.087@xmath123 & 1.29@xmath124&0.407@xmath125 + cosmos-15091 & 8.88@xmath550.37&7.92@xmath126 & 7.61@xmath127 & 0.169@xmath128 & 0.062@xmath129&0.077@xmath130 + cosmos-16286 & 9.22@xmath550.40&8.44@xmath131 & 7.92@xmath132 & 0.171@xmath133 & 0.317@xmath134&1.05@xmath135 + cosmos-16566 & 9.02@xmath550.37&8.60@xmath136 & 8.29@xmath137 & 0.137@xmath138 & 0.147@xmath139&0.037@xmath140 + cosmos-17118 & 9.86@xmath550.33&8.47@xmath141 & 7.66@xmath142 & 0.155@xmath143 & 0.752@xmath144&0.192@xmath145 + cosmos-17839 & 9.11@xmath550.34&8.19@xmath146&7.43@xmath147 & 0.091@xmath148 & 0.392@xmath149&0.037@xmath150 + cosmos-18358 & 9.54@xmath550.32 & 9.43@xmath100 & 8.12@xmath100 & 0.629@xmath100 & 1.11@xmath100&0.037@xmath100 + cosmos-19049 & 10.4@xmath550.30&9.85@xmath151 & 8.36@xmath112 & 0.081@xmath152&1.31@xmath153&1.72@xmath154 + cosmos-19077 & 9.40@xmath550.34&8.92@xmath155 & 8.05@xmath156 & 0.173@xmath157 & 0.692@xmath158&0.197@xmath159 [ fig:2a ] [ fig:2b ] samples of @xcite , @xcite , @xcite , and @xcite at @xmath160 . while variations in the stellar templates and the imf used can alter stellar mass estimates by @xmath120.3 dex , our sample still lies at considerably lower masses ( up to 2 orders of magnitude lower ) than the other samples.,scaledwidth=47.5% ] for a fixed @xmath161 of 1 kpc.,scaledwidth=47.5% ] as demonstrated in @xcite , there is a tight , linear relation between the total dynamical and stellar mass estimates . figure [ fig : dynmass ] may show that the stellar - to - dynamical mass ratio correlates with the ( mass - weighted ) age as well . this qualitatively agrees with the model in which a non - replenished reservoir of gas is turned into stars , such that the older systems have formed a proportionally larger number of stars . for further discussion , see section [ sec : discussion ] . to date , these systems represent the lowest dynamical masses ever measured at this epoch , and typically have lower stellar masses than those galaxies in similar star - forming spectroscopic surveys ( e.g. * ? ? ? * see also figure [ fig : masshist ] and section [ sec : comp ] ) . note that the systematic errors in @xmath162 dominate the size of the error bar . _ cosmos-12102 _ shows a broad and asymmetric line profile for which the line width most likely does not trace the underlying gravitational potential . the lack of a clear trend in the relationship between the two main `` observed '' quantities , @xmath163 and @xmath52 , can be seen in figure [ fig : sigmass ] , which suggests that any relationship between @xmath163 and @xmath162 must be driven by a relationship between the size of the galaxy and @xmath163 . figure [ fig : sizemass ] shows this relationship , as our sample lies on or below the observed size - mass relation for late - type galaxies at @xmath164 as found in @xcite in 3d - hst and candels . as found in @xcite . the emission line galaxies in our sample fall on or somewhat below the size - mass relation for normal star - forming galaxies at similar redshifts.,scaledwidth=47.5% ] , see the text for details ) for narrow - band selected star - forming galaxies in @xcite . the diagonal dotted line represents the nominal detection limit from the 3d - hst survey of 2.8 @xmath165 yr@xmath28 at @xmath166 . the ssfrs are averaged over 10 myr.,scaledwidth=47.5% ] the extreme nature of the star - formation in these systems is clearly seen in figure [ fig : ssfr ] . the stellar masses measured here are beginning to probe the same regime as @xcite and @xcite , with similar ssfr values . note that the ssfr values obtained from ` magphys ` are averaged over the last 10 myr : given the ephemeral nature of the bursts , the current ( s)sfr may not reflect the most vigorous period of star formation in these systems . these are more than an order of magnitude in excess of the ssfr values characteristic of the star - forming population of massive galaxies at similar redshifts , measured in h@xmath7 from @xcite . at lower redshifts 0.11 @xmath167 0.93 , @xcite have isolated a sample of eelgs selected on [ o iii ] flux that show remarkable similarities to our sample with sizes @xmath168 1.3 kpc , masses @xmath169 , ssfr values @xmath170 yr@xmath28 , and metallicity estimates of 0.05 @xmath3 0.6 @xmath8 ( some determined directly " using the [ o iii]@xmath94363 line as well , see section [ sec : metals ] ) . deep hst - acs @xmath171-band images reveal that most ( 80% ) of their eelg sample show non - axisymmetric morphologies indicative of recent mergers or interactions . only with samples that are complete over the redshift ranges in question will allow for a direct comparison between the two populations , which are currently only split artificially according to either optical or near - ir observations . however , caution must be taken in any interpretation : the much higher number density at @xmath1721.7 from @xcite than at very low - z from @xcite implies that there could be a very different mechanism involved to trigger the bursts . connecting the two populations in a qualitative sense is the subject of ongoing work . our current observations do not allow us to make strong conclusions about the internal dynamics of individual systems . currently , the only opportunities to study the internal dynamics of such small systems at high - redshift is with gravitational lensing : @xcite note that their @xmath173 strongly - lensed galaxies would resemble mergers or dispersion - dominated systems without the additional spatial resolution provided by the lensing . two objects similar to those presented here are _ macs j2135 - 0102 _ ( @xmath163=9.8 @xmath165 , z=3 , @xmath174=1.75 kpc , sfr=40 @xmath165 yr@xmath28 , @xmath175=54@xmath176 , @xmath177=67@xmath176 ) from @xcite and _ shizels-10 _ ( @xmath163=9.4 @xmath165 , z=1.45 , @xmath174=2.3 kpc , sfr=10 @xmath165 yr@xmath28 , @xmath175=64@xmath176 , @xmath177=26@xmath176 ) from @xcite , which both appear to have a ( relatively weak ) rotational component to their dynamics . further discussion of the dynamics of our present sample is deferred to section [ sec : discussion ] . several other studies have begun to characterize the starforming properties of high - z eelgs using various techniques . the most obvious comparable study to this work is the wfc3 infrared spectroscopic parallels survey ( wisp , * ? ? ? * ) , and specifically the study of @xcite . as a similar wfc3 grism survey , they are also able to select galaxies based on emission lines instead of photometric techniques , and thus can isolate a sample of high - ew elgs . @xcite present a sample of 26 such elgs with a median restframe [ o iii ] ew of 154 @xmath26 ( our median is 629 ) . these galaxies show similarly - low velocity dispersions ( @xmath12 70 @xmath176 ) and hence also have dynamical masses @xmath178 10@xmath179 . they derive stellar masses using an assumed m / l ratio and star - formation history in a similar fashion to @xcite , which have been shown to generally agree with sed - derived stellar masses @xcite , and have also begun to probe the @xmath180 regime . specific discussion of their metallicity estimates is deferred to section [ sec : metals ] . in addition to wisp , narrowband studies have also begun to uncover the general starforming population of galaxies at @xmath15 , probing stellar masses below 10@xmath181 . @xcite use data from the hizels survey to study h@xmath7 emitters at redshifts @xmath182 , and @xmath183 . the size of their survey area ( @xmath12 2 deg@xmath184 ) and the depth of the imaging allows them to isolate large and pure samples of h@xmath7 emitters down to a restframe h@xmath7+[nii ] ew of 25 , constraining the stellar mass function of star - forming galaxies down to 10@xmath181 at these redshifts . indeed , their sample also includes a number of objects with restframe h@xmath7+[nii ] ew values in excess of 300 @xmath26 and with stellar masses from sed - fitting below 10@xmath185 . their results for the characteristic ssfr ( i.e. the typical sfr for a galaxy at a given mass and redshift divided by its mass ) as a function of mass and redshift is shown in figure [ fig : ssfr ] . while some of our objects can be considered normal " at these redshifts according to this determination , a majority of them still have higher ssfrs than expected , albeit typically within 1 dex . this reinforces the notion that these objects are the high - ew tail of the total distribution of star - forming galaxies at these redshifts and do not comprise a genuinely separate population @xcite . so far , the main caveat is that we have assumed that star formation is primarily responsible for the strong line emission . the most compelling evidence to support this assumption is the relation between stellar mass and dynamical mass , and that the dynamical masses , with two exceptions , do not exceed @xmath186 @xcite . such a result would be entirely coincidental in the case that the emission lines are powered by agn since the width of agn emission lines is not coupled to the stellar mass of the host galaxy . in other words , we observe narrow emission lines in these small ( @xmath12 1 kpc ) systems , while typical agn narrow line regions have much larget emission - line widths @xmath187 @xcite . however , it is useful to look for evidence of agn contributions . although low - metallicity , low - mass agn are exceedingly rare in the local universe @xcite , there is some evidence that they may be more common at higher - z @xcite . such agn could cause large observed line fluxes . at @xmath188 , agn identification with a simple diagnostic such a high [ o iii]@xmath95007/h@xmath31 ratio becomes insufficient by itself , as shown by @xcite . we thus utilize the `` bpt '' diagnostics of @xcite for the objects in our sample with measurements of [ n ii ] and/or [ s ii ] in addition to h@xmath7 . given that these lines are typically quite weak in star forming galaxies and the strong influence of oh skylines in our nir spectra , in some instances we can only place an upper - limit on the ratios of those lines with h@xmath7 . these two bpt diagrams are shown in the left and central panels of figure [ fig : mex ] with contours showing galaxies from the sdss mpa - jhu value - added dr7 catalog . most of our points plausibly lie on an extension of the starforming region of the bpt diagrams and not on the extension of the agn region , or at least they lie far from the main locus of agn - powered emission lines . however , low - metallicity agn can lie in the starforming region as well @xcite , preventing us from completely ruling out the contribution of agn to our sample with these diagnostics . in some cases we observe higher [ o iii]/h@xmath31 ratios compared to the other starforming galaxies , but this can be explained simply by higher ionizations and lower metallicities in these systems compared to the low - z sample of sdss galaxies . @xcite find that starforming galaxies at @xmath15 are consistent with models that have more extreme ism conditions than those in the local universe . high electron temperatures ( discussed in the next section ) support such a hypothesis . this would be a further , unexplained coincidence if they are agn in addition to the low dynamical masses described previously . while @xcite suggest that agn are widespread in low - mass @xmath189 galaxies , the emission lines are not actually dominated by the agn . this is most evident in the @xmath190/@xmath191}$ ] relationship : the [ o iii ] lines have some agn contribution , but ( on average ) less than 50% . additionally , we utilize the mass - excitation ( mex ) diagnostic of @xcite , which combines the [ o iii]@xmath95007/h@xmath31 ratio with the stellar mass . @xcite conclude that this diagnostic also gives a meaningful probabilistic constraint on the agn / sf nature of galaxies at @xmath15 using a combination of the bpt diagnostics and x - ray data . while this diagnostic is easily applicable to our data , we can not constrain the mex agn / starforming _ probabilities _ for most of our sample given that it is not properly calibrated for objects with such low metallicities and high ssfrs , agn or otherwise : the five objects with a probability of star formation ( @xmath192 , compared to the probability that they are agn ; _ cosmos-18358 _ , _ cosmos-13184 _ , _ cosmos-10599 _ , _ uds-24154 _ , and _ uds-12435 _ ) have a median @xmath192 value of 0.940 , firmly placing them in the starforming regime . the remaining objects , while probabilistically unconstrained , still lie far from the population of agn in the mex diagnostic plot , as shown in figure [ fig : mex ] . we can place other constraints on the agn nature of these systems in much the same way as @xcite , i.e. independent of any measured emission line ratio(s ) . most objects in our sample do not have strong 24 @xmath37 m detections using spitzer - mips : _ cosmos-18358 _ has a mips 24@xmath37 m flux of 21.5@xmath559.0 @xmath37jy and clearly appears to be a merger ; _ cosmos-19049 _ has a flux of 14.1@xmath558.6 @xmath37jy , is physically large with @xmath193 kpc , and also has broad lines with @xmath194}=122 { \>{\rm km}\,{\rm s}^{-1}}$ ] ; _ cosmos-12102 _ has a flux of 55.3@xmath558.7 @xmath37jy and is further discussed below . while none of our objects have x - ray detections , goods - s is also the only field with sufficient depth to find @xmath195 agn which are not quasars . that being said , it is possible to hide even a rapidly accreting x - ray agn in a low mass galaxy @xcite . the consistent ( and resolved ) @xmath59- and @xmath60-sizes , as well as the sizes of the emission lines in the grism spectra , rule out the presence of a strong point source dominating the emission . an exception could be _ cosmos-12102 _ , illustrated by the star in figure [ fig : mex ] , which is also has the largest line width in our sample . @xcite find active black holes in similar mass dwarf galaxies with broad h@xmath7 emission in the local universe with @xmath196 . using their relation ( equation 5 ) of @xmath197 and fwhm@xmath198 to @xmath199 , _ cosmos-12102 _ has a black hole mass of @xmath200 , which is in their observed range . the mass determination is fraught with systematic uncertainties , such as variations in the geometry of the broad - line region and that at least some of the h@xmath7 luminosity comes from star formation , but we can not conclusively rule - out some agn contribution for this galaxy . @xmath96584/h@xmath7 , the bpt2 diagnostic of [ s ii]@xmath96718 + 6731/h@xmath7 , and the mex diagnostic @xcite of stellar mass . divisions between the agn and the star - forming regions in the bpt diagrams come from @xcite . in all cases , the gray contours represent data from the sdss mpa - jhu value - added dr7 catalog : these emission line and stellar mass measurements are described by @xcite and @xcite . arrows denote 3@xmath52 upper limits . each of the diagnostics point to star formation as the ionizing source with at most mild agn contribution , the possible exception being _ cosmos-12102 _ ( star symbol ) . the large uncertainties in [ o iii]/h@xmath31 are caused by very low and uncertain h@xmath31 fluxes , as [ o iii ] is robustly detected in all of these cases ; the true ratio could be even higher than the values plotted here . ] in order to measure the gas - phase oxygen abundances of these galaxies as a proxy for the metallicity , we first implement the so - called `` direct '' or @xmath201 method which requires a detection of the auroral [ o iii]@xmath94363 line as well as the [ o iii]@xmath564959 , 5007 and [ o ii]@xmath93727 lines . some of these lines lie in the uv / visible at these redshifts , so we can only apply this method to the x - shooter sample . using the calibrations outlined in @xcite , we convert the [ o iii ] emission - line ratios into an electron temperature ( with the electron density constrained by the ratio of [ o ii]@xmath93729 to @xmath93726 or [ s ii]@xmath96717 to @xmath96731 ) in the o@xmath202 region . the total oxygen abundance in the galaxy is o / h = o@xmath203/h@xmath203 + o@xmath202/h@xmath203 , assuming @xmath204o@xmath205o@xmath206 . two objects , _ uds-12539 _ and _ uds-6377 _ , have detections of @xmath94363 and an upper limit can be obtained for a third , _ goods - s-43928_. this object also lacks a detection of [ o ii]@xmath2073727 due to skyline contamination , so the o@xmath203/h@xmath203 component can not be constrained directly . the relative contribution of the o@xmath203/h@xmath203 component to the total oxygen abundance is 1.8% and 7.3% for the other two objects , so we neglect its contribution to the abundance of the third object . the n2 method of ( * ? ? ? * pp04 ) , which uses the ( log ) ratio [ n ii]@xmath96584/h@xmath7 , verifies this result . all derived temperatures are in excess of 20,000 k. lcc goods - s-43693&8.10@xmath550.34&n2 + goods - s-43928&7.70@xmath550.53&@xmath201/n2 + uds-6195&8.11@xmath550.23&o3n2 + uds-6377&7.52@xmath550.37&@xmath201 + uds-12539&7.45@xmath550.09&@xmath201 + uds-24154&8.25@xmath550.32&o3n2 + cosmos-19049&8.20@xmath550.20&n2 for the remainder of the sample , we must resort to other methods , namely the aforementioned n2 method and also the o3n2 method , both from pp04 ; o3n2 is the ( log ) ratio [ o iii]@xmath564959 , 5007/h@xmath31/n2 . for both methods , uncertainties include contributions from the error in the line ratios as well as the systematic scatter in the relations . if h@xmath31 is not detected at more than 1@xmath52 in our luci1 or x - shooter spectrum , we estimate the [ o iii]/h@xmath31 ratio from the grism spectrum . also note that in the two instances where the n2 method is used the @xcite calibration results in metallicities consistent with the pp04 values . results are displayed in table [ tab : z ] , with a median @xmath208 value of 7.90 ( 0.15 @xmath8 ) . this value agrees well with the median ` magphys`-derived metallicity of 0.17 @xmath8 for the full sample . given the relatively small number of direct measurements of the oxygen abundance in high - z galaxies , this provides an important piece of information . since the standard r23 diagnostic @xcite using [ o iii]@xmath56 4959 , 5007 + [ o ii ] @xmath56 3726 , 3729 + h@xmath31 is double - valued , it is important whether higher - z galaxies belong to the metal - rich upper branch or to the metal - poor lower branch . @xcite argue in favor of the upper - branch for galaxies with log ( @xmath163/@xmath165 ) @xmath25 8.2 at @xmath209 , and @xcite favor the upper - branch for galaxies with log ( @xmath163/@xmath165 ) @xmath25 8.8 at @xmath210 . our measurements suggest that these systems plausibly lie on the lower branch , even at log ( @xmath163/@xmath165 ) @xmath33 9.2 . in the left panel of figure [ fig : metallicity ] we show these results as well as those of @xcite and @xcite with the low - mass extrapolations to the mass - metallicity ( mz ) relations from @xcite for @xmath211 , @xcite for emission line galaxies at @xmath212 , and @xcite for starforming galaxies at @xmath213 utilizing the metallicity calibrations and functional parameterization of @xcite , as well as the results from @xcite . the right panel of figure [ fig : metallicity ] shows the same data plotted against the low - mass extrapolation of the `` fundamental metallicity relation '' ( fmr ) of @xcite and the relation for high - sfr galaxies at @xmath214 from @xcite . neither method is directly calibrated below @xmath32 , which is precisely the range we probe here . in the case of the mz relation , our results generally agree with the @xcite and @xcite extrapolations and typically have lower metallicities than objects with similar masses and star formation rates at lower redshifts . we see good agreement with distribution in @xmath208 with @xmath163 from @xcite as well . the slight overall trend observed where higher - mass objects have higher metallicities supports the idea that the lowest - mass objects are the youngest and therefore have formed less mass of heavy elements . the youngest objects should also have the highest ssfrs , which is known to drive much of the scatter in the mz relation . @xcite also explain that the steepness of the o / h versus @xmath163 relation has a direct theoretical impact on the role of outflows in these systems . in the fmr relation the solid points , denoting @xmath201-determined abundances , are considered more reliable at low metallicities @xcite and lie somewhat below the fmr , perhaps more in line with an extrapolation to the @xmath214 findings of @xcite . since @xcite do not see the higher - mass turnoff in the relation until @xmath215 , we may be observing the evolution of the fmr on the low - mass and/or high - sfr end , although we do not have the number statistics yet to quantify any offset . at the very least , we can conclusively rule - out that these objects lie above the fmr ( cf . * * who claim this relation is driven by the higher average sfrs of the systems probed at higher redshifts ) . in this section we use the observed velocity dispersions @xmath52 and sizes @xmath216 , combined with dynamical stability criteria , to constrain the gas fraction and its implications . we assume that the systems consist entirely of stars and gas : we neglect the contribution of dark matter to the total dynamical mass as measured within the central kpc . we also assume that these systems are isolated and not embedded in larger ( gaseous ) structures that exert pressure . we do not know the geometry of the systems , and therefore consider two extreme cases : for the case of a sphere with uniform density we calculate the jeans mass @xmath217 ; for the case of a thin rotating disk we calculate the toomre parameter @xmath218 . in both cases we assume that the gaseous body has the same extent as the stellar body , and that the velocity width of the nebular lines traces the total gas kinematics . for a uniform sphere the jeans mass @xcite is given by @xmath219 where we have equated the sound speed with observed velocity dispersion . this velocity represents the combined effect of all sources of pressure that act against collapse , which include thermal motions ( associated with the physical sound speed ) as well as turbulence and streaming motions . @xmath220 is the density of the gas , which is given by the gas mass ( @xmath221 ) and the size : @xmath222 where @xmath223 is the gas fraction . the total dynamical mass @xmath162 is taken from equation [ eqn : dyn ] , with a value of 5 for the proportionality constant for consistency with the case under consideration here : that of a sphere with uniform density . for typical values of @xmath58 ( 1 kpc ) and @xmath52 ( 50 @xmath176 ) we find that @xmath224 for @xmath225 . given that substantial star formation in these systems the gaseous body must be unstable : we conclude , assuming a homogeneous gaseous sphere , that @xmath226 . in order to address the question to what extent this conclusion is affected by the chosen geometry , we now consider the other ( opposite ) case , and assume that these systems are rotating disks , where instability can be described by the toomre parameter @xmath218 @xcite @xmath227 where @xmath228 is the velocity dispersion perpendicular to the disk , and @xmath229 is the average gas - mass surface density , given by @xmath230 where @xmath223 is the gas fraction . the epicyclic frequency , @xmath231 , in a rotating exponential disk is @xmath232 where @xmath233 is the circular velocity of the disk . while @xmath228 and @xmath233 are not observed directly , both contribute to the observed linewidth , @xmath52 . the contribution of rotation to @xmath52 , assuming an average inclination of 60 degrees and an exponential disk , is @xmath234 , such that we have @xmath235 which is empirically supported @xcite . combining the above , we solve for @xmath233 as a function of @xmath236 : @xmath237 such that an unstable system ( @xmath238 ) requires @xmath239 in order to produce a unique , physical solution . remarkably , regardless of whether we assume a homogeneous sphere or a rotating disk for the gas , we find that a high gas fraction is needed to explain the observed star formation activity . at the same time , the non - negligible contribution of the stellar mass to the total mass implies that @xmath223 can not be arbitrarily close to unity and should be @xmath240 ( see figure [ fig : dynmass ] ) . a significant caveat is that the proportionality constants in equations [ eqn : jeans ] , [ eqn : rho ] , [ eqn : dyn ] , [ eqn : kappa ] , [ eqn : sigmaobs ] , and [ eqn : vt ] depend on the details of the assumed geometry and dynamical structure ; their variation can alter the threshold value of @xmath223 . in addition , we ignore the stabilizing effect of the stellar disk on the gas disk , however this increases the implied gas fraction further . the median gas fraction in our sample is 72% ( i.e. the y - axis in figure [ fig : dynmass ] is a proxy for @xmath223 ) , in agreement with the theoretical calculation . while some objects are observed to have lower gas fractions , 18 out of our 22 objects are consistent with @xmath241 2/3 within 1@xmath52 . in the following we assume @xmath223=2/3 in order to constrain the star formation efficiency . in figure [ fig : gas ] we show the star formation rate surface density , assuming @xmath242 , versus the gas surface mass density @xmath229 ( from equation [ eqn : sigma ] ) . the implied gas depletion time scale ranges from @xmath243 to @xmath244 yr , with a median of @xmath245 yr . compared to normal present - day galaxies , gas is efficiently transformed into stars but , with the exception of a few objects , not as efficiently as observed in starbursting regions in the milky way and starbursting galaxies in the present - day or high - redshift universe ( i.e. * ? ? ? * ; * ? ? ? this mostly reflects the large gas fractions needed to produce systems that are unstable against star formation , rather than a modest level of star formation : the inverse ssfr ( stellar mass growth time scale ) of @xmath246 yr is very short , among the fastest ever measured . hence , the stellar mass grows at a dynamical time scale ( @xmath247 yr ) . the physical interpretation is that the star formation rate is not limited by availability of fuel but by the dynamical time scale . however , we have to keep in mind that we selected galaxies based on their ssfr : we are biased against objects that are older and/or have longer star formation time scales . further empirical investigation of lower levels of star formation in similarly massive galaxies is needed to address this issue . however , we propose that star formation will halt within @xmath248 myr , long before the gas reservoir is depleted . first , the stability arguments given above imply that the gas fraction only needs to be reduced by a small amount ( from the assumed @xmath249 to , say , @xmath250 ) . given the observed sfr , this takes @xmath1250 myr . second , feedback should play an important role in these low - mass systems with small escape velocities ( several 100 @xmath176 ) and high sfr : gas is easily transported out of the galaxy , and perhaps even out of the halo , preventing recycling . this paradigm is supported by @xcite , who see that higher mass starforming galaxies at z @xmath12 2 can support more extended rotationally - supported disks and are less efficient at driving outflows than their lower mass counterparts . in the above we have ignored the infall of cold gas , which could continue to feed and maintain the starburst . assuming that these galaxies reside in relatively low - mass halos ( @xmath251 ) the typical accretion rate onto the halo is several @xmath165/yr @xcite , somewhat lower than the sfr . the accretion rate onto the halo is not necessarily equal to the accretion rate onto the galaxy , and the latter is likely not constant . a period of above - average accretion for several 100 myr could , in principle , sustain the starburst . however , above - average accretion events are more likely to be of short duration , such that one such event can ignite the observed starburst by pushing the gas mass surface density above the threshold for star formation or by disturbing the already - present gas , creating an instability . hence , enhanced accretion could cause the star formation activity but not maintain it . based on the jeans and toomre instability arguments presented above , purely gaseous systems with velocity dispersions and sizes as observed become unstable once they reach a total mass of a few times @xmath252 , close to the observed masses of the objects in our sample . , which is inferred as a probable gas fraction given the star formation rates and @xmath163/@xmath162 ratios ( which imply @xmath253 ) in these systems . the solid black line shows the relation for local spiral galaxies from @xcite and the dotted line shows the result for local ( u)lirg and high - z smgs / qsos from @xcite . the location of the points suggests that these objects spend at least some of the time forming stars more efficiently than the normal , present - day spiral galaxies . our constraints are not stringent enough to confirm or rule out gas depletion timescales that are on par with or even shorter than more massive , starbursting systems.,scaledwidth=47.5% ] we presented near - infrared spectroscopy from the lbt / luci1 multi - object near - ir spectrograph and the vlt / x - shooter wide band spectrograph for a sample of hst / wfc3 grism - selected emission line objects with restframe equivalent widths of @xmath254 @xmath26 for [ o iii]@xmath95007 and/or h@xmath7 , and located in the redshift range @xmath0 . the observed emission lines are narrow , with measured velocity dispersions down to @xmath255 km s@xmath28 , implying low dynamical masses of @xmath256 , even for the lower - ew objects not included in @xcite . stellar masses determined using sophisticated ` magphys ` sed fitting to broadband magnitudes and the inclusion of line fluxes results in low stellar masses as well , @xmath257 . ratios of @xmath163 to @xmath162 range from 1/10 to 1 , which makes agn - dominated seds unlikely . emission - line ratios and the narrow line widths also suggest that agn do not significantly contribute to our sample , and therefore we conclude that the main ionizing source is hot , massive stars . direct probes of the oxygen abundances within these galaxies and [ o iii]/h@xmath31 line ratios of typically @xmath258 5 corroborate the expectation that these low mass systems have low metallicities , between 0.05 and 0.3 @xmath8 . they lie on or below the ( extrapolated ) mass - metallicity relationships for these redshifts @xcite which , combined with their young sed - derived ages , reinforces the notion that these are nascent galaxies undergoing their first major episode of star formation . measured ssfr values of @xmath259 yr@xmath28 for these galaxies are up to two orders of magnitude larger than those of typical @xmath186 starforming galaxies at @xmath260 @xcite , as well as comparable to or greater than the values from other high - ew systems as discovered in deep narrowband searches @xcite and in deep spectroscopic studies at both similar @xcite and lower redshifts @xcite . such high ssfr values have been difficult to reproduce in hydrodynamical simulations , but recently @xcite made significant progress by combining a high gas density threshold for star formation and a blastwave scheme for supernova feedback in their simulations of low - mass galaxies . such low mass systems , with observed velocity dispersions of @xmath261 and sizes of @xmath12 1 kpc are only unstable against star formation if their gas fractions are high ( above 2/3 ) , in agreement with the observed @xmath262 relation . the bursts are likely to be short - lived ( @xmath1250 myr ) , as , even in the absence of feedback , their intense star formation will rapidly build up stellar mass and lower their ssfr well before the gas depletion timescale ( @xmath12100 myr ) . these results strengthen the conclusions from @xcite , who argued that eelgs represent low - mass , starbursting galaxies . additionally , the existence of ( at least ) two strong galaxy - galaxy lenses in the candels/3d - hst fields where the background galaxy is an eelg at @xmath263 and @xmath264 ( * ? ? ? * ; * ? ? ? * respectively ) suggests that this type of object is common . the ubiquity of eelgs may be even more pronounced at high redshifts ( @xmath256 ; * ? ? ? such systems at @xmath265 1 @xmath3 2 thus may present an opportunity to study how star formation proceeded in the early universe before the advent of the next generation of observatories , such as the _ james webb space telescope_. the new generation of submillimeter observatories , such as alma , can provide direct estimates of the gas masses . searching for the presence of outflowing material would provide valuable clues about the feedback processes going on in these systems , which is especially relevant to support the hypothesis that these bursts can create the cored dark matter profiles observed in local dwarf galaxies ( e.g. * ? ? ? aird , j. , coil , a. l. , moustakas , j. , et al . 2012 , apj , 746 , 90 amorn , r. o. , perz - montero , e. , & vlchez , j. m. , 2010 , apj , 715 , l128 amorn , r. o. , vlchez , j. m. , hgele , g. f. , et al . 2012 , apj , 754 , l22 amorn , r. o. , prez - montero , e. , contini , t. , et al . 2014a , arxiv:1403.3441 amorn , r. o. , sommariva , v. , castellano , m. , et al . 2014b , arxiv:1403.3692 amorisco , n. c. , zavala , j. , & de boer , t. j. l. , 2014 , apj , 782 , l39 atek , h. , malkan , m. , mccarthy , p. , et al . 2010 , apj , 723 , 104 atek , h. , siana , b. , scarlata , c. , et al . 2011 , apj , 743 , 121 baldwin , j. , phillips , m. , & terlevich , r. , 1981 , pasp , 93 , 5 binney , j. , & tremaine , s. 2008 , galactic dynamics ( 2nd ed . ; princeton , nj : princeton univ . press ) bouwens , r. j. , oesch , p. a. , illingworth , g. d. , et al . 2013 , apj , 765 , l16 brammer , g. b. , van dokkum , p. , franx , m. , et al . 2012a , apjs , 200 , 13 brammer , g. b. , snchez - janssen , r. , labb , i. , et al . 2012b , apj , 758 , l17 brammer , g. b. , van dokkum , p. g. , illingworth , g. d. , et al . 2013 , apj , 765 , l2 cardamone , c. , schawinski , k. , sarzi , m. , et al . 2009 , mnras , 399 , 1191 chabrier , g. 2003 , , 115 , 763 charlot , s. , & longhetti , m. 2001 , mnras , 323 , 887 coe , d. , zitrin , a. , carrasco , m. , et al . 2013 , apj , 762 , 32 conroy , c , 2013 , ara&a , 51 , 393 contini , t. , garilli , b. , le fevre , o. , et al . 2012 , a&a , 539a , 91 curtis - lake , e. , mclure , r. j. , dunlop , j. s. , et al . 2013 , mnras , 429 , 302 da cunha , e. , charlot , s. , & elbaz , d. , 2008 , mnras , 388 , 1595 da cunha , e. , walter , f. , decarli , r. , et al . 2013 , apj , 765 , 9 daddi , e. , elbaz , d. , walter , f. , et al . 2010 , apj , 714 , l118 ellis , r. s. , mclure , r. j. , dunlop , j. s. , et al . 2013 , apj , 763 , l7 erb , d. k. , shapley , a. e. , pettini , m. , et al . 2006 , apj , 644 , 813 erb , d. k. , pettini , m. , shapley , a. e. , et al . 2010 , apj , 719 , 1168 frster - schreiber , n. m. , genzel , r. , bouch , n. , et al . 2009 , apj , 706 , 1364 fumagalli , m. , patel , s. g. , franx , m. , et al . 2012 , apj , 757 , l22 gallagher , j. s. , hunter , d. a. , & tutukov , a. v. 1984 , apj , 284 , 544 governato , f. , zolotov , a. , pontzen , a. , et al . 2012 , mnras , 422 , 1231 grebel , e. , 1997 , reviews in modern astronomy , 10 , 29 grogin , n. , kocevski , d. , faber , s. , et al . 2011 , apjs , 197 , 35 groves , b. a. , heckman , t. m. , & kauffmann , g. 2006 , mnras , 371 , 1559 henry , a. , martin , c. l. , finlator , k. , & dressler , a. 2013a , apj , 769 , 148 henry , a. , scarlata , c. , domnguez , a. , et al . 2013b , apj , 776 , l27 izotov , y. i. , stasiska , g. , meynet , g. , et al . 2006 , a&a , 448 , 955 izotov , y. i. & thuan , t. x. 2008 , apj , 687 , 133 izotov , y. i. , guseva , n. g. , & thuan , t. x. 2011 , apj , 728 , 161 jones , t. , swinbank , a. m. , ellis , r. s. , richard , j. , & stark , d. p. 2010 , mnras , 404 , 1247 juneau , s. , dickinson , m. , alexander , d. m. , & salim , s. , 2011 , apj , 736 , 104 karim , a. , schinnerer , e. , martnez - sansigre , a. , et al . 2011 , , 730 , 61 kauffmann , g. , heckman , t. m. , white , s. d. m. , et al . 2003 , mnras , 341 , 33 kassin , s. a. , weiner , b. j. , faber , s. m. , et al . 2007 , apj , 660 , l35 kennicutt , r. c. , jr . , calzetti , d. , walter , f. , et al . 2007 , apj , 671 , 333 kewley , l. j. , groves , b. , kauffmann , g. , & heckman , t. 2006 , mnras , 372 , 961 kewley , l. j. , maier , c. , yabe , k. , et al . 2013 , apj , 774 , l10 kewley , l. j. & ellison , s. l. 2008 , apj , 681 , 1183 koekemoer , a. , faber , s. , ferguson , h. , et al . 2011 , apjs , 197 , 36 law , d. r. , steidel , c. c. , erb , d. k. , et al . 2009 , apj , 697 , 2057 law , d. r. , steidel , c. c. , shapley , a. e. , et al . 2012 , apj , 756 , 29 lee , j. c. , kennicutt , r. c. , jos g. funes , s. j. , sakai , s. , & akiyama , s. 2009 , , 692 , 1305 maseda , m. v. , van der wel , a. , da cunha , e. , et al . 2013 , apj , 778 , l22 maiolino , r. , nagao , t. , grazian , a. , et al . 2008 , a&a , 488 , 463 mancini , c. , frster - schreiber , n. m. , renzini , a. , et al . 2011 , apj , 743 , 86 mannucci , f. , cresci , g. , maiolino , r. , et al . 2010 , mnras , 408 , 2115 masters , d. , mccarthy , p. , siana , b. , et al . 2014 , apj , in press mateo , m. l. 1998 , , 36 , 435 mcbride , j. , fakhouri , o. , & ma , c .- p . 2009 , mnras , 398 , 1858 navarro , j. , eke , v. , & frenk , c. , 1996 , mnras , 283 , 72 osterbrock , d. e. & mathews , w. g. 1986 , ara&a , 24 , 171 pacifici , c. , charlot , s. , blaizot , j. , & brinchmann , j. 2012 , mnras , 421 , 2002 pelupessy , f. i. , van der werf , p. p. , & icke , v. 2004 , , 422 , 55 pagel , b. e. j. , edmunds , m. g. , blackwell , d. e. , chun , m. s. , & smith , g. 1979 , mnras , 189 , 95 pettini , m. , & pagel , b. e. j. 2004 , mnras , 384 , l59 pontzen , a. , & governato , f. 2012 , mnras , 421 , 3464 reines , a. e. , greene , j. e. , & geha , m. 2013 , apj , 775 , 116 rix , h .- w . , guhathakurta , p. , colless , m. , & ing , k. 1997 , mnras , 285 , 779 sargent , w. l. w. & searle , l. 1970 , , 162 , l155 seifert , w. , appenzeller , i. , baumeister , h. , et al . 2003 , proc . spie , 4841 , 962 shen , s. , madau , p. , conroy , c. , governato , f. , & mayer , l. 2013 , arxiv:1308.4131 skelton , r. e. , whitaker , k. e. , momcheva , i. g. , et al . 2014 , apjs submitted smit , r. , bouwens , r. j. , labb , i. , et al . 2014 , apj , 784 , 58 sobral , d. , best , p. n. , smail , i. , et al . 2014 , mnras , 437 , 3516 stark , d. p. , schenker , m. a. , ellis , r. s. , et al . 2013 , apj , 763 , 129 stinson , g. s. , dalcanton , j. j. , quinn , t. , et al . 2007 , apj , 667 , 170 stott , j. p. , sobral , d. , bower , r. , et al . 2013 , mnras , 436 , 1130 straughn , a. n. , meurer , g. , pirzkal , n. , et al . 2008 , aj , 135 , 1624 straughn , a. n. , kuntschner , h. , kmmel , m. , et al . 2011 , aj , 141 , 14 swinbank , a. m. , sobral , d. , smail , i. , et al . 2012 , mnras , 426 , 935 toomre , a. 1964 , apj , 139 , 1217 tremonti , c. a. , heckman , t. m. , kauffmann , g. , et al . 2004 , apj , 613 , 898 trump , j. r. , weiner , b. j. , scarlata , c. , et al . 2011 , apj , 743 , 144 trump , j. r. , konidaris , n. p. , barro , g. , et al . 2013 , apj , 763 , l6 van dokkum , p. , brammer , g. b. , fumagalli , m. , et al . 2011 , apj , 743 , l15 van der wel , a. , straughn , a. , rix , h .- w . , 2011 , apj , 742 , 111 van der wel , a. , bell , e. , hussler , b. , et al . 2012 , apjs 203 , 24 van der wel , a. , van de ven , g. , maseda , m. , et al . 2013 , apj , 777 , l17 van der wel , a. , franx , m. , van dokkum , p. g. , et al . 2014 , apj , 788 , 28 vernet , j. , dekker , h. , dodorico , s. , et al . 2011 , a&a , 536a , 105 weisz , d. , dalcanton , j. , williams , b. , et al . 2011 , apj , 739 , 5 xue , y. q. , wang , s. x. , brandt , w. n. , et al . 2012 , apj , 758 , 129 zahid , h. j. , kashino , d. , silverman , j. d. , et al . 2013 , arxiv:1310.4950 zolotov , a. , brooks , a. m. , willman , b. , et al . 2012 , apj , 761 , 71

the discovery @xcite of afterglows to long - duration gamma - ray bursts ( grbs ) showed them to occur in star - forming galaxies at high redshifts . many aspects of the observed grb behaviour could be explained reasonably well by the relativistic fireball models , in which the prompt emission is largely produced by shocks internal to the outflow , and the long - lived afterglow by the outflow impacting and shocking the external ambient medium @xcite . however , it was early appreciated that if they emitted isotropically then the radiative energies implied for some bursts would be very large ( in excess of a solar rest mass in the case of grb990123 , for example ; @xcite ) . since no known mechanism can produce high - energy photons with efficiency approaching 100% , the total explosive energy required would be even greater , implausibly large for a stellar core - collapse powered event . the energetics argument led to the expectation that the grb outflow must be confined to a jet , reducing the overall energy requirements . in that case a sufficiently massive core collapsing to a black - hole might produce an outflow which could pierce a hydrogen - stripped envelope to still produce a relativistic jet @xcite . this picture received support from the observation of concurrent supernova events at grb sites for some ( low - redshift ) low - luminosity bursts @xcite . the class of grb - supernovae could be explained by the explosion of a wolf - rayet star , and it has been widely assumed that this progenitor must account for most if not all long - grbs . a strong prediction of all collimated models is that the observed light curve should exhibit an achromatic break , visible in both optical and x - ray light curves , when the relativistic outflow slows to the point that the doppler beaming angle becomes wider than the opening angle of the jet @xcite : the later the break time , the wider the jet . in the pre-_swift _ era a number of bursts with good afterglow observations showed breaks in the optical , at times ranging from a few hours to a few days . furthermore , it was contended that the implied jet opening angles anti - correlated with the isotropic - equivalent @xmath5-ray energies @xmath6 @xcite in other words , corrected for collimation nearly all bursts seemed to be accessing a rather standard energy reservoir ( to within about an order of magnitude ) , potentially usable as standard candles at high redshift . while the total rotational energies available from a collapsing core may be of order 10@xmath7 erg , the maximal extraction of energy is unlikely to exceed 10@xmath810@xmath9 erg @xcite . with this in mind , the standard energy reservoir with energies of @xmath1010@xmath8 erg appeared very attractive . unfortunately the x - ray data available pre-_swift _ only provided definitive evidence of ( single ) breaks for a few bursts . one of the expectations of _ swift _ was that many more bursts would be seen with clear breaks in the x - ray coincident with optical breaks . after more than four years of operations this is not what has been found . in fact , the situation is much more confusing , with x - ray light curves exhibiting complex behaviour including flares and plateaux @xcite . when late breaks ( beyond a few hours ) are seen they rarely coincide in time or degree with optical breaks , and are hard to reconcile with simple fireball model predictions ( e.g. * ? ? ? * ) , though there are notable exceptions . it is possible that from x - ray data alone a number of jet breaks , particularly those at late times or those which turn over rather smoothly , may go undetected through misinterpretation of their light curves . these so - called ` hidden ' breaks may be revealed on comparison of x - ray light curves with optical light curves @xcite , but the emphasis remains on x - ray light curves for break detection owing to the large numbers of well - sampled x - ray light curves now available . most crucially , in many grbs no temporal break is seen at all to late times , implying much more energy than expected . the most extreme example to - date is that of grb060729 , which continued a smooth x - ray power - law decline for 125 d , implying an opening angle greater than 28 degrees @xcite . this source continues to be monitored in the x - rays with _ chandra _ and was still detected one year after the grb event ( grupe et al . the light curve now shows a possible break at @xmath10290 d : the nature of this break is not yet known . while the modest luminosity and low redshift of grb060729 ( @xmath11 erg , @xcite ) make it unsuitable to test collapsar models to their limits , such observations provide a crucial guide to further theoretical developments . if such high collimation - corrected energies continue to be inferred , then the situation looks bleak for simple core - collapse models as the explanation for the high - luminosity long - grbs . in fact , it would suggest a different progenitor is required for high- and low - luminosity long - bursts . on the other hand , if late breaks are seen , although the jet model is not proven this could provide evidence for an upper envelope to the energy tapped by grbs which may still be consistent with that available in principle from core - collapse , but would require a new understanding of how such high efficiencies of radiative emission could be achieved . in this paper we follow a high redshift grb with a high isotropic energy out to late times to search for a jet break and thereby measure the total energy budget of the prompt @xmath5-ray emission . in section 2 we describe the @xmath5-ray , x - ray and optical observations and analyses . we derive the grb fluence in section 3 . in section 4 we characterise the afterglow spectra and light curves and compare the results to the fireball model in section 5 discussing the possibility of an early jet break . in section 6 we calculate limits on the jet opening angle and collimation - corrected energy of this grb . in section 7 we note the caveats associated with our results , discuss the energetics of this grb in relation to current popular progenitor models and compare this grb to other bursts of interest . in section 8 we conclude . throughout this paper we use the conventions @xmath12 for temporal and spectral power law models . errors are quoted at the 90% confidence level unless otherwise stated . the _ swift _ satellite triggered on grb080721 on 2008 july 21 at 10:25:16.979 ut ( t@xmath13 ) . the burst alert telescope ( bat ) @xmath5-ray light curve showed approximately six peaks , 3 strong followed by 3 weak , spanning t@xmath13 - 11 s to t@xmath13 + 370 s and the burst duration as defined by the t@xmath14 parameter was 16.2@xmath154.5 s @xcite . the bat data were reduced using the _ swift _ analysis software version 29 ( released 2008 june 29 , part of heasoft 6.5.1 ) . the _ konus - wind _ experiment triggered on the prompt emission of grb080721 beginning at 10:25:10.927 ut ( t@xmath16 ) . the propagation delay from _ swift _ to _ konus - wind _ is 3.375 s for this grb , i.e. , correcting for this factor one sees that the _ konus - wind _ trigger time corresponds to t@xmath13 - 9.427 s. from t@xmath16 to t@xmath16 + 482.560 s , 64 spectra in 101 channels ( from 20 kev to 15 mev ) were accumulated on time scales varying from 64 ms near the trigger time to 8.192 s by the time the signal became undetectable @xcite . data were processed using standard _ konus - wind _ analysis tools . the 20 kev7 mev peak flux is ( @xmath17)@xmath110@xmath18 erg @xmath19 s@xmath20 ( using the spectrum accumulated over the main peak , from t@xmath16 + 8.448 s to t@xmath16 + 10.494s ) . the corresponding isotropic peak luminosity is @xmath21 erg s@xmath20 . the @xmath5-ray light curves from both bat and _ konus - wind _ are shown in fig . [ batkonuslc ] . the _ swift _ x - ray telescope ( xrt ) slewed to the burst and began observations in windowed timing ( wt ) mode 113 s after the bat trigger , detecting the tail of the prompt emission and the x - ray afterglow , and transitioned to photon counting ( pc ) mode in the second orbit of data collection . with xrt we have monitored the bright afterglow out to 1.4@xmath110@xmath22 s or 16 d @xcite . the xrt data were reduced using the _ swift _ analysis software version 29 , as described in @xcite . light curves use dynamic binning with a minimum binsize of 5 s and spectra are grouped such that a minimum of 20 counts lie in each bin . we triggered our target - of - opportunity programme on the _ xmm - newton _ satellite to obtain continued observations of grb080721 once the x - ray afterglow flux had fallen below the sensitivity limits of the _ swift _ xrt . the afterglow is clearly detected with the european photon imaging cameras ( epic ) pn in a 12 ks observation performed on 2008 august 12 and again in a 73 ks observation beginning 2008 - 08 - 26 and ending 2008 - 08 - 27 . the _ xmm - newton _ data were reduced using the standard _ xmm - newton _ science analysis software version 20080701 - 1801 . the second observation is partially affected by high background so we removed periods of background with @xmath23 1 count s@xmath20 in the pn data leaving 46.5 ks of useful exposure time . we used the tool edetect - chain to identify the source position and measure a background - subtracted count rate . we then used especget to create spectra , and obtain the flux at each epoch by assuming a spectral shape identical to that of the best fit to the time - averaged _ swift _ xrt pc mode spectrum . the ultraviolet - optical telescope ( uvot ) on - board _ swift _ began observations 118 s after the bat trigger and found a bright optical afterglow in all optical filters , while the source was not detected in the uv filters @xcite . here we will focus on the @xmath24 and @xmath25 bands where temporal coverage was most comprehensive . we initiated ground - based observations once grb080721 became visible from the roque de los muchachos observatory , la palma . observations were taken using the nordic optical telescope ( not ) and the liverpool telescope ( lt ) , followed several hours later by observations from paranal observatory with the very large telescope ( vlt ) ut2 ( kueyen ) . after the initial observations , we initiated our programme for late - time light curve monitoring , obtaining further observations with the lt , the isaac newton telescope ( int ) , the william herschel telescope ( wht ) and the vlt ut1 ( antu ) . all optical afterglow photometry is listed in table [ optdata ] . ground - based optical photometry have been reduced in the standard fashion using iraf @xcite . for the photometric calibration , we used the zeropoints for the vlt fors2 , provided by the observatory , after verifiying the corresponding nights were photometric using standard stars . for the @xmath26 and @xmath27 filters , we converted the @xmath28 and @xmath29 zeropoints to sdss filters using the conversion at the sdss webpages . we adopt a galactic extinction of @xmath30 mag @xcite . a low resolution spectrum was taken at the not from which we measured a redshift of @xmath31 reported in @xcite , confirming an earlier redshift determination from independent data @xcite . starting on 2008 july 22.121 ( 0.7 d post - burst ) , we obtained a sequence of higher resolution spectra using the vlt focal reducer and low dispersion spectrograph ( fors ) 1 in long - slit spectroscopy mode with a 10 wide slit . the progression of grisms used was 300v ( 1200 s ) , 1200b ( 1800 s ) and 600v+gg435 ( 1800 s ) . the individual spectra were cosmic ray cleaned using the method of @xcite . the seeing stayed relatively stable during the observations , between 11 and 13 , yielding a spectral resolution of 13.4 ( 300v ) , 2.9 ( 1200b ) and 5.9 ( 600v+gg435 ) full width at half maximum ( fwhm ) . flux calibration was performed using an observation of the standard star bpm16274 . from the optical spectrum we derive a neutral hydrogen column density of log @xmath32(hi ) @xmath33 @xmath19 . a more detailed analysis of the optical spectra appears in @xcite . [ cols="<,<,<,<,<,<,<",options="header " , ] -ray ( grey , bat only ) , x - ray ( black , final 2 points from _ xmm - newton _ ) and optical ( colours : orange = @xmath24 , blue = @xmath34 and @xmath25 , red / pink = @xmath35 ) light curves shown together . the solid lines show a smoothly broken power law decay fitted to all bands from 430 s onwards ( see table [ table : lcfit ] ) . the optical magnitudes are corrected for galactic extinction , while the x - ray fluxes are 0.310 kev observed fluxes . we have transformed the @xmath5-ray fluxes to the x - ray band using an absorbed power law extrapolation.,width=321 ] we then concentrate our efforts on estimation of the light curve decay of the very late - time data only , where we can be more confident there is little to no influence of possible other components , including flares and energy injection . from 0.5 d onwards , we measure an x - ray decay of @xmath36 ( table [ table : lcfit ] ) . this is interestingly and significantly different from the @xmath37 found when we fit the whole dataset from 430 s onwards ( in which a single break is best fitted at 0.02 d ) , and indicates a change in slope which is not obvious when combined with earlier data . fitting the optical and x - ray decay indices independently rather than tied shows a difference of only @xmath38 , while any difference owing to a so - called cooling break ( see the following section ) between optical and x - rays should be @xmath39 in the fireball model ( e.g. * ? ? ? we thus conclude there is no such spectral break in this scenario . the question arises as to which of the two fits is actually better : the one containing data from 430 s onwards ( smoothly broken power law ) , or the one with data only after 0.5 d ( single power law ) . for this , we calculate the final @xmath40 value , only for the data past 0.5 d , for both fits . that is , the fits themselves are still done by minimizing @xmath40 calculated from the full applicable data range ( post 430 s and post 0.5 d , respectively ) , but we compare @xmath40 calculated by using only data past 0.5 d for each fit ( where likely only a power law applies ) . the @xmath40/dof for the smoothly broken power law in this case is rather bad , 111.1/55 , compared to 55.3/58 for a power law only . this would indicate that the relatively good total @xmath40 value for the smoothly broken power law fit arises largely from the early data between 430 s and 0.5 d. this could be the case if errors in the early section are slightly overestimated . it could also indicate a second break at x - ray wavelengths in this interval , which would account for the difference in final power law decays ( @xmath41 of 1.6 versus 1.4 ) . such a break would , however , increase the complexity of the underlying light curve model , and may make it more difficult to interpret the results . lastly , we could also be seeing effects from fitting a smoothly broken power law , where the second decay index is the actual final decay index , and thus may not be equal to the decay index measured along a small portion of the light curve ( see e.g. * ? ? ? we now investigate whether or not the data are consistent with the fireball model predictions . we use the closure relations ( found in , e.g. , table 1 of @xcite ) relating the optical and x - ray spectral and temporal slopes to determine how this grb afterglow may fit into the standard fireball model . in section 4.2 we have performed the same light curve fits to two subsets of the data : one beginning at 430 s since the grb and one beginning 0.5 d since the grb , with two different outcomes . following the reasoning in the last paragraph of section 4.2 , we examine the results obtained from fitting the data from 0.5 d to a single power law . the late - time temporal fits show that the slopes are essentially the same in the x - ray and optical bands , with @xmath42 , which indicates that both observing bands are in the same spectral regime , either both below or above the cooling break . in the case where they are both below the cooling break and the circumburst medium is homogeneous , the value of @xmath43 can be derived to be @xmath44 from @xmath45 ( well within the range of previously found values , @xcite ) , and the predicted temporal slope is @xmath46 , which is consistent with the observed value . a stellar - wind - like medium or a scenario in which both the optical and x - ray bands are situated above the cooling frequency can be ruled out by the standard closure relations between the spectral and temporal indices . the value for the x - ray spectral index , however , is somewhat at odds with the estimated @xmath47 , if there is no supposed break across the @xmath47 range . host galaxy extinction may mean the intrinsic optical magnitudes are brighter than those observed , allowing an increase in @xmath47 , which could resolve this apparent discrepancy . to investigate this , we can combine the @xmath48 optical photometry ( table [ optdata ] ) with an x - ray spectrum to estimate extinction from the broadband spectral energy distribution ( sed ) in the case that @xmath49 . the sed , centred at 1.725 d , was created and fitted via the method outlined in @xcite and adopts a single absorbed power law as the underlying model and small magellanic cloud - like host galaxy extinction . we find that extinction is required at the level of @xmath50 , corresponding to @xmath51 magnitudes . this value is one of the highest measured for a grb host galaxy @xcite , but such measurements are made only for grbs with detected optical afterglows and presumably extinction is much higher in the optically undetected dark bursts ( e.g. * ? ? ? * ; * ? ? ? for completeness we performed the same tests on the best - fitting smoothly broken power law model to the data from 430 s onwards . we find that the power law index of the electron energy distribution , @xmath43 , derived from the light curve predicts harder spectral indices than are observed . we investigated the possibility of an additional spectral component , caused by inverse compton ( ic ) emission , which would harden the spectrum ( as also proposed for e.g. grb990123 , @xcite and grb000926 , @xcite ) . following the equations in @xcite and @xcite , we find that the closure relations for the temporal and spectral indices still can not be satisfied . we note that in this scenario the break at 0.02 d may be considered an early jet break . there are two arguments against attribution of this feature to the jet break . firstly , after a jet break the temporal slopes are given by @xmath52 , regardless of whether the observing frequency is above or below the cooling frequency , so the optical and x - ray temporal slopes should be the same . this is not the case for this dataset . secondly , if this feature were a jet break it would imply an opening angle of 0.44@xmath53 ( calculated using the equations in the following section ) , far smaller than most previously determined values ( e.g. * ? ? ? * ; * ? ? ? * but see also @xcite ) . concluding , the results from fits to the late - time only data indicate that both observing bands are in the same spectral regime and in order to correctly predict the x - ray spectral slope the jet must be traversing a homogeneous medium . the difference between the optical - to - x - ray spectral index and the x - ray spectral slope is not predicted , but could be explained by the invocation of host galaxy extinction of @xmath54 . the results from using the full range of data can not be satisfactorily described by the fireball model and/or with an early jet break , and it is likely that in this case a single smoothly broken power law model is not applicable . these data have shown that a late - time temporal break at x - ray or optical wavelengths is not required . we can therefore use the latest observation time as a lower limit to any jet break time , and calculate the required limits on opening angle and total energy budget for the grb jet . to calculate the isotropic energy @xmath6 we use the following equation @xmath55 adopting the values listed in sections 2 and 3 ( @xmath31 , fluence @xmath56 erg @xmath19 ) and a luminosity distance @xmath57 of 6.5@xmath110@xmath58 cm ( where @xmath59 km s@xmath20 mpc@xmath20 , @xmath60 and @xmath61 , adopted in order to compare our value with the sample of @xcite ) . we find @xmath62 erg . we have not included a @xmath63 correction , because the energy range used encompasses the peak energy , and the measured fluence should be sufficiently close to the bolometric fluence erg @xmath19 corresponding to @xmath64 erg ] . using this value for @xmath6 we can go on to calculate an upper limit on the jet opening angle ( e.g. * ? ? ? * ) , since no jet break is seen up to our last _ xmm - newton _ observation . this is assuming we may apply the fireball model for the case of a uniform jet ( see section 5 for further discussion of this ) . @xmath65 adopting typical values for efficiency and density of @xmath66 and @xmath67 @xmath68 , we find @xmath69 radians or ( @xmath70 . + the lower limit on the total energy budget in @xmath5-rays for the grb is then obtained from the following equation which accounts for jet collimation using the newly derived jet opening angle upper limit , @xmath71 which gives @xmath72 erg , or more conservatively taking the lower limit for @xmath6 then @xmath73 erg using the end of the final _ xmm - newton _ observation ( 36.08 d or 3.12@xmath110@xmath2 s ) as our lower limit on any jet break time . we have followed the afterglow of the high @xmath6 , relatively high redshift grb080721 to beyond 10@xmath2 s at both x - ray and optical wavelengths . the late - time decay is well fitted with a single power law with no requirement for a break . this suggests that the jet break must lie beyond the detection limits of the intruments used , and from this we calculate a lower limit to the total prompt @xmath5-ray energy of the grb of @xmath73 erg and to the jet opening angle of @xmath74 . this relies on the assumption that the fireball model describes the physics of the jet , the jet is uniform in structure , and that the standard values for surrounding density and efficiency of energy conversion are applicable . we note that the dependence of the total energy on these latter two parameters is very weak ( equation 2 ) . we have compared the closure relations from the fireball model with the observed temporal and spectral slopes . the model and the observations can not be reconciled for the whole dataset , with departures from the model at early times ( before half a day ) , and different temporal slopes are derived depending on whether data before 0.5 d are included in the fit . it is likely that additional components combine with the afterglow emission at the onset of the afterglow , rendering the fireball model inapplicable , but at what time the afterglow begins to dominate is not known . this illustrates the difficulties of pinning down the details of the blastwave physics even when broad - band well - sampled data are available . the data either from 0.5 d or from 430 s onwards require that the grb jets traverse a homogeneous circumburst medium . the apparent inconsistency between the x - ray spectral index and the optical to x - ray spectral index may be caused by extinction in the host galaxy . we have also investigated the alternative scenario in which an earlier break in the light curve at @xmath100.02 d is the jet break , and found that this can not work within the fireball model . if we assume the fireball model , and therefore that one of the measured observational parameters is contaminated ( perhaps by an additional component or incorrect other assumption ) , the implied jet opening angle would be very small ( 0.44@xmath53 ) for a grb ( or indeed any known jet source ) . such collimation can not be explained by any current models and the chances of the narrow jet happening upon our line - of - sight are small . we do not prefer this scenario for these reasons , but we note that this explanation has been the preferred model for some other grbs which we discuss in section 7.3 . in the following subsection we summarise the main classes of long grb progenitor models , and discuss whether any of these can reach energies 10@xmath9 erg and beyond . in the collapsar scenario , two different ways to extract energy to make a grb are suggested in the literature , one using neutrinos and the other harnessing magnetic fields . in the neutrino - driven model , the energy that can be extracted depends sensitively on the accretion rate of collapsing material onto the newly formed black hole through the accretion disc and on the kerr parameter describing the rotation of the black hole , as seen in table 3 of popham et al . ( 1999 ; see also @xcite ) . for typical long grb durations ( tens to hundreds of seconds ) a grb total energy of 10@xmath9 erg is certainly possible within this model particularly for fast rotation rates . the observational testing of such models depends , therefore , not on measurements of grb total energies but on determination of the rotation and mass accretion rates . according to the recent grb progenitor models by @xcite , @xcite and @xcite , a lower metallicity star is preferred for a higher kerr parameter . @xcite and @xcite introduced models of chemically homogeneous evolution which produces helium stars with little or no hydrogen envelope that are metal - poor fast - rotators , and can therefore go on to form a grb . the complexities of stellar evolution are not all well understood , and make the determination of grb progenitors a difficult undertaking . however , the identification of the crucial role of metallicity opens up an avenue for observational constraints . we can measure the metallicity of the circumburst medium or host galaxy along the line - of - sight to grbs through afterglow spectroscopy , and for most long grbs metallicities of order a tenth the solar value have been inferred from this method @xcite , with the lowest measured value being @xmath75 solar for grb050730 @xcite though lines are often saturated hence likely to provide only lower limits on the metallicity @xcite . we can measure the metallicity , @xmath76 , of the environment of grb080721 from the column densities of neutral hydrogen ( hi ly@xmath77 ) and singly ionised silicon ( siii @xmath781808 ) in absorption ( fig . [ optspec ] ) of [ si / h]@xmath23 - 1.1 . all the other observed spectral lines are likely to be saturated , hence we have used the weakest detected metal line . this equates to @xmath79 times the solar value , within the typical distribution for grbs . while this should strictly be treated as a lower limit due to possible saturation of the siii @xmath781808 line , saturation is likely to be moderate at most and , therefore , the actual metallicity of the absorber should be of the order of one tenth of solar , unless significant dust depletion effects are at play ( as silicon is a mildly refractory element ) . if the metallicity is significantly lower than 0.1 solar , the kerr parameter may be close to @xmath80 @xcite , and an energy of 1@xmath110@xmath9 erg could be easily achieved according to the calculations of @xcite . a high metallicity could be an indication of large initial stellar masses , or simply that the models are not applicable . the magnetic - field - driven model uses the blandford - znajek mechanism to extract rotation energy from the black hole @xcite . this idea is indeed supported by the recent general relativistic magnetohydrodynamical simulations by @xcite and @xcite . the energies that can be extracted are up to 10@xmath7 erg at 100% efficiency . typical efficiencies are likely to be far lower than this ( but see * ? ? ? * ) , and with 1% efficiency the lower limit on total energy of grb080721 can be accommodated . here again the kerr parameter is an important factor in determining the available energy , in which case lower metallicity progenitor stars may be preferable . interestingly , the measured energy limit for grb080721 is comparable to the energy budget in the magnetar scenario @xcite , where the extraction of the rotational energy of a millisecond pulsar by strong magnetic fields is suggested @xcite . the magnetar scenario for long grbs has received renewed interest in recent observational works , where it has been suggested to explain some plateau phases and very steep x - ray declines ( * ? ? ? * ; * ? ? ? * specifically in e.g. grb070110 , @xcite ) . the claim of an 8 s periodicity in the highest energy bands for grb090709a suggests association with a magnetar for this source @xcite . the theoretical prediction for the upper limit on the energy of a maximally rotating neutron star is few@xmath110@xmath9 erg @xcite , and we note that @xmath1010@xmath9 erg is the lower limit we derive for this grb implying very high efficiency close to 100% . we therefore conclude that it is unlikely that grb080721 is produced via magnetar formation . the implied total energy for grb080721 based on the lower limit calculated here is not only a challenge to some progenitor models , but is a rare occurrence among grbs . we have compared this limit to the energies measured from detections of jet breaks in a _ swift _ sample of grbs published by @xcite in fig . [ egammaeisoswift ] . the lower limit for grb080721 is much larger than any energy measured from a jet break in this sample . comparing this source with the amati relation @xcite we see that it falls at the high @xmath81@xmath6 end of the correlation , in a continuation of the long grb population out to the most energetic sources . the _ xmm - newton _ observations have enabled a longer baseline for jet searches and this source was chosen specifically for its unusually high isotropic energy . log @xmath82 ( upper ) and log @xmath83 log @xmath6 ( lower ) plots for _ swift _ long grbs with jet breaks or possible jet breaks ( taken from table 2 of @xcite ; redshifts may be spectroscopic , photometric or from host association ) . jet break time is given in the observer frame . the arrows in the upper right of each plot show the limits set here on grb080721.,width=170 ] grb061007 @xcite was also a highly energetic grb with an isotropic energy of 1@xmath110@xmath7 erg @xcite . the closure relations for this source were also not entirely satisfied and no spectral evolution was found from 80 s to 3.1 d after the trigger . the high implied total energy , in both the wind and homogeneous medium cases considered in @xcite led the authors to conclude that an early jet break must have occurred , before t@xmath84 + 80 s , which implies a highly collimated outflow with opening angle @xmath85@xmath86 , four times smaller than any determined previously . this scenario , which we investigated for grb080721 in section 5 and subsequently ruled out , requires that the narrow jet happens to lie along our line - of - sight , may be supported by the rarity of such non - evolving afterglows and the unusually high initial optical brightness for grb061007 . however , we discussed long grb progenitor models in the previous section for which the large energies measured both for grb061007 and in this paper for grb080721 when assuming a very late , unobserved jet break with a more typical jet opening angle , are in fact achievable . another case for which the observations implied a high energy budget , while the standard fireball model was a poor description of the afterglow , was grb080319b @xcite . this collapsar scenario for this long grb is , however , supported by the detection of a supernova signature @xcite . to explain the complex afterglow behaviour , not smooth and featureless in this case but appearing to have a number of components including an initially extremely bright optical counterpart , @xcite introduced a second jet this 2-jet model consisted of an inner jet with @xmath82 = 0.03 d and opening angle 0.2@xmath53 and an outer jet more typical of grbs with @xmath82 = 11.6 d and opening angle 4@xmath53 . double and even triple jet systems are occasionally invoked when single jet models simply can not explain the observations ( e.g. grb030329 , berger et al . 2003 ; grb021004 , starling et al . 2005a ; grb050401 , kamble et al . 2009 ; grb050802 , oates et al . 2007 , de pasquale et al . 2009 ) , and while it is likely to be more physical to assume some structure across the jet ( e.g. * ? ? ? * ; * ? ? ? * ) , the interaction of multiple jets is not well known . we have observed the bright , highly energetic afterglow of grb080721 out to 36 days since the trigger in the optical and x - ray bands . we conclude that no jet break is present in the late - time light curve and we rule out a jet - break origin for the early light curve break , inferring a limit on the opening angle of @xmath74 and setting constraints on the total energy budget of the burst of @xmath87 erg within the fireball model . to obey the fireball model closure relations the grb jet must be expanding into a homogeneous medium and be extincted in the optical bands by approximately 0.6 magnitudes . the energy constraint we derive can be used as observational input for models of the progenitors of long gamma - ray bursts . we can likely rule out a magnetar progenitor for this grb as this would require close to 100% efficiency to reach the observed total energy . such high collimation - corrected energies could be accommodated with certain parameters of the standard massive star core - collapse models . one of the key observational parameters in distinguishing between various core - collapse models is the metallicity , diagnostic only if it lies outside of the typical range of @xmath88 . the metallicity we measure for grb080721 unfortunately does not allow a distinction between models to be made . the occurrence of such highly energetic or narrowly beamed grbs is rather rare , and for those which have the required measurements to allow a study such as this one on grb080721 we estimate there may be a handful per year among the 100130 well localised grbs currently triggering operational satellites . measurement of the high energy spectral peak , securing the redshift and monitoring the light curves for as long as possible ( preferably in multiple bands ) are all crucial , and these data have shown that a full understanding of the blastwave physics can be difficult to achieve even with a good quality , well - sampled , broad - band dataset . to determine whether these sources are true outliers from the grb population , perhaps requiring different or more extreme progenitors , will require dedicated efforts to build up small samples to compare with each other and with the growing database of more typical grbs . we acknowledge useful discussions with andrew read , jenny carter , daniele malesani and tayyaba zafar . rlcs , er , klp , kw , pac and jpo acknowledge financial support from stfc . ajvdh is supported by an appointment to the nasa postdoctoral program at the msfc , administered by orau through a contract with nasa . s - cy is supported by the doe scidac program ( doe de - fc02 - 06er41438 ) . pj acknowledges support by a marie curie european re - integration grant within the 7th european community framework program under contract number perg03-ga-2008 - 226653 , and a grant of excellence from the icelandic research fund . this work made use of data supplied by the uk _ swift _ science data centre at the university of leicester and _ xmm newton _ data taken under ao-7 proposal no . 055597 . the _ konus - wind _ experiment is supported by a russian space agency contract and rfbr grant 09 - 02 - 00166a . based on observations made with eso telescopes at the paranal observatory under programme i d 081.d-0853 . the int and wht are operated on the island of la palma by the isaac newton group in the spanish observatorio del roque de los muchachos of the instituto de astrofsica de canarias . the liverpool telescope is operated on the island of la palma by liverpool john moores university in the spanish observatorio del roque de los muchachos of the instituto de astrofisica de canarias with financial support from the uk science and technology facilities council . based on observations made with the nordic optical telescope , operated on the island of la palma jointly by denmark , finland , iceland , norway , and sweden , in the spanish observatorio del roque de los muchachos of the instituto de astrofsica de canarias . we thank birgitta nordstrom and edita stonkute for performing the not observations . references aloy m.a . , mller e. , ibez j.m . , mart j.m . , macfadyen a. , 2000 , apj , 531 , l119 amati l. , guidorzi c. , frontera f. , della valle m. , finelli f. , landi r. , montanari e. , 2008 , mnras , 391 , 577 band d. et al . , 1993 , apj , 413 , 281 barkov m.v . , komissarov s.s . , 2008 , mnras , 385 , l28 berger e. et al . , 2003 , nature , 426 , 154 berger e. , kulkarni s.r . , frail d.a . , 2003 , apj , 590 , 379 beuermann k. et al . , 1999 , a&a , 352 , l26 blandford r.d . , znajek r.l . , 1977 , mnras , 179 , 433 cenko s.b . et al . , 2009 , apj , 693 , 1484 chen h .- w . , prochaska j.x . , bloom j.s . , thompson b. , 2005 , apj , 634 , l25 chevalier r.a . , li z .- y . , 1999 , apj , 520 , l29 corsi a. et al . , 2005 , a&a , 438 , 829 costa e. et al . , 1997 , nature , 387 , 783 curran p.a . , van der horst a.j . , wijers r.a.m.j . , 2008 , mnras , 386 , 859 davanzo p. et al . , 2008 , gcn circular 7997 de pasquale m. et al . , 2009 , mnras , 392 , 153 duncan r.c . , thompson c. , 1992 , apj , 392 , l9 evans p.a . et al . , 2007 , a&a , 469 , 379 evans p.a . et al . , 2009 , mnras in press frail d.a . et al . , 2001 , apj , 562 , l55 fynbo j.p.u . et al . , 2006 , a&a , 451 , l47 fynbo j.p.u . et al . , 2009 , apjs submitted galama t. et al . , 1998 , nature , 395 , 670 golenetskii s. , aptekar r. , mazets e. , palshin v. , frederiks d , cline t. , 2006 , gcn circular 5722 golenetskii s. et al . , 2008 , gcn circular 7995 golenetskii s. et al . , 2009 , gcn circular 9647 gotz d. , mereghetti s. , von kienlin a. , beck m. , 2009 , gcn circular 9649 granot j. , sari r. , 2002 , apj , 568 , 820 grupe d. et al . , 2007 , apj , 662 , 443 harrison f.a . et al , 2001 , apj , 559 , 123 hjorth j. et al . , 2003 , nature , 423 , 847 jakobsson p. et al . , 2008 , gcn circular 7998 jhannesson g. , bjrnsson g. , gudmundsson e.h . , 2006 , apj , 640 , l5 kalberla p.m.w . , burton w.b . , hartmann d. , arnal e.m . , bajaja e. , morras r. , poppel w.g.l . , 2005 , a&a , 440 , 775 kamble a. , misra k. , bhattacharya d. , sagar r. , 2009 , mnras , 394 , 214 kocevski d. , butler n. , 2008 , apj , 680 , 531 komissarov s.s . , barkov m.v . , 2007 , mnras , 382 , 1029 komissarov s.s . , vlahakis n. , knigl a. , barkov m.v . , 2009 , mnras , 394 , 118 kulkarni s. et al . , 1999 , nature , 398 , 389 lyons n. , obrien p.t . , zhang b. , willingale r. , troja e. , starling r.l.c . , 2008 , mnras submitted macfadyen a. , woosley s. , 1999 , apj , 524 , 262 markwardt c.b . , gavriil f.p . , palmer d.m . , baumgartner w.h . , barthelmy s.d . , 2009 , gcn circular 9645 marshall f.e . , guidorzi c. , ward p.a . , barthelmy s. , burrows d.n . , roming p. , gehrels n. , 2008 , gcn report 156.1 mszros p. , rees m. , 1997 , apj , 476 , 232 mundell c.g . et al . , 2007 , apj , 660 , 489 nagataki s. , takahashi r. , mizuta a. , takiwaki t. , 2007 , apj , 659 , 512 nousek j.a . et al . , 2006 , apj , 642 , 389 oates s.r . et al . , 2007 , mnras , 380 , 270 obrien p.t . et al . , 2006 , apj , 647 , 1213 ohno m. et al . , 2009 , gcn circular 9653 paczynski b. , 1998 , apj , 494 , l45 panaitescu a. et al . , 2006 , mnras , 366 , 1357 panaitescu a. , 2007 , mnras , 380 , 74 pian e. et al . , 2006 , nature , 442 , 1011 popham , woosley s. , fryer c. , 1999 , apj , 518 , 356 prochaska j.x . , apj , 650 , 272 prochaska j.x . , chen h .- w . , dessauges - zavadsky m. , bloom j. , apj , 666 , 267 racusin j. et al . , 2008 , nature , 455 , 183 rhoads j. , 1997 , apj , 487 , l1 rossi e.m . , lazzati d. , rees m.j . , 2002 , mnras , 332 , 945 rosswog s. , ramirez - ruiz e. , 2003 , aipc , 727 , 361 sari r. et al . , 1998 , apj , 497 , l17 sari r. , esin a.a . , 2001 , apj , 548 , 787 schady p. et al . , 2007 , mnras , 377 , 273 schady p. et al . , 2007 , mnras , 380 , 1041 schlegel d.j . , finkbeiner d.p . , davis m. , 1998 , apj , 500 , 525 shen r .- f . , kumar p. , robinson e.l . , 2006 , mnras , 371 , 1441 stanek k.z . et al . , 2003 , apj , 591 , l17 starling r.l.c . , wijers r.a.m.j . , hughes m.a . , tanvir n.r . , vreeswijk p.m. , rol e. , salamanca i. , 2005a , mnras , 360 , 305 starling r.l.c . et al . , 2005b , a&a , 442 , l21 starling r.l.c . , wijers r.a.m.j . , wiersema k. , rol e. , curran p.a . , kouveliotou c. , heemskerk m.h.m . , 2007 , apj , 661 , 787 starling r.l.c . , van der horst a.j . , rol e. , wijers r.a.m.j . , kouveliotou c. , wiersema k. , curran p.a . , weltevrede p. , 2008 , apj , 672 , 433 tanvir n.r . et al . , 2008 , apj submitted , arxiv:0812.1217 thompson t.a . , chang p. , quataert e. , 2004 , apj , 611 , 380 tody d. , 1986 , instrumentation in astronomy vi ; proceedings of the meeting , tucson , az , mar . 4 - 8 , 1986 . part 2 ( a87 - 36376 15 - 35 ) . bellingham , wa , society of photo - optical instrumentation engineers , p.733 . troja e. et al . , 2007 , apj , 665 , 599 van dokkum p.g . , 2001 , pasp , 113 , 1420 van paradijs j. et al . , 1997 , nature , 386 , 686 wheeler c.j . , yi i. , hflich p. , wang l. , 2000 , apj , 537 , 810 willingale r. et al . , 2007 , apj , 662 , 1093 woosley s.e . , heger a. , 2006 , apj , 637 , 914 yoon s .- c . , langer n. , 2005 , a&a , 443 , 643 yoon s .- c . , langer n. , norman c. , 2006 , a&a , 460 , 199 zhang b. , mszros p. , 2001 , apj , 552 , l35 zhang b. , mszros p. , 2002 , apj , 571 , 876 zhang b. , mszros p. , 2004 , ijmpa , 19 , 2385

the study of stars is one of the most fascinating topics in astrophysics because it involves many areas of physics : gravitation , thermodynamics , hydrodynamics , statistical mechanics , quantum mechanics , relativity , ... and it has furthermore a very interesting history . in his classical monograph _ the internal constitution of the stars _ , eddington ( 1926 ) @xcite lays down the foundations of the subject and describes in detail the basic processes that govern the structure of ordinary stars . at that stage , only elements of `` classical '' physics are used and difficulties with such theories to account for the structure of high density stars such as the companion of sirius are pointed out . soon after the discovery of the quantum statistics by fermi ( 1926 ) @xcite and dirac ( 1926 ) @xcite , fowler ( 1926 ) @xcite uses this `` new thermodynamics '' to explain the puzzling nature of white dwarf stars . he understands that low mass white dwarf stars owe their stability to the quantum pressure of the degenerate electron gas . ) . the case of a partially degenerate self - gravitating fermi gas at arbitrary temperature has been discussed more recently by hertel & thirring ( 1971 ) @xcite and chavanis ( 2002 ) @xcite in the context of statistical mechanics . they describe the phase transition , below a critical temperature @xmath9 , from a gaseous configuration to a condensed state ( with a degenerate core surrounded by a halo ) . this provides a simple physical mechanism showing how the system can reach highly degenerate configurations as a result of gravitational collapse ; see chavanis ( 2006 ) @xcite for a review on phase transitions in self - gravitating systems . ] the resulting structure is equivalent to a polytrope of index @xmath10 so that the mass - radius relation of classical white dwarf stars behaves like @xmath11 ( chandrasekhar 1931a @xcite ) . the next step was made by chandrasekhar , aged only nineteen , who was accepted by the university of cambridge to work with fowler . in the boat that took him from madras to southampton @xcite , chandrasekhar understands that relativistic effects are important in massive white dwarf stars and that einstein kinematic must be introduced in the problem . in his first treatment ( chandrasekhar 1931b @xcite ) , he considers the ultra - relativistic limit and shows that the resulting structure is equivalent to a polytrope of index @xmath12 . applying the theory of polytropic gas spheres ( emden 1907 @xcite ) , this leads to a unique value of the mass @xmath13 that he interprets as a limiting mass , nowdays called the chandrasekhar limit . quantum mechanics can not prevent the system from collapsing to a point . however , he did not take this collapse very seriously and noted that , in reality , there exists stars with mass @xmath14 that do not show this `` ridiculous tendency '' , so that they must possess regions in which the laws of quantum mechanics are violated . ] the complete mass - radius relation of relativistic white dwarf stars was given later ( chandrasekhar 1935 @xcite ) and departs from fowler s sequence as we approach the limiting mass . at this critical mass , the radius of the configuration vanishes . above the critical mass , the equation of state of the relativistic degenerate fermi gas of electrons is not able to balance gravitational forces and , when considering the final evolution of such a star , chandrasekhar ( 1934 ) @xcite `` left speculating on other possibility '' . chandrasekhar s result was severly criticized by eddington ( 1935 ) @xcite who viewed this result as _ a reductio ad absurdum _ of the relativistic formula and considered the combination of special relativity and non - relativistic quantum theory as an `` unholly alliance '' . because of the querelle with eddington , it took some time to realize the physical implication of chandrasekhar s results . however , progressively , his early investigations on white dwarf stars were extended in general relativity to the case of neutron stars ( oppenheimer & volkoff 1939 @xcite ) and finally led to the concept of `` black holes '' , a term coined by wheeler in 1967 . without the dispute with eddington , this ultimate stage of matter resulting from gravitational collapse could have been predicted much earlier from the discovery of chandrasekhar @xcite . although these historic elements are well - known , it is less well - known that the concept of a maximum mass for relativistic white dwarf stars had been introduced earlier by anderson ( 1929 ) @xcite and stoner ( 1930 ) @xcite . these studies are mentioned in the early works of chandrasekhar but they have been progressively forgotten and are rarely quoted in classical textbooks of astrophysics . these authors investigated the equation of state of a relativistic degenerate fermi gas and predicted an upper limit for the mass of white dwarf stars . stoner ( 1930 ) @xcite uses a uniform mass density to model the star while chandrasekhar ( 1931 ) @xcite considers a more realistic @xmath12 polytrope . however , as noted in chandrasekhar ( 1931 ) @xcite , the value of the limiting mass found by stoner with his simplified model is relatively close to that obtained with the improved treatment . interestingly , nauenberg ( 1972 ) @xcite introduced long after a simplified treatment of relativistic white dwarf stars in order to obtain an analytical approximation of the mass - radius relation . it turns out that this model , which gives a very good agreement with chandrasekhar s numerical results ( up to some normalization factors ) , is similar to that introduced by stoner . leaving aside these interesting historical remarks , the result of chandrasekhar @xcite concerning the existence of a limiting mass is very profound because this mass can be expressed in terms of fundamental constants , similarly to the bohr radius of the hydrogen atom @xcite . hence , the mass of stars is determined typically by the following combination @xmath15 where @xmath16 is the constant of gravity , @xmath17 the planck constant , @xmath18 the velocity of light and @xmath19 the mass of the hydrogen atom ( @xmath20 is the solar mass ) . this formula results from the combination of quantum mechanics ( @xmath17 ) , special relativity ( @xmath18 ) and gravity @xmath21 . since dimensional analysis plays a fundamental role in physics , it is of interest to investigate how the preceding results depend on the dimension of space @xmath1 of the universe . in our previous investigations @xcite , we considered the case of classical white dwarf stars in @xmath1 dimensions and found that they become unstable in a space of dimension @xmath8 . in that case , the star either evaporates or collapses . therefore , quantum mechanics can not stabilize matter against gravitational collapse in @xmath8 contrary to what happens in @xmath4 @xcite . interestingly , this is similar to a result found by ehrenfest @xcite at the atomic level for coulomb forces in bohr s model and for the planetary motion ( kepler problem ) . the object of this paper is to extend these results to the case of relativistic white dwarf stars and exhibit particular dimensions of space which play a special role in the problem when we combine newtonian gravity , quantum mechanics and special relativity . we shall see that the problem is very rich and interesting in its own right . it shows to which extent the dimension @xmath4 of our universe is particular , with possible implications regarding the anthropic principle @xcite . we note that a similar problem has been considered in @xcite on the basis of dimensional analysis . our approach is more precise since we generalize the _ exact _ mathematical treatment of chandrasekhar @xcite to a space of dimension @xmath1 . a connection with other works investigating the role played by the dimension of space on the laws of physics is made in the conclusion . in appendix [ sec_stab ] , we propose different derivations of the stability limits of polytropic spheres and consider applications of these results to classical and relativistic white dwarf stars . following chandrasekhar @xcite , we model a white dwarf star as a degenerate gas sphere in hydrostatic equilibrium . the pressure is due to the quantum properties of the electrons and the density to the protons . in the completely degenerate limit , the electrons have momenta less than a threshold value @xmath22 ( fermi momentum ) and their distribution function is @xmath23 where @xmath17 is the planck constant . there can only be two electrons in a phase space element of size @xmath24 on account of the pauli exclusion principle . therefore , the number of electrons per unit volume is @xmath25 where @xmath26 is the surface of a unit sphere in @xmath1-dimensions . the mean kinetic energy per electron is given by @xmath27 where @xmath28 is the energy of an electron with impulse @xmath29 . in relativistic mechanics , @xmath30 the pressure of the electrons is @xmath31 using eq . ( [ es3 ] ) , the pressure can be rewritten @xmath32 finally , the mass density of the star is @xmath33 where @xmath19 is the mass of the proton and @xmath34 the molecular weight . if we consider a pure gas of fermions ( like , e.g. , massive neutrinos in dark matter models ) , we just have to replace @xmath35 by their mass @xmath36 . introducing the notation @xmath37 , we can write the density of state parametrically as follows @xmath38 where @xmath39 @xmath40 the function @xmath41 has the asymptotic behaviors @xmath42 @xmath43 the classical limit corresponds to @xmath44 and the ultra - relativistic limit to @xmath45 . explicit expressions of the function @xmath41 are given in appendix [ sec_exp ] for different dimensions of space . we provide here the explicit expression of the function @xmath41 defined by eq . ( [ es9 ] ) for different dimensions of space @xmath261 and @xmath262 , respectively : @xmath271 @xmath272 @xmath273 @xmath274 for a spherically symmetric distribution of matter , the equations of hydrostatic equilibrum are @xmath46 @xmath47 they can be combined to give @xmath48 expressing @xmath49 and @xmath50 in terms of @xmath51 and setting @xmath52 , we obtain @xmath53 we denote by @xmath54 and @xmath55 the values of @xmath51 and @xmath56 at the center . furthermore , we define @xmath57 @xmath58 note that the scale of length @xmath59 is independent on @xmath60 for @xmath2 . substituting these transformations in eq . ( [ c4 ] ) , we obtain the @xmath1-dimensional generalization of chandrasekhar s differential equation @xmath61 with the boundary conditions @xmath62 the radius @xmath63 of the star is such that @xmath64 . this yields @xmath65 the density can be expressed as @xmath66 where the central density @xmath67 finally , we find that the mass is related to @xmath55 by @xmath68 note that @xmath60 does not _ explicitly _ enter in this expression for @xmath4 but it is of course present implicitly . in the classical case @xmath44 , we find that the equation of state takes the form @xmath69 with @xmath70 therefore a classical white dwarf star is equivalent to a polytrope of index @xcite : corresponds to the value of the polytropic index of a classical white dwarf star in @xmath4 . the index @xmath71 refers to its corresponding value in @xmath1 dimensions . the same convention is adopted for the other indices @xmath72 , @xmath73 and @xmath74 . ] @xmath75 polytropic stars are described by the lane - emden equation @xcite . this can be recovered as a limit of the chandrasekhar equation . for @xmath44 , we have @xmath76 . we define @xmath77 . to leading order , @xmath78 . setting @xmath79 and combining the foregoing results , we find that eq . ( [ c7 ] ) reduces to the lane - emden equation with index @xmath80 : @xmath81 @xmath82 note that the condition at the origin is @xmath83 instead of @xmath84 as in the ordinary lane - emden equation . however , using the homology theorem for polytropic spheres @xcite , we can easily relate @xmath85 to @xmath86 , the solution of the lane - emden equation with index @xmath87 and condition at the origin @xmath84 . in an appropriate system of coordinates ( see @xcite for details ) . complete polytropes correspond to the terminal point in the series of equilibria . the series becomes dynamically unstable with respect to the euler - poisson system ( saddle point of the energy functional ) after the turning point of mass @xmath88 which appears for @xmath6 . thus , classical white dwarf stars are stable for @xmath89 and unstable for @xmath8 . , width=302 ] the structure and the stability of polytropic spheres in various dimensions of space has been studied by chavanis & sire @xcite . for @xmath90 , this study exhibits two important indices : @xmath91 according to this study , a classical white dwarf star is self - confined ( complete ) if @xmath92 , i.e. @xmath93 . in that case , the density vanishes at a finite radius @xmath63 identified as the radius of the star . on the other hand , it is nonlinearly dynamically stable with respect to the euler - poisson system if @xmath94 , and linearly unstable otherwise . therefore , classical white dwarf stars are stable for @xmath89 and unstable for @xmath8 ( see fig . [ alphac ] and appendix [ sec_stab ] ) . for @xmath95 , classical white dwarf stars are always self - confined and stable . using the results of @xcite , the mass - radius relation for complete polytropes with index @xmath87 in @xmath1 dimensions is @xmath96 where we have defined @xmath97 where @xmath98 is such that @xmath99 . using eq . ( [ class2 ] ) , we find that the mass - radius relation for classical white dwarf stars in @xmath1 dimensions is @xmath100 for @xmath101 , the mass @xmath102 decreases as the radius @xmath63 increases while for @xmath103 and for @xmath104 it increases with the radius ( see fig . [ classique ] ) . and the mass is independent on radius in @xmath6 . the dimensional factors @xmath105 and @xmath106 are defined in sec . [ sec_gen ] . , width=302 ] for @xmath6 the mass is independent on the radius and given in terms of fundamental constants by @xmath107 we recall that the value of the gravitational constant @xmath16 depends on the dimension of space so that we can not give an explicit value to this limiting mass . the central density is related to the radius by @xmath108 for @xmath3 , the radius is independent on mass and given in terms of fundamental constants by @xmath109 furthermore , for @xmath3 , the lane - emden equation ( [ class4 ] ) with index @xmath110 can be solved analytically , yielding @xmath111 and @xmath112 where @xmath113 is the first zero of bessel function @xmath114 . we obtain a density profile @xmath115 where the central density is related to the total mass by @xmath116 finally , for @xmath4 , using the mass - radius relation ( [ class9 ] ) , we find that the average density is related to the total mass by @xmath117 ( for @xmath118 ) . historically , this result was obtained by chandrasekhar ( 1931 ) @xcite who first applied the theory of polytropic gas spheres with index @xmath10 to classical white dwarf stars . it improves an earlier result @xmath119 obtained by stoner ( 1929 ) @xcite on the basis of his model of stars with uniform density ( see sec . [ sec_ana ] ) . in the ultra - relativistic limit @xmath45 , we find that the equation of state takes the form @xmath120 with @xmath121 therefore , an ultra - relativistic white dwarf star is equivalent to a polytrope of index @xmath122 this also directly results from the chandrasekhar equation . for @xmath45 , it reduces to @xmath123 @xmath124 where we have set @xmath125 and @xmath126 . this is the lane - emden equation with index @xmath127 . in an appropriate system of coordinates ( see @xcite for details ) . complete polytropes correspond to the terminal point in the series of equilibria . the series becomes dynamically unstable with respect to the euler - poisson system ( saddle point of the energy functional ) after the turning point of mass @xmath88 which appears for @xmath4 . thus , ultra - relativistic white dwarf stars are stable for @xmath128 and unstable for @xmath129 . , width=302 ] using the results of @xcite for @xmath90 , we deduce that an ultra - relativistic white dwarf star is self - confined if @xmath130 , i.e. @xmath131 . in addition , it is nonlinearly dynamically stable with respect to the euler - poisson system if @xmath132 and linearly unstable otherwise . therefore , ultra - relativistic white dwarf stars are stable for @xmath133 and unstable for @xmath129 ( see fig . [ alphar ] and appendix [ sec_stab ] ) . for @xmath95 , ultra - relativistic white dwarf stars are self - confined and stable . on the other hand , using the results of @xcite , the mass - radius relation for complete polytropes with index @xmath127 in @xmath1 dimensions is @xmath134 where we have defined @xmath135 using eq . ( [ ur2 ] ) , we find that the mass - radius relation for ultra - relativistic white dwarf stars in @xmath1 dimensions is @xmath136 for @xmath137 , the mass @xmath102 increases as the radius @xmath63 increases while for @xmath138 it decreases with the radius ( see fig . [ relativiste ] ) . and the mass is independent on radius in @xmath4 ( chandrasekhar s mass ) . , width=302 ] for @xmath2 , the radius is independent on mass and given in terms of fundamental constants by @xmath139 furthermore , for @xmath2 , the lane - emden equation ( [ ur4 ] ) with index @xmath140 can be solved analytically , yielding @xmath141 and @xmath142 . we obtain a density profile @xmath143 where the central density is related to the total mass by @xmath144 for @xmath4 , the mass is independent on radius and given in terms of fundamental constants by @xmath145 this is the chandrasekhar mass @xmath146 coming back to eq . ( [ c7 ] ) , we can show that for this limiting value , the radius @xmath63 of the configuration tends to zero ( see sec . [ sec_gen ] ) . historically , the existence of a maximum mass for relativistic white dwarf stars was first published by anderson ( 1929 ) @xcite who considered a relativistic extension of the model of stoner ( 1929 ) @xcite for classical white dwarf stars . he obtained a limiting mass @xmath147 ( for @xmath118 ) . the relativistic treatment of anderson was criticized and corrected by stoner ( 1930 ) @xcite who obtained a value of the limiting mass @xmath148 . the uniform density model of stoner was in turn criticized and corrected by chandrasekhar ( 1931 ) @xcite who applied the theory of polytropic gas spheres with index @xmath12 to relativistic white dwarf stars and obtained the value ( [ ur13 ] ) of the limiting mass @xmath149 . it seems that these historical details are not well - known because the references to the works of anderson and stoner progressively disappeared from the literature . collecting together the results of sec . [ sec_c ] , the mass - radius relation for relativistic white dwarf stars in @xmath1 dimensions can be written in the general case under the parametric form @xmath150 where we have defined @xmath151 and @xmath152 the mass @xmath105 and the radius @xmath106 can be expressed in terms of fundamental constants as @xmath153 @xmath154 we can now obtain the mass - radius curve @xmath155 by the following procedure . we fix a value of the parameter @xmath60 and solve the differential equation ( [ c7 ] ) with initial condition ( [ c8 ] ) until the point @xmath156 , determined by eq . ( [ c9 ] ) , at which the density vanishes . the radius and the mass of the corresponding configuration are then given by eq . ( [ gen1 ] ) . by varying @xmath55 , we can obtain the full curve @xmath157 parameterized by the value of the central density @xmath158 given by eq . ( [ c11 ] ) . to solve the differential equation ( [ c7 ] ) , we need the behavior of @xmath159 at the origin . expanding @xmath160 in taylor series and substituting this expansion in eq . ( [ c7 ] ) we obtain for @xmath161 : @xmath162 where @xmath163 we note in particular that @xmath164 . . the radius increases with the mass . the configurations are always stable and there exists a maximum radius achieved in the ultra - relativistic limit for @xmath165 . , width=302 ] . the configurations are always stable and there exists a maximum radius for partially relativistic distributions . , width=302 ] . the radius decreases with the mass . the configurations are always stable and there exists a maximum radius achieved in the classical limit for @xmath166 . , width=302 ] . the radius decreases with the mass . the configurations are always stable . , width=302 ] . the radius decreases with the mass . the configurations are always stable and there exists a limiting mass ( chandrasekhar s mass ) achieved in the ultra - relativistic limit for @xmath167 . for @xmath168 , quantum mechanics can not arrest gravitational collapse . , width=302 ] finally , we give the asymptotic expressions of the mass - radius relation . in the classical limit , using eq . ( [ class9 ] ) , we obtain @xmath169 in the ultra - relativistic limit , using eq . ( [ ur8 ] ) , we get @xmath170 in figs . [ chandra1]-[chandra4 ] , we plot the mass - radius relation of relativistic white dwarf stars ( full line ) for different dimensions of space . the asymptotic relations ( [ gen8 ] ) and ( [ gen9 ] ) valid in the classical ( c ) and ultra - relativistic ( r ) limits are also shown for comparison ( dashed line ) together with the analytical approximation ( dotted line ) derived in sec . [ sec_ana ] . . there exists a limiting mass for partially relativistic distributions . classical configurations are stable and the series of relativistic equilibria becomes unstable after the turning point of mass . thus , highly relativistic configurations in @xmath129 can not be in hydrostatic equilibrium.,width=302 ] . for @xmath171 the ultra - relativistic configurations ( equivalent to polytropes of index @xmath172 ) are not self - confined anymore and a spiral develops in the mass - radius relation . this is somehow similar to the classical spiral occurring in the @xmath173 plane in the thermodynamics of isothermal self - gravitating systems @xcite and to the spiral occurring in the @xmath174 plane in the general relativistic treatment of neutron stars @xcite . the series of equilibria becomes unstable at the first turning point of mass and new modes of instability occur at the secondary turning points @xcite.,width=302 ] . there exists a limiting mass achieved in the classical limit for @xmath175 . in fact , for @xmath8 , all the configurations ( classical and relativistic ) are unstable . quantum mechanics can not balance gravitational attraction even in the classical limit . , width=302 ] we present here an analytical approximation of the mass - radius relation in various dimensions of space based on the treatment by nauenberg @xcite in @xmath4 . this treatment amounts to considering that the density is uniform in the star and the mass - radius relation is obtained by minimizing the energy functional ( see appendix [ sec_ef ] ) with respect to the radius ( or to the density ) at fixed mass . as mentioned in the introduction , this is similar to the simplified model of relativistic white dwarf stars made by stoner @xcite before chandrasekhar s treatment @xcite . following nauenberg @xcite , we approximate the kinetic energy @xmath176 by the form @xmath177 where @xmath178 is the number of electrons and @xmath29 is an average over the star of the momentum of the electrons . ) of the kinetic energy in his treatment . this leads to more complicated expressions with , however , qualitatively similar conclusions . since the models are based on the ( unrealistic ) approximation that the stellar density is homogeneous , the results can not be expected to be more than qualitatively correct . therefore , the approximation ( [ ana1 ] ) made by nauenberg for the kinetic energy is sufficient for the purposes of this simplified approach . ] we assume that it is determined by an appropriate average value of the density by the relation @xmath179 based on the pauli exclusion principle . now , for the density , we write @xmath180 where @xmath181 is a dimensionless parameter . we also write the potential energy in the form @xmath182 where @xmath183 is another dimensionless parameter . by writing eq . ( [ ana4 ] ) , we have assumed that @xmath184 but we shall see that the following results pass to the limit for @xmath185 . we introduce two dimensionless variables @xmath186 and @xmath187 and two fixed constants @xmath188 and @xmath189 such that @xmath190 we determine @xmath191 and @xmath189 by the requirement that the relativity parameter @xmath192 have the form @xmath193 and that the potential energy can be written @xmath194 this yields @xmath195 @xmath196 comparing with eqs . ( [ gen4 ] ) and ( [ gen5 ] ) , we find that @xmath197 now , the energy @xmath198 of the star can be written @xmath199 in the classical limit @xmath44 , we get @xmath200 and in the ultra - relativistic limit @xmath45 , we obtain @xmath201 we shall consider the energy as a function of the radius @xmath63 , with the mass @xmath102 fixed . thus , the mass - radius relation will be obtained by minimizing the energy @xmath202 versus @xmath51 with fixed @xmath203 . writing @xmath204 , we obtain the equations @xmath205 @xmath206 defining the mass - radius relation in parametric form in the framework of the simplified model . note that @xmath207 represents the value of the density in units of @xmath208 , i.e. @xmath209 . therefore , eq . ( [ ana12 ] ) can be viewed as the relation between the mass and the density . eliminating the relativity parameter @xmath51 ( or density ) between eqs . ( [ ana12 ] ) and ( [ ana13 ] ) , we explicitly obtain @xmath210 @xmath211 @xmath212 @xmath213 in @xmath2 , there exists a maximum radius @xmath214 achieved for @xmath215 , and we have the scaling @xmath216 in @xmath3 , there exists a maximum radius @xmath214 achieved for @xmath217 , and we have the scaling @xmath218 in @xmath4 , there exists a maximum mass @xmath219 ( chandrasekhar s mass ) achieved for @xmath220 , and we have the scaling @xmath221 in @xmath6 , there exists a maximum mass @xmath219 achieved for @xmath222 , and we have the scaling @xmath223 in order to compare eqs . ( [ ana14])-([ana17 ] ) with the exact results , we need to estimate the values of the constants @xmath181 and @xmath183 . this can be done by considering the limiting forms of eqs . ( [ ana12 ] ) and ( [ ana13 ] ) . in the classical limit @xmath44 , we obtain @xmath224 and in the ultra - relativistic limit @xmath45 , we get @xmath225 comparing with eqs . ( [ gen8 ] ) and ( [ gen9 ] ) , we find that @xmath226 where the quantities @xmath227 and @xmath228 can be deduced from the numerical study of the lane - emden equation . the analytical approximations of the mass - radius relation ( [ ana14])-([ana17 ] ) are plotted in dotted lines in figs . [ chandra1 ] , [ chandra2 ] , [ chandra3 ] , and they give a fair agreement with the exact results ( full line ) . of course , they can not reproduce the spiral in @xmath6 which requires the resolution of the full differential equation ( [ c7 ] ) . let us address the stability of the configurations within this simplified analytical model . if we view a white dwarf star as a gas of electrons at statistical equilibrium at temperature @xmath229 , stable configurations are those that minimize the free energy @xmath230 at fixed mass , where @xmath202 is the total energy ( kinetic @xmath231 potential ) and @xmath232 is the fermi - dirac entropy ( see appendix [ sec_ef ] ) . for a completely degenerate gas at @xmath0 , stable configurations are those that minimize the energy @xmath202 at fixed mass . as we have seen previously , the first order variations @xmath233 determine the mass - radius relation . then , the configurations is stable if the energy is minimum , i.e. @xmath234 . now , we have @xmath235 for @xmath133 , all the solutions are stable while for @xmath8 , all the solutions are unstable . for @xmath236 , the solutions with @xmath237 are stable and the solutions with @xmath238 are unstable where @xmath239 we now show that the onset of instability precisely corresponds to the turning point of mass , i.e. to the point where the mass is maximum in the series of equilibria @xmath240 . in terms of reduced variables , this corresponds to @xmath241 , or equivalently @xmath242 . taking the logarithmic derivative of eq . ( [ ana12 ] ) , we have @xmath243 so that the condition @xmath242 yields @xmath244 . thus , instability sets in precisely at the maximum mass as could have been directly inferred from the turning point criterion @xcite . it is also useful to discuss the classical and the ultra - relativistic limits specifically . in the classical limit , using eq . ( [ ana11class ] ) , we get @xmath245 so that a classical white dwarf star is stable for @xmath89 and unstable for @xmath8 . in the ultra - relativistic limit , using eq . ( [ ana11relat ] ) , we get @xmath246 so that an ultra - relativistic white dwarf star is stable for @xmath133 and unstable for @xmath129 . this returns the results obtained in secs . [ sec_class ] and [ sec_ur ] . finally , considering the function @xmath247 given by eq . ( [ ana11 ] ) and assuming that the system evolves so as to minimize its energy ( this requires some source of dissipation ) , we have the following results : because they are elementary , but drawing them is helpful to visualize the results described in the text . ] for @xmath248 , there exists a stable equilibrium state ( global minimum of @xmath202 ) with radius @xmath63 for all mass @xmath102 . for @xmath4 , there exists a maximum mass @xmath13 so that : ( i ) for @xmath249 , there exists a stable equilibrium state ( global minimum of @xmath202 ) with radius @xmath250 ( ii ) for @xmath251 , the system collapses to a point ( @xmath167 ) but its energy remains finite ( lower bound ) ( iii ) for @xmath14 , the system collapses to a point @xmath252 and @xmath253 . for @xmath254 , there exists a maximum mass @xmath13 so that : ( i ) for @xmath249 , there exists a metastable equilibrium state ( local minimum of @xmath202 ) with @xmath237 and an unstable equilibrium state ( global maximum of @xmath202 ) with @xmath238 . the system can either reach the metastable state or collapse to a point ( @xmath252 , @xmath253 ) ; the choice probably depends on a notion of basin of attraction ( ii ) for @xmath255 , the system collapses to a point ( @xmath252 , @xmath253 ) . for @xmath256 , there exists a critical mass @xmath13 so that : ( i ) for @xmath249 , there exists an unstable equilibrium state ( global maximum of @xmath202 ) so the system either collapses ( @xmath252 and @xmath253 ) or evaporates ( @xmath175 and @xmath257 ) ( ii ) for @xmath14 , the system collapses to a point ( @xmath252 and @xmath253 ) . for @xmath258 , there exists an unstable equilibrium state ( global maximum of @xmath202 ) for all mass @xmath102 so the system either collapses ( @xmath252 and @xmath253 ) or evaporates ( @xmath175 and @xmath257 ) . _ `` why is our universe three dimensional ? does the dimension @xmath4 play a special role among other space dimensions ? '' _ several scientists have examined the role played by the dimension of space in determining the form of the laws of physics . this question goes back to ptolemy who argues in his treatise _ on dimensionality _ that no more than three spatial dimensions are possible in nature . in the @xmath259 century , kant realizes the deep connection between the inverse square law of gravitation and the existence of three spatial dimensions . interestingly , he regards the three spatial dimensions as a consequence of newton s inverse square law rather than the converse . in the twentieth century , ehrenfest @xcite , in a paper called `` in what way does it become manifest in the fundamental laws of physics that space has three dimensions ? '' argues that planetary orbits , atoms and molecules would be unstable in a space of dimension @xmath8 . this idea has been followed more recently by gurevich & mostepanenko @xcite who argue that if the universe is made of metagalaxies with various number of dimensions , atomic matter and life are possible only in @xmath260-dimensional space . other investigations on dimensionality are reviewed in the paper of barrow @xcite . we have found that the relativistic self - gravitating fermi gas at @xmath0 ( a white dwarf star ) possesses a rich structure as a function of the dimension of space . we have exhibited several characteristic dimensions @xmath2 , @xmath3 , @xmath4 , @xmath5 , @xmath6 and @xmath7 . for @xmath248 , there exists stable configurations for any value of the mass . for @xmath4 , the sequence of equilibrium configurations is stable but there exists a maximum mass ( chandrasekhar s limit ) above which there is no equilibrium state . for @xmath254 , the sequence of equilibrium configurations is stable for classical white dwarf stars but it becomes unstable for relativistic white dwarf stars with high density after the turning point of mass . therefore , the dimension @xmath4 is special because it is the largest dimension at which the sequence of equilibrium configurations is stable all the way long ; for @xmath129 , a turning point of mass appears so that ultra - relativistic white dwarf stars become unstable . therefore , the dimension @xmath4 is _ marginal _ in that respect . finally , for @xmath8 , the whole sequence of equilibrium configurations is unstable . therefore , as already noted in our previous papers @xcite , the dimension @xmath6 is critical because at that dimension quantum mechanics can not stabilize matter against gravitational collapse , even in the classical regime , contrary to the situation in @xmath4 @xcite . interestingly , this result is similar to that of ehrenfest @xcite although it applies to white dwarf stars instead of atoms . our exact description of @xmath1-dimensional white dwarf stars based on chandrasekhar s seminal paper @xcite shows that relativistic white dwarf stars become unstable in @xmath129 and that classical white dwarf stars become unstable in @xmath8 . therefore , for @xmath8 , a self - gravitating fermi gas forms a black hole or evaporates ( see appendix [ sec_stab ] ) . these conclusions have also been reached by bechhoefer & chabrier @xcite on the basis of simple dimensional analysis . this suggests that a @xmath8 universe is not viable ( see also appendix [ sec_fourd ] ) and gives insight why our universe in apparently three dimensional . we note that extra - dimensions can appear at the micro - scale , an idea originating from kaluza - klein theory . this idea took a renaissance in modern theories of grand unification which are formulated in higher - dimensional spaces . cm ) and it is a theoretical challenge to explain why only three dimensions are expanded while the others are compact . we note that a universe with @xmath129 dimensions would be very different from ours since compact objects like white dwarfs or neutron stars would be unstable and replaced by black holes . in fact , _ all the matter would collapse to a single point in a finite time _ ( see appendix [ sec_fourd ] ) . this observation may be connected to the fact that only three dimensions are extended . ] our approach shows that already at a simple level , the coupling between newton s equations ( gravitation ) , fermi - dirac statistics ( quantum mechanics ) and special relativity reveals a rich structure as a function of @xmath1 . in this respect , it is interesting to note that the critical masses ( [ class10 ] ) ( [ ur12 ] ) and radii ( [ class11 ] ) ( [ ur9 ] ) that we have found occur for simple _ integer _ dimensions @xmath261 and @xmath262 , which was not granted a priori . it is interesting to develop a parallel between the mass - radius relation @xmath240 of white dwarf stars and the caloric curve @xmath263 giving the temperature as a function of the energy in the thermodynamics of self - gravitating systems @xcite . in this analogy , the chandrasekhar mass in @xmath4 is the counterpart of the critical temperature for isothermal systems in @xmath3 ( in both cases , the equilibrium density profile is a dirac peak containing all the mass ) . for relativistic white dwarf stars in @xmath4 ( corresponding to a polytrope @xmath264 ) , the critical mass @xmath265 of bacterial populations described by the keller - segel model in @xmath3 ( corresponding to @xmath266 ) and the critical mass @xmath267 or critical temperature @xmath268 of self - gravitating isothermal systems in @xmath3 ( corresponding to @xmath266 ) @xcite . at these critical values @xmath251 or @xmath269 , the system forms a dirac peak . this analogy , sketched in sec . 8 of astro - ph/0604012v1 , is addressed specifically in @xcite . ] on the other hand , for @xmath129 , the mass - radius relation for white dwarf stars exhibits turning points , and even a spiraling behavior for @xmath270 , which is similar to the spiraling behavior of the caloric curve for isothermal systems in @xmath4 . in this analogy , the maximum mass , corresponding to a critical value of the central density which parameterizes the series of equilibria , is the counterpart of the antonov energy ( in the microcanonical ensemble ) @xcite or of the emden temperature ( in the canonical ensemble ) @xcite . the series of equilibria becomes unstable after this turning point . in addition , there is no equilibrium state above this maximum mass , or below the minimum energy or minimum temperature in the thermodynamical problem . in that case , the system is expected to undergo gravitational collapse . as a last comment ( notified by the referee ) , it should be emphasized that the conclusions reached in this paper concerning dimensionality implicitly assume that the laws of physics that we know remain the same in a universe of arbitrary dimension @xmath1 . this is of course not granted at all . there may be a new cosmological theory , a new theory of star formation and stellar evolution in higher dimensions . we also emphasize that our approach does not take into account general relativistic effects that can sensibly modify the results @xcite . _ acknowledgements : _ i am grateful to e. blackman and g. chabrier for pointing out their works after a first version of this paper was placed on astro - ph/0603250 . special thanks are due to n. barros for interesting discussions . we generalize in @xmath1 dimensions the usual stability criteria for polytropic gaseous spheres , and apply them to classical and ultra - relativistic white dwarf stars . let us consider the euler - poisson system @xcite : @xmath275 @xmath276 @xmath277 describing the dynamical evolution of a barotropic gas with an equation of state @xmath278 . the euler - poisson system conserves the mass @xmath102 and the energy @xcite : @xmath279 in the following , we shall essentially consider the case of a polytropic equation of state @xmath280 where @xmath281 . for @xmath282 , the energy functional @xmath283 is given by @xmath284 it can be rewritten @xmath285 for an isothermal equation of state @xmath286 , corresponding to @xmath287 or @xmath215 , we have @xmath288 for a general polytropic equation of state @xmath289 with arbitrary @xmath290 , it is convenient to write the energy of the gas in the form @xmath291 we have added a constant term @xmath292 proportional to the total mass ( which is a conserved quantity ) so as to recover the energy ( [ qw3 ] ) of an isothermal gas in the limit @xmath287 @xcite . we first consider the linear dynamical stability of a polytropic star described by the euler - poisson system and generalize the approach developed in chavanis @xcite . we consider a steady solution of the euler - poisson system satisfying @xmath293 and the condition of hydrostatic balance @xmath294 . for a polytropic equation of state @xmath295 , the equilibrium density profile is solution of the @xmath1-dimensional lane - emden equation @xcite . we consider _ complete polytropes _ such that the density vanishes at a finite radius @xmath63 . for @xmath296 , there exists a unique steady state for any mass @xmath102 . the mass - radius relation is @xcite : @xmath297 where @xmath298 is a constant given in terms of the solution @xmath299 of the lane - emden equation of index @xmath186 in @xmath1 dimensions . for the critical index @xmath300 , steady state solutions exist only for a unique value of the mass @xmath301 these solutions have the same mass @xmath13 but an arbitrary radius @xmath63 . we already anticipate that the index @xmath300 will correspond to a case of marginal stability separating stable and unstable solutions . linearizing the equations of motion ( [ stab1])-([stab3 ] ) around a stationary solution in hydrostatic balance and writing the perturbation in the form @xmath302 , we obtain after some calculations @xcite the eigenvalue equation @xmath303 where we have restricted ourselves to spherically symmetric perturbations and defined @xmath304 . for a polytropic gas with an equation of state @xmath289 , the foregoing equation becomes @xcite : @xmath305 the polytrope is stable if @xmath306 ( yielding oscillatory modes with pulsation @xmath307 ) and unstable if @xmath308 ( yielding exponentially growing modes with growth rate @xmath309 ) . considering the point of marginal stability ( @xmath310 ) and introducing the emden variables @xcite , eq . ( [ stab5 ] ) reduces to @xmath311 this equation has the exact solution @xcite : @xmath312 the point of marginal stability is then determined by the boundary conditions . one can show @xcite that the velocity perturbation is given by @xmath313 . therefore , if the density of the configuration vanishes at @xmath314 , one must have @xmath315 to avoid unphysical divergences . thus , if @xmath316 denotes the value of the normalized radius @xmath63 of the star @xcite such that @xmath317 , the natural boundary condition for the eigenvalue equation ( [ stab5 ] ) is @xmath318 . substituting this condition in eq . ( [ stab7 ] ) , we obtain the critical index @xmath319 corresponding to a marginally stable gaseous polytrope ( @xmath310 ) . our method gives the form ( [ stab7 ] ) of the neutral perturbation @xmath320 at the critical index @xmath300 . we conclude that the infinite family of steady state solutions with equal mass @xmath13 and arbitrary radius @xmath63 that exists for @xmath300 is marginally stable . we now consider the nonlinear dynamical stability problem . since the euler - poisson system ( [ stab1])-([stab3 ] ) conserves the mass @xmath102 and the energy @xmath283 , a maximum or a minimum of the energy functional @xmath321 $ ] at fixed mass @xmath322=m$ ] determines a steady state of the euler - poisson system that is formally nonlinearly dynamically stable @xcite . because of the presence of the kinetic term @xmath323=(1/2)\int \rho { \bf u}^{2}d{\bf r}$ ] , the functional @xmath321 $ ] has no absolute maximum . thus , we need to investigate the possible existence of a minimum of @xmath324 $ ] at fixed mass @xmath322=m$ ] . a barotropic star that minimizes the energy functional @xmath283 at fixed mass @xmath102 is nonlinearly dynamically stable with respect to the euler - poisson system . the cancellation of the first order variations @xmath325 , where @xmath326 is a lagrange multiplier , yields @xmath293 and the condition of hydrostatic balance @xmath294 . then , the condition of nonlinear dynamical stability is @xmath327 for all perturbations that conserve mass , i.e. @xmath328 . after some calculations @xcite , this can be put in the quadratic form @xmath329 we are led therefore to consider the eigenvalue problem @xmath330 if all the eigenvalues @xmath331 are negative , then @xmath332 and the configuration is a minimum of @xmath283 at fixed mass . this implies that it is nonlinearly dynamically stable . if at least one eigenvalue @xmath331 is positive , the configuration is a saddle point of @xmath283 and the star is dynamically unstable . the marginal case is when the largest eigenvalue @xmath331 is equal to zero . now , for @xmath310 , eqs . ( [ fdf2 ] ) and ( [ stab4 ] ) coincide . this implies that the conditions of linear and nonlinear dynamical stability are the same . in the case of polytropic stars , the case of marginal stability corresponds to the critical index @xmath300 . at that index , the equilibrium configurations with mass @xmath13 and radius @xmath63 _ all _ have the same value of energy @xmath333 ( see eq . ( [ stab22 ] ) later ) . this is therefore a very degenerate situation . the nonlinear dynamical stability of gaseous polytropes can also be investigated by plotting the series of equilibria @xmath334 ( mass vs central density ) of box - confined configurations and using the turning point argument of poincar @xcite . we can thus determine whether the last point on the series of equilibria , which corresponds to a _ complete polytrope _ whose density vanishes precisely at the box radius , is stable or not . for @xmath95 , there is no turning point of mass ( we restrict ourselves to @xmath335 ) , implying that the gaseous polytropes are always stable . for @xmath90 , a turning point of mass @xmath334 appears precisely for @xmath336 ( see fig . 5 of @xcite ) . this method shows that , for @xmath90 , complete polytropes with @xmath337 are nonlinearly dynamically stable ( they are minima of @xmath283 at fixed mass @xmath102 ) while complete polytropes with @xmath338 are dynamically unstable ( they are saddle points of @xmath283 at fixed mass @xmath102 ) . complete polytropes with @xmath336 and @xmath251 are marginally stable . we can also investigate the linear dynamical stability of gaseous polytropic spheres by using the method introduced by eddington @xcite and ledoux @xcite . if we introduce the radial displacement @xmath339 we can rewrite the eigenvalue equation ( [ stab5 ] ) in the form @xcite : @xmath340 this is the eddington equation of pulsations , which has been written here in the form of a sturm - liouville problem . it must be supplemented by the boundary conditions @xmath341 @xmath342 where @xmath343 is the lagrangian derivative of the pressure . since @xmath344 at the surface of the star , it is sufficient to demand that @xmath345 and @xmath346 be finite in @xmath314 . multiplying eq . ( [ stab10 ] ) by @xmath345 and integrating between @xmath347 and @xmath63 , we obtain @xmath348 the system is linearly dynamically stable if @xmath349 and unstable otherwise . since @xmath350 , a sufficient condition of stability is @xmath351 , i.e. @xmath352 it can be shown furthermore ( see below ) that the system is unstable if @xmath353 so that the criterion ( [ stab12 ] ) is a necessary and sufficient condition of dynamical stability ( the case @xmath354 is marginal ) . in terms of the index @xmath186 , a complete polytrope is stable with respect to the euler - poisson system in @xmath2 for @xmath335 and for @xmath355 , in @xmath3 for @xmath356 , in @xmath4 for @xmath357 and in @xmath90 for @xmath358 . from the theory of sturm - liouville problems , it is known that expression ( [ stab11 ] ) , which can be written @xmath359 , forms the basis of a variational principle . the function @xmath360 which maximizes the functional @xmath361 is the fundamental eigenfunction and the maximum value of this functional gives the fundamental eigenvalue @xmath362 . furthermore , any trial function under - estimates the value of @xmath362 so this variational principle may prove the existence of instability but can only give approximate information concerning stability . as shown by ledoux & pekeris @xcite , we can get a good approximation of the fundamental eigenvalue by taking @xmath360 to be a constant ( note that @xmath363 , i.e. @xmath364 , is the _ exact _ solution of the sturm - liouville equation ( [ stab10 ] ) at the point of marginal stability @xmath310 for a polytropic equation of state ) . for the trial function @xmath365 , expression ( [ stab11 ] ) gives @xmath366 according to the condition of hydrostatic balance , @xmath367 we have @xmath368 where we recognize the virial @xmath369 inserting @xmath370 in eq . ( [ stab16 ] ) , interchanging the dummy variables @xmath371 and @xmath372 and taking the half - sum of the resulting expressions , we get @xmath373 therefore , the virial can be written @xmath374 @xmath375 where @xmath376 is the potential energy and @xmath377 is the gravitational potential . therefore , we can rewrite eq . ( [ stab13 ] ) in the form @xmath378 where @xmath379 is the moment of inertia . since @xmath380 , we conclude that the system is unstable if @xmath381 , which completes the proof above . on the other hand , eq . ( [ stab17 ] ) provides an estimate of the pulsation period @xmath307 when the system is stable . this is the @xmath1-dimensional generalization of the ledoux stability criterion . note finally that , for a spherically symmetric system , using the gauss theorem @xmath382 we get @xmath383 this expression may be useful to calculate the potential energy . the virial theorem associated with the barotropic euler - poisson system ( [ stab1])-([stab3 ] ) is @xcite : @xmath384 where @xmath385 is the kinetic energy of the macroscopic motion and @xmath386 . for a polytropic equation of state , the energy functional ( [ wkre ] ) , which is a conserved quantity , can be written @xmath387 where @xmath388 is the internal energy . we note that @xmath389 for a polytrope . at equilibrium , @xmath390 , we get @xmath391 and @xmath392 . for @xmath3 , this implies that @xmath393 . for @xmath394 , this implies that @xmath395 where we recall that @xmath396 for @xmath90 and @xmath335 , the system is dynamically stable if @xmath397 and unstable otherwise ( poincar argument @xcite ) . since @xmath398 , we find from eq . ( [ stab22 ] ) that the polytropic star is stable if and only if @xmath399 . at the point of marginal stability @xmath300 , where several steady configurations exist with the same mass @xmath13 given by eq . ( [ dim13b ] ) and an arbitrary radius @xmath63 , the energy of these configurations is @xmath333 for all @xmath63 . using @xmath400 and eliminating the pressure term in eq . ( [ vir1 ] ) , the virial theorem for @xmath184 can be put in the form @xmath401 alternatively , eliminating the kinetic energy in eq . ( [ vir1 ] ) , the virial theorem for @xmath184 can be written @xmath402 for @xmath403 , corresponding to classical white dwarf stars , the virial theorem becomes @xmath404 for @xmath405 , corresponding to relativistic white dwarf stars , we have @xmath406 finally , considering the polytropic index @xmath300 of marginal stability , the virial theorem ( [ vir2 ] ) takes the form @xmath407 if we consider the dimension @xmath6 , it reduces to @xmath408 this equation describes the case of classical white dwarf stars at the critical dimension @xmath6 where @xmath409 . it can be integrated into @xmath410 for @xmath411 , we find that @xmath412 for @xmath413 so that the system evaporates . alternatively , for @xmath414 , we find that @xmath415 in a finite time , so that the system collapses and forms a dirac peak in a finite time . if we consider the dimension @xmath4 , eq . ( [ vir3 ] ) reduces to @xmath416 this equation describes the case of relativistic white dwarf stars at the critical dimension @xmath4 where @xmath417 . for @xmath411 , we find that @xmath412 for @xmath413 indicating that the system evaporates . finally , we show that the instability criterion for polytropic stars can be obtained from simple dimensional analysis . we shall approximate the internal energy ( [ vir1c ] ) and the potential energy ( [ stab19b ] ) by @xmath418 @xmath419 where @xmath102 is the total mass of the configuration and @xmath63 its radius ( @xmath183 and @xmath181 are dimensionless parameters ) . for homogeneous spheres , the values of @xmath181 and @xmath183 are given by eq . ( [ dim1c ] ) . with these expressions , the energy functional ( [ qw1 ] ) becomes @xmath420 we now need to minimize this functional with respect to @xmath63 for a given mass @xmath102 ( a minimum necessarily requires @xmath421 ) . we first look for the existence of critical points ( extrema ) . the cancellation of the first order variations @xmath422 yields the mass - radius relation @xmath423 this relation determines the radius @xmath63 of the star as a function of its mass @xmath102 . this expression is consistent with the exact mass - radius relation ( [ dim13 ] ) deduced from the lane - emden equation ( of course , our simple approach can only model compact density profiles corresponding to @xmath424 for @xmath90 ) @xcite . for @xmath296 , there is one , and only one , extremum of @xmath425 for each mass @xmath102 . in order to have a true minimum , we need to impose @xmath426 therefore , the system is stable if @xmath427 and unstable otherwise . this simple dimensional analysis returns the exact stability criterion ( [ stab12 ] ) . we can be a little more precise . for @xmath428 , the functional @xmath425 has a global minimum reached for a finite , and non - zero , value of @xmath63 . this solution is stable . for @xmath429 , the functional @xmath425 has an unstable global maximum : for @xmath103 , @xmath430 when @xmath175 ( evaporation ) and @xmath431 when @xmath252 ( collapse ) ; for @xmath3 , @xmath430 when @xmath252 and @xmath175 ; for @xmath90 and @xmath432 , @xmath430 when @xmath252 and @xmath433 ; for @xmath90 and @xmath434 , @xmath430 when @xmath252 and @xmath431 when @xmath433 . since we have constructed a particular configuration ( homogeneous sphere ) which makes the energy @xmath283 diverge to @xmath435 , the above arguments prove that the exact functional @xmath436 $ ] given by eq . ( [ wkre ] ) has no absolute minimum at fixed mass for @xmath429 . since the functional @xmath324 $ ] has only one critical point ( cancelling the first variations ) at fixed mass , we conclude that , for @xmath429 , this critical point is a saddle point . therefore , there is no stable steady state of polytropic spheres for @xmath429 . for the critical index @xmath300 , the relation ( [ dim3 ] ) shows the existence of a critical mass @xmath437 the functional ( [ dim1 ] ) can be rewritten @xmath438 for @xmath251 , @xmath439 for all @xmath63 ( at equilibrium @xmath421 ) . this determines an infinite family of solutions with the same mass @xmath13 and different radii . these solutions have the same energy and are marginally stable . this returns the result of the exact model where the configurations are solution of the lane - emden equation ( see sec . [ sec_eigen ] ) . for @xmath249 , the function @xmath425 is monotonically decreasing with @xmath63 ( so the system tends to evaporate ) : for @xmath103 , @xmath425 goes from @xmath347 to @xmath435 and for @xmath90 , @xmath425 goes from @xmath440 to @xmath347 . for @xmath14 , the function @xmath425 is monotonically increasing with @xmath63 ( so the system tends to collapse ) : for @xmath103 , @xmath425 goes from @xmath347 to @xmath440 and for @xmath90 , @xmath425 goes from @xmath435 to @xmath347 . we can obtain a simple dynamical model by using the virial theorem ( [ vir1 ] ) . at equilibrium ( @xmath390 ) , we have @xmath441 inserting the expressions ( [ dim1a ] ) and ( [ dim1b ] ) in eq . ( [ dim6 ] ) , we recover the mass - radius relation ( [ dim3 ] ) . for @xmath296 and for any given mass @xmath102 , there is only one steady state . its radius @xmath106 given by eq . ( [ dim3 ] ) and its energy @xmath442 is given by eq . ( [ stab22 ] ) . it corresponds to the extremum value of @xmath425 . estimating the moment of inertia by @xmath443 and inserting the expressions ( [ dim1a ] ) and ( [ dim1b ] ) in eq . ( [ vir2b ] ) , we obtain @xmath444 this equation determines the evolution of the radius of the star for a fixed mass @xmath102 and a fixed energy @xmath283 . the evolution of the kinetic energy is then given by @xmath445 . the solution of eq . ( [ dim8 ] ) depends on two control parameters @xmath102 and @xmath283 and on the initial condition @xmath446 and @xmath447 . we shall consider the case where @xmath283 is equal to the value @xmath442 corresponding to the steady state , such that the r.h.s . of the above equation is equal to zero at equilibrium . then , we can rewrite eq . ( [ dim8 ] ) as @xmath448 where @xmath106 is the radius of the star at equilibrium . since the dynamics is non - dissipative , in the euler equation ( [ stab2 ] ) @xcite . in that case @xmath449 , so that the system evolves so as to minimize the functional @xmath321 $ ] at fixed mass . ] the system does not evolve towards the minimum of @xmath425 ( unless @xmath450 initially ) . the system can either oscillate around the minimum ( stable ) or evolve away from it ( unstable ) . considering small perturbations around equilibrium , writing @xmath451 with @xmath452 , and linearizing the foregoing equation , we obtain @xmath453 where we have used the mass - radius relation ( [ dim3 ] ) to simplify the expression . this is the equation for a harmonic oscillator with pulsation @xmath454 the star is stable if @xmath455 and unstable otherwise . this returns the exact stability criterion ( [ stab12 ] ) . furthermore , using eqs . ( [ dim1b ] ) and ( [ dim7 ] ) , the pulsation can be rewritten in the form @xmath456 which exactly coincides with the ledoux formula ( [ stab17 ] ) . therefore , our simple dimensional model allows to obtain a lot of interesting results . the case of arbitrary perturbations around equilibrium will be considered in a future work . finally , we note that for @xmath300 , eq . ( [ dim8 ] ) becomes @xmath457 and eq . ( [ dim9 ] ) corresponds to @xmath333 and @xmath251 yielding @xmath458 . hence , @xmath459 corresponding to a marginal evolution . in our dimensional analysis , the constants @xmath181 , @xmath183 and @xmath326 are dimensionless parameters that could be chosen to fit the exact results at best . alternatively , we can try to obtain quantitative predictions by calculating their values for homogeneous spheres . this yields @xmath460 note that the potential energy of a homogeneous sphere in @xmath1 dimensions can be easily calculated from eq . ( [ stab19e ] ) . on the other hand , in @xmath3 , a direct calculation gives @xmath461 where @xmath462 is a reference radius where @xmath463 . all the results given above pass to the limit @xmath185 provided that we take @xmath464 . using eq . ( [ dim1c ] ) , we find that the approximate value of the pulsation ( [ dim11 ] ) becomes @xmath465 on the other hand , if we compare the approximate mass - radius relation ( [ dim3 ] ) with the exact mass - radius relation ( [ dim13 ] ) , we obtain an estimate of the constant @xmath466 in the form @xmath467 this has to be compared with the exact value ( [ dim15 ] ) given in terms of the solution @xmath299 of the emden equation of index @xmath186 in @xmath1 dimensions @xcite . let us consider the case @xmath4 . for @xmath10 , corresponding to classical white dwarf stars , we find @xmath468 instead of the exact value @xmath469 . for @xmath12 , corresponding to relativistic white dwarf stars , we find @xmath470 instead of the exact value @xmath471 this suggests that the homogeneous star model will provide a fair description of relativisitic white dwarf stars and a poorer description of classical white dwarf stars . we shall come back to these different issues ( static and dynamics ) in a future work . in this appendix , we show that the condition of thermodynamical stability in the canonical ensemble is equivalent to the condition of nonlinear dynamical stability with respect to the barotropic euler - poisson system . we apply this result to white dwarf stars . we consider a barotropic gas with an equation of state @xmath278 described by the euler - poisson system @xcite . we introduce the energy functional @xmath472 the first term @xmath473 is the work @xmath474 done in compressing the system from infinite dilution , the second term @xmath475 is the gravitational energy and the third term @xmath476 is the kinetic energy associated with the mean motion . for a gas in local thermodynamic equilibrium , the equation of state is @xmath477 or @xmath478 and the first law of thermodynamics can be written @xmath479 where @xmath480 is the entropy density and @xmath481 is the density of internal energy . for a gas without interaction ( apart from the long - range gravitational attraction ) , the internal energy is equal to the kinetic energy . there are two important cases where the gas is barotropic . the first case is when @xmath482 . this concerns either adiabatic ( or isentropic ) fluids so that @xmath483 or fluids at zero temperature so that @xmath0 . when @xmath482 , the first law of thermodynamics reduces to @xmath484 . it can be integrated into @xmath485d\rho'$ ] . then , the work @xmath473 done by the pressure force ( first term in eq . ( [ ef1 ] ) ) coincides with the internal energy @xmath486 of the gas . in that case , we get @xmath487 and the energy functional ( [ ef1 ] ) can be written @xmath488 thus , at @xmath0 or for an adiabatic evolution , the total energy of the gas @xmath202 is conserved by the euler - poisson system ( since @xmath283 is conserved ) . alternatively , for an isothermal gas @xmath489 , the first law of thermodynamics @xmath479 can be written @xmath490 . it can be integrated into @xmath491d\rho'$ ] . therefore , the work @xmath473 done by the pressure force ( first term in eq . ( [ ef1 ] ) ) coincides with the free energy @xmath492 of the gas . in that case , we get @xmath493 and the energy functional ( [ ef1 ] ) can be written @xmath494 thus , for an isothermal evolution , the free energy of the system @xmath495 is conserved by the euler - poisson system ( since @xmath283 is conserved ) . at @xmath0 , we recover the conservation of the energy @xmath202 . let us apply these results to white dwarf stars . we can view a white dwarf star as a barotropic gas described by an equation of state @xmath278 . according to the discussion of appendix [ sec_eigen ] ( see also @xcite ) , it is nonlinearly dynamically stable with respect to the euler - poisson system if it is a minimum of the energy functional @xmath283 at fixed mass . if a minimum exists , it is necessary that @xmath496 . as a result , a steady state of the euler - poisson system is nonlinearly dynamically stable if , and only if , it is a minimum of @xmath497={\cal w}[\rho,{\bf u}]-\theta[{\bf u}]$ ] at fixed mass @xmath102 . in conclusion , the condition of nonlinear dynamical stability can be written @xmath498\quad |\quad m[\rho]=m\rbrace . \label{ef4}\ ] ] for a white dwarf star at zero temperature ( @xmath0 ) , according to eqs . ( [ ef2 ] ) and ( [ ef3 ] ) , the functional @xmath499 reduces to the energy @xmath500 , where @xmath486 is the kinetic energy . we can also view a white dwarf star as a gas of self - gravitating relativistic fermions at statistical equilibrium in the canonical ensemble . it is thermodynamically stable if , and only if , it is a minimum of free energy @xmath495 at fixed mass @xmath102 . the free energy is given by @xmath501 where @xmath502 is the maximum value of the distribution function fixed by the pauli exclusion principle . for a white dwarf star at zero temperature ( @xmath0 ) , the free energy @xmath495 reduces to the energy @xmath500 . the condition of thermodynamical stability in the canonical ensemble can be written @xmath503\quad |\quad m[f]=m\rbrace . \label{ef6}\ ] ] the critical points of free energy at fixed mass , determined by the variational principle @xmath504 , correspond to the relativistic mean - field fermi - dirac distribution @xmath505 where @xmath506 is a lagrange multiplier ( related to @xmath326 ) determined by the mass @xmath102 and @xmath28 is given by eq . ( [ es3 ] ) . using eqs . ( [ es1 ] ) , ( [ es4 ] ) and ( [ ef7 ] ) , the density and the pressure are of the form @xmath507 and @xmath508 . eliminating @xmath509 between these two expressions , we find that @xmath278 so that the gas is barotropic . the equation of state is parameterized by @xmath229 and is fully determined by the entropic functional in eq . ( [ ef5 ] ) , which is here the fermi - dirac entropy . at @xmath0 we obtain the explicit relations of sec . [ sec_es ] . furthermore , the condition that @xmath510 is a function of the energy @xmath511 implies that the corresponding barotropic gas is at hydrostatic equilibrium . indeed , using eq . ( [ es5 ] ) , @xmath512 so that @xmath513 the relativistic mean - field fermi - dirac distribution ( [ ef7 ] ) is just a critical point of free energy at fixed mass @xmath102 . to determine whether it corresponds to a true minimum of free energy , we can proceed in two steps . we first minimize @xmath514 $ ] for a fixed density profile @xmath515 . since the potential energy @xmath516 $ ] and the mass @xmath322 $ ] are entirely determined by the density profile , this is equivalent to minimizing @xmath517-ts[f]$ ] at fixed @xmath515 , where @xmath486 is the kinetic energy . this gives a distribution @xmath518 where @xmath519 is a local lagrange multiplier determined by the density @xmath515 , using @xmath520 . since @xmath521 , the distribution ( [ ef8 ] ) is a true minimum of @xmath514 $ ] at fixed @xmath515 . substituting the optimal distribution function ( [ ef8 ] ) in eq . ( [ ef5 ] ) , we obtain a functional @xmath522\equiv f[\tilde{f}]$ ] of the density @xmath515 . using eqs . ( [ es1 ] ) , ( [ es4 ] ) and ( [ ef8 ] ) , we note that the density and the pressure are of the form @xmath523 and @xmath524 . eliminating @xmath519 between these expressions , we find that @xmath278 where the equation of state is the same as the one determined from the fermi - dirac distribution ( [ ef7 ] ) at equilibrium . now , we can show that @xmath525=\tilde{\cal w}[\rho ] , \label{ef9}\end{aligned}\ ] ] where @xmath526 $ ] is the functional defined in sec . [ sec_efa ] . the relation ( [ ef9 ] ) can be obtained by an explicit calculation which extends the proof given in appendix b of @xcite for classical particles ( this will be shown in a future work ; see also particular cases in sec . [ sec_efc ] ) but it also results from a straightforward argument . we note that the distribution function ( [ ef8 ] ) corresponds to a condition of local thermodynamical equilibrium with uniform temperature and zero average velocity . thus , it locally satisfies the first law of thermodynamic @xmath479 with @xmath489 . integrating this relation like in sec . [ sec_efa ] , we find that @xmath527-ts[\rho]={\cal w}_{1}[\rho]$ ] where @xmath527=u[\tilde{f}]$ ] and @xmath528=s[\tilde{f}]$ ] . this directly yields the identity ( [ ef9 ] ) . we can now conclude that @xmath514 $ ] has a minimum @xmath529 at fixed mass @xmath102 if , and only if , @xmath522=\tilde { \cal w}[\rho]$ ] has a minimum @xmath530 at fixed mass @xmath102 . in that case , @xmath529 is given by eq . ( [ ef8 ] ) where @xmath531 is determined by @xmath530 , writing @xmath532 . therefore , a system is thermodynamically stable in the canonical ensemble if , and only if , the corresponding barotropic gas with the same equilibrium distribution is nonlinearly dynamically stable with respect to the barotropic euler - poisson system . said differently , a system that minimizes the functional ( [ ef9 ] ) is ( i ) thermodynamically stable in the canonical ensemble and ( ii ) nonlinearly dynamically stable with respect to the barotropic euler - poisson system . this result applies to white dwarf stars at arbitrary temperature . although the above results are general , it may be useful to explicitly compute the functionals ( [ ef1 ] ) and ( [ ef5 ] ) for white dwarf stars at @xmath0 and check the relation ( [ ef9 ] ) . if we view a white dwarf star as a barotropic gas described by an equation of state @xmath533 , its energy @xmath499 can be written @xmath534 with @xmath535 a white dwarf star is nonlinearly dynamically stable with respect to the euler - poisson system if it is a minimum of @xmath499 at fixed mass @xmath102 . at @xmath0 , the equation of state is given by eq . ( [ es7 ] ) . in the classical limit ( polytrope @xmath87 ) we have @xmath536 and in the ultra - relativisitic limit ( polytrope @xmath127 ) we have @xmath537 in the general case , ( [ es7 ] ) , the function ( [ ef11 ] ) can be written @xmath538 integrating by parts and using eqs . ( [ es8 ] ) and ( [ es9 ] ) , we find after straightforward calculations that @xmath539 alternatively , we can view a white dwarf star as a gas of self - gravitating relativistic fermions at statistical equilibrium in the canonical ensemble . at @xmath0 , its free energy @xmath230 coincides with its energy @xmath202 . using eq . ( [ ef5 ] ) , it is given by @xmath540 with @xmath541 a white dwarf star at @xmath0 is thermodynamically stable if it is a minimum of the energy @xmath202 at fixed mass @xmath102 . in the classical limit , @xmath542 and @xmath543 so @xmath544 therefore , the free energy can be written @xmath545 using eq . ( [ class1 ] ) , this is equivalent to eq . ( [ ef12 ] ) . in the ultra - relativistic limit , @xmath546 and @xmath547 so @xmath548 . therefore , the free energy can be written @xmath549 using eq . ( [ ur1 ] ) , this is equivalent to eq . ( [ ef13 ] ) . in the general case , we have @xmath550 using eq . ( [ es3 ] ) , it can be put in the form @xmath551 now , it is straightforward to check that the two expressions in the r.h.s . ( [ ef15 ] ) and ( [ ef21 ] ) are equal so that @xmath552 implying the relation ( [ ef9 ] ) . finally , if we consider a classical isothermal self - gravitating gas at temperature @xmath229 with an equation of state @xmath286 , its free energy ( [ ef5 ] ) can be written @xmath553 comparing with eq . ( [ qw3 ] ) , we find that relation ( [ ef9 ] ) is indeed satisfied . therefore , a classical isothermal self - gravitating gas that minimizes the functional ( [ qw3 ] ) or ( [ ef22 ] ) at fixed mass is ( i ) thermodynamically stable in the canonical ensemble and ( ii ) nonlinearly dynamically stable with respect to the barotropic euler - poisson system . we had already made this observation in @xcite . in @xcite , this result was extended to an arbitrary form of entropic functional , including the fermi - dirac entropy , for non - relativistic systems . the present paper shows that , due to relation ( [ ef9 ] ) , the equivalence between nonlinear dynamical stability with respect to the barotropic euler - poisson system and thermodynamical stability in the canonical ensemble is general . if we consider , in a @xmath1-dimensional universe , a hamiltonian system of self - gravitating classical point masses whose dynamics is described by the newton equations @xmath554 the scalar virial theorem reads @xcite : @xmath555 where @xmath176 is the kinetic energy and @xmath556 the trace of the potential energy tensor for the @xmath178-body system @xcite . for @xmath184 , using eq . ( [ stab18 ] ) and introducing the total energy @xmath198 , which is a constant of the motion for an isolated system , we have @xcite : @xmath557 we note that the dimension @xmath6 is critical . in that case , @xmath558 which yields after integration @xmath559 . for @xmath560 , @xmath561 indicating that the system evaporates . for @xmath562 , @xmath563 goes to zero in a finite time , indicating that the system forms a dirac peak ( `` black hole '' ) in a finite time . more generally , for @xmath8 , since @xmath564 , we have @xmath565 so that the system forms a dirac peak in a finite time if @xmath562 . therefore , self - gravitating systems with @xmath562 are not stable in a space of dimension @xmath8 . the study of the present paper indicates that this observation remains true if quantum ( pauli exclusion principle for fermions ) and relativistic effects are taken into account . in this sense , a universe with @xmath8 is not viable .

the concept of solvent mediated interaction among colloidal particles provides an invaluable tool for the physical understanding of the behavior of complex fluids . when two hard are immersed in a molecular solvent , the simple excluded volume constraint is enough to cause an effective attraction between them driven by purely entropic effects , as shown in the seminal work by asakura and oosawa ( ao ) @xcite . the depletion mechanism has been extensively studied , mostly in hard sphere fluids @xcite , and represents a very successful paradigm for the interpretation of colloidal aggregation . according to the ao prediction , which neglects correlations , the strength of attraction is proportional to the fluid density and its range coincides with the diameter of the solvent molecules . however , both numerical simulations and liquid state theories proved that the effective interaction acquires a remarkable structure when the density of solvent molecules is increased , due to the presence of coordination shells in liquids . attractive interactions between solvent molecules reduce particle accumulation near the walls giving rise to smoother solvent mediated forces @xcite . in the neighborhoods of the critical point of the fluid the critical casimir effect @xcite is expected to give rise to long range tails in the solvent mediated attraction between colloidal particles : the effective force between two bodies acquires the range of the correlation length of the underlying critical solvent and is described by a universal scaling law which depends on a very limited number of features ( e.g. the geometry of the bodies , the boundary conditions at the colloidal surfaces , the system dimensionality ) . this remarkable phenomenon has been indeed experimentally observed by use of advanced optical techniques in a lutidine solvent close to its critical point @xcite . numerical simulations have been performed for two representative models of solvent : a fluid and a fluid of hard spheres with three attractive patches . in both cases the emergence of a long range attractive tail in the effective interaction between two large spheres immersed in the solvent is clearly visible @xcite . an enhanced rate of aggregation among colloidal particles is expected to occur as a result of critical casimir effect , and indeed also this consequence has been directly observed experimentally @xcite . these investigations suggest a novel way to control colloidal aggregation by tuning the solvent temperature . critical casimir effect is due to the confinement of order parameter fluctuations induced by the presence of external boundaries . as such , it governs the long range properties of the effective interaction , while at short range excluded volume effects are still present , providing the essential ingredient for the occurrence of the depletion phenomenon . the interplay between these two quite different physical mechanisms which act at different has not been deeply investigated . numerical simulations showed that , on approaching the critical point , the effective force indeed loses the short range oscillations due to the liquid layering near the surface of the bodies and acquires a smooth attractive form @xcite but no theoretical study of this effect has been attempted yet . more generally , a thorough investigation of the form of solvent mediated interactions in the whole phase diagram of a correlated fluid has not been performed . such a study would clarify the emergence of the long range tails in the effective interactions on approaching the critical region , as well as the coexistence of the depletion mechanism , acting at short distances , and the critical casimir effect , governing the physics at large distances . in this paper we tackle this problem by examining the solvent mediated forces between two hard walls immersed in a hard core yukawa fluid . the theoretical investigation is performed in the region in order to exclude wetting phenomena , which deserve a separate analysis . preliminarily , we must choose the most appropriate theoretical tool for such a study : a natural choice is density functional theory ( dft ) , which proved accurate in the description of confined fluids . however , no specific implementation of dft has been investigated in the critical region and most of the liquid state theories , which provide the theoretical basis for the practical formulation of dft , give a poor description of the critical regime . in section ii we propose a weighted density approximation ( wda ) especially designed for being accurate also near the critical point . this theory is validated against other dft prescriptions as well as available numerical simulations in several states . in section iii this approach is applied to the critical region of a yukawa fluid : the density profiles and the solvent mediated force between two walls are determined by numerical minimization of the density functional . the development of long range tails in both the density profile and the effective force is quantitatively investigated , confirming the trends already shown in simulations . in section iv the universal properties of the effective interaction in the critical region are discussed . the emergence of scaling laws is investigated in the critical and regime . we also show that , within our approximate dft formulation , the critical casimir scaling function is deeply related to the universal bulk properties of the critical fluid . a simple prescription for the theoretical evaluation of the critical and casimir scaling function is compared to the numerical results and to available simulations in ising systems . section v offers some final comments ad perspectives . the purpose of this work is the investigation of the solvent mediated interaction between two planar parallel walls immersed in a classical fluid in a wide range of temperatures and densities , including the critical region . in order to achieve this goal we need an accurate description of the properties of a confined fluid . the most successful theoretical approach to study inhomogeneous systems is density functional theory @xcite . although alternative techniques , like integral equations or scaled particle theory , have been proposed @xcite , dft is generally considered to be the most accurate and versatile tool for dealing with inhomogeneous systems and has been applied in several frameworks : from the study of fluids in nanopores , to the structure of the liquid - vapor interface , to the theory of freezing ( see e.g. @xcite ) . according to mermin s extension @xcite of the hohenberg and kohn density functional theorem @xcite , the equilibrium density profile @xmath0 of a confined fluid can be found by minimizing a suitably defined density functional @xmath1 $ ] at fixed chemical potential @xmath2 and temperature @xmath3 . @xmath1 $ ] can be conveniently expressed in terms of the external potential @xmath4 and the intrinsic free energy functional @xmath5 $ ] as : @xmath6=\mathcal{f}[n(\bm{r})]-\int \mathrm{d}\bm{r}\ , n(\bm{r } ) \bigl ( \mu -\phi(\bm{r})\bigr ) . \label{eq : omega}\ ] ] the intrinsic free energy functional is exactly known only in the ideal gas limit @xmath7= \int \mathrm{d}\bm{r}\ , n(\bm{r})\bigl(\log \left({\lambda}^3 n(\bm{r})\right)-1 \bigr)\ ] ] ( here @xmath8 and @xmath9 is the thermal de broglie wavelength ) , while , in an interacting system , it is customary to separate this contribution , splitting @xmath5 $ ] as the sum of the ideal and the excess term @xmath10 $ ] : @xmath11=\mathcal{f}^{\mathrm{id } } [ n(\bm{r})]+\mathcal{f}^{\mathrm{ex}}[n(\bm{r})].\ ] ] several approximations for the excess part have been proposed over the years for dealing with specific problems . hard sphere fluids are successfully described by rosenfeld s fundamental measure theory ( fmt ) @xcite also at bulk densities close to the solid transition . even if the fmt is widely used and it its implementation is straightforward , at least for the planar geometry , rosenfeld s approximation for the excess free energy holds in principle only for fluids of purely hard particles of any shape . attractive contributions in the interparticle potential are generally added as a mean field perturbation to the reference hard sphere excess free energy functional . however , such an approach gives only qualitative predictions , particularly in the critical region , because it does not take into account correlations arising from the attractive tail of the potential . also the weighted density approximations ( wda ) by tarazona @xcite and curtin and ashcroft @xcite , whose prediction of the density profile of the hard sphere fluid are rather accurate , at least at moderate densities , includes the effects of an attractive tail in the potential only at the mean field level . furthermore , approaches based on a truncation of the gradient expansion of the excess free energy functional , such as the square - gradient approximation , can describe only the slowly varying part of the density profile but fail in dealing with the short range structure close to the wall . in order to describe the microscopic properties of the solvent mediated forces in the whole phase diagram of a fluid , including the liquid - vapor transition , we need an approximation of the excess free energy functional which , in the uniform limit , provides an accurate description of the homogeneous fluid both in the dense and in the critical regime . unfortunately such an implementation of dft has not been devised yet . a novel wda - based approximation for the excess free energy functional based on the hierachical reference theory of fluids @xcite will be now introduced . according to the weighted density approximation , the excess free energy functional is expressed in terms of a weighted density @xmath12 as @xmath13=\int \mathrm{d}\bm{r}\ , n(\bm{r}){\psi}^{\mathrm{ex } } \bigl(\bar{n}(\bm{r } ) \bigr ) , \label{eq : wda_excfreeenergy}\ ] ] where @xmath14 is the excess free energy per particle of the homogeneous system evaluated at bulk density @xmath15 and the weighted density @xmath12 is written as a local average of the density profile , in terms of an isotropic weight function @xmath16 : @xmath17 the weight function can be generally dependent on the local value of an auxiliary reference density @xmath18 and has to satisfy the normalization requirement @xmath19 to ensure that in the homogeneous limit the weighted density coincides with the actual density of the fluid . in order to get a reliable approximation for the excess free energy functional we have to choose properly the two key ingredients which characterize our wda ansatz , namely the homogeneous free energy @xmath20 and the weight function @xmath21 . the only available microscopic liquid state theory which is able to account both for non critical and critical properties of a homogeneous fluid is the hierarchical reference theory ( hrt ) @xcite , which will be therefore adopted in this work for the evaluation of the excess free energy @xmath22 of the uniform fluid . although the general formalism of hrt can be applied to fluids and mixtures with arbitrary pair interactions , quantitative results for specific models require the closure of the exact hrt equations by introducing some approximation . a closure which proved remarkably accurate has been implemented in the case of a hard core yukawa ( hcy ) fluid @xcite , because the resulting hrt equations considerably simplify by use of the known solution of the ornstein - zernike equation available for this interaction . the hcy potential is defined as the sum of a pure hard core term of diameter @xmath23 and an attractive yukawa tail of inverse range @xmath24 : @xmath25 where the parameter @xmath26 , which defines the energy scale , is positive . in the following we will investigate this model taking @xmath23 and @xmath27 as the units of length and temperature respectively . having established the form of our excess functional in the homogeneous limit , we proceed with the explicit definition of @xmath10 $ ] for general density profiles . we first focus our attention on the effects of the attractive part of the potential . it is well known that its main contribution to the internal energy is given by the hartree term : this circumstance has been extensively recognized in the previous treatments , where attractive interactions were included just through such a contribution . we are therefore led to isolate this term in the excess free energy , by writing : @xmath13=\mathcal{f}^{\mathrm{ex}}_{\mathrm{r}}[n(\bm{r})]+ \mathcal{f}^{\mathrm{ex}}_{\mathrm{h}}[n(\bm{r } ) ] , \label{eq : split}\ ] ] where the hartree contribution is given by @xmath28= \frac{1}{2}\int\mathrm{d}\bm{r}^\prime\int \mathrm{d}\bm{r}^{\prime\prime } \ , n(\bm{r}^\prime ) v_{\mathrm{y}}(|\bm{r}^\prime-\bm{r}^{\prime\prime}| ) n(\bm{r}^{\prime\prime}),\ ] ] whereas the reference term @xmath29 $ ] , defined by eq . ( [ eq : split ] ) , contains both the entropic contribution to the free energy , arising from hard core repulsion , and the correlations induced by the attractive interaction . our choice is then to use wda to represent only the part of the intrinsic free energy functional , retaining the exact description of the hartree energy : @xmath30=\int \mathrm{d}\bm{r}\ , n(\bm{r}){\psi}^{\mathrm{ex}}_{\mathrm{r}}\bigl(\bar{n}(\bm{r } ) \bigr),\ ] ] where @xmath31 the form of the weight function can be determined following the strategy put forward by tarazona @xcite , by requiring that the two particle direct correlation function reduces , in the homogeneous limit , to the known form of the underlying bulk liquid state theory which , in our case , is the direct correlation function @xmath32 predicted by hrt : @xmath33}{\delta n(\bm{r } ) \delta n(\bm{r}^\prime)}\right|_{n(\bm{r})=\rho_b } \hskip -5 mm & = & c(|\bm{r}-\bm{r}^\prime|,{\rho}_b)+ \beta \ , v_{\mathrm{y}}(|\bm{r}-\bm{r}^\prime| ) \nonumber \\ & \equiv & c_{\mathrm{r}}(|\bm{r}-\bm{r}^\prime|,{\rho}_b ) , \label{eq_constraint_wda_r}\end{aligned}\ ] ] where the last equality is a definition . this constraint can be fulfilled only by a density dependent weight function . in a previous work curtin and ashcroft @xcite proposed that the weight function should depend on the local value of the weighted density itself , i.e. @xmath34 . unfortunately , even for a purely repulsive hard sphere system , this hypothesis leads to difficulties in the solution of the nonlinear differential equation for the weight function which must be imposed in order to implement the constraint ( [ eq_constraint_wda_r ] ) . this problem can be overcome choosing a position independent auxiliary density @xmath18 , as suggested by leidl and wagner @xcite . in our application of dft to the description of confined systems , we assume that the auxiliary density actually coincides with the homogeneous density at the same chemical potential ( also referred to as _ bulk density _ ) , i.e. we set @xmath35 it is straightforward to obtain from eq . ( [ eq_constraint_wda_r ] ) an algebraic equation for the fourier transform of the weight function : @xmath36 which , at least in the cases examined in this work , always admits real solutions . the physical root can be determined enforcing the normalization condition ( [ eq : normalization_w ] ) , recalling that the compressibility sum rule , satisfied by the hrt direct correlation function , requires @xmath37 , @xmath38 ) is very close to the critical point . , width=264 ] as in other implementations of wda , the range of the weight function always remains comparable to the size of the molecules . this feature is preserved also in the critical region , as shown in fig . [ fig : weightfunction ] . once the intrinsic density functional has been determined , the approximate equilibrium properties , such as the density profile @xmath0 and the grand canonical potential @xmath39 $ ] , can be obtained minimizing the functional ( [ eq : omega ] ) at fixed temperature and chemical potential . in a hcy fluid , the value of the chemical potential is related to the bulk density by : @xmath40 a central quantity for the present investigation is the force acting on the two planar hard walls in the hcy fluid . in this geometry , symmetry requires that all the local properties may depend on the single coordinate @xmath41 , orthogonal to the two plates , placed at @xmath42 and @xmath43 respectively . remarkably , if the wall separation @xmath44 is greater than @xmath23 , the force per unit surface @xmath45 acting on the plates , sometimes called _ solvation force _ , can be expressed as a pressure difference @xcite : @xmath46}{\sigma } \right)\right|_{t,\mu}-p_b(t,\mu ) , \label{eq : solvation_derivata}\ ] ] where @xmath47/\sigma$ ] is the grand canonical potential per unit surface of the fluid confined in the region @xmath48 $ ] , determined by the minimization of the approximated grand canonical functional at fixed @xmath49 and @xmath50 is the pressure of the fluid at the same values of temperature @xmath3 and chemical potential @xmath51 . on the other hand , when @xmath52 there are no particles between the walls and the attractive force per unit surface acting on the walls arises uniquely from the presence of the fluid in the regions @xmath53 and @xmath54 : the first term in eq . ( [ eq : solvation_derivata ] ) vanishes and the force is given by @xmath55 the solvation force @xmath45 defined above is a difference of pressures and goes to zero in the limit @xmath56 . by means of standard functional identities it is possible to express , without any further approximation , the derivative of the grand potential per unit surface in terms of the contact density : @xmath57 and the solvation force can be finally written as @xcite : @xmath58 we also remark that this version of wda exactly satisfies the contact value theorem @xcite @xmath59 leading to the more suggestive identity @xmath60 the above results show that , in order to obtain the force acting on the walls we just need to perform the minimization of the functional in the region @xmath48 $ ] in order to evaluate the contact density @xmath61 . the minimization has been be carried out by a simple iterative ( picard ) method , taking advantage of the exact implicit relation for the approximate equilibrium density profile @xmath62 given by : @xmath63 } , \label{eq : iteration_minimization}\ ] ] where the potential of mean force @xmath64 is defined as : @xmath65}{\delta n(\bm{r})}\bigg|_{\rho(z ) } + \int \mathrm{d}\bm{r}^\prime \rho(z^\prime)v_{\mathrm{y}}(|\bm{r}-\bm{r}^\prime|).\ ] ] a spatial @xmath66 has been generally used in the numerical minimization , while for bulk densities @xmath67 and close to the critical point @xmath68 has been reduced up to two orders of magnitude . typically , up to few thousand picard iterations were necessary to achieve a precision of one part in @xmath69 for the density profile and one part in @xmath70 in the grand potential . the minimization of the previously defined grand canonical functional allows to evaluate the equilibrium properties of the confined fluid . within this approach , the relevant quantities can be found at every temperature and bulk density of interest , also in the vicinity and below the critical point of the hcy fluid . most of the calculations refer to a yukawa fluid with range @xmath71 , where several simulation results are available . we performed minimizations of the functional for values of the temperature @xmath3 above the critical point ( @xmath72 ) and bulk reduced densities @xmath73 up to @xmath74 . in the high temperature limit our model reduces to a hard sphere fluid , whose properties have been extensively investigated by numerical simulations . figure [ fig : hs_simulation_fmt ] shows the density profiles @xmath62 of a hard sphere fluid near a hard wall at two different values of @xmath75 . the agreement of the wda prediction with the monte carlo ( mc ) data of ref . @xcite is very good up to reduced densities of the order of @xmath76 , while at higher values the phase of oscillations in the density profile are correctly captured , although a slight underestimation of the peak value is observed . the comparison of our density profiles with those predicted by the `` white bear '' version of the fmt @xcite shows , as expected , that rosenfeld s theory gives more accurate estimates of the oscillation peaks , particularly at high density . for a hard sphere fluid between two hard walls ( distance @xmath77 ) at different values of the bulk reduced density @xmath73 . mc simulation data ( symbols ) are taken from ref . we have obtained the fmt density profiles ( lines ) from the minimization of the approximated fmt - wb functional @xcite . to enhance visual clarity the density profiles at @xmath78 and @xmath79 are shifted upward by @xmath80 and @xmath81 respectively.,width=264 ] when the temperature is decreased the contribution of the yukawa tail to the density profile becomes relevant . we compared the density profile obtained within our dft approximation with the mc simulation data for @xmath71 at temperature @xmath82 and for @xmath83 at @xmath84 . figure [ fig : t2_diversirho ] shows that the wda estimate is remarkably accurate at reduces densities @xmath80 and @xmath85 ; small deviations from the mc simulation data appear at reduced density @xmath86 . we note that at @xmath87 the contact reduced density is overestimated of about @xmath88 with respect to simulation data , even if the contact theorem is verified with a relative error of the order of @xmath89 . this disagreement in the contact value is due to different estimates of the grand canonical potential per unit volume of the homogeneous fluid and it is compatible with the the spread in the values of the bulk pressure obtained within different simulation techniques @xcite . of a yukawa hard sphere fluid ( @xmath90 ) at reduced temperature @xmath82 confined between two hard walls at different values of bulk reduced density @xmath73 . lines represent the predictions of the present wda . points are mc data from ref . @xcite ( reduced density @xmath86 ) and ref . @xcite ( reduced density @xmath80 and @xmath85 ) . the distance between the walls is @xmath91.,width=264 ] in panel a ) of figure [ fig : profili_forze_louis ] we compare our results for the density profile of the hcy fluid characterized by @xmath83 with the recent mc simulations from ref . this figure shows that at relatively low densities the agreement between our approximation and mc predictions is remarkable also for attractive potentials of shorter range . in particular we predict accurately the kink in the density profile at @xmath92 , which is only qualitatively reproduced within mean field approximation @xcite . of a hcy fluid at a single hard wall ( @xmath83 ) for bulk reduced density @xmath93 and temperature @xmath84 . lines represent the predictions of the present wda . points are mc data from ref . panel b ) : force per unit surface acting between two infinite parallel hard walls immersed in a hcy fluid ( @xmath83 ) in the same thermodynamic state of panel a ) . lines are the predictions of the wda from eq . ( [ eq : solvation_rho ] ) , whereas points represent the md simulation of ref . @xcite.,width=264 ] the minimization of the grand canonical functional provides both the value of the contact density @xmath94 and of the grand free energy . it is therefore possible to obtain the depletion force either by calculating the derivative w.r.t . @xmath95 of the grand potential , as in eq . ( [ eq : solvation_derivata ] ) , or by making use of eq . ( [ eq : solvation_rho ] ) . the consistency between the two estimates is a good check for the accuracy of the numerical procedure . to give an example , for a hard sphere fluid , the relative difference between the two results is less than @xmath96 if the absolute value of the force per unit surface and @xmath97 is larger than @xmath98 . nonetheless at smaller values of the force the result obtained by differentiation of the grand free energy is less stable , due to errors introduced by the discretization . the solvent mediated force acting on two parallel hard walls immersed in a hard sphere fluid , is compared with the monte carlo data of wertheim et al . @xcite as well as the predictions based on fmt in fig . [ fig : forza_hs_piani ] , showing a nice agreement also at relatively high densities . at reduced density @xmath99 the force maximum per unit surface is of the order of @xmath100 , and the oscillations due to the packing of the hard spheres are damped within two or three diameters . furthermore , at this value of the reduced density , a small deviation between wda and fmt is present only at the first minimum . at reduced density @xmath76 the force at distances of the order of the hard sphere diameter @xmath23 is hundred times larger than at @xmath101 . even this feature is well reproduced by both wda and fmt , as can be seen in the inset . the strong oscillating behavior of the mc data is captured by wda with a correct phase , even if the peak values are a little underestimated , whereas fmt behaves considerably better . between two infinite planar hard walls immersed in a fluid of hard spheres of diameter @xmath23 for two values of the bulk reduced density . the value of the force within the wda approximation has been obtained via eq . ( [ eq : solvation_rho ] ) . the fmt - wb result comes from the minimization of the `` white bear '' version of the fmt functional @xcite via eq . ( [ eq : solvation_rho ] ) . the mc data is taken from ref . the inset highlights the behavior of the force at @xmath102 at small distances.,width=264 ] the solvent mediated force per unit area between two walls in a hcy fluid has not been extensively investigated by numerical simulations . in panel b ) of fig . [ fig : profili_forze_louis ] we show a comparison between our results and the monte carlo data of ref . @xcite for a hcy fluid of inverse range @xmath83 at @xmath84 and @xmath103 . even if the net force is quite small , our prediction agrees very well with the numerical simulations at all values of the wall separation . we stress that , particularly at small distances , the wda force is much more accurate than any mean field perturbation method ( see ref . @xcite ) . the detailed comparisons of our novel dft with both numerical simulations and state of the art theories allow to conclude that in the high temperature limit the present wda is able to correctly reproduce the density profile and the effective interactions between the hard walls with a very satisfactory accuracy up to reduced densities of about @xmath85 . moreover , this formulation of wda appears to be the best available dft for a hcy fluid at finite temperature . we performed the minimization of the wda density functional at several values of temperatures and reduced bulk densities for a hcy fluid of inverse range @xmath71 confined between two hard walls . at high temperatures the system behaves like a hard sphere fluid , whereas when the temperature is decreased the contribution of the yukawa tail becomes more and more relevant , and the strongly oscillating character of both the density profiles and the solvent mediated force is lost . figure [ fig : profili_05 ] shows the dependence of the density profile on temperature at fixed bulk reduced density @xmath104 . at reduced temperature @xmath105 the system behaves like a hard sphere fluid . as the temperature is lowered towards its critical value , the density profile gradually becomes monotonic losing the oscillating features typical of hard spheres and the density at contact assumes values four times lower than the bulk density . the range of the perturbation produced by the wall extends at larger and larger distances as the temperature approaches @xmath106 , giving rise to a region where a kind of drying of the wall can be observed . ) at a single hard wall at bulk reduced density @xmath107 and different values of the temperature . the inset shows the same data , also including the density profile at temperature @xmath105 using a different scale . , width=264 ] the attractive tail in the pair interaction of the hcy fluid smoothes the density profile reducing the layering of particles . as a consequence , the effective force between the two walls loses the strongly repulsive peak present at @xmath108 when the interaction between the fluid particles is purely hard sphere ( see fig . [ fig : forza_hs_piani ] ) . fig . [ fig : forze_rhocritico ] shows the force per unit surface between the walls for different values of the temperature at the critical bulk reduced density @xmath109 . the repulsive contribution to the interaction force , present at @xmath105 gradually disappears at lower temperatures and the force becomes purely attractive and monotonic , confirming the findings of the numerical simulations in a different model @xcite . by approaching the critical temperature ( @xmath110 ) the effective force becomes weaker and weaker at short distance , as can be seen in the right panel of fig . [ fig : forze_rhocritico ] : its amplitude reduces almost by a factor two due to a @xmath111 change in temperature . however , a closer look to the long distance tail of the solvent mediated force shows that its range indeed increases close to the critical temperatures , as expected on the basis of scaling arguments . however this occurs at very large separations ( @xmath112 for the data shown in the figure ) . ) obtained from the minimization of the wda functional using eq . ( [ eq : solvation_rho ] ) at @xmath113 and different values of the temperature.,width=264 ] fig . [ fig : forze_tcritico ] shows the force per unit surface between the walls in different density regimes when the value of the temperature is close to @xmath106 . we note that when the bulk density is higher than @xmath114 the force is an order of magnitude larger than for @xmath115 . the force is monotonic and purely attractive for all values of the reduced density and its range grows near @xmath114 , as expected . ) obtained from the minimization of the wda functional using eq . ( [ eq : solvation_rho ] ) at @xmath116 and different values of the reduced density.,width=264 ] in lattice fluid models the coexistence curve is symmetric about the critical temperature and the critical isochore coincides with the locus @xmath117 of the maxima of the isothermal susceptibility . for such systems , according to the renormalization group terminology , the path to the critical point orthogonal to the relevant odd operator coincides with the critical isochore . in a real fluid the coexistence curve is asymmetric about the critical isochore . in this case a good approximation for the same path is given by the line @xmath118 in the phase diagram defined as the locus of the points @xmath119 such as : @xmath120 where @xmath121 is the isothermal compressibility . [ fig : fit_profili_rhocritico ] shows the density profile of the hcy fluid at a hard wall along the line @xmath118 . its behavior at distances larger than the bulk correlation length is well fitted by an exponential of the form : @xmath122 where @xmath75 is the bulk density of the fluid , @xmath123 is a negative amplitude factor and @xmath124 is the bulk correlation length . this exponential decay of the density profile is observed also if the system is far from the critical region and is probably related to the location in the bulk phase diagram of the point we investigated with respect to the line @xcite . following the argument of ref . @xcite , we expect that the exponential long range behavior of the density profile reflects in an analogous exponential decay of the force between the two walls . provided we do not cross the line , this decay should be present both far from the critical point and in the critical region , where it agrees with the predictions of the theory of the critical casimir effect @xcite : @xmath125 the exponential decay of the solvation force is indeed confirmed by the exact solution of a @xcite ising slab under symmetry breaking boundary conditions and was observed in monte carlo simulations of three dimensional simple fluids @xcite . [ fig : fit_forze_rhocritico ] shows the long distance exponential decay of the force per unit surface and @xmath97 between two planar hard walls mediated by a hcy fluid along the critical line @xmath118 . the force obtained with our approach is very well fitted by eq . ( [ eq : fit_forza ] ) at large distances ( @xmath126 ) , whereas at short distances , where depletion effects become relevant , the solvent mediated force , always attractive , displays a plateau ( see inset ) . ) at a single hard wall in the critical region . here points : fits of the density profiles according to eq . ( [ eq : fit_profilo ] ) performed at distances larger than three times the correlation length of the homogeneous hcy fluid at the same temperature and bulk density . @xmath75 , @xmath123 and @xmath124 are free fitting parameters and the results obtained for @xmath75 and @xmath124 agree well with the bulk values of density and correlation length respectively ( the accuracy is better than @xmath128 for the bulk density and @xmath129 for the correlation lenght ) . the bulk correlation lengths obtained from the fitting procedure are @xmath130 , from the lowest to the highest reduced temperature . the inset shows a magnification of the same density profiles at short distances . , width=264 ] ) at a single hard wall along the line @xmath118 . here points : fits of the force according to eq . ( [ eq : fit_forza ] ) performed at distances larger than four times the bulk correlation length of the homogeneous hcy fluid at the same temperature and bulk reduced density . @xmath131 , and @xmath124 are free parameters in the fitting procedure . the result obtained for @xmath124 agrees well with the bulk value of the correlation length ( the accuracy of the fit is better than @xmath132 ) . the bulk correlation lengths obtained from the fitting procedure are @xmath133 , from the lowest to the highest reduced temperature . the inset shows the force profile at short distance.,width=264 ] the same wda formalism previously introduced may be generalized to other interesting geometries , in addition to the planar one . most importantly , it can be used to evaluate the effective interaction between two spherical particles in a solvent , with obvious applications to the study of aggregation in colloidal suspensions . the direct minimization of the dft , although numerically feasible , represents a task considerably more complex than in planar geometry . therefore , in this first application of the formalism , we have chosen to resort to the simple but effective derjaguin approximation @xcite , which allows to express the interaction between two convex objects starting from the knowledge of the force between two planar walls , independently on the physical origin of the force . according to this approximation , the force @xmath134 between two spheres of radius @xmath135 can be written as : @xmath136 where @xmath95 is the minimal surface to surface distance between the sphere and @xmath137 is the force per unit surface between the two walls at distance @xmath41 . this approximation gives accurate results provided @xmath138 and if the interaction potential between the two walls decays rapidly at large distances . when the force between the walls is mediated by a hard sphere fluid with particles of diameter @xmath23 it is possible to show that derjaguin s expression is the best approximation of the true depletion interaction without taking in account curvature effects @xcite and it is accurate in the limit of @xmath139 , where @xmath140 is the size ratio . the limits of derjaguin approximation when applied to depletion interactions is a debated issue . particularly , it is a matter of discussion the size ratio at which derjaguin s theory starts to fail , and how its accuracy depends on the concentration of depletant . in figure [ fig : depletion_sferedure ] we compare the prediction for the depletion potential @xmath141 between two big hard spheres in a fluid of smaller hard spheres obtained both by derjaguin approximation and mc simulations at two different values of the bulk density of the smaller particles . the predictions for the depletion potential obtained by derjaguin approximation are accurate at @xmath142 only at values of the packing fraction of the small spheres lower than @xmath143 . at @xmath144 derjaguin approximation overestimates by about @xmath145 both the contact value and the first repulsive peak in the potential , while the oscillations at larger values of distance are underestimated when compared to mc data . the rather poor performance of derjaguin approximation is probably due to the presence of a strong repulsive peak in the solvation force between the two walls at @xmath146 ( see fig . [ fig : forza_hs_piani ] ) which appears to be a peculiarity of the slab geometry . we expect that , for smoother inter - wall effective interactions , the agreement would be considerably better . between two hard spheres in a fluid of smaller hard spheres ( size ratio @xmath142 ) at bulk packing fraction @xmath147 equal to @xmath88 ( left panel ) and @xmath148 ( right panel ) . mc ( points ) and fmt predictions obtained with the insertion trick ( purple lines ) are taken from ref . the red line at @xmath149 is obtained by use of derjaguin approximation starting from the solvation force between two planar hard walls evaluated from the fmt - wb @xcite approximation . the green lines represent the depletion force obtained within derjaguin approximation when the solvation force between the walls is given by the present wda approximation.,width=264 ] the results presented above show that derjaguin approximation can be safely adopted is the size ratio between the depletant and the colloid is sufficiently small and up to moderate densities of depletant ( i.e. @xmath150 and @xmath151 ) , and we expect that similar considerations apply when the depletant is a yukawa hard core fluid . within the limits of validity of the derjaguin approximation , we can determine the solvent mediated potential @xmath152 between two hard spheres of radius @xmath135 immersed in a yhc fluid in order to examine the phase stability of such a colloidal suspension . according to noro - frenkel extended law of corresponding states @xcite , fluids characterized by short ranged interaction potentials obey the same equation of state , when expressed in terms of reduced variables . in particular , it was observed that the dimensionless second virial coefficient @xmath153 where @xmath154 is the second virial coefficient of a hard sphere system with particles of radius @xmath135 , assumes a value of about @xmath155 at the critical point independently on the particular form of the interaction and that its value remains constant in a relatively large density range across the critical point @xcite . it is therefore possible to estimate the the gas - liquid spinodal line for a system of hard sphere colloidal particles dispersed in a hcy fluid , by evaluating their reduced second virial coefficient . ) in the @xmath156 plane . the hrt results for the coexistence curve @xcite are shown by a black line . green dots represent the mc data from ref . blue points show the aggregation boundary of two big hard spheres ( size ratio @xmath157 ) predicted on the basis of our wda plus derjaguin approximation and noro - frenkel criterion . the dashed line connecting the points is a guide to the eye : aggregation takes place on the right of this boundary . the inset shows the same phase diagram in a wider density interval.,width=264 ] the blue points in figure [ fig : aggregation ] identify the phase separation line of a hs fluid induced by the presence of a depletant modeled as a hcy fluid . the size ratio between the depletant and the guest hs particles is @xmath142 . at high values of the reduced temperature the phase separation occurs at reduced density of about @xmath80 , as expected in the limit of hs depletant . when the temperature decreases , the concentration of depletant needed to induce phase separation decreases monotonically . at the depletant critical temperature we observe phase separation at depletant concentrations @xmath158 much lower than the critical one . this implies that , at this value of the size ratio @xmath159 , the phase separation is not related to the presence of long range tails in the effective force , which characterizes the critical region of the solvent , but is still mainly due to the short range attraction generated by the depletion mechanism . we note that whenever a direct short range repulsion is present between the colloidal particles , as for the case of charged systems , the strong attraction due to depletion is severely weakened and particle aggregation takes place at considerably larger solvent densities . instead , in the critical region , the long range tails of the solvent mediated ( casimir ) force is not effectively contrasted by the additional short range repulsion . in extreme circumstances ( i.e. when the direct repulsion between particles is sufficiently strong ) , ordinary depletion may be fully screened and phase separation inhibited . however , aggregation is generally expected in a small pocket within the critical region , due to the emergence of long range casimir forces . for repulsive wall - solvent interactions , this pocket will be centered at solvent densities larger than the critical one , due to the strong asymmetry of the critical casimir forces ( see e.g. figs . [ fig : forze_tcritico ] and [ fig : scala_micro_tcritica ] ) . the aim of this section is to evaluate the solvent mediated interaction induced between two walls when the depletant is in the critical regime . thermal fluctuations in fluids occur on a characteristic range , determined by the correlation length @xmath124 , which usually is comparable with the molecular diameter or with the range of the interactions . when approaching a second order phase transition , the range of the fluctuations of the order parameter ( the particle density in the case of a simple fluid ) grows larger up to diverging at the critical point . in this regime the effective force between two bodies immersed in the critical fluid acquires a universal form and obeys scaling laws , as many physical properties near criticality do . in 1978 @xcite fisher and de gennes first recognized that a confinement of the critical fluctuations of the order parameter gives rise to an universal , long ranged interaction which they named critical casimir force . according to the scaling approach , the universal contribution to the force per unit surface @xmath160 acting between two infinite plates confining a critical fluid can be written as @xcite : @xmath161 where @xmath95 is the distance between the two walls . the upper sign refers to the temperature ( while the lower to the one ) and the scaling variables ( @xmath162 ) @xmath163 are defined in terms of the two scaling fields @xmath164 here @xmath165 is the bulk correlation length at @xmath166 , @xmath167 is a non - universal metric factor and @xmath168 , @xmath169 and @xmath170 are the usual critical exponents . the function @xmath171 is usually referred to as the scaling function of the critical casimir force in planar geometry . this function is universal in sense that it depends only on the bulk universality class , on the boundary conditions imposed at the confining surfaces and on the geometry of the system ( which in this case is @xmath172 ) . we remark that there is no extra metric factor associated with @xmath173 and that there is a dependence on the sign of the field @xmath44 because boundary conditions at the walls break the bulk symmetry @xmath174 . according to the theory of scaling , eq . ( [ eq : casimir_film ] ) represents the asymptotic decay of the solvent mediated force as @xmath175 and for @xmath176 . the bulk yukawa fluid under investigation belongs to the 3d ising universality class and the boundary conditions are determined by the affinity of the wall surfaces with the fluid particles : if the contact density is less than the bulk density the boundary condition is of type @xmath177 , otherwise of type @xmath178 . in this work we only deal with temperatures ( @xmath179 ) and with symmetric @xmath180 boundary conditions , which arise for purely repulsive interactions between the fluid particles and two identical confining hard walls . the analysis of the critical casimir force and of the related universal scaling function @xmath181 is a rather difficult task both experimentally , for the small forces involved , and theoretically , for the lack of an accurate description of critical fluids in confined geometries . most of the results present in the literature deal with the temperature dependence of the casimir fluctuation induced interaction at zero magnetic field . an indirect estimate of the scaling function for the film geometry and the 3d ising universality class , at @xmath166 under @xmath182 and @xmath183 boundary conditions was given in ref . @xcite monitoring the thickness of a binary fluid film at different temperatures near @xmath106 . the first direct evaluation of the critical casimir force was performed in 2008 @xcite for a system consisting of a colloidal particle close to a wall immersed in a binary mixture ( geometry ) of water and lutidine at different compositions and for both the relevant boundary conditions . this experiment , however , allowed to probe only the exponential tail of the scaling function . a mc study of the solvent mediated potential between two spherical particles in a simple fluid along the critical isochore has been performed in refs . @xcite with different boundary conditions . however , the determination of the full casimir scaling function could not be obtained in the temperature range examined in the simulations . along the symmetry line ( @xmath166 ) , more precise estimates of the universal casimir scaling functions for the 3d ising universality class and film geometry have been obtained via mc simulations of the ising model @xcite . few theoretical approaches were devised to address this problem : in addition to the mean field results @xcite , its is worthwhile mentioning the extended local functional method @xcite and a long wavelength analysis of density functional theory @xcite . the latter investigations have been also extended away from the symmetry line ( @xmath184 ) providing predictions on the shape of the critical casimir scaling function in the case @xcite . monte carlo simulations at @xmath185 were recently performed in ref . @xcite . the wda approach developed above allows the study of this problem starting from the microscopic hcy fluid model confined between two walls . according to the scaling hypothesis , the effective force per unit surface between the two walls @xmath160 should depend on the physical control parameters @xmath186 only through the combination ( [ eq : casimir_film ] ) , implying the collapse of different data sets onto the same universal curve . figure [ fig : scala_micro_rhocritica ] shows the scaling function obtained from independent calculations at different temperatures along the previously defined critical line @xmath187 ( i.e. @xmath188 ) . note that , even at reduced temperature @xmath127 as low as @xmath189 , our estimates show a marked temperature dependence , and the data along different isotherms do not collapse as we expected . at the lowest temperature we investigated , a significant difference between our prediction and the mc simulations of ref . @xcite suggests the presence of strong corrections to scaling . we also remark that the curves at the lowest temperatures develop a kink at small values of @xmath173 , due to the singular behavior of the scaling function at @xmath190 . in fact , at any given reduced temperature @xmath191 , the quantity @xmath192 tends to zero as @xmath193 , forcing the estimate of the scaling function to vanish . ) from the microscopic force obtained within the present wda approximation . the corresponding bulk correlation lengths are @xmath194 , from the lowest to the highest reduced temperature . the lines connecting the points are a guide to the eye . the mc data are taken from ref . @xcite and refer to @xmath180 boundary conditions and the different sets correspond to different estimates of the corrections to scaling.,width=264 ] figure [ fig : scala_micro_tcritica ] shows the scaling function at fixed temperature near @xmath106 for different values of the scaling variable @xmath195 , corresponding to different bulk reduced densities . the scaling function is always negative and a strong asymmetry is evident between the curves at density above and below @xmath114 . for positive values of the scaling field @xmath44 ( i.e. @xmath196 in our case ) , the magnitude of the force becomes larger and larger and the peak is shifted towards small values of @xmath173 . and for different values of the scaling variable @xmath197 . the lines connecting the points ar a guide to the eye . the largest correlation length , @xmath198 , corresponds to the critical reduced density @xmath113.,width=264 ] although the direct numerical evaluation of the critical casimir scaling function predicted by this class of dft is not conclusive , due to severe corrections to scaling , an accurate estimate of the asymptotic behavior can be obtained by a long wavelength ( lw ) analysis of the dft equations . in fact , following ref . @xcite we note that : * the density profile @xmath62 displays reflection symmetry about @xmath199 , limiting the range of interesto to @xmath200 $ ] ; * when the walls are far apart ( @xmath201 ) , the difference between the density profile corresponding to a wall to wall distance @xmath95 and its single wall limit , reached for @xmath202 , is significant only for @xmath203 ; * as a consequence , the effective force per unit surface @xmath204 ( hence the casimir scaling function ) just depends on the long distance tail of the density profile , which is expected to be a slowly varying function of the coordinate @xmath41 . therefore our wda intrinsic free energy functional can be approximated by keeping only the lowest term in a gradient expansion about the bulk density @xmath75 : @xmath205}{\sigma}=l\,\varphi(\rho_b ) + \nonumber \\ & & \,\int { \rm d } z \ , \bigg [ \frac{b}{2 } \left ( \frac{{\rm d}\delta n(z ) } { { \rm d } z}\right)^2 + \varphi(\rho_b+\delta n(z ) ) -\varphi(\rho_b ) \bigg ] , \label{eq : longw}\end{aligned}\ ] ] where @xmath206 is the free energy density in the bulk , times @xmath169 . this expression coincides with the long wavelength limit of our wda functional , the stiffness @xmath207 being related to the range of the direct correlation function in the homogeneous system @xmath208 : @xmath209 in the presence of short range interactions , the direct correlation function is analytic in @xmath210 away from the critical point , where it displays a @xmath211 singularity . however , within our approximate closure of the hrt equations , the critical exponent @xmath212 and analyticity is preserved also at criticality @xcite , keeping the stiffness @xmath207 finite in the whole phase diagram . this implies that the long wavelength limit of the structure factor of the homogeneous fluid follows the orstein - zernike ansatz : @xmath213 with @xmath214^{-1}\ ] ] and @xmath215 . close to the critical point , the hrt bulk free energy density @xmath206 acquires a scaling form : @xmath216 where @xmath114 is the critical density and @xmath217 is the chemical potential ( the temperature dependence of these quantities is understood ) , while @xmath218 and @xmath219 are non universal metric factors and @xmath127 is the reduced temperature . in the following it will be convenient to express the universal quantities in terms of the scaling field @xmath220 instead of the previously defined variable @xmath197 . within our hrt closure , the critical exponents are @xmath221 , @xmath222 , @xmath223 in @xmath224 , which agree within @xmath111 with the accepted values . the metric factors appearing in the scaling function are implicitly defined by the requirement that @xmath225 has the following expansion at small @xmath226 @xcite : @xmath227 in fig . [ fig : scalapsi_g4 ] the asymptotic hrt scaling function @xmath225 is shown together with a parameterization of the exact result for the 3d ising universality class . although the two curves are indistinguishable on this scale , calculations at different reduced temperatures , also shown , suggest the presence of important corrections to scaling . as predicted by hrt compared with a parameterization of the exact result from ref . @xcite ( black curve ) . the red line , showing the hrt asymptotic result , is identical to the exact result on the scale of the figure . a few rescaled free energies obtained from the integration of the hrt equations at different reduced temperatures are also shown . inset : universal amplitude ratio @xmath228 for a hcy fluid at the critical density as a function of the reduced inverse correlation length ( green points ) . the usually accepted value @xcite is shown by a blue triangle whereas the red square represents the asymptotic hrt value . , width=264 ] the minimization of the long wavelength functional ( [ eq : longw ] ) in slab geometry gives rise to a differential equation whose solution allows to evaluate the asymptotic decay of the effective force between two hard walls in a critical fluid . the derivation , already detailed in ref . @xcite and not repeated here , provides a closed form for the critical casimir scaling function in terms of two universal quantities : the bulk free energy scaling function @xmath225 and the universal amplitude ratio @xmath228 . defining the auxiliary quantity @xmath229 by the implicit relations : @xmath230 the critical casimir scaling function in three dimensions is given by : @xmath231 the universal amplitude ratio @xmath228 is expressed in terms of the non universal metric factors previously introduced as @xmath232 again , the evaluation of @xmath228 from the hrt equations displays severe correction to scaling in a hcy fluid , as shown in the inset of fig . [ fig : scalapsi_g4 ] . more importantly , the usually quoted `` exact '' value @xcite @xmath233 turns out to differ significantly from the hrt prediction @xmath234 . the asymptotic study of the dft equations allows to extract the critical casimir scaling function just from bulk quantities via eqs . ( [ sigma1]-[eq : scalingc ] ) . it is then instructive to contrast these predictions with the outcome of the direct minimization of the hrt functional , already shown in fig . [ fig : scala_micro_rhocritica ] . such a comparison can be found in fig . [ fig : micro_vs_long ] , where the scaling functions obtained from the microscopic dft at a few reduced temperatures @xmath235 in the critical region are shown to agree remarkably well with the predictions of the long wavelength analysis , provided both the scaling function for the free energy @xmath225 and the universal amplitude ratio @xmath228 are consistently evaluated at the same reduced temperature @xmath235 . and @xmath228 evaluated at the same reduced temperature as the dft calculation . black line : asymptotic limit of the critical casimir scaling function . points : prediction of @xmath181 from monte carlo simulations @xcite . , width=264 ] the presence of strong corrections to scaling in both @xmath225 and @xmath228 , already highlighted , induces strong effects in the critical casimir scaling function which , at reduced temperatures lower than @xmath189 , is still quite far from its asymptotic limit . the main effect is due to the growth of the amplitude ratio , which , as shown in fig . [ fig : scalapsi_g4 ] , appears to reach its universal value only extremely close to the critical point , according to the prediction of hrt for the model of critical fluid investigated here . we remark that , due to the already quoted difference between the hrt estimate of the universal amplitude ratio @xmath228 and the value obtained via series expansions , the critical casimir scaling function predicted by our dft significantly differs from the one obtained in mc simulations , as can be seen in fig . [ fig : micro_vs_long ] . we presented a novel density functional , based on the weighted density paradigm , able to describe classical inhomogeneous fluids in a large portion of their phase diagram , critical point included . this is the first attempt to describe the effects of correlations induced by attractive interactions in confined fluids . the theory is based on the description of the uniform system provided by the hierarchical reference theory , one of the few liquid state approaches able to cope with long range density fluctuations . this technique , applied to the evaluation of the effective interaction between two hard walls in a fluid , allowed for the investigation of the crossover between a depletion - like mechanism at high temperatures and the critical casimir effect emerging near the critical point . our method does not rely on a long wavelength approximation and provides a complete picture of the solvent mediated force for any wall separation , displaying the presence of important non universal contributions in the effective interaction at short distances , even in the critical region . we believe that this dft will be useful in investigating other correlated systems , when density fluctuations are expected to play an important role . we showed that , at large separations , the solvent mediated force per unit surface between the walls decays exponentially on the scale of the correlation length in the whole portion of the phase diagram to the left of the line . such a behavior can not be considered as a signature of the onset of critical casimir effect : only the product between the _ amplitude _ of the long range exponential tail and the cube of the correlation length is a genuine universal quantity . our microscopic approach allows for the determination of the universal quantities characterizing the critical casimir effect , namely the scaling function @xmath236 , both along the critical isochore ( @xmath188 ) and in the regime . strong corrections to scaling have been observed in the hcy fluid we investigated : the universal features appear to emerge only in a narrow neighborhood of the critical point , at least in the model we examined . it would be useful to compare this prediction with numerical simulations for the hcy fluid model as well as with theoretical investigations of other systems , like the ising model , where the correction to scaling may be weaker . these studies will hopefully clarify the origin of the discrepancy between the hrt estimate of the universal amplitude ratio @xmath228 and the commonly accepted value . the approach presented in this work allows for further improvements . in our density functional , the ideal gas term and the hartree contribution to the internal energy have been treated exactly , while the remaining term has been approximated by use of a weighted density trick . the next step will be to treat the hard sphere term by the fundamental measure theory , known to be very accurate in dealing with excluded volume effects , limiting the weighted density contribution only for the residual correlation term . this adjustment is expected to increase the accuracy of the theory at high density , without however modifying the description of the universal properties of the critical casimir effect . in this first application we just considered a planar geometry , whose implications for the phenomenon of colloidal aggregation depend upon further assumptions , namely the derjaguin approximation , which however turns out to be rather inaccurate when the two external bodies are not very close . a natural further step will be to perform the functional minimization in cylindrical geometry , appropriate for dealing with two spherical particles thereby avoiding any mapping . finally , a very interesting application of this formalism will be the investigation of temperatures , where wetting phenomena are expected close to the first order transition boundary . the hierarchical reference theory of fluids provides a consistent description of the full transition line and then it appears to be the natural starting point for the development of a microscopic theory , beyond mean field , for the study of phase coexistence near a wall .

the present knowledge of solar and stellar magnetism derives essentially from the interpretation of polarization in spectral lines due to the zeeman effect . in the presence of a weak magnetic field , the radiative transfer equation for polarized light can be solved analytically and it is known as the `` weak field approximation '' . although very simple , it relies on a set of strong assumptions that we discuss later . however , the weak field approximation is applicable in many scenarios and gives very good results for the inference of magnetic fields as compared with more elaborate methods . in solar physics , it is at the heart of the success of a large number of synoptic magnetographs , like those of big bear @xcite . it has been also used recently to produce vector magnetograms with the imaging magnetograph experiment ( imax ) instrument @xcite onboard the sunrise balloon @xcite . in stellar spectropolarimetry , the weak field approximation is at the base of the least - squares deconvolution ( lsd ; * ? ? ? * ) , the most successful technique used to detect and measure magnetic fields in solar - type stars , or in some recent works on central stars of planetary nebulae @xcite , white dwarfs @xcite , pulsating stars @xcite , hot subdwarfs @xcite and ap and bp stars @xcite . most of the scientific cases commented so far deal with unresolved structures which means that the weak field approximation has to be understood in terms of fluxes of the stokes parameters instead of specific intensities . recently , @xcite derived the weak field expressions in terms of fluxes , particularizing to the case of the stellar dipole . they neglected rotation to simplify the equations and to derive analytical expressions for the inference of the magnetic field . when rotation is taken into account ( in other words , there is a correlation between the magnetic field and the line - of - sight velocity ) , the problem becomes much more challenging and it is usually solved numerically ( e.g. , * ? ? ? * for a recent effort ) . however , we show in this paper that it is possible to obtain analytical expressions for the stokes vector fluxes in the weak field approximation . there are two interests on this effort . first , having an analytical expression for the stokes flux might be of help to investigate the interplay between rotation and the magnetic field . second , it can be used to speed up inversion codes based on the weak - field approximation because no numerical integration is needed . this will facilitate the application of bayesian inference codes , like the one developed by @xcite , that are based on markov chain monte carlo methods that need to carry out the synthesis thousands of times to sample the posterior distribution . this is the first paper of a series dealing with analytical expressions for the fluxes of the stokes parameters in the weak field limit . this paper deals with the expressions for a rotating star with a magnetic field on its surface . the results depend on the correlation between the magnetic field and the velocity and are general for any rotation profile or any magnetic field configuration . we particularize the equations to the case of a rotating magnetic stellar dipole . let us consider a star with arbitrary velocity and magnetic fields on its surface . we define both fields for an arbitrary point @xmath0 in the surface using the reference frame represented in fig . [ fig : fig_geom ] . the line - of - sight ( los ) is along the @xmath1 axis , while we choose the axis @xmath2 as the one defining the reference direction for positive stokes @xmath3 . axis . the direction @xmath2 defines the reference for @xmath4 . the vector @xmath0 defines any position on the stellar surface.,scaledwidth=50.0% ] our aim is to compute the local stokes profiles emerging from every point of the visible surface and compute its integral . as explained in the introduction , a simple ( although subject to approximations ) way to relate the emergent stokes profiles with the magnetic field vector is the weak - field approximation . this approximation ( combined with the weak - line approximation for linear polarization ) leads to an analytical solution to the radiative transfer equation . it is based on the following three fundamental assumptions : i ) the magnetic field vector is constant along the line of sight on the region of the stellar atmosphere where the spectral line is formed , ii ) the magnetic field intensity is sufficiently weak so that the zeeman splitting is smaller than any other line broadening mechanism and iii ) the los velocity and any broadening mechanism are constant with height in the line formation region . under these simplifying conditions , the magnetic field can be considered as a perturbation to the zero - field case . this approximation is not strictly valid in general , but it is a good approximation whenever the magnetic field strength is weak in the line formation region in the absence of strong gradients along the los . quantitatively , the approximation holds whenever @xmath5 , where @xmath6 ( see eq . ( 3.14 ) in * ? ? ? * ) is the zeeman splitting that is proportional to the field strength , @xmath7 is the dominant broadening mechanism and @xmath8 represents the effective land factor of the line , which quantifies the magnetic sensitivity of the line and that only depends on the quantum numbers of the transition ( see , e.g. , * ? ? ? * ) . from its definition , it is clear that the weak field regime occurs at different field strengths for different spectral lines , for different stellar objects and for different spectral resolutions . for instance , for a magnetically sensitive iron spectral line with @xmath9 at 5000 with a standard thermal broadening at @xmath10 k , the weak - field regime is applicable for fields below @xmath11600 g if the spectral resolution of the spectrograph is smaller than the line broadening . under the previous simplifying assumptions , and to first order in @xmath6 , the intensity profile @xmath12 of a spectral line formed in a weak magnetic field is fully insensitive to the magnetic field . in other words , it fulfills the standard radiative transfer equation in the absence of a magnetic field ( e.g. , * ? ? ? . likewise , at first order in @xmath6 , the circular polarization profile , i.e. , the stokes @xmath13 profile for a given spectral line , has the following expression : @xmath14 where the symbol @xmath15 stands for the longitudinal component of the magnetic field ( the projection along the los in the reference system shown in fig . [ fig : fig_geom ] ) . the symbol @xmath16 stands for the wavelength while @xmath17 is the central wavelength of the transition . the constant @xmath18 . at first order in @xmath6 , linear polarization is zero . in order to obtain an expression for the stokes profiles characterizing linear polarization we have to expand the radiative transfer equation to second order in @xmath6 and assume that the spectral line is not saturated and weak . under these assumptions , a good approximation to the intensity profile emerging from the stellar atmosphere is to consider it , in general , proportional to a voigt function ( see eqs . ( 9.84 ) and ( 9.86 ) in * ? ? ? for the sake of simplicity , we consider the case of a gaussian line profile . all our results can be derived for a voigt profile but the final expressions are more complicated without any fundamental difference . this assumption somehow reduces the scope of applicability of our results for linear polarization because the line shape self - similarity is only valid for lines whose line depth is smaller than @xmath1140% according to the results of @xcite . with this assumption , the weak - field approximation for the stokes @xmath19 and @xmath20 , defined in the @xmath21 reference system shown in fig . [ fig : fig_geom ] , are ( see * ? ? ? * ; * ? ? ? * ) : @xmath22 where the symbol @xmath23 plays the role of the effective land factor for linear polarization and quantifies the sensitivity of linear polarization to the magnetic field . again , it is only a function of the quantum numbers of the transition @xcite . the symbols @xmath24 and @xmath25 stand for the components of the magnetic field vector along the @xmath2 and @xmath26 axes , respectively . the stokes profiles in any other reference system that is defined as a rotation around the @xmath1 axis by an angle @xmath27 are given by the usual rotation @xcite : @xmath28 where the quantities with the subindex `` new '' refer in this case to the new reference frame . when observing unresolved objects , the detected polarized radiation is obtained as an integration in the plane of the sky of the individual stokes parameters at each point of the stellar surface ( @xmath29 ) , all of them referred to a common reference system . we describe the stellar disk using the polar coordinates @xmath30 and @xmath31 , where @xmath30 is normalized to the stellar radius , so that its range is from 0 to 1 ( see * ? ? ? therefore , the average of any arbitrary function @xmath32 on the surface @xmath29 will be given by : @xmath33 as a consequence , it is obvious that the observed values of the stokes flux vector depends on the surface distribution of the magnetic field , on the centre - to - limb variation ( clv ) of the radiation , and on the doppler effect due to the rotation of the star . following standard approximations and assuming that the star has azimuthal symmetry , @xcite parameterized the intensity at each point of the star as a product of two functions : @xmath34 , that only depends on the wavelength ( the intensity at the disc centre since they neglect rotation ) , and @xmath35 , the clv that depends on the position on the disk and is wavelength independent ( e.g. , * ? ? ? * ; * ? ? ? * ) : @xmath36 the clv is given in terms of @xmath37 , where @xmath38 is the astrocentric angle between the normal to a point in the stellar surface and the line of sight ( see fig . [ fig : fig_geom ] ) . the parameters @xmath39 and @xmath40 have values between 0 and 1 and are supposed to be constant along the spectral line ( although they can vary from line to line ) . with this assumption , we also force the line depth to be @xmath41-independent . note that the values of @xmath39 and @xmath40 have to fulfill the condition that the intensity has to be positive . this simple parameterization allowed them to express the weak field equations in terms of the observed fluxes , thus mimicking the expressions for the specific stokes vector . this was possible because the intensity flux is assumed to be the product of the wavelength - dependent intensity at disk center and the wavelength independent clv . using this relationship , the weak - field and weak - line approximation for the integrated polarized flux remains formally the same than for the specific stokes vector but the components of the magnetic field appear weighted by the clv law . when the velocity field is included , the situation turns out to be much more complicated because the intensity itself depends on the local projection of the velocity along the @xmath1 axis ( @xmath42 ) , which subsequently depends on the position along the stellar disk in a potentially complicated way . in this case , the fluxes of the stokes vector are expressed as : @xmath43\frac{\partial^2 { i}(\lambda , v_z(\rho,\theta),\mu)}{\partial \lambda^2 } \nonumber \\ f_{u}(\lambda)&=&-\frac{{c}^2}{2}\lambda_0 ^ 4 g \int_0 ^ 1 \rho d\rho \int_0^{2\pi } d\theta~b_x(\rho,\theta)b_y(\rho,\theta ) \frac{\partial^2 { i}(\lambda , v_z(\rho,\theta),\mu)}{\partial \lambda^2 } \nonumber \\ f_{v}(\lambda)&=&-{c}\lambda_0 ^ 2 g \int_0 ^ 1 \rho d\rho \int_0^{2\pi } d\theta~b_z(\rho,\theta ) \frac{\partial { i}(\lambda , v_z(\rho,\theta),\mu)}{\partial \lambda},\end{aligned}\ ] ] the immediate consequence of the dependence on @xmath44 is that the integrals can not be separated . to proceed , it is crucial to decouple the rotation velocity and the wavelength to end up with expressions for the weak field approximation in terms of the observed fluxes . in order to do that , we assume that the shape of the local absorption profile is independent of the position on the stellar surface , apart from the doppler shift introduced by @xmath42 , and that the clv is wavelength independent . furthermore , if we assume that the absorption profile is gaussian , we end up with : @xmath45\right ] , \label{eq : intensity_gaussian}\ ] ] where @xmath46 stands for the continuum intensity , @xmath47 for the line depth , and @xmath48 for the line width . this approximation is not appropriate for stars with large radial temperature and density gradients because the lines close to the limb will differ strongly from the lines at the disk center . apart from the gaussian shape , it is possible to work with voigt functions but the subsequent expressions turn out to be more complicated and no new physics is gained , so we prefer to stick with the gaussian shape . the gaussian shape is also consistent with the assumption of an unsaturated and weak spectral line that is necessary for eqs . ( [ eq : q_u ] ) to hold . we can carry out the taylor expansion of @xmath49 around @xmath50 ( we drop now the @xmath41-dependence ) to obtain the following expression for the intensity at each point of the stellar surface : @xmath51 defining the emission line profile at zero velocity as @xmath52,\ ] ] the absorption profile at an arbitrary velocity shift can be written as : @xmath53 taking into account that the gaussian function is the generating function of the hermite polynomials : @xmath54=(-1)^n\exp\left[-\frac{x^2}{2 } \right ] \textit{he}_n(x),\ ] ] where @xmath55 is the hermite polynomial of order @xmath56 with weight function @xmath57 , also known as chebyshev - hermite polynomials ( see * ? ? ? if a voigt function is used instead for the local profile , the derivatives are more complex and the recurrence relations of @xcite can be used . taking the previous definition into account , the derivatives ( @xmath58 ) of the intensity profile can be expressed as : @xmath59 so that the intensity profile of eq . ( [ eq : intensity_gaussian ] ) is given by the following gram - charlier series : @xmath60,\ ] ] where we have extended the summation to @xmath61 to include the profile at zero velocity , since @xmath62 . the previous formula shows that the intensity is given as the addition of infinite terms ( potentially a large number of them for obtaining convergence ) given by the product of a function that depends on the wavelength and another one that depends on the velocity along the line of sight . the latter is the only one that changes on the stellar surface , so that this variable separation allows us to carry out the surface integration . in order to simplify the following computations , it is interesting to define the function @xmath63 so that the intensity profile and its first and second derivatives with respect to the wavelength are given by : @xmath64 \nonumber \\ { i}'(\lambda , v_z , \mu)&= & -\psi'(\lambda ) f(\mu ) \mathcal{h}(\lambda , v_z ) - \psi(\lambda ) f(\mu ) \mathcal{h}'(\lambda , v_z ) \nonumber \\ { i}''(\lambda , v_z , \mu)&= & -\psi''(\lambda ) f(\mu ) \mathcal{h}(\lambda , v_z ) - 2\psi'(\lambda ) f(\mu ) \mathcal{h}'(\lambda , v_z ) -\psi(\lambda ) f(\mu ) \mathcal{h}''(\lambda , v_z),\end{aligned}\ ] ] where the @xmath65 and @xmath66 symbols represent the first and second derivatives with respect to the wavelength . applying the previous developments and dropping the dependences of the variables to avoid crowding , the polarized flux in the weak - field weak - line approximation can be written as : @xmath67 \nonumber \\ f_{u}(\lambda ) & = & \frac{{c}^2}{2}\lambda_0 ^ 4 g \left [ \psi''(\lambda ) \overline{b_xb_yf\mathcal{h } } + 2\psi'(\lambda ) \overline{b_xb_yf\mathcal{h } ' } + \psi(\lambda ) \overline{b_xb_yf\mathcal{h } '' } \right ] \nonumber \\ f_{v}(\lambda)&=&{c}\lambda_0 ^ 2 g \left [ \psi'(\lambda ) \overline{b_zf\mathcal{h } } + \psi(\lambda ) \overline{b_zf\mathcal{h } ' } \right ] , \label{eq : fi_star}\end{aligned}\ ] ] where we remind that the overlines represent integration in the plane of the sky , as indicated in eq . ( [ eq : integration_sky ] ) . it is important to note that the area asymmetry of the polarized flux is zero as can be verified by computing the wavelength integral of the previous equations ( it is easy to verify that the integral is zero by using the properties of the hermite polynomials ) . the symbol @xmath68 stands for the continuum intensity flux and it is trivially given by : @xmath69 as expected , the stokes parameters depend on quantities averaged over the stellar surface . using the expansion on hermite polynomials of the absorption spectral line , the terms appearing in the previous equations can be written as the following summations : @xmath70 the functions @xmath71 are given by : @xmath72 it is important to note that eqs . ( [ eq : fi_star ] ) and ( [ eq : averages_surface ] ) are formally invariant for any magnetic field configuration , rotation profile and center - to - limb variation once the assumptions stated before ( sufficiently weak magnetic field , wavelength independent clv , unsaturated and weak gaussian line ) are fulfilled . the information about the magnetic configuration , rotation and the clv is encoded in the @xmath71 functions , which may depend on additional external parameters . this is the case of the dipolar field with solid body rotation ( see sec . [ sec : dipole ] ) , in which these functions depend on the orientation of the dipole and the angular velocity vector . the integrated stokes parameters also depend on terms that include the wavelength derivative of the @xmath73 function . thanks to the effective separation between wavelength and velocity , these are easily obtained from the previous ones by using known properties of the hermite polynomials . for instance : @xmath74 and the rest of terms are computed following the same strategy . when the star is not rotating , the only terms contributing to the summations in eq . ( [ eq : averages_surface ] ) are those with @xmath61 . likewise , the terms with derivatives of eqs . ( [ eq : averages_surface_d ] ) are always zero . the polarized fluxes of eqs . ( [ eq : fi_star ] ) can be used to explain stellar observations using a standard optimization method to fit the functional forms to the observations . however , we note that the expressions for the polarized fluxes depend on @xmath47 and @xmath48 , properties of the local line profiles to which we do not have direct access . it is possible , though , to drop the dependence on @xmath47 from the equations if we use the fact that @xmath75 , @xmath76 and @xmath77 can be written in terms of @xmath78 , @xmath79 and @xmath80 , which are , otherwise , observed quantities . this is the strategy followed by @xcite to write the weak - field weak - line approximation for a non - rotating star in terms of the observed fluxes ( and the parameters of the clv and the gaussian profile ) . so : @xmath81 which gives , after isolation of @xmath75 , @xmath76 and @xmath77 : @xmath82 \nonumber \\ f_{u}(\lambda ) & = & -\frac{{c}^2}{2 } \lambda_0 ^ 4 g \left [ \frac{\overline{b_xb_y f \mathcal{h}}}{\overline{f\mathcal{h } } } f''_\mathcal{i } + 2\left ( \frac{\overline{b_xb_y f \mathcal{h}'}}{\overline{f\mathcal{h } } } - \frac{\overline{b_xb_y f \mathcal{h}}~\overline{f\mathcal{h}'}}{\overline{f\mathcal{h}}^2 } \right ) f'_{i}\right . \nonumber \\ & + & \left . \left ( \frac{\overline{b_xb_y f \mathcal{h}''}}{\overline{f\mathcal{h } } } - \frac{\overline{b_xb_yf\mathcal{h}}~\overline{f\mathcal{h}''}}{\overline { f\mathcal{h}}^2 } + 2\frac{\overline{b_xb_yf\mathcal{h}}~\overline{f\mathcal{h}'}^2}{\overline { f\mathcal{h}}^3 } -2\frac{\overline{b_xb_y\mathcal{h}'}~\overline{f\mathcal{h}'}}{\overline { f\mathcal{h}}^2 } \right ) ( f_c - f_{i})\right ] \nonumber \\ f_{v}(\lambda ) & = & -{c } \lambda_0 ^ 2 g \left [ \frac{\overline{b_z f \mathcal{h}}}{\overline{f\mathcal{h } } } f'_{i } + \left ( \frac{\overline{b_z f \mathcal{h}}~\overline{f\mathcal{h}'}}{\overline{f\mathcal{h}}^2 } - \frac{\overline{b_z f \mathcal{h}'}}{\overline{f\mathcal{h}}}\right ) ( f_c - f_{i } ) \right ] \label{eq : fi_star_observer}\end{aligned}\ ] ] these expressions are the generalization of the weak - field weak - line approximation to an unresolved star with an arbitrary velocity and magnetic field on the surface . the dependency on the local absorption @xmath47 has disappeared once the observed @xmath83 is used . however , the local broadening is still present and has to be either inferred from the observations or fixed to a reasonable value . the set of eqs . ( [ eq : fi_star_observer ] ) represent a description of the stokes parameters in terms of observed quantities and also in terms of wavelength - dependent unknowns that are explicit functions of the field and velocity distribution on the star ( e.g. , @xmath84 ) . both contributions are conveniently separated . in principle , one could use these expressions to obtain information about the magnetism of the star following two strategies . the first one is to parameterize these terms by fixing a model for the magnetic structure and infer the values of the parameters . this is the approach we follow in the next section . the second one is to non - parametrically infer these wavelength - dependent unknowns directly using a suitable inversion procedure . since this would require additional regularization techniques , a full treatment of the problem would have to be done in a bayesian framework . an interesting property of the integrated stokes parameters when rotation is taken into account is that the circular polarization flux depends on the derivative with wavelength of the flux ( @xmath85 ) but also on the flux itself ( @xmath83 ) . likewise , the linear polarization fluxes depend on the flux and its first and second order derivative with respect to the wavelength . when particularizing the polarized flux of eq . ( [ eq : fi_star_observer ] ) to the non - rotating star , we recover the equations developed by @xcite . this is a consistency check of our equations . let us consider a star that is rotating with an angular velocity vector @xmath86 , as shown in fig . [ fig : fig_geom_dipole ] . in this section we use two different reference systems for our computations . the arbitrary reference system @xmath87 , shown in the left panel of fig . [ fig : fig_geom_dipole ] , is the one associated with an observer . the line - of - sight ( los ) is along the @xmath1 axis and the axis @xmath88 ( arbitrarily chosen by the observer ) defines the reference direction for positive stokes @xmath3 . the angular velocity vector is defined in this reference system by its modulus ( @xmath89 ) , the inclination angle ( @xmath90 ) and the azimuth ( @xmath27 ) with respect to the @xmath88 axis . the second reference system , @xmath21 , is obtained after a rotation of an angle @xmath27 around the @xmath1 axis , so that the angular velocity vector is now contained in the @xmath91 plane . this second reference system will help us in obtaining the final analytical expressions . one of the simplest non - trivial configuration of a stellar magnetic field is that of a dipole , whose axis is determined by a unit vector @xmath92 along the dipole moment and an intrinsic dipolar field . the three components of the unit vector @xmath92 on the cartesian system @xmath21 are : @xmath93 where @xmath94 is the inclination of the dipolar moment with respect to the los and @xmath95 is its azimuth with respect to the @xmath2 axis . in the arbitrary reference system @xmath87 , the azimuth is transformed to @xmath96 . the emergent stokes parameters @xmath97 and @xmath98 are then transformed to the arbitrary reference system following eq . ( [ eq : rotation_stokes ] ) . note that our notation can be translated to that used by @xcite making the transformations ( in format ours@xmath99theirs ) : @xmath100 , @xmath101 , @xmath102 , @xmath103 . therefore , the cosine of the inclination of the dipole axis with respect to the rotational axis , usually termed @xmath104 , is plainly given by the dot product @xmath105 , i.e. , @xmath106 . the magnetic field at the surface of the star of a dipole placed at the center of the star can be expressed in cartesian coordinates in the @xmath21 system as : @xmath107 \nonumber \\ b_y&=&-\frac{b_p}{2}\left [ e_y - 3y(x e_x+y e_y+z e_z ) \right ] \nonumber \\ b_z&=&-\frac{b_p}{2}\left [ e_z - 3z(x e_x+y e_y+z e_z ) \right],\end{aligned}\ ] ] where @xmath108 is the intensity of the magnetic field at the poles . an arbitrary point @xmath109 in the visible surface of the star can be written in terms of the polar coordinates on the plane of the sky @xmath29 as : @xmath110 where we have selected the positive solution for @xmath111 to isolate the visible half of the star . for simplicity , we assume that the star is rotating as a solid body because it greatly simplifies the equations ( the introduction of a non - rigid rotation is formally identical , except for the computation of the @xmath71 integrals ) , so that : @xmath112 in terms of the polar coordinates on the stellar surface , the component of the velocity along the los in the @xmath21 reference system is given by : @xmath113 taking all these definitions into account and after some tedious algebra , it is possible to find relatively simple expressions for the @xmath71 integrals , which now depend on the set of parameters @xmath114 . the procedure is to plug the functional form of the dipolar magnetic field and carry out the angular and radial integrals on the visible disk . we end up with : @xmath115 \right.\nonumber \\ & + & b_1(i , j,\alpha,\omega , n ) \left [ ( 1-a - b ) f_3(n ) + a f_4(n ) + b f_5(n)\right ] \nonumber \\ & + & c_1(i , j,\alpha,\omega , n ) \left [ ( 1-a - b ) f_2(n+1 ) + a f_3(n+1 ) + b \left . f_4(n+1)\right ] \right\ } \nonumber \\ x_2(i , j,\alpha,\omega , n , a , b ) & = & ( \omega \sin \alpha)^n \left\{a_2(i , j,\alpha,\omega , n ) \left [ ( 1-a - b ) f_1(n+2 ) + a f_2(n+2 ) + b f_3(n+2)\right ] \right . + & b_2(i , j,\alpha,\omega , n ) \left [ ( 1-a - b ) f_1(n+4 ) + a f_2(n+4 ) + b f_3(n+4)\right ] \nonumber \\ & + & c_2(i , j,\alpha,\omega , n ) \left [ ( 1-a - b ) f_2(n+1 ) + a f_3(n+1 ) + b f_4(n+1)\right ] \nonumber \\ & + & d_2(i , j,\alpha,\omega , n ) \left [ ( 1-a - b ) f_2(n+3 ) + a f_3(n+3 ) + b f_4(n+3)\right ] \nonumber \\ & + & \left . e_2(i , j,\alpha,\omega , n ) \left [ ( 1-a - b ) f_1(n ) + a f_2(n ) + b f_3(n)\right ] \right\ } \nonumber \\ x_3(i , j,\alpha,\omega , n , a , b ) & = & ( \omega \sin \alpha)^n \left\ { a_3(i , j,\alpha,\omega , n ) \left [ ( 1-a - b ) f_1(n+2 ) + a f_2(n+2 ) + b f_3(n+2)\right ] \right . \nonumber \\ & + & b_3(i , j,\alpha,\omega , n ) \left [ ( 1-a - b ) f_1(n+4 ) + a f_2(n+4 ) + b f_3(n+4)\right ] \nonumber \\ & + & c_3(i , j,\alpha,\omega , n ) \left [ ( 1-a - b ) f_2(n+1 ) + a f_3(n+1 ) + b f_4(n+1)\right ] \nonumber \\ & + & d_3(i , j,\alpha,\omega , n ) \left [ ( 1-a - b ) f_2(n+3 ) + a f_3(n+3 ) + b f_4(n+3)\right ] \nonumber \\ & + & \left . e_3(i , j,\alpha,\omega , n ) \left [ ( 1-a - b ) f_1(n ) + a f_2(n ) + b f_3(n)\right ] \right\ } \nonumber \\ x_4(i , j,\alpha,\omega , n , a , b ) & = & ( \omega \sin \alpha)^n a_4(i , j,\alpha,\omega , n ) \left [ ( 1-a - b ) f_1(n ) + a f_2(n ) + b f_3(n)\right ] . \label{eq : x_functions}\end{aligned}\ ] ] these functions depend on the orientation of the dipole given by the angles @xmath94 and @xmath95 , the angular velocity parameters and the parameters of the clv and the index @xmath56 of the summation . the following functions take into account the dependence on the orientation of the dipole and only depend on angular integrals : @xmath116 -6\sin^2 i \cos^2 j s(n ) + 6 s(n+2 ) \nonumber \\ b_2(i , j,\alpha,\omega , n ) & = & 9 \cos^2 i \left [ s(n+2 ) - s(n ) + 2s(n+4 ) \right ] \nonumber \\ & + & 9 \sin^2 i \cos^2 j \left [ s(n ) - 4 s(n+2 ) + 4s(n+4)\right ] \nonumber \\ & + & 9s(n+2 ) - 18s(n+4 ) \nonumber \\ c_2(i , j,\alpha,\omega , n ) & = & 6 \cos i \sin i \sin j s(n+1 ) \nonumber \\ d_2(i , j,\alpha,\omega , n ) & = & 18 \cos i \sin i \sin j \left [ s(n+1 ) - 2s(n+3 ) \right ] \nonumber \\ e_2(i , j,\alpha,\omega , n ) & = & \sin^2 i \cos ( 2j ) s(n ) \nonumber \\ a_3(i , j,\alpha,\omega , n ) & = & -3 \sin^2 i \cos j \sin j s(n ) \nonumber \\ b_3(i , j,\alpha,\omega , n ) & = & 18 \sin^2 i \cos j \sin j \left [ s(n+2 ) - s(n+4 ) \right ] \nonumber \\ c_3(i , j,\alpha,\omega , n ) & = & -3 \cos i \sin i \cos j s(n+1 ) \nonumber \\ d_3(i , j,\alpha,\omega , n ) & = & 18 \cos i \sin i \cos j \left [ s(n+1 ) - s(n+3 ) \right ] \nonumber \\ e_3(i , j,\alpha,\omega , n ) & = & \sin^2 i \cos j \sin j s(n ) \nonumber \\ a_4(i , j,\alpha,\omega , n ) & = & s(n ) , \label{eq : abcde_functions}\end{aligned}\ ] ] with the angular integral @xmath117 given by : @xmath118 \sqrt{\pi } \gamma \left ( \frac{1+n}{2 } \right ) \gamma^{-1 } \left ( 1+\frac{n}{2 } \right ) , \label{eq : s_functions}\ ] ] which is valid for @xmath119 . likewise , the radial integrals @xmath120 are independent on the orientation of the dipole and are computed from : @xmath121 which are valid if @xmath122 . once the expressions for @xmath71 have been particularized to the case of a rotating dipole , it is possible to analyze in detail the shape of the profiles according to eq . ( [ eq : fi_star ] ) . we analyze the symmetry properties of the circular polarization profile , although this can be extended easily to the linear polarization profiles . recall that the profile @xmath75 is a symmetric function of wavelength with respect to the rest wavelength of the line , while @xmath76 is antisymmetric . according to the properties of the hermite polynomials , the terms with @xmath123 in the expression for @xmath124 in eq . ( [ eq : averages ] ) are symmetric , while those with @xmath125 are always antisymmetric . likewise , for @xmath126 , the terms with @xmath123 are antisymmetric , while those with @xmath125 are symmetric . consequently , the polarized flux @xmath127 can be written as the addition of two functions , one symmetric and the other one antisymmetric : @xmath128 where @xmath129 \nonumber \\ f_{v}^\mathrm{antisym}(\lambda ) & = & { c}\lambda_0 ^ 2 g \left [ \psi'(\lambda ) \overline{b_zf\mathcal{h}}_\mathrm{sym } + \psi(\lambda ) \overline{b_zf\mathcal{h}'}_\mathrm{antisym } \right].\end{aligned}\ ] ] taking into account the definition of the @xmath130 function of eq . ( [ eq : x_functions ] ) , it is easy to verify that the symmetric contribution to @xmath127 depends only on @xmath131 , while the antisymmetric contribution depends only on @xmath132 . consequently , we will find standard antisymmetric circular polarization profiles whenever @xmath133 , while they will be fully symmetric when @xmath134 . stellar rotation is taken into account by keeping fixed the los and rotating the vector @xmath92 around the angular rotation velocity @xmath86 . this is accomplished by computing the new rotated dipolar moment unit vector as : @xmath135 where the rotation matrix is given by : @xmath136,\end{aligned}\ ] ] with @xmath137 indicating the stellar rotation phase and taking values between @xmath138 and @xmath139 . note that our initial of phases is an arbitrary position of the vectors , while using the standard notation , the initial phase coincides with the @xmath86 and @xmath92 vectors lying in the same plane parallel to the @xmath1 axis . this implies a phase shift between our @xmath137 angle and the standard definition that is given by : @xmath140 finally , the inclination and azimuth angles with respect to the @xmath21 reference system are obtained using the standard relations , taking into account that @xmath141 is a unit vector : @xmath142 , \qquad j_\mathrm{rot } = \arctan \frac{(e_\mathrm{rot})_y}{(e_\mathrm{rot})_x}\ ] ] all the functions needed in the previous equations are well behaved and can be computed numerically without much complexity . the main difficulty resides on the fact that , if the rotation velocity @xmath143 is large , the convergence of the expansion in eqs . ( [ eq : averages_surface ] ) is slow and oscillating , thus making it hard to compute . the main problem is that the hermite polynomials grow in amplitude while the coefficients decrease slow . only after many terms ( the number of which depend on the ratio between the rotation velocity and the broadening ) , the coefficients start to compensate for the increase in the amplitude of the hermite polynomials and convergence is eventually obtained . this is a well - known defect of the gram - charlier expansion ( e.g. , * ? ? ? when the rotation velocity is much larger than the doppler width of the local spectral line ( larger than a factor @xmath144 ) , the compensation between the amplitude increase of the hermite polynomials and the coefficients of the series expansion occurs for very large values of @xmath56 [ see eqs . ( [ eq : averages_surface ] ) ] . at this moment , the double precision computations start to become unprecise and the terms in the expansion do not compensate adequately . this purely numerical drawback can be overcome if arbitrary precision methods are used or if more elaborate methods like the edgeworth expansion are adapted to our problem @xcite . in any case , we have been able to numerically compute expansions up to @xmath145 . in other words , our method with a double precision summation of the expansions shown in eqs . ( [ eq : averages_surface ] ) is presently applicable for rotation velocities not larger than 8 times the doppler width of the local spectral line . in any case , it is advantageous to use the following set of modified hermite polynomials : @xmath146 which fulfill the following recursion formula : @xmath147 computing the hermite polynomials with this recursion formula , the expansions of eqs . ( [ eq : averages_surface ] ) can be extended to much larger values of @xmath56 . the advantage of using the modified @xmath148 hermite polynomials is that their amplitude grow much slower with @xmath56 than @xmath55 due to the inclusion of the term @xmath149 and @xmath150 . this facilitates the computation of each term in the expansions without numerical overflows . we note that the hermite polynomials are computed using their recursion relations @xcite . from the computational point of view , once the rotation angular velocity of the star is set , the wavelength dependence of eqs . ( [ eq : averages_surface ] ) and their derivatives in terms of the hermite polynomials can be precomputed once and used in all subsequent calculations . this greatly accelerates the process because only the @xmath151 functions have to be computed for a given orientation of the dipolar moment and added until convergence . a free computer program in idl is available to compute the observed stokes fluxes for a rotating dipole using the expressions in this paper . in this non - optimized version of the code , we synthesize profiles evaluated at 500 wavelength points with a step of 8 m in @xmath1135 - 40 ms . given our assumptions ( weak - field , weak - line , fixed center - to - limb variation , global dipolar field ) , one of the most direct applications of our equations is to study the magnetism of hot stars . for this reason , we use typical values of the broadening and rotation velocity of such stars ( e.g. , * ? ? ? * ) . we show in fig . [ fig : examples ] two examples for a dipolar magnetic field with different field configurations . both computations share the parameters for the clv ( @xmath152 and @xmath153 ) , the magnetic field strength in the pole ( @xmath154 g ) and the rotation velocity on the surface ( @xmath155 km s@xmath156 ) . the artificial selected spectral line is centered at @xmath157 and has an effective land factor of @xmath9 and its equivalent for linear polarization of @xmath158 . since the magnetic field takes its maximum value at the poles , the spectral line can be safely considered in the weak - field regime . the line broadening is set to 0.1 , typical of an observation at a spectral resolution of @xmath159 of a line with a microturbulent width of @xmath115 km s@xmath156 . these values of the broadening and rotation velocity are close to the numerical limit that we presently have for computing the summations of eqs . ( [ eq : averages ] ) . given this limitation , our scheme is presently not applicable to fast rotators with narrow local absorption profiles . the computed intensity flux normalized to the continuum flux is shown in the upper panel of each figure . it is independent of the phase since , under the weak field regime , the intensity does not depend on the magnetic field and we have assumed that the star is structureless . however , it takes into account the rotation broadening and the presence of a clv . concerning the linear polarization fluxes , we see maximum amplitudes around 0.02% , while this increases to values around 100 times larger for circular polarization . this is a direct consequence of the fact that linear polarization appears only in second order . we have generalized the expressions for the stokes parameters in the weak - field weak - line regime to the stellar case in which rotation is taken into account . this produces an intensity flux profile that is broadened by rotation while it is insensitive to the magnetic field . as a consequence , our formalism can only be applied to stars in which the intensity profile is not observed to be modulated by rotation . however , the circular and linear polarized fluxes are sensitive to the topology of the magnetic field . we have demonstrated that the inclusion of doppler shifts generates a dependence of circular polarization of the intensity flux itself , and not only on its wavelength derivative . likewise , the linear polarization profiles depend now on the first and second wavelength derivative and on the intensity flux itself . we have verified also that the expressions developed in this paper simplify to those obtained by @xcite in the non - rotating dipole case . the general weak - field expressions have been particularized to a rotating dipole . the extension to a more general multipolar magnetic field is relatively easy and can be done fast if analytical expressions for the angular and radial integrals exist . in subsequent papers of this series we will deal with thin magnetized disks in keplerian rotation and unresolved structures in the quiet sun . likewise , we will also explore the possibility of using more elaborate expansions to improve the convergence . we thank r. manso sainz for fruitful discussions . financial support by the spanish ministry of economy and competitiveness through projects aya2010 - 18029 ( solar magnetism and astrophysical spectropolarimetry ) and consolider - ingenio 2010 csd2009 - 00038 is gratefully acknowledged . , v. , del toro iniesta , j. c. , lvarez - herrero , a. , domingo , v. , bonet , j. a. , gonzlez fernndez , l. , lpez jimnez , a. , pastor , c. , gasent blesa , j. l. , mellado , p. , piqueras , j. , aparicio , b. , balaguer , m. , ballesteros , e. , belenguer , t. , bellot rubio , l. r. , berkefeld , t. , collados , m. , deutsch , w. , feller , a. , girela , f. , grauf , b. , heredero , r. l. , herranz , m. , jernimo , j. m. , laguna , h. , meller , r. , menndez , m. , morales , r. , orozco surez , d. , ramos , g. , reina , m. , ramos , j. l. , rodrguez , p. , snchez , a. , uribe - patarroyo , n. , barthol , p. , gandorfer , a. , knoelker , m. , schmidt , w. , solanki , s. k. , & vargas domnguez , s. 2011 , sol . phys . , 268 , 57 , s. k. , barthol , p. , danilovic , s. , feller , a. , gandorfer , a. , hirzberger , j. , riethmller , t. l. , schssler , m. , bonet , j. a. , martnez pillet , v. , del toro iniesta , j. c. , domingo , v. , palacios , j. , knlker , m. , bello gonzlez , n. , berkefeld , t. , franz , m. , schmidt , w. , & title , a. m. 2010 , apjl , 723 , l127 , j. 2011 , in astronomical society of the pacific conference series , vol . 437 , solar polarization 6 , ed . j. r. kuhn , d. m. harrington , h. lin , s. v. berdyugina , j. trujillo - bueno , s. l. keil , & t. rimmele , 83

in this paper we are concerned with the study of the three - dimensional map @xmath6 defined by @xmath7 with three parameters , @xmath8 , @xmath9 , and @xmath10 . this map is a natural generalization of the famous two - dimensional hnon map @xcite ; indeed , like the latter , the map ( [ eq:3hm1 ] ) is quadratic , has constant jacobian @xmath11 and moreover , when @xmath12 , reduces to the two - dimensional hnon map . therefore , it is natural to call ( [ eq:3hm1 ] ) the _ 3d hnon map_. " we will study a number of properties of ( [ eq:3hm1 ] ) , including : 1 . its nontrivial dynamics and , most importantly , the existence of _ wild - hyperbolic _ strange attractors ; 2 . its origination in homoclinic bifurcations ; 3 . its structure as a member of the family of three - dimensional quadratic maps with constant jacobian . three - dimensional maps have not been as widely studied as one- and two - dimensional maps , though in recent years there has been a considerable increase in their study . one of important reasons for this fact is that multi - dimensional dynamical systems ( the dimension of the phase space is at least four for flows and three for maps ) can exhibit complicated dynamics that is decidedly distinct from lower - dimensional cases . in particular they can possess a new variety of strange attractor called a _ wild - hyperbolic attractor _ @xcite . wild - hyperbolic attractors have no uniform hyperbolic structure ; moreover , they may contain homoclinic tangencies and coexisting periodic orbits of different types . however , these attractors contain no stable invariant subsets such as attracting periodic orbits . most crucially , the latter property holds for open regions in the space of smooth dynamical systems . thus , wild - hyperbolic attractors are very interesting objects and their detection in concrete models is a fundamental problem of nonlinear dynamics . most of the strange attractors occurring in two - dimensional , smooth maps ( or three - dimensional flows ) are not structurally stable since they possess homoclinic tangencies . for such systems , even though the orbits are observed to be quite chaotic for some parameter values , very small changes in the parameters may destroy the aperiodicity and give rise to periodic behavior . such parameter regions are called _ stability windows_. even if these windows are not observable `` in numerical computations , their appearance ( under arbitrarily small perturbations ) is inevitable because of the prevalence of homoclinic tangencies . such attractors were called _ quasi - attractors _ in @xcite . one of the most challenging problems in dynamical systems is to distinguish aperiodic motion from extremely long periodic motion . moreover , since most of the chaotic attractors we see in applications are definitely quasi - attractors , it is unclear in which sense they are chaotic , or how one can define and measure the ' ' probability " for a quasi - attractor to be chaotic . however , it is well - known that there are strange attractors that are free of these problems ; for example , hyperbolic attractors and the lorenz attractor . though the latter is not structurally stable @xcite,@xcite ( in contrast to hyperbolic attractors ) it contains no stable periodic orbits , every orbit in it has positive maximal lyapunov exponent , etc . moreover , these properties are robust @xcite,@xcite . the reason is that lorenz attractor possesses a pseudo - hyperbolic structure . in terms of turaev and shilnikov @xcite this means that the following properties hold : 1 . there is a direction in which the flow is strongly contracting ( `` strongly '' means that any possible contraction in transverse directions is always strictly weaker ) ; and 2 . transverse to this direction the flow expands areas . the robustness of pseudo - hyperbolicity ensures the robustness of the chaotic behavior of orbits in the lorenz attractor . moreover , pseudo - hyperbolicity is maintained even for ( small , indeed ) time - periodic perturbations so that a periodically forced lorenz attractor is an example of a genuinely strange " attractor since it contains no stable invariant subsets ( for more discussion see @xcite,@xcite,@xcite ) . however it is necessary to note that homoclinic tangencies arise inevitably under such perturbations . consequently systems with periodically forced lorenz attractors fall into newhouse regions and demonstrate an extremely rich dynamics @xcite,@xcite . such attractors were called wild - hyperbolic attractors in @xcite ; in this paper the theory of wild - pseudo - hyperbolic attractor was discussed and an example of a wild - hyperbolic spiral attractor was constructed . generalizing these ideas , we say that wild - hyperbolic attractors possess the following two main properties that distinguish them from other attractors : ( 1 ) wild - hyperbolic attractors allow homoclinic tangencies ( hence , they belong to newhouse regions ) ; ( 2 ) every such attractor and all nearby attractors ( in the @xmath13-topology with @xmath14 ) have no stable periodic orbits . one of the most felicitous properties of wild - hyperbolic attractors is that they can be created in _ local bifurcations _ of periodic orbits ; in particular , when the orbit has three or more floquet multipliers on the unit circle @xcite . consequently , the construction of wild - hyperbolic attractors in concrete models is possible when the model contains sufficiently many parameters to permit such a degeneracy . indeed , the map ( [ eq:3hm1 ] ) is sufficiently complex and has been shown to possess wild - hyperbolic attractors @xcite . moreover , numerical simulations have shown that these attractors persist in quite big parameter domains and can be order - one in size @xcite . in [ sec : attractors ] we will show that ( [ eq:3hm1 ] ) has a small , wild - hyperbolic lorenz attractor in a ( small ) parameter domain near the point @xmath15 at which a fixed point with multipliers @xmath16 exists . the general case of fixed points with such a triplet of multipliers was considered in @xcite , where it was shown that two distinct cases occur . one of these cases has a local flow normal form that coincides with the shimitsu - marioka system ( see ( [ = sm ] ) ) . this system has a lorenz attractor in some parameter domains @xcite,@xcite . as was noted in @xcite a periodically forced lorenz system ( when the perturbation is rather small ) can give rise to a wild - hyperbolic attractor . moreover , this does not destroy the pseudo - hyperbolicity of the attractor @xcite,@xcite . note that the flow normal form only approximates the map and , since the residual terms are exponentially small in time , it is reasonable that the map will have a wild - hyperbolic attractor ( which is small , in fact , because the approximation must be very good ) , see more discussions in @xcite . the second property of ( [ eq:3hm1 ] ) that we mentioned above is that this map can arise as a local normal form for multi - dimensional systems with a homoclinic tangency . this shows that such maps are not exotic and gives a reason for an extensive study of their properties . it is well - known that quadratic maps appear naturally when studying bifurcations of quadratic homoclinic tangencies @xcite,@xcite,@xcite,@xcite,@xcite . they appear as normal forms of rescaled first return maps to the neighborhood of a point on the homoclinic orbit . there are four standard forms for such one- and two - dimensional homoclinic maps : " 1 . @xmath17 ( the logistic map ) ; 2 . @xmath18 ( the standard hnon map ) ; 3 . @xmath19 ( the mira map ) ; 4 . @xmath20 ( a generalized hnon map ) . the third case is a two - dimensional endomorphism and is one of the well - studied polynomial maps ( see , e.g. @xcite ) ; we call it the mira map " following @xcite ) . the fourth case can be obtained as a small perturbation ( due to @xmath21-terms ) of the standard hnon map . this map was obtained in a study of bifurcations of two - dimensional diffeomorphisms with a homoclinic tangency to a saddle fixed point ( with unit jacobian ) @xcite,@xcite . the 3d hnon map ( [ eq:3hm1 ] ) is absent from this list , and in [ sec : local][sec : rescaling ] we will fill this gap . in [ sec : local ] we study bifurcations of a three - dimensional diffeomorphism with quadratic homoclinic tangency to a saddle - focus fixed point . for a three - dimensional map , such points can be of two types : @xmath22 or @xmath23 ( sometimes called _ type a _ and _ type b _ , respectively ) . recall that @xmath24 saddle is a fixed point with @xmath25 , @xmath26 , and @xmath27 . in our case the two - dimensional manifold will be assumed to correspond to a complex conjugate pair of multipliers . we will consider the case where the the product the absolute value of the multipliers , @xmath28 is one . the bifurcations in the general case , where @xmath29 , were studied in @xcite,@xcite . we will use their rescaling method " to study our case as well . the essence of this method is that a suitable change of coordinates and parameters in a neighborhood of the homoclinic orbit is found to bring the map into a standard form . if the homoclinic tangency is assumed to be quadratic the resulting map is also typically quadratic . for example , for a type @xmath22 saddle - focus with @xmath30 , the rescaled first return maps coincide with the standard hnon map up to asymptotically small terms ( when the time of return tend to infinity ) . for a saddle - focus of type @xmath23 , the corresponding rescaled first return map is the mira map . note the condition @xmath31 leads to maps that are effectively two - dimensional ; consequently all three - dimensional volumes near the fixed point are collapsed . this is not the case when @xmath32 for the unperturbed map when three - dimensional rescaled maps can appear a priori . indeed , we will show that when @xmath33 the map ( [ eq:3hm1 ] ) occurs as the normal form of the rescaled map for type @xmath22 saddle - focus ; and for the type @xmath23 case the corresponding normal form is ) is also the rescaled first return map for a four - dimensional diffeomorphism that has a quadratic homoclinic tangency to a saddle - focus fixed point with two pairs of complex conjugate leading multipliers @xcite,@xcite,@xcite . ] @xmath34 both ( [ eq:3hm1 ] ) and ( [ eq:3hm2 ] ) are quadratic three - dimensional maps with constant jacobian @xmath35 . it is natural to call them the three - dimensional hnon maps of the _ first _ and _ second kind _ , respectively . the quadratic 3d hnon maps are elements of the group of polynomial automorphisms . the study of the dynamics of polynomial mappings has a long history applied dynamics ; for example , they often used in the study of particle accelerators @xcite . the study of _ cremona maps _ , i.e. , polynomial maps with constant jacobian , also has intrinsic mathematical interest @xcite . an interesting mathematical problem is to obtain a normal form for an arbitrary degree-@xmath36 cremona map . this problem is unsolved ; the major obstruction is that it is not known in general that such a map has a polynomial inverse . this is the content of the much studied , jacobian conjecture " : let @xmath37 be a cremona map . then @xmath38 is bijective and has a polynomial inverse . this conjecture is still open . the set of polynomial maps with polynomial inverses is called the _ affine cremona group_. for the case of the plane , the structure of this group was obtained by jung @xcite . jung s theorem states that planar polynomial diffeomorphisms are `` tame ; '' that is , they can be written as a finite composition of `` elementary '' ( or triangular ) and affine maps . more recently friedland and milnor @xcite showed that any map in this group is either conjugate to a composition of generalized hnon maps , @xmath39 where @xmath40 is a polynomial and @xmath41 , or else the map is dynamically trivial . not much is known about the structure of the affine cremona group for higher dimensions , though it was recently proved that the three - dimensional cremona map @xmath42 is not tame @xcite . thus the normal form for the 3d affine cremona group must consist of more than tame maps . however , it is still possible that the nontame maps are dynamically trivial , like the above example . another interesting problem concerns the approximation of smooth diffeomorphisms by polynomial diffeomorphisms . one result along this direction is for the class of @xmath43 symplectic diffeomorphisms , which , as turaev has shown @xcite , can be approximated by compositions of the symplectic cremona maps @xmath44 where @xmath45 and @xmath46 is a polynomial . related approximation methods are used in the study of particle accelerators @xcite , and results have also been obtained for analytic symplectic maps @xcite . some work has been done on finding normal forms for quadratic symplectic and quadratic volume - preserving maps . moser showed that every quadratic symplectic map in @xmath47 can be written as the composition of an affine symplectic map and a symplectic shear @xcite . for the four - dimensional case , he obtained a normal form . for the case of volume - preserving maps , it was shown in @xcite,@xcite that any quadratic three - dimensional , volume - preserving map whose inverse is also quadratic and whose dynamics is non - trivial can be transformed ( by affine transformations ) to the form @xmath48 where @xmath49 , @xmath50 and one of the parameters @xmath51 or @xmath52 can be eliminated ( by a coordinate shift ) . by analogy with the planar results , there are two additional cases but they have trivial dynamics ( see [ sec : quadratic ] ) . we also note that it is not hard to construct three - dimensional quadratic diffeomorphisms that do not fall into this classification because they have an inverse of higher degree . for example , the diffeomorphism @xmath53 has a quartic inverse . note that both of the maps ( [ eq:3hm1 ] ) and ( [ eq:3hm2 ] ) fall into the form ( [ eq:3hm3 ] ) . thus the results of @xcite show that when @xmath54 both of the 3d hnon maps occur as normal forms for quadratic elements of the affine cremona group . moreover , even when @xmath55 the map ( [ eq:3hm2 ] ) is a normal form for a certain class of quadratic maps @xcite . tatjer showed that every quadratic map , @xmath38 , for which @xmath56 is also quadratic , but has no invariant linear foliation is linearly conjugate to ( [ eq:3hm2 ] ) . in [ sec : statement ] , we will generalize the results of @xcite to show that ( [ eq:3hm3 ] ) is the the normal form for quadratic diffeomorphisms with quadratic inverses when @xmath55 as well . consider a three - dimensional diffeomorphism @xmath57 for some @xmath58 that satisfies the following conditions ( see fig . [ fig : type21 ] ) : 1 . @xmath0 has a type @xmath22 saddle - focus fixed point @xmath59 with multipliers @xmath60 where @xmath61 ( w.l.o.g . @xmath62 ) , ; 2 . @xmath63 . 3 . the stable , @xmath64 , and unstable , @xmath65 , invariant manifolds of the fixed point @xmath59 have a quadratic homoclinic tangency at the points of a homoclinic orbit @xmath66 . saddle - focus for the map @xmath0 under assumputions ( a)-(c ) . [ fig : type21],width=288 ] diffeomorphisms that are close to @xmath0 and satisfy conditions ( b ) and ( c ) comprise , in the space of @xmath13 diffeomorphisms , a codimension - two , locally - connected surface @xmath67the bifurcation surface . our main problem is to study bifurcations of periodic orbits in families of diffeomorphisms that transversely cross @xmath67 . we denote such a family by @xmath4 so that @xmath0 satisfies ( a)-(c ) . the first question that arises is the optimal choice of the governing parameters @xmath68 . it is evident that the set of such parameters must include both a parameter @xmath69 of the splitting of the initial quadratic homoclinic tangency and a parameter @xmath70 , similar to @xmath71 , that varies the jacobian of the saddle - focus fixed point . however , it was shown in @xcite that diffeomorphisms on @xmath67 possess continuous ( topological ) conjugacy invariants on the set of non - wandering orbits , so - called @xmath72-moduli . the most important modulus is the angular argument @xmath73 of the complex multipliers of the saddle - focus ( see [ sec : rescaling ] ) . by definition , @xmath72-moduli are natural governing parameters . therefore , we include the @xmath72-modulus @xmath73 into the set of governing parameters . thus , we take as the third parameter @xmath74 . more generally , we must consider a three - parameter family @xmath75 , @xmath76 , that is transversal to @xmath67 at @xmath5 and such that @xmath77 consider a sufficiently small and fixed neighbourhood @xmath78 of the orbit @xmath66 . it can be represented as a union of a small neighbourhood @xmath79 of @xmath59 with a number of small neighbourhoods of those points of @xmath66 which do not belong to @xmath79 . in the family @xmath75 we study bifurcations of _ single - round periodic orbits _ from @xmath78that is orbits periodic that enter @xmath79 exactly once . every point of such an orbit of period @xmath80 can be considered as a fixed point of the corresponding first return map @xmath81 along orbits lying in @xmath78 . in [ sec : rescaling ] , we will prove the following result about the form of this first return map in some domains of the parameters . [ tm:1 ] let @xmath75 be the three parameter family defined above . then , in any neighborhood of the origin , @xmath5 , of the space of parameters , there exist infinitely many open domains , @xmath82 , such that the first return map @xmath81 takes the form @xmath83 where @xmath84 are rescaled coordinates and the domains of definition of the new coordinates and the parameters @xmath85 are arbitrarily large and cover , as @xmath86 , all finite values . furthermore , the jacobian , @xmath10 , is positive or negative depending on the orientability of the map @xmath81 . the terms denoted @xmath87 are functions ( of the coordinates and parameters ) that tend to zero as @xmath88 along with all derivatives up to order @xmath89 . theorem [ tm:1 ] shows that the 3d hnon map ( [ eq:3hm1 ] ) is a normal form for this homoclinic bifurcation , since it appears when we omit the @xmath87-terms in ( [ eq : t07 + 1 ] ) . we will see that the new parameters @xmath90 and @xmath10 can be given as functions of @xmath91 by @xmath92 where @xmath93 is some coefficient ( an invariant of @xmath0 ) and @xmath94 . the exact relations will be given in formulas ( [ mpar * ] ) , ( [ eq : t04 ] ) and ( [ eq : t06 ] ) below . thus , we can study the bifurcations of single round periodic orbits in the family @xmath75 , when @xmath95 , by studying the bifurcations of fixed points of the 3d hnon map ( [ eq:3hm1 ] ) . certainly , the problem of studying bifurcations in the 3d hnon map is much simpler than the corresponding problem for the original first return maps . moreover , the 3d hnon map provides a simple , standard model for which various well - tried methods ( including numerical ones ) can be applied . however , the bifurcation problem as whole for the 3d hnon map is very complicated and here we focus only on two aspects . in [ sec : attractors ] we first give the simple construction of bifurcation surfaces , lines and points corresponding , respectively , to codimension one , two and three bifurcations of the fixed points . then we consider the problem of the existence of wild - hyperbolic strange attractors . we will obtain the following result . [ tm:2 ] let @xmath75 be the three parameter general family defined as before . then , in any neighborhood of the origin @xmath5 of the space of parameters , there exist infinitely many open domains @xmath96 such that the diffeomorphism @xmath75 of the family at @xmath97 has a wild - hyperbolic strange attractor of lorenz - type . the proof of this theorem is given in [ sec : attractors ] following the analysis of @xcite . the main idea of the proof is to apply the results of paper @xcite where it was shown that maps having a fixed point with the triplet @xmath98 of multipliers can , in some cases , be reduced to a flow normal form that coincides with the shimitsu - marioka system ( see ( [ fl1 ] ) ) which possesses a lorenz attractor for some parameter domains @xcite,@xcite . our verification of these facts for the corresponding fixed point of the 3d hnon map allows us to obtain th . [ tm:2 ] . a justification of the fact that the attractor in the 3d hnon map is actually the wild - hyperbolic lorenz - type attractor is given in @xcite . consequently , we will omit these details . solving the bifurcation problem for the family @xmath75 allows us to solve almost automatically an analogous problem for a diffeomorphism @xmath99 ( @xmath100 with a type @xmath23 saddle focus . specifically , assume that @xmath101 satisfies ( see fig . [ fig : type12 ] ) : 1 . @xmath101 has a type @xmath23 fixed point @xmath102 with multipliers @xmath103 where @xmath104 ( w.l.o.g . @xmath105 ) ; 2 . @xmath106 ; and 3 . the stable and unstable invariant manifolds of @xmath102 have a quadratic homoclinic tangency at the points of some homoclinic orbit @xmath107 . saddle - focus for the map @xmath101 under assumptions ( a@xmath108)-(c@xmath108 ) . [ fig : type12],width=288 ] as in the case of the type @xmath22 saddle - focus , we consider a three - parameter family @xmath109 where @xmath110 . we suppose here that @xmath111 is a splitting parameter , @xmath112 . [ tm:3 ] let @xmath109 be the three - parameter general family defined above . then , in any neighborhood of @xmath113 there exist infinitely many open domains , @xmath114 , such that the first return map @xmath115 takes the form @xmath116 where @xmath84 are rescaled coordinates and the domains of definition of the new coordinates and the parameters @xmath85 are arbitrarily large and cover , as @xmath86 , all finite values . furthermore , the jacobian , @xmath10 , is positive or negative depending on the orientability of the map @xmath117.the terms denoted @xmath87 are functions ( of the coordinates and parameters ) that tend to zero as @xmath88 along with all derivatives up to order @xmath89 . the proof immediate from th . indeed , the inverse @xmath118 is , evidently , a diffeomorphism satisfying the original conditions ( a(c ) . thus , the map ( [ eq : t07 + 1 ] ) is , in this case , the rescaled map for the inverse first return map @xmath119 . the last can be be easily written in form ( [ eq : t07 + 2 ] ) , as we show explicitly below . the inverse of ( [ eq:3hm1 ] ) is of interest in any case . since ( [ eq:3hm1 ] ) has the constant jacobian @xmath10 it is invertible whenever @xmath120 . this inverse is @xmath121 this map can be rewritten in the `` standard '' form of @xcite , if we define new variables @xmath122 and new parameters @xmath123 then the resulting map has precisely the form ( [ eq:3hm2 ] ) . as we stated above map is well - known in homoclinic dynamics : it was found to be the rescaled first return map for four - dimensional diffeomorphisms that have a quadratic homoclinic tangency to a saddle - focus fixed point with two pairs of complex conjugate leading multipliers @xcite,@xcite,@xcite . the dynamics of each of the maps ( [ eq:3hm1 ] ) and ( [ eq:3hm2 ] ) is extremely rich . however , for the volume - contracting case , we note that these maps demonstrate distinct types of stable dynamics . for example , ( [ eq:3hm1 ] ) gives rise to attractors similar to those of the lorenz system , as we will see in [ sec : attractors ] . however , ( [ eq:3hm2 ] ) need not have attractors of this wild " type , although certainly , a `` lorenz repeller '' will be present . nevertheless , we can expect existence of extremely small " lorenz - type attractors in ( [ eq:3hm2 ] ) by the following argument . as we will see in [ sec : attractors ] , the map ( [ eq:3hm2 ] ) can have a fixed point with multipliers @xmath124 at certain values of its parameters . the flow normal form near this point can have a homoclinic loop to a saddle - focus and , thus , ( [ eq:3hm2 ] ) can have a fixed point @xmath59 with multipliers @xmath125 , where @xmath126 and @xmath127 , such that the manifolds @xmath128 and @xmath129 will have a ( quadratic ) homoclinic tangency . consequently , we can apply th . [ tm:2 ] to this situation giving immediately the promised extremely small " attractors . nevertheless , as it is shown in @xcite the lorenz - type wild - hyperbolic attractor for ( [ eq:3hm1 ] ) can be quite big and were observed to have a large basin . the proof of existence of big , globally attracting wild - hyperbolic attractors in map ( [ eq:3hm2 ] ) is still open problem . finally , we consider the more general structure of polynomial diffeomorphisms of @xmath130 . as we remarked in the introduction , the 3d hnon maps ( [ eq:3hm1 ] ) and ( [ eq:3hm2 ] ) have the form of ( [ eq:3hm3 ] ) when @xmath54 . as we will shown in [ sec : quadratic ] , this result has the following generalization ; [ tm:4 ] let @xmath131 be a quadratic diffeomorphism with a quadratic inverse and constant jacobian @xmath132 . then , after an affine coordinate transformation @xmath38 can be written in the form ( [ eq:3hm3 ] ) or else the dynamics of @xmath38 is trivial . note that both of the maps ( [ eq:3hm1 ] ) and ( [ eq:3hm2 ] ) fall into this classification . the classification of the dynamics of the trivial cases can be done from the normal forms ( [ eq : qforms ] ) that will be obtained in [ sec : quadratic ] . we begin the proof of th . [ tm:1 ] by constructing a local map @xmath133 , valid in a neighborhood @xmath79 of the fixed point @xmath134 , that simplifies the description of the local unstable @xmath135 and local stable @xmath136 manifolds of @xmath59 . then we obtain a global map @xmath137 that takes a neighborhood of a point on @xmath138 to a neighborhood of a point of @xmath139 . by @xcite,@xcite , one can introduce @xmath140 coordinates @xmath141 in a neighborhood @xmath78 of the fixed point @xmath59 such that the local map @xmath142 can be written , for small enough @xmath68 , in the form ) holds for general case , including @xmath33 . for the volume - preserving case the analogous form was established in @xcite ] @xmath143 where @xmath144 is the rotation by angle @xmath73 , and @xmath145 here the coefficients , @xmath146 , and @xmath147 as well as @xmath148 and @xmath149 depend ( smoothly ) on the parameters @xmath68 . the form ( [ eq : t0 ] ) will be called the _ main normal form _ ; it is very convenient for calculations . in particular , the local stable and unstable manifolds of @xmath59 are @xmath150 and @xmath151 in the neighborhood @xmath78 . when @xmath152 , choose a homoclinic point @xmath153 on the homoclinic orbit @xmath66 and a corresponding point @xmath154 that is an image of @xmath155 : @xmath156 for some @xmath157 , recall fig . [ fig : type21 ] . let @xmath158 and @xmath159 be some small neighborhoods of the points @xmath160 and @xmath155 . the global map @xmath161 is well defined for small enough @xmath68 , and can be written in the coordinates of ( [ eq : t0 ] ) as @xmath162 where the @xmath163 matrix @xmath164 , the vectors @xmath165 , @xmath166 , and @xmath167 , the coefficients @xmath168 and @xmath36 , and the higher order terms all depend smoothly on @xmath68 . we noted especially the dependence of the coefficient @xmath169 that will become the splitting parameter for the manifolds @xmath64 and @xmath65 near the homoclinic point @xmath160 . as discussed in the previous section , we therefore denote it by @xmath170 and consider as the first of the governing parameters . the assumption ( c ) that the tangency between @xmath65 and @xmath64 is quadratic implies that @xmath171 in addition , since @xmath0 is assumed to be a diffeomorphism , and since the jacobian @xmath172then we must have @xmath173 . indeed if @xmath174 is not identically zero , then one can use the linear coordinate change , @xmath175 where @xmath176 , ( [ eq : rotation ] ) , is a rotation by an angle @xmath177 , to transform the map ( [ eq : t1 ] ) to a new map of the same form such that the vector @xmath166 becomes @xmath178 here the new matrix @xmath164 , point @xmath165 , and vector @xmath167 are modified from the corresponding coefficients in the original ( [ eq : t1 ] ) . now the nonvanishing of the determinant of the jacobian ( [ eq : j1 ] ) implies that @xmath179 one of merits of the main normal form ( [ eq : t0 ] ) is that the iterations @xmath180 for any @xmath80 can be calculated quite easily , namely , in a form close to the form when @xmath133 is linear . in particular for the linear case we have @xmath181 this is the so - called _ cross - form _ of map @xmath182 . we can write @xmath182 in an analogous cross - form when @xmath133 is in the form ( [ eq : t0 ] ) [ lem2 ] for any positive integer @xmath80 and for any sufficiently small @xmath68 the map @xmath183 can be written in the cross - form @xmath184 where the vector @xmath185 and the function @xmath186 are uniformly bounded along with all derivatives up to order @xmath187 . in addition , all of the derivatives of order @xmath188 for the functions @xmath189 and @xmath190 tend to @xmath191 as @xmath192 . let @xmath193 denote the orbit : @xmath194 . then it is easy to see that the @xmath195 iterate of ( [ eq : t0 ] ) is @xmath196 the solution of ( [ eq:3 * ] ) for any given sufficiently small @xmath197 and large @xmath80 ( the so - called boundary value problem ) can be found by successive approximations @xcite if the zeroth approximation is taken as the corresponding solution of the linear problem . using the fact that @xmath198 , we obtain the following estimates @xmath199 where @xmath200 and @xmath201 are some positive constants estimating the norms for derivatives of the functions @xmath148 and @xmath149 in the domain @xmath78 , from ( [ eq : t0 ] ) . these inequalities evidently imply the desired result for the map . estimates for the derivatives can be found analogously . general results of this form for near saddle fixed point were given in @xcite,@xcite . here we give sharper estimates that apply in this particular case . the main task of this section is to obtain an analytical form for the first return map for diffeomorphisms of the family @xmath75 to a neighborhood of the homoclinic point @xmath160 . to construct this map we first find an image of the local map @xmath133 for which @xmath202 this will be true for each @xmath203 where @xmath204 is some sufficiently large integer . for each such @xmath80 there are strips @xmath205 and @xmath206 such that @xmath207 . a first return map is obtained by composing @xmath182 with the global map @xmath208 , to obtain a map on a neighborhood of @xmath160 , recall fig . [ fig : type21 ] . thus the first return map will be of the form @xmath209 under the global map @xmath137 the strip @xmath210 is transformed into three - dimensional horse - shaped region @xmath211 that has non - empty intersection with @xmath210 when @xmath68 is small enough . thus , the map @xmath212 is a generalized , three - dimensional , horseshoe map . we will show that these horseshoe maps can possess rich dynamics ; for example , they can lorenz - like , wild - hyperbolic strange attractors . the goal is the prove the following lemma : [ rescaling lemma . ] [ lm : rescaling ] for any sufficiently large @xmath80 such that @xmath213 the first return map @xmath212 can be brought , by means of an affine transformation of coordinates , to the following form @xmath214 where @xmath215 are some coefficients , @xmath216 . \end{array}\ ] ] also , in ( [ eq : t07 ] ) the @xmath217-terms have the indicated asymptotic behavior for all derivatives up to order @xmath187 . using formulas ( [ eq : t1 ] ) , ( [ eq : t0kk ] ) and conditions ( [ eq : van3 ] ) and ( [ eq : b1 ] ) we can write the first return map @xmath218 in the form @xmath219 where the @xmath163 matrix @xmath220 and @xmath221 introduce new coordinates @xmath222 where @xmath223 , in order to eliminate the affine terms in the first two equations and the term linear in @xmath224 in the third equation of ( [ eq : van8 ] ) . then , the map ( [ eq : van8 ] ) is rewritten as @xmath225 where @xmath226 and @xmath227 . \end{array}\ ] ] next , we rescale the map ( [ van9 ] ) introducing new coordinates ( x , y , z ) defined through @xmath228 where we suppose that values of @xmath73 are such that ( [ eq : t02 ] ) holds . then , map ( [ van9 ] ) is rewritten in the new coordinates in form ( [ eq : t07 ] ) . now theorem [ tm:1 ] easily follows from this lemma . indeed , we have only to indicate those domains @xmath82 of the initial parameters where the coefficients in the third equation of ( [ eq : t07 ] ) are finite . denote the coefficient of the third equation of ( [ eq : t07 ] ) in front of @xmath229 as @xmath9 , i.e. @xmath230 since @xmath231 and @xmath232 , the coefficients @xmath8 and @xmath9 can take arbitrary finite values when varying parameters @xmath233 ( for @xmath8 ) and @xmath73 ( for @xmath9 ) , see formulas ( [ mpar * ] ) and ( [ eq : t04 ] ) . moreover , the finiteness of @xmath9 means that the corresponding values of the trigonometrical term @xmath234 are asymptotically small and , in any case , do not exceed @xmath235 in order . for such values of @xmath73 , we obtain @xmath236 thus , by ( [ van7 ] ) , @xmath237 in this case . denote as @xmath10 the coefficient in front of @xmath238 from the third equation of ( [ eq : t07 ] ) . by virtue of ( [ eq : t05 ] ) and ( [ van7 ] ) we have @xmath239 since @xmath240 at @xmath5 , it follows from ( [ eq : t06 ] ) that @xmath10 can take ( when @xmath80 is large ) any arbitrary finite value ( positive or negative depending on the sign of @xmath241 ) as the value of the jacobian @xmath242 of the local map @xmath133 varies . this completes the proof of theorem [ tm:1 ] . 1 . we noted that at least three parameters must be used to study bifurcations of a homoclinic orbit for a three - dimensional saddle - focus . the governing parameters that we use are @xmath243 corresponding to the splitting distance between the stable and unstable manifolds , the phase @xmath244 of the focus multiplier , and the jacobian @xmath245 of the saddle - focus . if the map @xmath0 is exactly volume preserving , i.e. , @xmath246 , then the resulting map ( [ eq : t07 ] ) is as well , i.e. @xmath247 . such maps were studied in @xcite . moreover , th . [ tm:2 ] implies that there are nearby maps ( by unfolding the parameter @xmath10 to make the map volume contracting ) that have strange attractors . in this section we will study the dynamics of the 3d hnon map ( [ eq:3hm1 ] ) and prove th . the dynamics can be complex , especially when @xmath248 is close to one , essentially because when @xmath249 the map has , for certain values of the parameters , fixed points with three multipliers are on the unit circle . it is well known that bifurcations of such points can lead to appearance of homoclinic tangencies , invariant circles , and even strange attractors ( as @xmath248 decreases below one ) or repellers ( as @xmath248 increases above one ) @xcite,@xcite . we are interested primarily in the stable dynamics of the map ( [ eq:3hm1 ] ) , i.e. , in characterizing its attractors . one way to find attractors with complex dynamics is to study the local normal forms that arise near fixed points that have three multipliers that are @xmath250 . we follow here the analysis of @xcite where local normal forms of the equilibrium of a flow with three zero eigenvalues and with additional symmetries were analyzed and conditions for the existence of strange attractors were given . these results can be applied to the analysis of bifurcations of codimension - three fixed points of three - dimensional maps . in this case we can often construct some approximated flow normal form whose time - one shift map coincides ( up to terms of a certain order ) with some power of the original map restricted on a neighborhood of the fixed point . usually , the presence of multipliers @xmath251 or @xmath252 implies that the corresponding approximated flow possesses some symmetry , such as a reflection @xmath253 , or a rotation , @xmath254 . one of the most interesting of their results is that lorenz attractors can appear by local bifurcations of an equilibrium with three zero eigenvalues when there is a reflection symmetry through the origin @xcite,@xcite,@xcite . these results can be applied to the local bifurcation analysis of ( [ eq:3hm1 ] ) when the fixed points has multipliers @xmath255 . as is true for the general volume - preserving quadratic diffeomorphism of lomel and meiss @xcite , the 3d hnon map ( [ eq:3hm1 ] ) has at most two fixed points on the diagonal : @xmath256 providing that the discriminant , @xmath257 , is nonnegative , @xmath258 the primary bifurcations of these fixed points occur when they have multipliers on the unit circle . these bifurcations are easily determined using the characteristic polynomial of the jacobian of ( [ eq:3hm1 ] ) , @xmath259 the first codimension - one case corresponds to a single multiplier one , where @xmath260 , or equivalently to the surface where @xmath257 vanishes : @xmath261 this is simply a saddle - node bifurcation that creates the pair of fixed points ( [ eq : fp ] ) in the domain @xmath262 . similarly , a period - doubling bifurcation occurs when there is a multiplier @xmath251 , i.e. , where @xmath263 , or equivalently for @xmath264 , which gives @xmath265 note that this bifurcation occurs for the point @xmath266 when @xmath267 , and for the point @xmath268 when @xmath269 . the final codimension - one bifurcation corresponds to the hopf bifurcation when pair of multipliers is on the unit circle , @xmath270 , so that @xmath271 or equivalently on the surface @xmath272 here we have written the equation for the surface @xmath273 in parametric form using the phase @xmath274 . when @xmath275 , this bifurcation only occurs for the fixed point @xmath266 . there are three codimension - two curves in the @xmath85 parameter space . the hopf surface , @xmath276 terminates at @xmath277 when the multipliers are @xmath278 along the curve @xmath279 . it also terminates on @xmath280 where the multipliers are @xmath281 along the curve @xmath282 . finally , the period - doubling and saddle - node curves are tangent when the multipliers are @xmath283 on the curve @xmath284 . the bifurcation surfaces are most easily viewed in two - dimensional slices through constant @xmath10 planes . an example of the resulting bifurcation curves are shown when @xmath285 in fig . [ fig : type21bifb06fp1 ] and fig . [ fig : type21bifb06fp2 ] . note that @xmath266 is stable inside the curvilinear triangle bounded by @xmath277 , @xmath280 and @xmath273 , when @xmath286 . the second fixed point , @xmath268 , is never stable when @xmath275 . ) for the fixed point @xmath266 when @xmath287 . fixed points do not exist in the shaded region . the solid curve is the saddle - node @xmath277 , the dotted curve is the period doubling , @xmath280 , and dot - dash curve is the hopf bifurcation , @xmath273 . the fourth curve ( dashed ) corresponds to the existence of a double eigenvalue . there are five corresponding regions , and a representative configuration of the multipliers in the complex-@xmath288-plane is shown for each . the codimension - two bifurcations are shown as the square " ( period - doubling - hopf ) , triangle " ( saddle - node - period doubling ) and circle " ( saddle - node - hopf ) points . [ fig : type21bifb06fp1],width=432 ] ) in for the fixed point @xmath268 when @xmath287 . the curves have the same meaning as in fig . [ fig : type21bifb06fp1 ] . for this fixed point there is no hopf bifurcation . [ fig : type21bifb06fp2],width=384 ] a similar analysis can be used to derive the bifurcation surfaces for the map ( [ eq:3hm2 ] ) . indeed , the formula for the fixed points , and consequently the saddle - node surface is the same as for ( [ eq:3hm1 ] ) . the period - doubling and hopf surfaces are slightly modified . an example of a slice through these surfaces for @xmath285 is shown in fig . [ fig : type12bifb06fp1 ] and fig . [ fig : type12bifb06fp2 ] . ) for the fixed point @xmath266 when @xmath287 . the curves have the same meaning as in fig . [ fig : type21bifb06fp1 ] . [ fig : type12bifb06fp1],width=432 ] ) in for the fixed point @xmath268 when @xmath287 . the curves have the same meaning as in fig . [ fig : type21bifb06fp1 ] . for this fixed point there is no hopf bifurcation . [ fig : type12bifb06fp2],width=384 ] we will now focus on the point @xmath266 and expand around the codimension - three bifurcation point where the multipliers are @xmath289 , that is about the point @xmath290 where @xmath291 . note that on the parameter slice @xmath54 , the hopf curve ( [ eq : hopf ] ) coincides with the saddle - node curve ( [ eq : sn ] ) so that there is no hopf bifurcation . moreover the codimension - two points in fig . [ fig : type21bifb06fp1 ] corresponding to the multipliers @xmath292 ( square ) and @xmath293 ( triangle ) collide at the codimension - three point . otherwise the bifurcation diagram for @xmath266 looks similar the slice in the figure . our goal is to prove th . this can be done with the aid of the following lemma . [ smlem ] in a region of the parameters @xmath294 near the point @xmath295 the second iterate of ( [ eq:3hm1 ] ) in some neighborhood of @xmath266 is approximately the time - one map of the the three - dimensional flow @xmath296 up to arbitrary small correction terms . moreover the coordinates @xmath297 and parameters @xmath288 and @xmath298 can take arbitrary finite values . shifting coordinates @xmath299 brings map ( [ eq:3hm1 ] ) to the form @xmath300 introduce new ( small ) parameters @xmath301 by the following formulas @xmath302 note that @xmath303 the first inequality means that @xmath275 , and the second is fulfilled automatically when the fixed points exists , e.g. to the right of the @xmath277 curve in fig . [ fig : type21bifb06fp1 ] . we have also the following relation @xmath304 between the new parameters and the parameter @xmath8 . next , we bring map ( [ hen1 ] ) to the following form @xmath305 where @xmath306 note that the jacobian of the map ( [ res1 ] ) is close to the jordan normal form for the multipliers @xmath289 when @xmath307 are small . the first equation of ( [ res1 ] ) does not contain any high order terms , and in the second and third components , we take into account only quadratic terms . moreover , these quadratic terms have been further simplified by standard normal form analysis , so that only the resonant ones remain . the second power of the map ( [ res1 ] ) can be embedded into a flow . namely , using the picard iteration procedure ( of the second order ) , we can see that the map ( [ res1 ] ) coincides , up to terms of order @xmath308 , with the time - one map of the following flow @xmath309 after a smooth change of coordinates of the form @xmath310 equations ( [ fl1 ] ) are written as follows @xmath311 where @xmath312 flow ( [ flow ] ) is the same that was considered @xcite and , in our case we have also that @xmath313 . then , we can introduce new time @xmath314 and new rescaled coordinates @xmath297 , by formulas @xmath315 @xmath316 such that system ( [ flow ] ) is reduced to the following one @xmath317 where @xmath318 since values of @xmath319 can be arbitrarily small , it implies that the rescaled coordinates @xmath297 and new parameters @xmath288 and @xmath298 can take arbitrarily finite values . note that if we omit the @xmath320 terms in the system ( [ sm ] ) , we obtain system ( [ = sm ] ) , which is the well - known , shimizu - morioka model . this system was intensively studied in @xcite,@xcite where , in particular , the existence of the lorenz attractor was proved for some domain of the parameters . for example , the domain of the existence of the lorenz attractor in shimizu - morioka model is known to be contained in a rectangle @xmath321 ( @xcite , fig . 13 ) . using this fact we can pointed out the following boundaries of values of the parameters @xmath322 : @xmath323 ( and , respectively , of the parameters @xmath85 ) such that in the corresponding domain of the initial parameters @xmath85 our 3d hnon map has a wild - hyperbolic lorenz - type attractor . this completes the proof of the main theorem . suppose that @xmath324 is a quadratic cremona map , i.e. , is a diffeomorphism with constant jacobian @xmath325 . then it is easy to see that without loss of generality we can write it in the form @xmath326 where @xmath327 is the diagonal scaling map with @xmath328 , @xmath200 is an affine volume - preserving map , @xmath329 is a vector of quadratic forms so that the map @xmath330 is a volume - preserving _ quadratic shear _ @xcite . as in th . [ tm:4 ] we now assume that the map @xmath38 has a polynomial inverse ( i.e. , it is an element of the affine cremona group ) and that the inverse is also a quadratic map . the same property then will also hold for the volume - preserving map @xmath331 . it was shown in @xcite that maps of the form @xmath332 can be transformed , by an affine change of variables , to one of the three following forms : @xmath333 where @xmath334 is a scalar quadratic form . consequently @xmath335 for some affine transformation @xmath164 , and @xmath336 , @xmath337 , or @xmath338 . using the same transformation on @xmath38 gives @xmath339 but the scaling map @xmath340 commutes with @xmath164 , so @xmath341 finally we can apply an additional scaling transformation to the variables and parameters to transform @xmath342 to one of the three forms @xmath343 where the parameters have been rescaled . the dynamics of @xmath344 and @xmath345 are trivial . indeed for @xmath344 the @xmath346 dynamics decouples from the remaining dynamics , and for @xmath345 the @xmath347 dynamics is linear and decouples from the @xmath348 dynamics . thus the only dynamically interesting case is @xmath349 . since this map is the same as ( [ eq:3hm3 ] ) , we have proved th . [ tm:4 ] . an example of this analysis consider the act - map ( arneodo - coullet - tresser map ) which was introduced by arneodo , coullet and tresser in their study of strange attractors in a family of differential equations on @xmath130 with homoclinic points of shilnikov type @xcite . @xmath350 where @xmath351 are real parameters with @xmath352 , @xmath353 , and @xmath354 an integer . the act - map always has a fixed point at the origin . nevertheless , when @xmath355 it is a quadratic family with a quadratic inverse , so it follows from th . [ tm:4 ] that it is affinely conjugate to ( [ eq:3hm3 ] ) with @xmath356 . to see this explicitly define new coordinates using the transformation @xmath357 then , the map is rewritten as @xmath358 where where @xmath359 . for the case @xmath355 we can rescale the coordinates by a factor @xmath360 to set the coefficient of @xmath361 to one . thus the map is written in the promised form ( [ eq:3hm3 ] ) . note that if we do one additional transformation and shift the variables by @xmath362 we can also obtain the standard form ( [ eq:3hm2 ] ) . here we define @xmath363 , and @xmath364 . the inequality @xmath365 holds guaranteeing the existence of a fixed point for the map . we have seen that many phenomena in the dynamics of three - dimensional hnon maps are markedly different from the possible phenomena predicted by our knowledge of two - dimensional dynamics . primarily , we have seen these new phenomena when the jacobian , @xmath10 , is close to @xmath366 . two natural problems appear here , regarding the understanding of bifurcation scenarios that lead to chaotic three - dimensional dynamics . the first is to understand how 2d dynamics is transformed into 3d dynamics " upon the variation of @xmath10 from values near zero to those close to @xmath367 . the second problem which is , in a sense , the inverse of the first one is to understand how 3d volume - preserving dynamics is transformed into dissipative ( and chaotic ) dynamics " as @xmath10 moves away from @xmath366 . in the current paper ( and also in @xcite ) , some partial results related to the second problem have been obtained ; however , the first problem is as of yet unstudied . however , our analysis implies that these global problems have no unified , qualitative solution . one general conclusion is that the stable dynamics of the hnon maps of the first and second kind , ( [ eq:3hm1 ] ) and ( [ eq:3hm2 ] ) , have unmistakably different characters . the bifurcations in ( [ eq:3hm1 ] ) and ( [ eq:3hm2 ] ) both lead eventually to the formation of three - dimensional versions of smale horseshoes . however , these have different characters : the horseshoes of ( [ eq:3hm1 ] ) contain type @xmath22 saddles of , while those of ( [ eq:3hm2 ] ) contain type @xmath23 saddles . thus , a third global problem appears : to understand how the character of the attractors changes as one changes type of nonlinearity in 3d hnon families . " in this case , it seems that using the family ( [ eq:3hm3 ] ) with the parameters @xmath368 determining the type of nonlinearity may be very profitable . while the problems posed above are of general interest , there are also many partial problems relating directly to the dynamics of quadratic maps . some of these may also be solved in the the near future : 1 . does the hnon map of the second kind , ( [ eq:3hm2 ] ) , have a large scale , wild - hyperbolic attractor ? 2 . what are the dynamical properties of the local normal form for ( [ eq:3hm2 ] ) near fixed points with three unit multipliers ? + this is related to the `` reverse - shimitsu - marioka '' system since ( [ eq:3hm2 ] ) is the inverse of ( [ eq:3hm1 ] ) . thus we will obtain the model ( [ = sm ] ) , with @xmath369 replaced by @xmath370 . consequently , the question is to determine the properties of the stable dynamics of this system . what are main bifurcations scenarios the 3d hnon map ( [ eq:3hm1 ] ) that lead from simple dynamics to a wild - hyperbolic , lorenz attractor ? + there are two distinct , but useful approaches : ( a ) to fix @xmath10 and consider the corresponding two - parameter family ; ( b ) to consider @xmath10 as the main governing parameter and to move for given values of @xmath371 from small values of @xmath10 to values close to @xmath366 . what role is played by the codimension - two bifurcations ( recall [ sec : bifurcations ] ) in the formation of attracting invariant sets ? + the analysis of codimension - one bifurcations alone gives us only a representation of the main hyperbolic elements of the dynamics . a detail analysis of codimension - three bifurcations ( as in [ sec : bifurcations ] ) allows us to find in local normal forms ( using the flow embedding approach ) and verify the existence of small , wild - hyperbolic attractors . analogous methods can be applied to the study of codimension - two bifurcations . we believe that this approach will provide an understanding of principal mechanisms for the creation of wild - hyperbolic attractors . in summary , three - dimensional hnon maps have an abundance of new dynamical behavior which , it seems , has an important place in the theory of multidimensional dynamics as whole . we believe that they form a natural direction for a program of investigation similar to the well developed programs in one and two - dimensional dynamics . this is especially true since these maps cover ( up to the affine conjugacy ) all possible quadratic three - dimensional automorphisms . dragt and d.t . symplectic maps and computation of orbits in particle accelerators . in _ integration algorithms and classical mechanics ( toronto , on , 1993 ) _ , pages 5985 . soc . , providence , ri , 1996 . dynamics and moduli of @xmath375-conjugacy of 4d - diffeomorphisms with a structurally unstable homoclinic orbit to a saddle - focus fixed point . , 200:107134 , 2000 . providence , ri . bifurcation and chaos in the marioka - shimizu system . in _ methods of qualitative theory of differential equations _ , pages 180193 . gorky , 1986 . english translation in selecta math . soviet . 10 ( 1991 ) , 105 - 117 .

mass - loss from massive stars in the red supergiant phase is fundamental to the chemical enrichment of the interstellar medium @xcite . red supergiants , with initial masses of 10@xmath250@xmath3 , are known to lose mass at rates of up to 10@xmath4/yr @xcite . mass - loss rates are commonly estimated indirectly , using observations of infrared continuum excesses ( e.g. , * ? ? ? * ) or measurements of co rotational line intensities ( e.g. , * ? ? ? outflowing material is also directly surveyed through observations of systematic spectral line displacements @xcite and interferometric studies over multiple epochs @xcite . the physical processes involved in red supergiant mass - loss are however not well understood . while mass - loss from less massive stars , such as agb stars , is understood in terms of significant radial pulsations and winds driven by radiation pressure on dust grains , such mechanisms are not applicable to red supergiants @xcite , except those in a superwind phase @xcite . rotation in stars that are spun - up as they condense from protostellar clouds could play a role in driving mass - loss @xcite , as could alfvn and acoustic waves ( e.g. , * ? ? ? in the case of betelgeuse ( @xmath5 orionis , spectral type m2iab ) , the elevation of cool photospheric gas @xcite and gas at chromospheric temperatures @xcite has been attributed to large - scale surface convective activity @xcite . betelgeuse is perhaps the archetypal red supergiant . its large angular diameter ( 0.047 , * ? ? ? * ) and high apparent magnitude ( @xmath6 ) make it an excellent subject for high spatial resolution studies . the continuum angular diameter of betelgeuse has been measured using interferometric techniques at radio @xcite , mid - ir ( 8@xmath212@xmath0 m , * ? ? ? * ; * ? ? ? near - ir ( e.g. , * ? ? ? * ; * ? ? ? * ) , optical ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and ultraviolet @xcite wavelengths , and significant variation of the diameter with wavelength is observed . the apparent mid - ir diameter is also observed to vary significantly on timescales of years @xcite . numerous interferometric campaigns , from optical to mid - ir wavelengths , have further revealed up to three hotspots on the stellar surface ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? these spots range in intensity from a few to a few tens of percent of the intensity of the stellar disk . the detection of these spots is generally interpreted as direct evidence for the presence of a few giant convection cells on the stellar surface ( e.g. , * ? ? ? a profusion of circumstellar components have been identified surrounding betelgeuse . furthest from the star , an analysis of mid - ir spectra and interferometric data by @xcite showed the presence of two thin shells of dust at approximately 40 and 80 stellar radii ( @xmath7 ) , with temperatures of @xmath1400k and @xmath1200k respectively . later work by @xcite revealed a third , more compact dust shell at 4@xmath7 that was not apparent in previous data . the dust condensation zone and acceleration mechanism are not identified . radio continuum observations by @xcite at wavelengths between 0.7 cm and 6 cm , probing spatial scales of @xmath8 , revealed the presence of neutral gas at temperatures between 3500k and 1300k . these temperatures are consistent with the temperature profile of the betelgeuse atmosphere in these regions , as derived from mean temperature and density models @xcite . at @xmath9 , the gas distribution is asymmetric , indicating an anisotropic mechanism for elevating the gas above the photosphere . ultraviolet @xcite and h@xmath5 @xcite imaging observations have shown that this cool gas co - exists with asymmetrically - distributed hot chromospheric material . observations of [ fe ii ] lines by @xcite that correspond to the dominant cool atmospheric component showed , surprisingly , that this material was at rest with respect to the star . the circumstellar envelope as viewed across the _ jhk _ wavelength bands was imaged using a ` lucky - imaging ' approach combined with adaptive optics by @xcite , and was found to include a plume extending to 6@xmath7 . closer to the star , strong evidence exists for the presence of a cool molecular layer or shell directly above the photosphere within @xmath10 ( e.g. * ? ? ? * ; * ? ? ? this layer is optically thick at mid - ir wavelengths but is optically thin in the near - ir . spectral signatures of water from betelgeuse and the red supergiant @xmath0 cephei , found using data from the infrared space observatory ( iso ) , were interpreted by @xcite as originating from a 1500k layer surrounding these stars . modeling of near- and mid - ir interferometric data by @xcite confirmed the presence of such a layer , with a mid - ir optical depth of @xmath12 and a temperature of @xmath12000k . further work by @xcite combining both interferometric and spectroscopic data placed the layer surrounding betelgeuse at beyond @xmath11 , with a temperature of 2250k . the layer was shown not to be solely composed of water by @xcite , who suggested the presence of amorphous alumina ( al@xmath12o@xmath13 ) to explain the high mid - ir opacity . the spectrum of water lines from betelgeuse is indeed more complex than revealed by the iso data ; signatures of water at photospheric temperatures were detected by @xcite and @xcite . most recently , wide - band , spectrally dispersed mid - ir interferometric observations by @xcite were fitted by a layer containing sio , amorphous alumina and water . high resolution measurements of the co overtone band near 2.3@xmath0 m @xcite and one - dimensional imaging @xcite were modeled with a large patch of gas near @xmath14 with motions of 10@xmath215kms@xmath15 with respect to the star . this emerging model of a photosphere and overlying layer for a selection of red supergiant stars allows a better understanding of dust formation and red supergiant wind acceleration ( e.g. * ? ? ? * ) , leading towards a convergent picture of red supergiant mass loss . studies of the extended betelgeuse atmosphere form a basis for a better understanding of the atmospheric dynamics of red supergiant stars . the existence of a cool layer with high opacity at mid - ir wavelengths directly above the photosphere is now established for betelgeuse . the size , temperature and composition of the layer have been investigated , assuming spherical symmetry . little , however , is known about the spatial structure and time - variability of this layer . consistent instrumentation and modeling procedures have also not been applied over multiple observation epochs . in this paper , we present observations of the mid - ir continuum of betelgeuse using the berkeley infrared spatial interferometer ( isi , * ? ? ? interferometric data using three telescopes were gathered during each year between 2006 and 2010 , with angular resolutions sufficient to model the stellar shape and potential large - scale features . we describe the observations and image - modeling processes in 2 , and perform fits to a photosphere - and - layer model for our observations in 3 . in 4 , we discuss the nature of the observed non - uniformities in our model images . we examine the implications of our results for models of the betelgeuse atmosphere and red supergiant mass - loss in 5 , and present our conclusions in 6 . the isi consists of three 1.65 m aperture telescopes with a heterodyne detection system at each telescope . the double - sideband detection systems are sensitive to @xmath16@xmath0 m radiation within single sideband widths of @xmath12.7ghz . the @xmath17co@xmath12 laser local oscillators can operate at numerous wavelengths corresponding to vibrational molecular transitions between wavelengths of 9@xmath0 m and 12@xmath0 m . for the observations reported here , the local oscillators at each telescope were tuned to a wavelength of 11.15@xmath0 m , chosen so as to exclude strong stellar and telluric spectral lines from the detection band @xcite . ccc 2006 & 8 , 9 , 10 nov ; 7 dec & 4.2@xmath180.2 + 2007 & 14 , 15 , 16 nov & 4.1@xmath180.3 + 2008 & 22 , 23 , 24 sep & 3.8@xmath180.5 + 2009 & 4 , 5 , 7 , 9 , 18 , 20 , 22 nov & 4.4@xmath180.1 + 2010 & 5 , 17 , 18 nov ; 9 , 13 dec & 4.9@xmath180.3 the van cittert - zernicke theorem equates the complex cross - correlation of the electric fields detected at two spatially - separated telescopes with the two - dimensional fourier transform of the source brightness distribution within the telescope fields of view . instead of measuring complex cross - correlations , the isi , like all optical / ir interferometers , measures the normalized cross - correlation amplitude , known as the michelson fringe visibility . the michelson fringe visibility , @xmath19 , is given by @xmath20 where the fringe power , @xmath21 is the cross - correlation amplitude between signals measured at two telescopes , @xmath22 and @xmath23 , and @xmath24 and @xmath25 are the source powers measured at telescopes @xmath22 and @xmath23 respectively . visibility amplitudes are recorded for each isi telescope baseline pair . the visibility phase in the mid - ir for a given baseline is , however , strongly affected by randomly varying atmospheric pathlengths above each telescope , and can not be recovered from the data . the isi instead measures the closure phase , or the sum of the measured visibility phases for each baseline . the closure phase is independent of optical pathlength differences between pairs of telescopes @xcite and it provides information about the point - inversion symmetry of the source . the spatial frequency sampled by the projection of a baseline of length @xmath26 , perpendicular to the direction of a source , at a wavelength @xmath27 , is proportional to @xmath28 . by obtaining measurements at different source hour - angles , multiple spatial frequencies at different position angles are sampled . visibility and closure phase data were recorded for betelgeuse over timespans of five weeks or less each year between 2006 and 2010 ( see table 1 ) . during this time period , the three isi telescopes were configured in an approximately equilateral triangle with baseline lengths between 34 m and 40 m . for betelgeuse , this configuration allows spatial frequencies between 20 and 37sfu ( spatial frequency unit ; 1sfu@xmath29cyclesradian@xmath15 ) to be sampled . for all observations , tip - tilt corrections were applied to center the source on _ k_-band guider cameras operating at frame rates of 27hz . a standard observing sequence consisted of 4.5-minute dataset intervals of staring at the source and position - switching between the source and blank fields on either side of the source . the position - switching measurements moved the telescopes between the source and the sky every 15s . the source powers were measured using additional 140hz chopping between the source , or the sky , and cold loads with temperatures near that of the sky . a single visibility measurement consisted of either : ( 1 ) a fringe power measured during a chopping / position - switching dataset , normalized using the source powers during that dataset ; or ( 2 ) a fringe power measured during a staring dataset , normalized by weighted averages of source powers measured during the nearest chopping / position - switching datasets . closure phases were calculated for each triplet of visibility measurements . visibility magnitudes and closure phase zero - point offsets were calibrated using observations of aldebaran ( @xmath5 taurus ) conducted on the same night . aldebaran was assumed to have a diameter of 20mas and zero closure phase . the visibility calibrations for the two kinds of visibility measurements were different because of the different times spent on - source during chopping and staring datasets . data that were affected by systematic errors or poor atmospheric conditions were discarded . the total flux densities of betelgeuse during each observing epoch were also estimated using aldebaran , assumed to have a flux density at 11.15@xmath0 m of 615jy @xcite , as a calibrator . these total flux density values are given in table 1 , and represent the total power of betelgeuse within the 5@xmath305 fields of view of the isi detectors . image models were fitted to the calibrated visibility and closure phase datasets for each epoch . the data , analytic fits , and coverage of the @xmath31 spatial frequency plane , are shown in figure 1 . effective model - independent image reconstruction was not possible because of the sparsity of the @xmath31 coverage in all epochs . our model images include a centered uniform disk ( ud ) , and up to two point sources offset from the image centers . the uniform disks represent the apparent stellar surface , and the point sources represent localised intensity fluctuations in this surface . emission from the dust surrounding the star at large angular scales , while contributing to the total detected intensity , was assumed not to contribute to the visibility at the sampled spatial frequencies . the intensity of each image component was fitted to a fraction of the total intensity . other free parameters included the ud radius and the positions of the point sources . the model complex visibility at a point @xmath32 in the @xmath31 plane was given by @xmath33 where @xmath22 is the fraction of the total intensity contributed by the ud , @xmath34 is the uniform disk angular radius , @xmath35 denotes a bessel function of the first kind and of order unity , @xmath36 represents the fractions of the total intensity of the point sources , and @xmath37 and @xmath38 represent angular offsets to the west and north for the point sources , with @xmath39 . ccccccc 2006 & 0.51@xmath180.02 & 24.5@xmath180.8 & 0.04@xmath180.01 & -2.4@xmath180.1 & -23.7@xmath180.2 & 2.683 + 2007 & 0.54@xmath180.01 & 24.8@xmath180.3 & - & - & - & 2.053 + 2008 & 0.48@xmath180.02 & 24.7@xmath180.8 & - & - & - & 2.416 + 2009 & 0.557@xmath180.008 & 26.4@xmath180.2 & 0.009@xmath180.001 & -25.0@xmath180.3 & 10.5@xmath180.2 & 4.565 + & & & 0.007@xmath180.001 & 18.3@xmath180.3 & -20.2@xmath180.2 & + 2010 & 0.61@xmath180.03 & 26.4@xmath180.3 & 0.025@xmath180.009 & -29.7@xmath180.3 & -9.3@xmath180.1 & 3.65 + & & & 0.009@xmath180.004 & 8.3@xmath180.3 & -25.9@xmath180.1 & we used a weighted least - squares fitting technique for each epoch , with each baseline and the closure phase data given equal weights . the differing quality of data from different epochs necessitated using models with different numbers of free parameters to fit different epochs . data from 2006 were fitted with a single point source in addition to a ud , data from 2007 and 2008 were only fitted with uds , and data from 2009 and 2010 were each fitted with a ud and two point sources . an additional point source was added to the fit for a given epoch only if the reduced @xmath40 values calculated for the full dataset , and for the closure phase data alone , were both decreased upon the addition of the point source . we also attempted to fit other image models . these included uniform ellipses , uniform ellipses with point sources , and uds with offset uniform disks or offset two - dimensional gaussian profiles in place of the point sources . none of these trials , however , converged to a satisfactory image , and the angular sizes of the fitted point sources are not determined . the fractions of the total intensity contributed by the point sources were not constrained to be positive , as our data are sensitive to both bright and dark features on the stellar disk . simulations were conducted to assess possible degeneracies in fitted image model parameters given the sparsity of the @xmath31 coverage of our datasets , in order to ensure that our data were not being over - fitted . no significant degeneracies were found for any epoch . for example , the 2006 , 2009 , and 2010 datasets constrain both the locations and brightnesses of the point sources . closure phase measurements at multiple array position angles constrain the position angles of the point sources , and both the closure phase measurements and the visibility measurements constrain the angular offsets of the point sources from the image centers . the results of the fits , as well as the reduced @xmath40 for each epoch , are given in table 2 . all the fits have reduced @xmath40 values that are greater than unity . this implies that the data contain minor features besides gaussian noise that are not being fitted by the free parameters of the image models . isi data for betelgeuse obtained during 2006 were also analyzed by @xcite . despite our choice of different observing dates , a different method of calibration and slightly different image fitting procedures , our results match those of @xcite within the margins of error . we however do not adopt tatebe et al.s results for the 2006 epoch in order to maintain the same analysis procedure across all epochs . using isi observations over the period 1993@xmath22009 , @xcite reported a systematic decrease of 15% in the mid - ir continuum ud diameter of betelgeuse . here , we augment this finding with new measurements of the ud diameter of betelgeuse . figure 2 shows the data plotted in figure 1 of @xcite , with the new results also included . though the present work re - analyzes some previous observations examined by @xcite , the diameter measurements are obtained from different data sets and analyses from those reported by @xcite . the measurement of the ud diameter of betelgeuse at 11.15@xmath0 m by @xcite , obtained using the very large telescope interferometer in 2003 , is also plotted . recently , @xcite showed that k - band diameter measurements , made over a similar time period , vary less than the mid - ir diameters plotted in figure 2 . while our results from 2006@xmath22008 are consistent to within the margins of error with the trend towards a smaller size @xcite , the ud diameters of betelgeuse in 2009 and 2010 are significantly larger than those during 2006@xmath22008 . our analysis method for the 2007 and 2008 datasets essentially matched that of @xcite , in that only uds were fitted to the data . our technique however differed somewhat in the methods of calibration and fitting . we consider the variability in the apparent size of betelgeuse in the mid - ir further in 5.2 . effective temperatures of stellar surfaces are among the fundamental parameters of stars @xcite . the isi measurements of the flux density of betelgeuse , and of the ud sizes and fractions of the total intensity , allow the effective temperature of the 11.15@xmath0 m apparent stellar surface to be determined at each observing epoch . we use a planck function for the power emitted by a blackbody to express the effective temperature , @xmath41 , as @xmath42 where @xmath43 is planck s constant , @xmath44 is the observing frequency , @xmath45 is boltzmann s constant , @xmath46 is the solid angle subtended by the stellar disk , @xmath47 is the vacuum speed of light , and @xmath48 is the measured power from the stellar disk in units of wm@xmath49hz@xmath15 . we do not use the widely - applied technique of calculating effective temperatures using estimates for the bolometric flux from the star ( e.g. , * ? ? ? * ; * ? ? ? * ) for two reasons : radiation at mid - ir wavelengths is strongly affected by thermal emission from circumstellar dust shells and bolometric correction factors at these wavelengths are unique to individual stars ; and also because the correspondance between the 11.15@xmath0 m scattering surface and the photosphere of betelgeuse is not certain . the calculated effective temperatures for the observed stellar disk of betelgeuse during each observing epoch are plotted in figure 3 . we do not account for attenuation by the interstellar medium , estimated as modifying the observed flux density by a factor of 0.975 @xcite . we also do not account for absorption of the stellar radiation by the dust that surrounds betelgeuse at angular scales of 1 or larger , or for the emission from this dust in the line of sight to the stellar face . this is because the dust shells are extremely optically thin at mid - ir wavelengths when viewed radially , with a total line - of - sight optical depth of approximately 0.0065 @xcite . our measured effective temperatures of the mid - ir surface of betelgeuse are consistently lower than the value of 3641k derived by @xcite for the photospheric component of a photosphere - and - layer model . furthermore , significant variability is evident between the effective temperatures measured at different epochs . we therefore consider sources of mid - ir continuum opacity in the atmosphere of betelgeuse in an attempt to elucidate the scattering surface probed by our observations and the cause of the observed variability . stellar atmospheres have opacities that vary with wavelength . stars appear smaller at wavelengths where the atmospheric opacity is smaller . the surface of a star , when observed at a given wavelength , can be approximated as the location where the optical depth at that wavelength , measured from the observer , is approximately unity . the photosphere of a star , however , is the apparent stellar surface observed at a wavelength where the opacity matches the rosseland mean opacity @xcite . the rosseland mean opacity , @xmath50 , is given by @xmath51 where @xmath52 is the intensity of radiation from a point in a stellar atmosphere , @xmath53 is the temperature at that point , and @xmath54 is the opacity at a frequency @xmath44 . regions of a stellar spectrum with large radiation intensity but low opacity contribute most to determining the rosseland mean opacity . @xcite concluded that the mid - ir continuum opacity of a gas with solar atomic abundances under surface conditions present on red giant stars closely matches the rosseland mean opacity of such a gas . molecular layers surrounding red giant stars have , however , been shown to significantly change the mid - ir opacity from that of a solar abundance atomic gas , while remaining relatively transparent in shorter wavelength ir bands @xcite . the @xcite model for betelgeuse , combining @xmath55-band visibility data and a 11.15@xmath0 m isi visibility curve @xcite , includes a shell with a @xmath55-band optical depth of 0.06 and a 11.15@xmath0 m optical depth of 2.33 . @xcite showed that this shell was composed of water , amorphous alumina dust and sio gas , and that the 11.15@xmath0 m opacity was dominated by alumina . the presence of alumina in a layer surrounding betelgeuse , within 1.5@xmath7 , was suggested by @xcite as the only dust species that could condense at the temperatures present in this region . alumina dust grains forming close to stars possibly play a role as nucleation sites for dust species that condense at lower temperatures ( e.g. * ? ? ? bremsstrahlung interactions between electrons and neutral hydrogen atoms also contribute to the mid - ir continuum opacity in red giant stars @xcite . this emission mechanism was adduced by @xcite to explain the variation with wavelength of the apparent sizes of the radio photospheres of long - period variables , and ascribed to betelgeuse by @xcite . we discuss this opacity model further in 3.4 . prior work , the dramatic variations in apparent mid - ir diameters of betelgeuse , and the low effective temperatures all provide clear motivation to model the new isi observational results in terms of a photosphere and a molecular layer . our method is similar to that of @xcite in that we modeled a spherically symmetric layer surrounding an opaque stellar photosphere with an assumed diameter of 0.04371 and an assumed surface temperature of 3641k @xcite . visibility and closure phase models for any observed point sources were subtracted from the data during each observation epoch in order to allow the model to be spherically symmetric . we did not attempt to fit an asymmetric layer because of the large number of free parameters involved . the free parameters of our model were : the temperature , @xmath53 , of the layer ; the optical depth , @xmath56 , of the layer along the line of sight to the center of the star ; and , the outer radius , @xmath57 , of the layer . the layer was assumed to have uniform temperature and opacity , and was also assumed to have an inner radius that corresponds to the photospheric radius . radiation within the isi band was assumed to arise only from blackbody emission by the layer . the modeling procedure involved fitting visibilities derived from a model image , created using a trial set of free parameters , to the measured visibilities . the measured visibilities for each epoch , and their errors , were scaled by the measurements of the measured power emitted by the ud surface of betelgeuse . the flux density of a pixel in the model image grid was calculated as : + for @xmath58 , @xmath59 and , for @xmath60 , @xmath61 here , @xmath62 describes the optical path length through the layer , @xmath63 is the absorption coefficient , @xmath64 is the radial distance of a pixel from the image center , @xmath46 is the solid angle subtended by a single pixel , @xmath65 is the specific intensity radiated by a blackbody of temperature @xmath53 at frequency @xmath44 , and @xmath66 is the assumed temperature of the photosphere . @xmath62 and @xmath63 are given by @xmath67 the pixel size of the model images was @xmath68 milli - arcseconds . for each iteration of the fit , a model image was generated and integrated along one dimension , and a one - dimensional fourier transform was performed in order to generate a set of visibilities . each model visibility was compared to measured visibilities that were closest to it ( i.e. , within @xmath69 ) in spatial frequency . a levenberg - marquardt fitting algorithm was employed to calculate the free parameter set that minimized the global @xmath40 . the results of the fit for each observing epoch are given in table 3 . errors for each parameter were calculated using the resulting covariance matrix . in order to confirm the validity of our analyses , we added the visibility and closure phase models of the point sources , that were earlier subtracted from the visibility data from each epoch , to the best - fit model visibilities . @xmath40 values relating this sum to the original visibility data were calculated for each epoch , and were found to be comparable to the @xmath40 values of the fits to uds and point sources shown in table 2 . the changes with time of the fitted layer temperatures and outer radii are similar to the changes with time in the effective stellar temperatures and the apparent ud radii respectively . all our fitted layer radii and temperatures are roughly consistent with the 1.3@xmath21.5@xmath7 ( 0.057 to 0.065diameter ) and 1500@xmath22000k layers modeled by most authors @xcite . the values we calculate for the 11.15@xmath0 m optical depth of @xmath11 are half of those resulting from the models of @xcite and @xcite . cccc 2006 & 1.9@xmath180.7 & 0.8@xmath180.1 & 28.6@xmath180.7 + 2007 & 2.5@xmath180.9 & 0.9@xmath180.1 & 25.5@xmath180.9 + 2008 & 2@xmath181 & 1.2@xmath180.2 & 25@xmath181 + 2009 & 2.3@xmath180.8 & 0.99@xmath180.06 & 26.8@xmath180.8 + 2010 & 2.8@xmath180.6 & 1.05@xmath180.03 & 29.8@xmath180.6 the lack of wideband mid - ir spectral information during our observing epochs limits modeling of molecular gas and dust species that contribute to the layer opacity . we are able , however , to investigate some basic physical characteristics of the layer using the opacity mechanism described by @xcite . this work showed that , for solar abundance gas at temperatures similar to those in red giant atmospheres , collision processes between neutral hydrogen atoms and free electrons dominate the mid - ir opacity . for temperatures of 2500k to 3000k , the primary sources of free electrons were found to be thermally ionized na and al . the abundances of h@xmath70 , h@xmath12 and h@xmath71 relative to h were found to be too low to contribute to the opacity through collisions with electrons , and photoionization and photodissociation processes were also found to be insignificant . the absorption coefficient due to electron - hydrogen atom collisions , first derived by @xcite , is given in simplified form as @xmath72 where @xmath73 is the electron number density , @xmath74 is the hydrogen atom number density , and @xmath22 is expressed as a third - order polynomial in @xmath53 @xcite . this opacity mechanism is used to explain the sizes of the radio photospheres of red giant stars @xcite , at different wavelengths . for example , @xcite found that the radio photosphere of betelgeuse appeared larger , yet cooler , for longer wavelengths . we consider the implications of this mechanism dominating the mid - ir continuum opacity . equation 9 can be used to calculate hydrogen atom number densities from the fitted layer optical depths . the electron number densities were obtained from table 2 of @xcite , which gives the fractional ionizations , relative to @xmath74 , of various metals at different temperatures . we further used the hydrogen number densities to calculate the total layer mass , @xmath75 , for each epoch , using a distance of @xmath76pc to betelgeuse @xcite . our results are given in table 4 . we find masses for the layer of between @xmath77 and @xmath78 . our values for the hydrogen number density of between @xmath79@xmath80 and @xmath81@xmath80 , using a hydrogen atomic mass of @xmath82 g , are equivalent to mass densities of @xmath83g@xmath80 . such values are consistent with the hydrogen density derived for the betelgeuse layer modeled by @xcite , where they derived their hydrogen density by assuming a ratio between the h@xmath12o number density and the hydrogen number density of @xmath84 @xcite and using their calculated values of @xmath85 for the layer . our values for the hydrogen density are therefore also consistent with the values for the water density derived by @xcite , though they result from very different methods of calculation . ccc 2006 & 5@xmath181 & 6@xmath181 + 2007 & 4.6@xmath180.9 & 3@xmath181 + 2008 & 9@xmath182 & 4@xmath182 + 2009 & 4.3@xmath180.8 & 3@xmath181 + 2010 & 3.5@xmath180.4 & 5.0@xmath180.9 we now consider a physical interpretation of the point sources used to model the isi interferometric data on betelgeuse . figure 4 shows images of the intensity distribution of betelgeuse during each observing epoch , calculated from the photosphere and layer models , with the point source locations indicated by filled white disks . the color brightness of the images is consistent with the same flux density scale , also depicted in figure 4 , and the areas of the disks representing the point sources indicate their flux densities calculated from their fitted fractions of the total intensity . all point sources are positive in intensity with respect to the star . non - uniformities on the observed stellar disk of betelgeuse at optical and near - ir wavelengths have been reported by several authors ( e.g. , * ? ? ? * ; * ? ? ? these asymmetries are generally modeled either as point sources , or as uds or two - dimensional gaussian profiles of fixed , unresolved size . the intensities of these ` spots ' are generally found to be between 5% and 20% of the intensity of the stellar disk . the distribution and strengths of the spots are found to vary significantly across optical and near - ir bands during contemporaneous observations @xcite , with the star sometimes appearing featureless in one band and highly non - uniform in another . the spots are also found to vary on timescales of less than 8 weeks @xcite . one such spot was recently resolved in the _ h_-band by @xcite , who reported a diameter of 0.011 for this spot . these results are consistent with a prediction by @xcite for extremely large - scale convection on red giants . red supergiants are suggested to have only a few tens of convection cells on their surfaces , as opposed to the thousands present on the surface of the sun . similar predictions also result from three - dimensional numerical simulations ( e.g. , * ? ? ? * ; * ? ? ? * ) , and an initial attempt at matching simulations to interferometric observations was made by @xcite . @xcite predicted temperature fluctuations of up to @xmath86k with respect to the photospheric temperatures of red giant stars . this is consistent with the intensities of spots observed in the visible and near - ir . the point sources fitted to our mid - ir data do not arise directly from large - scale convective features on the surface of betelgeuse . if we assume a 0.01 diameter for the weakest of our point sources ( see table 2 ) , which contributes 0.7% of the 4400jy flux density of betelgeuse in 2009 , its temperature of @xmath1660k above the effective surface temperature , calculated using equation 2 , is consistent with a photospheric hotspot of @xmath11000k behind the @xmath87 layer . if we assign a similar diameter to any of the other point sources we observe , even greater temperatures would result for associated photospheric hotspots . the mid - ir point sources are hence more intense than expected from surface temperature fluctuations caused by giant convection cells @xcite . furthermore , every one of our point sources is located beyond the edge of the photospheric disk ( see figure 4 ) . it is clear therefore that the observed asymmetry of betelgeuse in the mid - ir is caused by a non - uniform intensity distribution in the layer . we do not have sufficient sampling of the @xmath31 plane in any epoch to identify whether the layer is in fact spherically symmetric , but with temperature fluctuations , or whether the layer is extended in particular directions . the point sources can also be explained in terms of opacity inhomogeneities in the layer . in the rayleigh - jeans approximation to the planck law , the effective radiation temperature , @xmath88 , of a column of material with optical depth @xmath56 and temperature @xmath53 is given by @xmath89 for example , increasing the value of @xmath56 from 1 to 3 increases @xmath88 by a factor of 1.5 . non - uniform spatial , temperature and opacity distributions of the betelgeuse layer could all explain the observed mid - ir non - uniformities . molecular layers directly surrounding the photospheres of red supergiant stars , and in particular betelgeuse , have been modeled by a variety of authors under differing assumptions for the layer geometry and the dominant opacity source . a layer of water with a column density of 10@xmath90cm@xmath49 at a temperature of @xmath11500k was first suggested by @xcite for betelgeuse , based on modeling of near - ir spectra . a combination of spectral and interferometric modeling by @xcite , assuming a layer geometry identical to ours , suggested a layer temperature of 2050k and an outer radius of 1.45@xmath7 , while modeling the mid - ir opacity as being dominated by a continuum of water emission and absorption lines . under the assumption of a layer with a well - defined inner radius , @xcite modeled a shell beyond 1.3@xmath7 with a temperature of 2250k , again dominated by water opacity . more advanced spectral modeling by @xcite revealed that water could not play a dominant role in determining the mid - ir continuum opacity , and instead suggested the presence of amorphous alumina in the molecular layer . this model was consolidated by @xcite to suggest a shell between 1.32@xmath7 and 1.42@xmath7 at a temperature of 1520k . the properties we derive for a layer surrounding betelgeuse , summarised in table 3 for each observing epoch , are largely consistent with these results . we find temperatures that vary between 1900k and 2800k and outer radii that vary between 1.14@xmath7 and 1.36@xmath7 . our modeling technique most closely matched that of @xcite in assuming thermal continuum emission from the layer within our observing band . this assumption is justified by the featureless spectrum of betelgeuse within , and around , our observing band and by the lack of any significant water lines within and surrounding our band @xcite . the results are consistent with previous models of the betelgeuse layer . a consistent picture of the composition of the layer has , however , not been achieved . while spectral modeling has uncovered the presence of numerous molecular species , including h@xmath12o , co , sio and possibly cn , as well as alumina dust @xcite , the opacity of the layer as a function of wavelength is not yet fully determined . our results for the optical depth of the layer of @xmath11 at 11.15@xmath0 m wavelength are not in agreement with those of @xcite and @xcite , who obtain optical depths of @xmath12 . the number density of atomic hydrogen , based on the 11.15@xmath0 m opacity arising from electron - hydrogen atom collisions , is @xmath110@xmath92@xmath80 . this density , given the derived layer size , corresponds to h@xmath12o column densities of 10@xmath93cm@xmath49 ; this result is consistent with h@xmath12o column densities measured by @xcite and those derived from spectral modeling ( e.g. , * ? ? ? we therefore suggest that electron - hydrogen atom collisions play a significant role in determining the mid - ir continuum opacity of the betelgeuse layer , and that this mechanism must also be accounted for in models for the layer composition . we have shown that the layer of material surrounding the photosphere of betelgeuse has a non - uniform intensity distribution . a similar result was recently reported by @xcite based on interferometric measurements of the one - dimensional intensity and velocity profile of co gas in the inner atmosphere of betelgeuse . @xcite also reported a change in the one - dimensional co intensity and velocity profile between two epochs spaced by a year . we find that the intensity distribution of the layer varies between each the five observing epochs we present , which are also spaced by approximately one year . the non - uniformities that we observe in the layer reflect , on much smaller spatial scales , the non - uniformities observed in various components of the circumstellar environment of betelgeuse . a gaseous component at similar temperatures to the betelgeuse layer was shown to have an asymmetric intensity distribution by @xcite . our observations and those of @xcite could be linked by the opacities in both wavelength bands being partly caused by electron - hydrogen atom collisions . the opacity due to this mechanism is proportional to @xmath94 , implying that the observed stellar size decreases with shorter observing wavelengths . material at chromospheric temperatures is also observed to be asymmetrically distributed @xcite , and observations of a circumstellar envelope by @xcite at near - ir wavelengths also reveal significant asymmetry . asymmetry in the large - scale ( @xmath11 ) dust distribution surrounding betelgeuse was also observed by @xcite . molecular layer models have been applied to the few red supergiant stars that have been observed with sufficient spatial resolution , such as @xmath0 cephei @xcite and @xmath5 herculis @xcite . similar layers are identified around agb stars @xcite . the properties of individual red supergiant layers vary significantly from star to star , and a greater sample of measurements is required to understand what determines individual layer properties . the processes by which red supergiant stars form and accelerate dust , while crucial to mass - loss scenarios , are little understood . existing models of stellar atmospheres ( e.g. , * ? ? ? * ; * ? ? ? * ) can not predict locations for dust condensation , and attempts to model dust acceleration through radiation pressure have also so far been unsuccessful ( e.g. , * ? ? ? our multi - epoch study of a layer directly surrounding betelgeuse provides new insight into the processes involved in red supergiant mass - loss . m flux density , of betelgeuse over the period 2006@xmath22010 . the visual magnitude measurements , joined by a solid line , are 30-day averages of data obtained from the aavso online archive . the 11.15@xmath0 m flux density measurements , also listed in table 1 , are shown here as crosses . the timespans of each isi observing epoch are bounded by vertical dotted lines . ] we have shown that the basic properties of the layer , given in table 3 , exhibit striking variability from year to year . the layer is found to decrease in size and increase in optical depth over the years @xmath95 , and increase in size and decrease in optical depth between @xmath96 . the layer temperature does not show such systematic variability , although the largest layer size ( in 2010 ) does correspond to the highest temperature . the variability in the apparent mid - ir size of betelgeuse reported by @xcite can also be interpreted to indicate variations in the layers . intriguingly , a similar phenomenon was reported by @xcite , who noted a decrease in the size of betelgeuse at visible wavelengths followed by an increase . figure 5 shows measurements of the visual magnitude of betelgeuse during between 2006 and 2010 , as well as the isi measurements of the total 11.15@xmath0 m flux density ( also listed in table 1 ) . the timespans of the isi observing epochs are also plotted . the visual magnitude measurements were obtained from the online database of the american association of variable star observers ( aavso ) . the data , collected by a multitude of volunteer observers , were averaged over 30-day timespans ; each timespan with less than 10 measurements was discarded . the visual variability of betelgeuse over our observing epochs was approximately 0.9 magnitudes . the visual intensity of betelgeuse is not likely to represent that of the photosphere because the apparent stellar diameter at visible wavelengths is greater than that at near - ir wavelengths @xcite . also , using contemporaneous interferometric observations of betelgeuse at different wavelengths , @xcite found that hotspots evident in the visible were not present in the near - ir . observations of betelgeuse at visible wavelengths could represent regions within the layer observed at mid - ir wavelengths . it is interesting that the mid - ir flux density of betelgeuse , as well as the effective temperatures plotted in figure 3 , and the visual magnitude appear somewhat correlated , particularly in their increase during the 2009 and 2010 epochs . previous authors @xcite have concluded that the visual variability of betelgeuse is probably indicative of quasi - periodic pulsations , and is too great to be explained by variations in the hotspot distribution . the layer of material surrounding betelgeuse is a candidate location for the onset of dust formation . a number of authors have identified the presence of alumina dust directly above the photosphere @xcite . the derived physical parameters of the layer , such as the temperatures and hydrogen mass densities , are suitable for the formation of amorphous alumina @xcite . while some aspects of the modeled variability of the layer might be due to variations in the photosphere , measurements of changes in the apparent mid - ir appearance of betelgeuse must be largely caused by variations in the layer . if alumina contributes significantly to the mid - ir continuum opacity , these variations would represent variations in the dust content of the layer . our estimates of changes in the mass of the layer , on the order of @xmath97 ( see table 4 ) , might indicate evolution in the layer . the point sources we observe in the layer of material surrounding betelgeuse and their changes suggest that the layer dynamics are anisotropic . asymmetries observed in the circumstellar environment of betelgeuse are generally linked to the presence of giant convection cells on the stellar surface . such convection cells are thought to shape the circumstellar environment of betelgeuse by elevating cool photospheric material @xcite and driving large - scale chromospheric motions @xcite . our results indicate that giant convection cells have a significant role in shaping the betelgeuse layer . this is because the mechanism that elevates the layer material above the photosphere is clearly anisotropic . the large intensities of the observed point - source components imply the action of large - scale photospheric features , and their variability between observing epochs demonstrates the transient nature of the cause of the point sources . changes in the distribution of convection cells on the photosphere of betelgeuse could also influence the overall layer properties . while the nature of the non - uniform brightness distribution of the layer is yet to be determined , our observations represent evidence for the role of giant convection cells in shaping the inner atmosphere of betelgeuse . observations of a sample of red supergiant stars with high spatial resolution and wide bandwidth , over multiple epochs , are necessary to form a general picture for the role of molecular layers in mass - loss processes . more detailed interferometric imaging of betelgeuse at mid - ir wavelengths is also required to understand the true role of giant convection cells in shaping the layer . simultaneous visible , near - ir , and mid - ir interferometry could also provide interesting information on correspondences between different regions of the layer . interferometric studies with both high spatial resolution and high spectral resolution are planned for the isi @xcite , and have been performed at near - ir wavelengths by @xcite and @xcite . such investigations are aimed at revealing the molecular compositions , kinematic structures and true extents of red supergiant layers . more detailed modeling of layer opacities is also required , taking into account contributions from molecular absorption and emission , dust and electron - hydrogen atom collisions . upcoming optical and infrared interferometer systems , such as the magdalena ridge observatory interferometer and the matisse instrument at the very large telescope interferometer , as well as high spatial resolution observations with the next generation of extremely large telescopes could finally clarify our understanding of red supergiant mass loss . the isi has been used to conduct observations of betelgeuse over 2006@xmath22010 in the mid - ir continuum with high spatial resolution . by fitting visibility and closure phase measurements with image models , the apparent stellar size was measured at each observing epoch , and up to two point sources were sometimes detected at the edge of the stellar disk . the results were interpreted in terms of an absorbing layer surrounding an assumed 3641k photosphere . we have shown that : 1 . observations during every epoch can be modeled in terms of a layer surrounding the photosphere , with optical depths of @xmath11 at 11.15@xmath0 m wavelength , outer sizes between 1.16@xmath7 and 1.36@xmath7 , and temperatures between 1900k and 2800k . electron - hydrogen atom collisions contribute to the opacity of the layer . 3 . assuming a non - varying photosphere , the layer exhibits significant variability . 4 . the layer has a non - uniform intensity distribution that also varies from year to year . w. fitelson , b. walp , c.s . ryan , d.d.s . hale , a.a . chandler , k. reichl , r.l . griffith , v. toy , as well as many undergraduate researchers all participated in these observations , and their excellent help is greatly appreciated . this research made use of the simbad database . fitting programs used the mpfit idl routines of c. b. markwardt . we are grateful for support from the gordon and betty moore foundation , the office of naval research , and the national science foundation . adams , w. s. , & maccormack , e. 1935 , , 81 , 119 antia , h. m. , chitre , s. m. , & narasimha , d. 1984 , , 282 , 574 bester , m. , danchi , w. c. , hale , d. , townes , c. h. , degiacomi , c. g. , mekarnia , d. , & geballe , t. r. 1996 , , 463 , 336 bloemhof , e. e. , danchi , w. c. , & townes , c. h. 1985 , , 299 , l37 bonneau , d. , & labeyrie , a. 1973 , , 181 , l1 boyajian , t. s. , et al . 2009 , , 691 , 1243 buscher , d. f. , baldwin , j. e. , warner , p. j. , & haniff , c. a. 1990 , , 245 , 7p chiavassa , a. , plez , b. , josselin , e. , & freytag , b. 2009 , , 506 , 1351 chiavassa , a. , haubois , x. , young , j. s. , plez , b. , josselin , e. , perrin , g. , & freytag , b. 2010 , , 515 , a12 dalgarno , a. , & lane , n. f. 1966 , , 145 , 623 danchi , w. c. , bester , m. , degiacomi , c. g. , greenhill , l. j. , & townes , c. h. 1994 , , 107 , 1469 de beck , e. , decin , l. , de koter , a. , justtanont , k. , verhoelst , t. , kemper , f. , & menten , k. m. 2010 , , 523 , a18 dirks , u. , schirrmacher , v. , & sedlmayr , e. 2008 , , 491 , 643 freytag , b. , steffen , m. , & dorch , b. 2002 , astronomische nachrichten , 323 , 213 gehrz , r. 1989 , in interstellar dust , eds . l. j. allamandola & a. g. g. m. tielens , iau symp . 135 , pg . 445 gilliland , r. l. , & dupree , a. k. 1996 , , 463 , l29 gustafsson , b. , edvardsson , b. , eriksson , k. , jrgensen , u. g. , nordlund , . , & plez , b. 2008 , , 486 , 951 hale , d. d. s. , et al . 2000 , , 537 , 998 harper , g. m. , brown , a. , & lim , j. 2001 , , 551 , 1073 harper , g. m. , brown , a. , & guinan , e. f. 2008 , , 135 , 1430 harper , g. m. , richter , m. j. , ryde , n. , brown , a. , brown , j. , greathouse , t. k. , & strong , s. 2009 , , 701 , 1464 hartmann , l. , & macgregor , k. b. 1980 , , 242 , 260 haubois , x. , et al . 2009 , , 508 , 923 hebden , j. c. , eckart , a. , & hege , e. k. 1987 , , 314 , 690 heger , a. , & langer , n. 1998 , , 334 , 210 humphreys , r. m. 1986 , in luminous stars and associations in galaxies , eds . c. w. h. de loore , a. j. willis , & p. laskarides , iau symp . 116 , pg . 45 jennings , d. e. , & sada , p. v. 1998 , science , 279 , 844 jennison , r. c. 1958 , , 118 , 276 josselin , e. , & plez , b. 2007 , , 469 , 671 kervella , p. , verhoelst , t. , ridgway , s. t. , perrin , g. , lacour , s. , cami , j. , & haubois , x. 2009 , , 504 , 115 kiss , l. l. , szab , g. m. , & bedding , t. r. 2006 , , 372 , 1721 lim , j. , carilli , c. l. , white , s. m. , beasley , a. j. , & marson , r. g. 1998 , , 392 , 575 michelson , a. a. , & pease , f. g. 1921 , , 53 , 249 monnier , j. d. , geballe , t. r. , & danchi , w. c. 1998 , , 502 , 833 ohnaka , k. 2004 , , 421 , 1149 ohnaka , k. , et al . 2005 , , 429 , 1057 ohnaka , k. , et al . 2009 , , 503 , 183 ohnaka , k. , et al . 2011 , arxiv:1104.0958 pease , f. g. 1922 , , 34 , 346 perrin , g. , ridgway , s. t. , coud du foresto , v. , mennesson , b. , traub , w. a. , & lacasse , m. g. 2004 , , 418 , 675 perrin , g. , ridgway , s. t. , verhoelst , t. , schuller , p. a. , coud du foresto , v. , traub , w. a. , millan - gabet , r. , & lacasse , m. g. 2005 , , 436 , 317 perrin , g. , et al . 2007 , , 474 , 599 plez , b. 2010 , in hot and cool : bridging gaps in massive star evolution , eds . c. leitherer , p. bennett , p. morris , & j. van loon , asp conf . ser . 425 , pg . 124 reid , m. j. , & menten , k. m. 1997 , , 476 , 327 rosseland , s. 1924 , , 84 , 525 ryde , n. , harper , g. m. , richter , m. j. , greathouse , t. k. , & lacy , j. h. 2006 , , 637 , 1040 salpeter , e. e. 1977 , , 15 , 267 schwarzschild , m. 1975 , , 195 , 137 stencel , r. e. , pesce , j. e. , & bauer , w. h. 1989 , , 97 , 1120 tatebe , k. , & townes , c. h. 2006 , , 644 , 1145 tatebe , k. , chandler , a. a. , wishnow , e. h. , hale , d. d. s. , & townes , c. h. 2007 , , 670 , l21 townes , c. h. , wishnow , e. h. , hale , d. d. s. , & walp , b. 2009 , , 697 , l127 tsuji , t. 2000 , , 538 , 801 tsuji , t. 2000 , , 540 , l99 tsuji , t. 2006 , , 645 , 1448 van belle , g. t. , et al . 2010 , arxiv:1012.1146 verhoelst , t. , et al . 2006 , , 447 , 311 verhoelst , t. , van der zypen , n. , hony , s. , decin , l. , cami , j. , & eriksson , k. 2009 , , 498 , 127 weiner , j. , danchi , w. c. , hale , d. d. s. , mcmahon , j. , townes , c. h. , monnier , j. d. , & tuthill , p. g. 2000 , , 544 , 1097 weiner , j. , hale , d. d. s. , & townes , c. h. 2003 , , 588 , 1064 weiner , j. , hale , d. d. s. , & townes , c. h. 2003 , , 589 , 976 weiner , j. 2004 , , 611 , l37 wilson , r. w. , dhillon , v. s. , & haniff , c. a. 1997 , , 291 , 819 wishnow , e. h. , mallard , w. , ravi , v. , lockwood , s. , fitelson , w. , wertheimer , d. , & townes , c. h. 2010 , , 7734 woitke , p. 2006 , , 452 , 537 woitke , p. 2007 , in why galaxies care about agb stars : their importance as actors and probes , eds . f. kerschbaum , c. charbonnel , & r. f. wing , asp conf . ser . 378 , pg . 156 woodruff , h. c. , et al . 2009 , , 691 , 1328 yoon , s .- c . , & cantiello , m. 2010 , , 717 , l62 young , j. s. , et al . 2000 , , 315 , 635

in this paper , we make some progress on a problem that has essentially been open since 1902 @xcite . it is well known that for each monic polynomial @xmath3 one can construct a matrix @xmath4 , such that @xmath5 . one of the options is a companion matrix ( see , e.g. , @xcite ) @xmath6.\ ] ] thus , we can numerically compute the zeros of the polynomial @xmath7 as eigenvalues of the corresponding companion matrix @xmath8 using tools from numerical linear algebra . this approach is used in many numerical packages , for instance in the ` roots ` command in matlab @xcite . the aim of this paper is to find a similar elegant tool for finding the zeros of a system of two bivariate polynomials of degree @xmath1 @xmath9 an approach analogous to the univariate case would be to construct matrices @xmath10 , @xmath11 , @xmath12 , @xmath13 , @xmath14 , and @xmath15 of size @xmath16 such that @xmath17 \det(a_2+x b_2 + y c_2)&=q(x , y ) . \end{split}\ ] ] this would give an equivalent two - parameter eigenvalue problem @xcite @xmath18 ( a_2+x b_2 + y c_2 ) \ , u_2&=0 \end{split}\ ] ] that could be solved by the standard tools like the qz algorithm , see @xcite for details . this idea looks promising , but there are many obstacles on the way to a working numerical algorithm that could be applied to a system of bivariate polynomials . although it is known for _ more than a century _ @xcite that such matrices of size @xmath19 exist , so far there are no efficient numerical algorithms that can construct them . even worse , it seems that the construction of such matrices might be an even harder problem than finding zeros of polynomials @xmath7 and @xmath20 . there exist simple and fast constructions @xcite that build matrices of size @xmath21 that satisfy ( [ eq : detrep ] ) , where the resulting two - parameter eigenvalue problem ( [ eq : twopar ] ) is singular ; we will discuss more details in section [ sec : multipar ] . recent results @xcite show that it is possible to solve singular two - parameter eigenvalue problems numerically for small to medium - sized matrices . however , the @xmath21 size of the matrices pushes the complexity of the algorithm to the enormous @xmath22 and it is reported in @xcite that this approach to compute zeros is competitive only for polynomials of degree @xmath23 . the construction of @xcite yields matrices that are of asymptotic order @xmath24 , while those of @xcite are of asymptotic order @xmath25 . in this paper we give two new representations . the first one uses the tree structure of monomials in @xmath26 and @xmath27 . the resulting matrices are smaller than those of @xcite , with the same asymptotic order @xmath25 . this representation can be used for bivariate polynomials as well as for polynomial multiparameter eigenvalue problems @xcite ; that is , for polynomials with matrix coefficients . the second representation is even more condensed , with asymptotic order @xmath28 , and can be applied to scalar bivariate polynomials . although the size of the matrices asymptotically still grows quadratically with @xmath1 , the smaller size renders this approach attractive for polynomials of degree @xmath29 , or for larger @xmath1 if the polynomials have only few terms . this already is an interesting size for a practical use and might trigger additional interest in such methods that could culminate in even more efficient representations . moreover , as we will see , for modest @xmath1 , the order of the matrices is only roughly @xmath30 . furthermore , for polynomials of degree 3 , we present a construction of matrices of order ( exactly ) 3 . there are other ways to study a system of polynomials as an eigenvalue problems , see , e.g. , @xcite and @xcite , but they involve more symbolic computation . in @xcite an algorithm is proposed that only requires to solve linear systems and check rank conditions , which are similar tools that we use in the staircase method @xcite to solve the obtained singular two - parameter eigenvalue problem . of course , there are many numerical methods that can be applied to systems of bivariate polynomials , two main approaches are the homotopy continuation and the resultant method , see , e.g. , @xcite and the references therein . there are also many methods which aim to compute only real solutions of a system of two real bivariate polynomials , see , e.g. , @xcite . we compare our method with two existing approaches , mathematica s nsolve @xcite and phcpack @xcite in section [ sec : numex ] , and show that our approach is competitive for polynomials up to degree @xmath31 . let us mention that another advantage of writing the system of bivariate polynomials as a two - parameter eigenvalue problem is that then we can apply iterative subspace numerical methods such as the jacobi davidson method and compute just a small part of zeros close to a given target @xmath32 @xcite ; we will not pursue this approach in this paper . the rest of this paper is organized as follows . in section [ sec : motiv ] we give some applications where bivariate polynomial systems have to be solved . in section [ sec : detrep ] we introduce determinantal representations . section [ sec : multipar ] focuses on two - parameter eigenvalue problems . in section [ sec : lin1 ] we give a determinantal representation that is based on the `` tree '' of monomials , involves no computation , and is suitable for both scalar and matrix polynomials . the matrices of the resulting representation are asymptotically of order @xmath25 . in section [ sec : lin2 ] we give a representation with smaller matrices , of asymptotic order @xmath28 , that involves just a trivial amount of numerical computation ( such as computing roots of low - degree univariate polynomials ) and can be computed very efficiently . this representation may be used for scalar polynomials . we end with some numerical experiments in section [ sec : numex ] and conclusions in section [ sec : conc ] . in delay differential equations , determining critical delays in the case of so - called commensurate delays may lead to a problem of type @xcite . the simplest example is of the form @xmath33 , where @xmath34 is the delay ; asked are values of @xmath35 that results in periodic solutions . this yields @xmath7 and @xmath20 of degrees 2 and 3 , respectively . more delay terms with delays that are multiples of @xmath35 gives polynomials of higher degree . polynomial systems of form arise in numerous applications and fields , such as signal processing @xcite and robotics @xcite . in computer aided design , one may be interested in the intersections of algebraic curves , such as ellipses @xcite . in two - dimensional subspace minimization @xcite , such as polynomial tensor optimization , one is interested in two - dimensional searches @xmath36 , where @xmath37 , @xmath26 is the current point , and @xmath38 and @xmath39 are search directions ; see @xcite and the references therein . in systems and control the first - order conditions of the @xmath40-approximation problem of minimizing @xmath41 , for a given impulse response @xmath42 of degree @xmath1 , and @xmath43 , lead to a system of type @xcite . when considering quadratic eigenvalue problems in numerical linear algebra , it is of interest to determine @xmath44 , as an approximate eigenvalue for a given approximate eigenvector @xmath45 , which gives a system of degree 3 in the real and imaginary part of @xmath46 ( * ? ? ? generalizations to polynomial eigenvalue problems give rise to polynomials @xmath7 and @xmath20 of higher degree . also , there has been some recent interest in this problem in the context of the chebfun2 project @xcite . in chebfun2 , nonlinear real bivariate functions are approximated by bivariate polynomials , so solving is relevant for finding zeros of systems of real nonlinear bivariate functions and for finding local extrema of such functions . in this section we introduce determinantal representations and present some existing constructions . the difference between what should theoretically be possible and what can be done in practice is huge . the algorithms we propose reduce the difference only by a small ( but still significant ) factor ; there seems to be plenty of room for future improvements . we say that a bivariate polynomial @xmath47 has degree @xmath1 if all its monomials @xmath48 have total degree less or equal to @xmath1 , i.e. , @xmath49 , and if at least one of the monomials has total degree equal to @xmath1 . we say that the square @xmath50 matrices @xmath51 , and @xmath52 form a _ determinantal representation _ of the polynomial @xmath7 if @xmath53 . as our motivation is to use eigenvalue methods to solve polynomial systems , we will , instead of determinantal representation , often use the term _ linearization _ since a determinantal representation transforms an eigenvalue problem that involves polynomials of degree @xmath1 into a linear eigenvalue problem . a definition of linearization that extends that for the univariate case ( see , e.g. , @xcite ) is the following . a linear bivariate pencil @xmath54 of size @xmath50 is a linearization of the polynomial @xmath47 if there exist two polynomial matrices @xmath55 and @xmath56 such that @xmath57 and @xmath58.\ ] ] we are interested not only in linearizations of scalar polynomials but also in linearizations of matrix bivariate polynomials of the form ( cf . ) @xmath59 where the @xmath60 are @xmath61 matrices . in line with the above , a linear pencil @xmath62 of matrices of size @xmath50 presents a linearization ( determinantal representation ) of the matrix polynomial @xmath63 if there exist two polynomial matrices @xmath55 and @xmath56 such that @xmath57 and @xmath64.\ ] ] in this case @xmath65 . each linearization of a matrix polynomial gives a linearization for a scalar polynomial , as we can think of scalars as of @xmath66 matrices ; the opposite is not true in general . dixon @xcite showed that for every scalar bivariate polynomial @xmath47 of degree @xmath1 there exists a determinantal representation with symmetric matrices of size @xmath16 . dickson @xcite later showed that this result can not be extended to general polynomials in more than two variables , except for three variables and polynomials of degree two and three , and four variables and polynomials of degree two . although they both give constructive proofs , there does not seem to exist an efficient numerical algorithm to construct the determinantal representation with matrices of size @xmath16 for a given bivariate polynomial of degree @xmath1 . in recent years , the research in determinantal representations is growing , as determinantal representations for a particular subset of polynomials , _ real zero polynomials _ , are related to linear matrix inequality ( lmi ) constraints used in semidefinite programming sdp . for an overview see , e.g. , @xcite ; here we give just the essentials for bivariate polynomials that are related to our problem . we say that a real polynomial @xmath47 satisfies the _ real zero condition _ with respect to @xmath67 if for all @xmath68 the univariate polynomial @xmath69 has only real zeros . a two - dimensional _ lmi set _ is defined as @xmath70 where @xmath51 , and @xmath52 are symmetric matrices of size @xmath50 and @xmath71 stands for positive semidefinite . in sdp we are interested in convex sets @xmath72 that admit an lmi representation , i.e. , @xmath73 is an lmi set for certain matrices @xmath51 and @xmath52 . such sets are called _ spectrahedra _ and helton and vinnikov @xcite showed that such @xmath73 must be an algebraic interior , whose minimal defining polynomial @xmath7 satisfies the real zero condition with respect to any point in the interior of @xmath73 . their results state that if a polynomial @xmath47 of degree @xmath1 satisfies real zero condition with respect to @xmath32 , then there exist symmetric matrices @xmath51 , and @xmath52 of size @xmath16 such that @xmath74 and @xmath75 . matrices @xmath51 , and @xmath52 thus form a particular determinantal representation for @xmath7 . the problem of constructing an lmi representation with symmetric or hermitian matrices @xmath76 and @xmath52 for a given spectrahedron @xmath73 raised much more interest than the related problem of generating a determinantal representation for a generic bivariate polynomial . there exist procedures , which rely heavily on slow symbolic computation or other expensive steps , that return an lmi representation with hermitian matrices for a given spectrahedron , but they are not efficient enough . for instance , a method from @xcite , based on the proof from @xcite , does return @xmath16 matrices for a polynomial of degree @xmath1 , but the reported times ( 10 seconds for a polynomial of degree 10 ) show that it is much too slow for our purpose . as a first step of the above method is to find zeros of a system of bivariate polynomials of degree @xmath1 and @xmath77 , this clearly can not be efficient enough for our needs . in addition , we are interested in determinantal representations for polynomials that do not necessary satisfy the real zero condition . in sdp and lmi the matrices have to be symmetric or hermitian , which is not required in our case . we need a simple and fast numerical construction of matrices that satisfy ( [ eq : detrep ] ) and are as small as possible ideally their size should increase linearly and not quadratically with @xmath1 . if we look at the available determinantal representations for generic bivariate polynomials , we first have the linearization by khazanov with matrices of size @xmath78 @xcite . in ( * appendix ) , a smaller linearization for bivariate matrix polynomials is given with block matrices of order @xmath79 . the linearization uses all monomials of degree up to @xmath77 and contains a direct expression for the matrices @xmath80 , @xmath81 and @xmath52 such that @xmath82 . similar to @xcite , it can be applied to matrix polynomials . we give an example for a general matrix polynomial of degree @xmath83 , from which it is possible to deduce the construction for a generic degree . this linearization will be superseded in section [ sec : lin1 ] by a more economical one . [ ex : lin0 ] ( * ? ? ? * appendix ) we take a matrix bivariate polynomial of degree @xmath83 @xmath84 if @xmath45 is a nonzero vector , then @xmath85 if and only if @xmath86 , where @xmath87}\ ] ] and @xmath88^t.\ ] ] we have @xmath89 and @xmath90 is a linearization of @xmath63 . we remark that quarez @xcite also gives explicit expressions for determinantal representations . he is interested in symmetric representations and is able to construct , for a bivariate polynomial of degree @xmath1 such that @xmath91 , a linearization with symmetric matrices of size @xmath92 , where @xmath93 this has asymptotically the same order as the linearization that we give in section [ sec : lin1 ] . let us also remark that in the phase , when we are solving a two - parameter eigenvalue problem to compute the zeros of a system of two bivariate polynomials , we can not exploit the fact that the matrices are symmetric , so this is not important for our application . there are some other available tools , for instance it is possible to construct a determinantal representation using the package ncalgebra @xcite for noncommutative algebra that runs in mathematica @xcite , but this does not give satisfactory results for our application as the matrices that we can construct have smaller size . in this section we briefly present the two - parameter eigenvalue problem and the available numerical methods . a motivation for the search for small determinantal representations is that if we transform a system of bivariate polynomials into an eigenvalue problem , then we can apply existing numerical methods for such problems . a _ two - parameter eigenvalue problem _ has the form where @xmath94 , and @xmath95 are given @xmath96 complex matrices . we are looking for @xmath97 and nonzero vectors @xmath98 , @xmath99 , such that ( [ eq : twopar ] ) is satisfied . in such case we say that a pair @xmath100 is an eigenvalue and the tensor product @xmath101 is the corresponding eigenvector . if we introduce the so - called operator determinants , the matrices @xmath102 then the problem ( [ eq : twopar ] ) is related to a coupled pair of generalized eigenvalue problems @xmath103 \label{drugi}\\[-1.5ex ] \delta_2 \ , w&=y \ \delta_0 \ , w\nonumber\end{aligned}\ ] ] for a decomposable tensor @xmath104 . if @xmath105 is nonsingular , then atkinson @xcite showed that the solutions of ( [ eq : twopar ] ) and ( [ drugi ] ) agree and the matrices @xmath106 and @xmath107 commute . in the nonsingular case the two - parameter problem ( [ eq : twopar ] ) has @xmath108 eigenvalues and we can numerically solve it with a variant of the qz algorithm on ( [ drugi ] ) from @xcite . ideally , if we could construct a determinantal representation with matrices @xmath19 for a bivariate polynomial of degree @xmath1 , this would be the method that we would apply on the `` companion '' two - parameter eigenvalue problem to get the zeros of the polynomial system . as matrices @xmath109 , and @xmath110 have size @xmath111 , the computation of all eigenvalues of a nonsingular two - parameter eigenvalue problem has time complexity @xmath112 , which would lead to @xmath113 algorithm for a system of bivariate polynomials . of course , for this approach we need a construction of a determinantal representation with matrices @xmath16 that should not be more computationally expensive than the effort to solve a two - parameter eigenvalue problem . unfortunately , all practical constructions for determinantal representations ( including the two presented in this paper ) return matrices that are much larger than @xmath16 . if we have a determinantal representation with matrices larger than the degree of the polynomial , then the corresponding two - parameter eigenvalue problem is singular , which means that both matrix pencils ( [ drugi ] ) are singular , and we are dealing with a more difficult problem . there exists a numerical method from @xcite that computes the regular eigenvalues of ( [ eq : twopar ] ) from the common regular part of ( [ drugi ] ) . for the generic singular case it is shown in @xcite that the regular eigenvalues of ( [ eq : twopar ] ) and ( [ drugi ] ) do agree . for other types of singular two - parameter eigenvalue problems the relation between the regular eigenvalues of ( [ eq : twopar ] ) and ( [ drugi ] ) is not completely known , but the numerical examples indicate that the method from @xcite can be successfully applied to such problems as well . however , the numerical method , which is a variant of a staircase algorithm @xcite , has to make a lot of decisions on the numerical rank and a single inaccurate decision can cause the method to fail . as the size of the matrices increases , the gaps between singular values may numerically disappear and it may be difficult to solve the problem . this is not the only issue that prevents the use of determinantal representations to solve a bivariate system . the algorithm for the singular two - parameter eigenvalue problems still has complexity @xmath112 , but the fast determinantal representations that we are aware of return matrices of size @xmath21 instead of @xmath114 . this is what pushes the overall complexity to @xmath22 and makes this approach efficient only for polynomials of small degree . nonetheless , at complexity so high , each construction that gives a smaller determinantal representation can make a change . in view of this , we propose two new linearizations in the next two sections . we are interested in linearizations of the matrix polynomial @xmath115 of degree @xmath1 , where @xmath60 are square matrices . our goal is to find square matrices @xmath51 , and @xmath52 as small as possible such that @xmath116 . also , we need a relation that @xmath85 if and only if @xmath117 , where @xmath118 is a tensor product of @xmath45 and a polynomial of @xmath26 and @xmath27 . the linearization in this section also applies to scalar bivariate polynomials , where all matrices are @xmath66 and @xmath119 . in section [ sec : detrep ] we have given a linearization with block matrices of order @xmath79 . we can view this linearization in the following way . if @xmath85 for @xmath120 , then @xmath121 , where the vector @xmath118 has the form @xmath122^t.\ ] ] this means that @xmath118 always begins with the initial block @xmath45 and then contains all blocks of the form @xmath123 where @xmath124 . to simplify the presentation we will usually omit @xmath45 when referring to the blocks of the vector ( [ eq : blockvec ] ) . the blocks are ordered in the degree negative lexicographic ordering , i.e. , @xmath125 if @xmath126 , or @xmath127 and @xmath128 . the above block structure of vector @xmath118 is defined in the rows of the matrix from the second one to the last one ( see example [ ex : lin0 ] ) . for each block @xmath129 of ( [ eq : blockvec ] ) such that @xmath130 there always exists a preceding block @xmath20 of the grade @xmath131 such that either @xmath132 or @xmath133 ( when @xmath134 and @xmath135 both options are possible ) . suppose that @xmath132 , @xmath136 , and @xmath137 , where function @xmath138 returns the index of a block . then the matrix @xmath139 has block @xmath140 on position @xmath141 and block @xmath142 on position @xmath143 . these are the only nonzero blocks in the block row @xmath144 . a similar construction with @xmath140 replaced by @xmath145 is used in the case @xmath133 . the first block row of the matrix @xmath139 is used to represent the matrix polynomial @xmath63 . one can see that there exist linear pencils @xmath146 , @xmath147 , such that @xmath148 \underline{u},\ ] ] where @xmath149 is the number of blocks in ( [ eq : blockvec ] ) . the pencils in ( [ eq : firstrow ] ) are not unique . for instance , a term @xmath150 of @xmath63 can be represented in one of up to the three possible ways : 1 . if @xmath151 , we can set @xmath152 where @xmath153 , 2 . if @xmath154 , we can set @xmath155 where @xmath156 , 3 . if @xmath157 , we can set @xmath158 where @xmath159 . based on the above discussion we see that not all the blocks in ( [ eq : blockvec ] ) are needed to represent a matrix polynomial @xmath63 . what we need is a minimal set of monomials @xmath160 , where @xmath151 , that is sufficient for a matrix polynomial of degree @xmath1 . we can formulate the problem of finding the smallest possible set for a given polynomial as a graph problem . for the polynomial of degree @xmath161.[fig : graph ] ] we can think about all possible terms @xmath162 , where @xmath151 , as of nodes in a directed graph @xmath163 with the root @xmath164 and a directed edge from node @xmath165 to node @xmath166 if @xmath167 or @xmath168 ( see figure [ fig : graph ] for the case @xmath169 ) . now , we are looking for the smallest connected subgraph @xmath170 with a root @xmath164 that can represent a given polynomial . equivalently , we are looking for a minimal directed rooted tree . let us remember that for each term @xmath171 of the polynomial @xmath63 there are up to three possible nodes in the graph @xmath163 that can be used to represent it . it is sufficient that one of these nodes is in a minimal tree @xmath170 . furthermore , if @xmath172 , then we can assume that we always use a node of degree @xmath131 to represent @xmath171 and then there are only one or two options for a given term . all together , each nonzero term @xmath150 , where @xmath173 , in the polynomial @xmath174 defines one of the following rules for the subgraph @xmath170 : * if @xmath175 then @xmath176 has to be in the subgraph @xmath170 , * if @xmath177 then @xmath178 has to be in the subgraph @xmath170 , * if @xmath154 and @xmath157 then at least one of @xmath179 or @xmath180 has to be in the subgraph @xmath170 . the term @xmath181 can be presented by the root @xmath164 , which is always present in the subgraph @xmath170 . finding a minimal tree for a given polynomial is not an easy problem : it can be formulated as an np - hard directed steiner tree problem ( dst ) ( see , e.g. , @xcite ) , where one has a directed graph @xmath182 with nonnegative weights on edges and the goal is to find the minimum weight directed rooted tree that connects all terminals @xmath183 to a given root @xmath184 . suppose that we are looking for a minimal representation tree for a polynomial @xmath63 of degree @xmath1 . in the graph @xmath163 , which contains all nodes @xmath162 for @xmath151 ( see figure [ fig : graph ] for the case @xmath169 ) , we put weight @xmath164 on all directed @xmath26 and @xmath27 edges . now we add a new vertex for each monomial @xmath162 that is present in @xmath63 and connect it with zero weight edges from all possible nodes in @xmath163 that could be used to represent the monomial in the linearization . we make a dst problem by taking node @xmath164 as a root and all newly added vertices as terminals . from a solution of the dst problem the minimal representation tree can be recovered . although this is an np - hard problem , there exist some polynomial time approximation algorithms that give a solution close to the optimal one and could be used to construct a small determinantal representation for a given polynomial with small number of nonzero terms . for the latest available algorithms , see , e.g. , @xcite and the references therein . [ ex : minlin ] we are interested in a minimal tree for the matrix polynomial @xmath185 nonzero terms in ( [ eq : pol6 ] ) define the nodes that have to be present in the minimal subgraph . they are either strictly defined as are the nodes @xmath164 , @xmath186 , and @xmath187 , or come in pairs where at least one element of each pair has to be present in the subgraph . such pairs are @xmath188 , and @xmath189 . the situation is presented in figure [ fig : graphex ] , where nodes and pairs , such that either left or right node has to be included , are shadowed green . the nodes of the minimal connected subgraph that includes all required nodes are colored red . in a dst formulation each green shadow presents a terminal linked by zero weight edges to one or two nodes that are included in the region . on all other edges we put weight 1 and then search for the minimum weight directed rooted tree that connects all terminals to the root @xmath164 . for the matrix polynomial ( [ eq : pol6 ] ) of degree @xmath190.[fig : graphex ] ] matrix polynomial ( [ eq : pol6 ] ) can thus be represented with matrices of block size @xmath191 . if we order the nodes of the subgraph in the degree negative lexicographic ordering , then @xmath192 has the form @xmath193^t\ ] ] and a possible first block row of @xmath62 has the form @xmath194.\ ] ] in the subsequent block rows , the matrix @xmath62 has only 20 nonzero blocks , 10 of them are identity blocks on the main diagonal . the remaining nonzero blocks are @xmath140 on block positions @xmath195 , @xmath196 , @xmath197 , @xmath198 , @xmath199 , @xmath200 , @xmath201 and blocks @xmath145 on positions @xmath202 . if we have a generic matrix polynomial @xmath63 , whose terms are all nonzero , then it is easy to see that the subgraph that contains all terms @xmath162 , where @xmath151 and either @xmath175 or @xmath203 is even , is minimal . the detailed situation for the case @xmath204 is presented in figure [ fig : mingraph ] , and representation trees for polynomials of degree from 1 to 8 are presented in figure [ fig : lin1first8 ] . counting the number of nodes in the tree gives the following result @xmath205 if we compare this with the linearization from example [ ex : lin0 ] that has matrices of block size @xmath79 , we see that the new linearization uses matrices of roughly half size . the size of the matrices is also smaller than ( [ eq : quarez ] ) from @xcite , which has the same asymptotic order . for a generic polynomial of degree @xmath161.[fig : mingraph ] ] = [ circle , draw , fill = blue!40 , inner sep=0pt , minimum size=5pt , node distance=0.35 cm ] \(00 ) ; ; \(00 ) ; ( 10 ) [ below left of=00 ] ; ( 01 ) [ below right of=00 ] ; ; ( 00 ) to ( 10 ) ; ( 00 ) to ( 01 ) ; \(00 ) ; ( 10 ) [ below left of=00 ] ; ( 20 ) [ below left of=10 ] ; ( 01 ) [ below right of=00 ] ; ( 02 ) [ below right of=01 ] ; ; ( 00 ) to ( 01 ) to ( 02 ) ; ( 00 ) to ( 10 ) to ( 20 ) ; \(00 ) ; ( 10 ) [ below left of=00 ] ; ( 20 ) [ below left of=10 ] ; ( 30 ) [ below left of=20 ] ; ( 01 ) [ below right of=00 ] ; ( 02 ) [ below right of=01 ] ; ( 03 ) [ below right of=02 ] ; ( 12 ) [ below left of=02 ] ; ; ( 00 ) to ( 01 ) to ( 02 ) to ( 03 ) ; ( 00 ) to ( 10 ) to ( 20 ) to ( 30 ) ; ( 02 ) to ( 12 ) ; \(00 ) ; ( 10 ) [ below left of=00 ] ; ( 20 ) [ below left of=10 ] ; ( 30 ) [ below left of=20 ] ; ( 40 ) [ below left of=30 ] ; ( 50 ) [ below left of=40 ] ; ( 60 ) [ below left of=50 ] ; ( 70 ) [ below left of=60 ] ; ( 01 ) [ below right of=00 ] ; ( 02 ) [ below right of=01 ] ; ( 12 ) [ below left of=02 ] ; ( 22 ) [ below left of=12 ] ; ( 32 ) [ below left of=22 ] ; ( 42 ) [ below left of=32 ] ; ( 52 ) [ below left of=42 ] ; ( 03 ) [ below right of=02 ] ; ( 04 ) [ below right of=03 ] ; ( 05 ) [ below right of=04 ] ; ( 06 ) [ below right of=05 ] ; ( 16 ) [ below left of=06 ] ; ( 07 ) [ below right of=06 ] ; ( 15 ) [ below left of=04 ] ; ( 25 ) [ below left of=15 ] ; ( 35 ) [ below left of=25 ] ; ; ( 00 ) to ( 01 ) to ( 02 ) to ( 03 ) to ( 04 ) to ( 05 ) to ( 06 ) to ( 07 ) ; ( 00 ) to ( 10 ) to ( 20 ) to ( 30 ) to ( 40 ) to ( 50 ) to ( 60 ) to ( 70 ) ; ( 02 ) to ( 12 ) to ( 22 ) to ( 32 ) to ( 42 ) to ( 52 ) ; ( 04 ) to ( 15 ) to ( 25 ) to ( 35 ) ; ( 06 ) to ( 16 ) ; \(00 ) ; ( 10 ) [ below left of=00 ] ; ( 20 ) [ below left of=10 ] ; ( 30 ) [ below left of=20 ] ; ( 40 ) [ below left of=30 ] ; ( 50 ) [ below left of=40 ] ; ( 60 ) [ below left of=50 ] ; ( 01 ) [ below right of=00 ] ; ( 02 ) [ below right of=01 ] ; ( 12 ) [ below left of=02 ] ; ( 22 ) [ below left of=12 ] ; ( 32 ) [ below left of=22 ] ; ( 42 ) [ below left of=32 ] ; ( 03 ) [ below right of=02 ] ; ( 04 ) [ below right of=03 ] ; ( 05 ) [ below right of=04 ] ; ( 06 ) [ below right of=05 ] ; ( 15 ) [ below left of=04 ] ; ( 25 ) [ below left of=15 ] ; ; ( 00 ) to ( 01 ) to ( 02 ) to ( 03 ) to ( 04 ) to ( 05 ) to ( 06 ) ; ( 00 ) to ( 10 ) to ( 20 ) to ( 30 ) to ( 40 ) to ( 50 ) to ( 60 ) ; ( 04 ) to ( 15 ) to ( 25 ) ; ( 02 ) to ( 12 ) to ( 22 ) to ( 32 ) to ( 42 ) ; \(00 ) ; ( 10 ) [ below left of=00 ] ; ( 20 ) [ below left of=10 ] ; ( 30 ) [ below left of=20 ] ; ( 40 ) [ below left of=30 ] ; ( 50 ) [ below left of=40 ] ; ( 01 ) [ below right of=00 ] ; ( 02 ) [ below right of=01 ] ; ( 12 ) [ below left of=02 ] ; ( 22 ) [ below left of=12 ] ; ( 32 ) [ below left of=22 ] ; ( 03 ) [ below right of=02 ] ; ( 04 ) [ below right of=03 ] ; ( 14 ) [ below left of=04 ] ; ( 05 ) [ below right of=04 ] ; ; ( 00 ) to ( 01 ) to ( 02 ) to ( 03 ) to ( 04 ) to ( 05 ) ; ( 00 ) to ( 10 ) to ( 20 ) to ( 30 ) to ( 40 ) to ( 50 ) ; ( 02 ) to ( 12 ) to ( 22 ) to ( 32 ) ; ( 04 ) to ( 14 ) ; \(00 ) ; ( 10 ) [ below left of=00 ] ; ( 20 ) [ below left of=10 ] ; ( 30 ) [ below left of=20 ] ; ( 40 ) [ below left of=30 ] ; ( 01 ) [ below right of=00 ] ; ( 02 ) [ below right of=01 ] ; ( 12 ) [ below left of=02 ] ; ( 22 ) [ below left of=12 ] ; ( 03 ) [ below right of=02 ] ; ( 04 ) [ below right of=03 ] ; ; ( 00 ) to ( 01 ) to ( 02 ) to ( 03 ) to ( 04 ) ; ( 00 ) to ( 10 ) to ( 20 ) to ( 30 ) to ( 40 ) ; ( 02 ) to ( 12 ) to ( 22 ) ; [ thm : one ] we can linearize each matrix polynomial @xmath63 of degree @xmath1 with matrices of block size @xmath206 from ( [ eq : sizeg ] ) using a minimal tree @xmath170 that contains the terms @xmath162 , where @xmath151 and either @xmath175 or @xmath203 is even . we order all nodes of a minimal tree @xmath170 in the degree negative lexicographic ordering and form the block matrix @xmath55 in the following way . all diagonal blocks of @xmath55 are @xmath142 . if a node with index @xmath7 is connected to a node with index @xmath20 with an @xmath26 or @xmath27 edge , then we put @xmath140 or @xmath145 in the block position @xmath207 , respectively . because of the ordering , the matrix @xmath208 is block lower triangular and nonsingular . its inverse @xmath209 is therefore also a lower triangular matrix with diagonal identity blocks . let @xmath210 be the number of nodes in @xmath170 . if follows from @xmath211 that the first block column of @xmath209 has the form @xmath212^t,\ ] ] where @xmath213 is the monomial in the @xmath203th node of @xmath170 for @xmath214 ( @xmath215 ) . now we will construct the linearization of the matrix polynomial @xmath63 . we need a block matrix @xmath216 , whose elements are linear pencils in @xmath26 and @xmath27 . we take @xmath217 and adjust the first block row @xmath218 , where we put linear pencils such that @xmath219^t)=p(x , y).\ ] ] this is always possible as for each term @xmath150 in the polynomial @xmath63 there exists a term @xmath220 in @xmath170 such that @xmath221 is one of the following three options : @xmath222 , @xmath223 , or @xmath224 . the product @xmath225 is an upper block triangular matrix of the form @xmath226,\ ] ] where @xmath227 are matrix polynomials . if we introduce the matrix polynomial @xmath228,\ ] ] then it follows that @xmath229\ ] ] and since @xmath230 , this proves that @xmath216 is indeed a linearization of the matrix polynomial @xmath63 . [ ex::cubic1 ] as an example we consider the scalar bivariate polynomial @xmath231 which was already linearized in @xcite with matrices of size @xmath232 ( we can also get a @xmath232 linearization if we insert the coefficients in matrix ( [ eq : ex2 ] ) of example [ ex : lin0 ] ) . now we can linearize it with matrices of size @xmath233 as @xmath234 , where @xmath235}.\ ] ] in the next section we will further reduce the size of the matrices to @xmath236 and @xmath237 . we will upgrade the approach from the previous section and produce even smaller representations for scalar polynomials . as before , representations have a form of the directed tree , but instead of using only @xmath26 and @xmath27 , an edge can now be any linear polynomial @xmath238 such that @xmath239 . these additional parameters give us enough freedom to produce smaller representations . the root is still @xmath164 while the other nodes are polynomials in @xmath26 and @xmath27 that are products of all edges on the path from the root to the node . in each node all monomials have the same degree , which is equal to the graph distance to the root . before we continue with the construction , we give a small example to clarify the idea . = [ ellipse , draw , fill = red!20 , font= , text width=4.2em , text centered , rounded corners , minimum width=4.1em , minimum height=1em , node distance=2 cm ] = [ rectangle , draw , fill = red!20 , font= , text width=1.6em , text centered , rounded corners , minimum width=1.6em , minimum height=1.4em , node distance=1.6 cm ] ] [ ex : lin2 ] a linearization of a polynomial of degree @xmath83 with matrices of size @xmath240 is presented in figure [ fig : ex4 ] . let us explain the figure and show how to produce the matrices from the representation tree . the nodes in the representation tree are the following polynomials : @xmath241 q_3(x , y)= ( x+y ) \ , q_2(x , y ) = x^2-y^2 , & & q_4(x , y)= ( 2x - y ) \ , q_1(x , y ) = 2x -y . \end{array}\ ] ] the polynomial of degree @xmath83 is then a linear combination of nodes in the representation tree and coefficients @xmath242 which are polynomials of degree @xmath164 contained in the ellipses . this gives @xmath243 similar as in section [ sec : lin1 ] , we can write the matrices by putting the linear coefficients in the first row and relations between the polynomials @xmath244 to @xmath245 in the subsequent rows . for each edge of the form @xmath246 we put @xmath247 in the position @xmath248 in the matrix @xmath249 and @xmath164 in the position @xmath250 . in the first row we put @xmath251 in the position @xmath252 if @xmath253 is the linear factor that multiplies the polynomial @xmath254 in the linearization . the matrix @xmath255 that corresponds to figure [ fig : ex4 ] , such that @xmath256 , is @xmath257.\ ] ] 1 . compute @xmath1 zeros @xmath258 of the polynomial @xmath259 . 2 . form a branch of the tree with the root @xmath260 and nodes @xmath261 , where @xmath262 is a successor of @xmath263 and the edge from @xmath263 to @xmath262 contains the factor @xmath264 for @xmath265 . 3 . compute linear coefficients @xmath266 for nodes @xmath267 in the following way : 1 . take @xmath268 , 2 . take @xmath269 , where @xmath270 is a coefficient of @xmath254 at @xmath271 , for @xmath272 , 3 . take @xmath273 . 4 . compute the remainder @xmath274 which has the form @xmath275 , where @xmath276 is a polynomial of degree @xmath277 . if @xmath278 then stop and return the tree . add node @xmath279 and connect it to the root by an edge having the factor @xmath27 . if @xmath276 is a nonzero constant @xmath280 , then use @xmath281 as a coefficient for the node @xmath279 , stop , and return the tree . recursively call the same algorithm to obtain a representation tree with the root @xmath282 for the polynomial @xmath276 . connect @xmath279 to @xmath282 by an edge with a factor @xmath27 and return the tree with the root @xmath283 . in example [ ex : lin2 ] we showed how to construct the bivariate pencil @xmath284 from a representation tree and the corresponding linear coefficients . the outline of an algorithm that constructs a representation tree and the corresponding linear coefficients for a given polynomial @xmath47 is presented in algorithm [ alg : lin2 ] . in the following discussion we give some missing details and show that the algorithm indeed gives a linearization . * the nodes @xmath261 that we construct in step 2 are polynomials of the form @xmath285 for @xmath286 . all monomials in @xmath263 have degree @xmath287 and the leading term is @xmath288 . * each product @xmath289 for @xmath286 is a polynomial with monomials of exact degree @xmath290 , while @xmath291 is a polynomial of degree @xmath164 . the linear factors @xmath292 in step 3 are constructed so that : * * leading two monomials ( @xmath293 and @xmath271 ) of @xmath294 agree with the part @xmath295 of the polynomial @xmath47 for @xmath272 , * * the product @xmath296 agrees with the part of @xmath47 composed of all monomials of degree exactly @xmath1 , * * the product @xmath297 agrees with the part of @xmath47 composed of all monomials of degree up to @xmath164 . + as a result , the remainder in step 4 has the form @xmath298 , where @xmath276 is a polynomial of degree @xmath277 . the situation at the end of step 4 is presented in figure [ fig : lin22 ] . the remainder @xmath299 is a polynomial of the form @xmath300 , where @xmath276 is a polynomial or degree @xmath277.[fig : lin22 ] ] * if the coefficient @xmath301 is zero , then we can apply a linear substitution of @xmath26 and @xmath27 of the form @xmath302 and @xmath303 , where we pick @xmath304 such that @xmath305 this ensures that the substituted polynomial in @xmath306 and @xmath307 will have a nonzero coefficient at @xmath308 . after we complete the representation tree for the substituted polynomial in @xmath306 and @xmath307 , we perform the substitution back to @xmath26 and @xmath27 . * if the polynomial @xmath276 in step 4 is not a constant , then we obtain a representation subtree for @xmath276 by calling recursively the same algorithm . in order to obtain the final representation tree , we then join the existing branch to the representation subtree for the polynomial @xmath276 . we do this by introducing a new node @xmath279 in step 6 that is linked to the root by the edge with the factor @xmath27 . to this new node we link the root @xmath282 of the subtree for the polynomial @xmath276 in step 9 , again using the edge with the factor @xmath27 . as @xmath282 is linked to the root by two edges @xmath27 , this multiplies all nodes in the subtree by @xmath186 and , since the subtree is a representation for @xmath276 , this gives a representation for the remainder @xmath309 from step 4 . the situation after step 9 with the final representation tree for the polynomial @xmath47 is presented in figure [ fig : lin23 ] . ] from the output of algorithm [ alg : lin2 ] , matrices @xmath310 such that @xmath74 can be obtained in the same way as in example [ ex : lin2 ] . let us remark that the zeros @xmath258 in step 1 can be complex , even if polynomial @xmath7 has real coefficients . thus , in a general case a linearization produced by algorithm [ alg : lin2 ] has complex matrices @xmath51 , and @xmath52 . [ ex::pol2 ] we apply algorithm [ alg : lin2 ] on @xmath311 from example [ ex::cubic1 ] . first , we compute the roots @xmath312 of the polynomial @xmath313 . the zeros are ordered so that @xmath314 . in exact computation the order is not important , but in numerical tests we experience better results with this order . this gives the polynomials in the first branch of the representation tree : @xmath315 and we can compute the corresponding coefficients @xmath316 for the remainder @xmath317 we need just one additional node @xmath318 with the coefficient @xmath319 . the determinantal representation with @xmath240 matrices is @xmath320 , where @xmath321},\quad b = { \footnotesize\left[\begin{array}{rrrr } 2 & 4 & 7 & 0\cr -1 & 0 & 0 & 0 \cr 0 & -1 & 0 & 0\cr 0 & 0 & 0 & 0\end{array}\right ] } , \quad \text{and } \\[1 mm ] c & = { \footnotesize \left[\begin{matrix } 3 & 4.9681 + 4.5036i & 7.8882 & 0.88972 + 5.5576i\cr -0.0079857 + 1.1259i & 0 & 0 & 0 \cr 0 & -0.0079857 - 1.1259i & 0 & 0\cr -1 & 0 & 0 & 0\end{matrix}\right]}.\end{aligned}\ ] ] = [ circle , draw , fill = blue!40 , inner sep=0pt , minimum size=5pt , node distance=0.35 cm ] \(00 ) ; ; \(00 ) ; ( 10 ) [ below left of=00 ] ; ; ( 00 ) to ( 10 ) ; \(00 ) ; ( 10 ) [ below left of=00 ] ; ( 20 ) [ below left of=10 ] ; ( 01 ) [ below right of=00 ] ; ; ( 00 ) to ( 01 ) ; ( 00 ) to ( 10 ) to ( 20 ) ; \(00 ) ; ( 10 ) [ below left of=00 ] ; ( 20 ) [ below left of=10 ] ; ( 30 ) [ below left of=20 ] ; ( 01 ) [ below right of=00 ] ; ( 02 ) [ below right of=01 ] ; ; ( 00 ) to ( 01 ) to ( 02 ) ; ( 00 ) to ( 10 ) to ( 20 ) to ( 30 ) ; \(00 ) ; ( 10 ) [ below left of=00 ] ; ( 20 ) [ below left of=10 ] ; ( 30 ) [ below left of=20 ] ; ( 40 ) [ below left of=30 ] ; ( 50 ) [ below left of=40 ] ; ( 60 ) [ below left of=50 ] ; ( 70 ) [ below left of=60 ] ; ( 01 ) [ below right of=00 ] ; ( 02 ) [ below right of=01 ] ; ( 12 ) [ below left of=02 ] ; ( 22 ) [ below left of=12 ] ; ( 32 ) [ below left of=22 ] ; ( 42 ) [ below left of=32 ] ; ( 03 ) [ below right of=02 ] ; ( 04 ) [ below right of=03 ] ; ( 15 ) [ below left of=04 ] ; ; ( 00 ) to ( 01 ) to ( 02 ) to ( 03 ) to ( 04 ) ; ( 00 ) to ( 10 ) to ( 20 ) to ( 30 ) to ( 40 ) to ( 50 ) to ( 60 ) to ( 70 ) ; ( 02 ) to ( 12 ) to ( 22 ) to ( 32 ) to ( 42 ) ; ( 04 ) to ( 15 ) ; \(00 ) ; ( 10 ) [ below left of=00 ] ; ( 20 ) [ below left of=10 ] ; ( 30 ) [ below left of=20 ] ; ( 40 ) [ below left of=30 ] ; ( 50 ) [ below left of=40 ] ; ( 60 ) [ below left of=50 ] ; ( 01 ) [ below right of=00 ] ; ( 02 ) [ below right of=01 ] ; ( 12 ) [ below left of=02 ] ; ( 22 ) [ below left of=12 ] ; ( 32 ) [ below left of=22 ] ; ( 03 ) [ below right of=02 ] ; ( 04 ) [ below right of=03 ] ; ; ( 00 ) to ( 01 ) to ( 02 ) to ( 03 ) to ( 04 ) ; ( 00 ) to ( 10 ) to ( 20 ) to ( 30 ) to ( 40 ) to ( 50 ) to ( 60 ) ; ( 02 ) to ( 12 ) to ( 22 ) to ( 32 ) ; \(00 ) ; ( 10 ) [ below left of=00 ] ; ( 20 ) [ below left of=10 ] ; ( 30 ) [ below left of=20 ] ; ( 40 ) [ below left of=30 ] ; ( 50 ) [ below left of=40 ] ; ( 01 ) [ below right of=00 ] ; ( 02 ) [ below right of=01 ] ; ( 12 ) [ below left of=02 ] ; ( 22 ) [ below left of=12 ] ; ( 03 ) [ below right of=02 ] ; ; ( 00 ) to ( 01 ) to ( 02 ) to ( 03 ) ; ( 00 ) to ( 10 ) to ( 20 ) to ( 30 ) to ( 40 ) to ( 50 ) ; ( 02 ) to ( 12 ) to ( 22 ) ; \(00 ) ; ( 10 ) [ below left of=00 ] ; ( 20 ) [ below left of=10 ] ; ( 30 ) [ below left of=20 ] ; ( 40 ) [ below left of=30 ] ; ( 01 ) [ below right of=00 ] ; ( 02 ) [ below right of=01 ] ; ( 12 ) [ below left of=02 ] ; ; ( 00 ) to ( 01 ) to ( 02 ) ; ( 00 ) to ( 10 ) to ( 20 ) to ( 30 ) to ( 40 ) ; ( 02 ) to ( 12 ) ; representation trees for polynomials of degree from 1 to 8 are presented in figure [ fig : small ] . if we compare them to the determinantal representations from section [ sec : lin1 ] in figure [ fig : lin1first8 ] , then we see that representations obtained by algorithm [ alg : lin2 ] are much smaller . the following lemma shows that asymptotically we use @xmath322 fewer nodes than in section [ sec : lin1 ] . algorithm [ alg : lin2 ] returns representation tree @xmath163 for the linearization of a polynomial @xmath47 of degree @xmath1 of size @xmath323 it follows from the recursion in the algorithm ( see figure [ fig : lin23 ] ) that the number of nodes satisfies the recurrence equation @xmath324 the solution of this equation with the initial values @xmath325 , @xmath326 , and @xmath327 is ( [ eq : sizeg2 ] ) . for generic polynomials of degrees @xmath328 and @xmath329 it turns out to be possible to modify the construction and save one node in the representation tree . the main idea is to apply a linear substitution of variables @xmath26 and @xmath27 in the preliminary phase to the polynomial @xmath47 to eliminate some of the terms . this implies that the resulting matrices are of order @xmath83 ( @xmath328 ) and @xmath161 ( @xmath329 ) , instead of order @xmath330 and @xmath190 as seen before and , it also reduces the size of the matrices for @xmath331 and @xmath332 by 1 . we give details in the following two subsections . let us consider a cubic bivariate polynomial @xmath333 , where we can assume that @xmath334 . we introduce a linear substitution of the form @xmath335 and @xmath336 , where @xmath165 is such that @xmath337 and @xmath338 . the substitution is well defined if @xmath165 is a single root of ( [ eq : sc31 ] ) and the only situation , where this is not possible , is when @xmath42 has a triple root . the above substitution transforms @xmath47 into a polynomial @xmath339 such that its coefficients @xmath340 and @xmath341 are both zero . if we apply algorithm [ alg : lin2 ] to @xmath339 and choose @xmath342 for the first zero , then the remainder in step 4 is zero and we get @xmath343 matrices @xmath344 , @xmath345 , and @xmath346 such that @xmath347 . now , it is easy to see that for @xmath348 , @xmath349 , and @xmath350 , @xmath82 . [ ex::pol3 ] we take the recurrent example @xmath311 ( see examples [ ex::cubic1 ] and [ ex::pol2 ] ) . if we take @xmath351 ( see ( [ eq : sc31 ] ) ) and @xmath352 , then substitution @xmath335 and @xmath336 changes @xmath47 into a polynomial @xmath353 algorithm [ alg : lin2 ] gives @xmath343 matrices @xmath344 , @xmath345 , and @xmath346 such that @xmath347 , from which matrices @xmath354,\quad b = { \footnotesize \left[\begin{matrix } 1.5317 & -2.4833 & 7\cr -1 & 0 & 0\cr 0 & -1 & 0\end{matrix}\right ] } , } \quad \text{and } \\[1 mm ] c & = { \footnotesize \left[\begin{matrix } 2.2189 & 2.8587 & 0.00559 + 7.8813i\cr -1.1269 & 0 & 0\cr 0 & -0.0079857 + 1.1259i & 0\end{matrix}\right]}\end{aligned}\ ] ] such that @xmath355 , are obtained and we have a @xmath343 linearization . before we give a construction for a generic quartic bivariate polynomial , let us consider a particular case , when a polynomial @xmath356 of degree 4 is such that @xmath357 . in this case @xmath161 nodes are enough to represent the polynomial @xmath47 . the representation tree for the polynomial @xmath47 is presented in figure [ fig : ex4new ] , where @xmath358 and @xmath359 are the zeros of @xmath360 . of degree 4 such that @xmath357.[fig : ex4new ] ] for a generic quartic polynomial we first transform it into one with zero coefficients at @xmath361 , @xmath362 , @xmath363 , and @xmath364 . except for very special polynomials , we can do this with a combination of two linear substitutions . similar as in case @xmath328 , we first introduce a linear substitution of the form @xmath335 and @xmath336 , where @xmath165 is such that @xmath365 and @xmath366 . the substitution is well defined if @xmath165 is a single root of @xmath367 . after the substitution we have a polynomial @xmath339 such that its coefficients @xmath368 and @xmath340 are both zero . on this polynomial we apply a new substitution @xmath369 and @xmath370 , where @xmath371 and @xmath372 . this substitution is well defined if @xmath45 is a single root of @xmath373 ; therefore , both substitutions exist for a generic polynomial of degree @xmath330 . after the second substitution we get a polynomial @xmath374 such that its coefficients @xmath375 , @xmath376 , @xmath377 , and @xmath378 are all zero . for such polynomial we can construct a representation with matrices @xmath240 as presented in figure [ fig : ex4new ] . this gives @xmath240 matrices @xmath379 , @xmath380 , and @xmath381 such that @xmath382 . finally , if we take @xmath383 then @xmath384 . if we add the constructions from subsections [ ss : case3 ] and [ ss : case4 ] as special cases to algorithm [ alg : lin2 ] , then we save one node for all generic polynomials of degree @xmath385 or @xmath386 . although this advantage seems to be modest , numerical results in the following section point out that for small @xmath1 this does speed up the computation of the zeros considerably ( for instance , for @xmath204 the corresponding @xmath387-matrices are of order @xmath388 instead of @xmath389 ) . determinantal representations from sections [ sec : lin1 ] and [ sec : lin2 ] can be used to numerically solve a system of two bivariate polynomials . we first linearize the problem as a two - parameter eigenvalue problem and then solve it with the method for singular two - parameter eigenvalue problems from @xcite . subsequently , we refine the solutions by two steps of newton s method . we refer to the numerical methods that use linearizations from sections [ sec : lin1 ] and [ sec : lin2 ] as lin1 and lin2 , respectively . in the first example we take polynomials with random coefficients , while the second example considers some challenging benchmark polynomials . [ ex : randtest]we compare lin1 and lin2 to nsolve in mathematica 9 and phclab 1.02 @xcite running phcpack 2.3.84 on systems of full bivariate polynomials of the same degree , whose coefficients are random real numbers uniformly distributed on @xmath390 $ ] or random complex numbers , such that real and imaginary part are both uniformly distributed on @xmath390 $ ] . the results are presented in table [ tbl : one ] . for each @xmath1 we run all methods on the same set of 20 random polynomial systems and measure the average time . lin1 and mathematica s nsolve work faster for polynomials with real coefficients while this does not make a change for lin2 and phclab , therefore , the results in the table for lin2 and phclab are an average of 20 real and 20 complex examples . clearly , if lin1 is applied to a polynomial with real coefficients , then matrices @xmath105 , @xmath391 , and @xmath110 are real . if we apply lin2 then the matrices are complex in general as roots of univariate polynomials are used in the construction . although the complex arithmetic is more expensive than the real one , complex eigenproblems from lin2 are so small that they are solved faster than the larger real problems from lin1 . r|cccccc|rr & & + @xmath1 & lin1 ( @xmath392 ) & lin1 ( @xmath393 ) & lin2 & phclab & nsolve ( @xmath393 ) & nsolve ( @xmath392 ) & lin1 & lin2 + ' '' '' 3 & 0.01 & 20.01 & @xmath394 & 0.18 & 0.25 & 0.04 & 25 & 9 + 4 & 0.02 & 20.02 & 0.01 & 0.21 & 0.42 & 0.07 & 64 & 25 + 5 & 0.03 & 20.05 & 0.02 & 0.26 & 0.67 & 0.17 & 121 & 64 + 6 & 0.08 & 20.18 & 0.05 & 0.34 & 1.04 & 0.22 & 225 & 100 + 7 & 0.23 & 20.57 & 0.16 & 0.44 & 2.75 & 0.61 & 361 & 169 + 8 & 0.67 & 22.04 & 0.54 & 0.59 & 2.17 & 0.88 & 576 & 289 + 9 & 2.25 & 26.21 & 1.33 & 0.80 & 5.53 & 1.48 & 841 & 400 + 10 & 6.24 & 16.62 & 3.38 & 1.05 & 8.12 & 3.85 & 1225 & 576 + computational times for lin1 , lin2 , and phclab are very similar for each of the 20 test problems of the same degree . on the other hand , nsolve needs substantially more time for certain problems . for example , for complex polynomials of degree @xmath395 , nsolve needed approximately @xmath396 for 16 of the 20 examples , and @xmath397 for the additional 4 examples . that explains why the average time for nsolve ( @xmath393 ) is larger in case @xmath395 than in case @xmath398 . beside the computational time , accuracy and reliability are another important factors . nsolve is the only method that finds all solutions in all examples , but on the other hand , the results are on average less accurate than with other methods . as a measure of accuracy we use the maximum value of @xmath399 where @xmath400 is a jacobian matrix of @xmath401 and @xmath402 at @xmath32 , over all computed zeros @xmath32 . @xmath403 is an absolute condition number of a zero @xmath32 and we assume that in random examples all zeros are simple . for a good method ( [ eq : condn ] ) should be as small as possible . for real or complex systems of degree @xmath404 , lin2 is the fastest method and usually the most accurate one . it is never significantly less accurate than the others , so it clearly wins in this case . for real polynomials of degree @xmath398 computational times of all methods are very close . nsolve is the fastest method with a tight margin in 17 out of 20 cases , but is also several orders of magnitude less accurate . phclab is faster and slightly less accurate than lin2 in 5 out of 40 cases , but , in one of them it fails to compute all the solutions . for @xmath405 and @xmath406 phclab becomes the fastest method , but is less reliable . in many cases it does not compute all the solutions . for @xmath405 this happens in 14 out of 40 times and for @xmath406 in @xmath407 out of @xmath408 cases . as phclab is using random initial systems , a possible remedy is to run phclab several times . also lin2 fails to compute solutions for @xmath409 real examples for @xmath406 . a remedy for lin2 in these cases is to interchange variables @xmath26 and @xmath27 . lin1 is competitive in particular for real systems of degree @xmath404 . for @xmath398 it misses one solution in one example and in two examples for @xmath406 we have to adapt the criteria for detecting a numerical rank in the staircase algorithm to get the correct number of solutions . let us remark that each node less in the representation tree really does make a difference . for instance , if we do not apply the special case for @xmath329 in subsection [ ss : case4 ] , then the @xmath387 matrices for lin2 for polynomial systems of degree @xmath406 are of size @xmath410 instead of @xmath411 and the average computational time rises from @xmath412 to @xmath413 . [ ex:25]we test lin1 and lin2 on 25 examples ex001 to ex025 from @xcite . this set contains challenging benchmark problems with polynomials of small degree from @xmath414 to @xmath415 that have many multiple zeros and usually have less solutions than a generic pair of the same degrees . lin1 and lin2 performed satisfactorily on most examples , but , they also failed on some . instead of giving the details for all 25 examples , we give the key observations . * multiple zeros can present a problem for the algorithm from @xcite that is used to solve the projected regular problem @xmath416 , @xmath417 that is obtained from ( [ drugi ] ) by the modified staircase algorithm from @xcite . the qz algorithm is first applied to @xmath416 and then @xmath417 is multiplied by @xmath418 and @xmath419 . the eigenvalues are clustered along the diagonal so that multiple eigenvalues @xmath26 should be in the same block . for several of the 25 examples with eigenvalues of high multiplicity the clustering criteria have to be adapted otherwise the results are not so accurate . * lin2 is faster , but the accuracy can be lost if the polynomial in step 1 of algorithm [ alg : lin2 ] has multiple zeros , an example is @xmath402 from ex008 . the method fails for ex014 , ex018 , and ex020 . * we get very good results in example ex005 with the system @xmath420 and @xmath421 using lin2 . in this case lin2 returns optimal determinantal representations with matrices of size @xmath422 and @xmath423 , respectively . the obtained two - parameter eigenvalue problem is not singular and we get the solutions in @xmath424 , while phclab and nsolve need @xmath425 . for comparison , lin1 , applied to the same problem , returns @xmath387 matrices of size @xmath426 , while lin2 gives @xmath387 matrices of size @xmath427 . * lin1 is slower but can be more accurate . because there is no computation in the construction , no errors are introduced in the construction of the linearization . lin1 manages to solve 22 out of 25 examples ( in some examples the parameters have to be adapted to make it work ) , but fails for ex007 , ex016 , and ex018 . * nsolve always finds all solutions but is slower than lin1 and lin2 except for ex014 where polynomials are of degrees @xmath428 and @xmath429 . phclab usually finds just one instance of a multiple eigenvalue and thus returns much less zeros . [ ex : tri ] encouraged by the good results for ex005 in example [ ex:25 ] , we carry out some experiments with polynomials of form @xmath430 , where @xmath431 is a polynomial of small degree @xmath432 . for such polynomials algorithm [ alg : lin2 ] returns matrices of size @xmath433 or even smaller . for example , it is easy to see that for @xmath434 we get linearization of the smallest possible size @xmath16 . we compared lin2 , phclab , and nsolve . computational times for random polynomials with complex coefficients of the above form are presented in table [ tbl : two ] . as @xmath1 increases , phclab becomes faster then lin2 , but in most cases it does not compute all solutions . for instance , for @xmath435 it misses 16 and 20 zeros for @xmath434 and @xmath436 , respectively . therefore , lin2 might be the preferred method for such polynomial systems . cc|lcc & + @xmath1 & @xmath437 & lin2 & phclab & nsolve + ' '' '' 15 & 1 & 10.48 & 11.9 & 13.7 + 15 & 3 & 10.87 & 11.8 & 14.1 + 20 & 1 & 12.1 & 14.5 & 10.8 + 20 & 3 & 13.8 & 15.0 & 11.3 + 25 & 1 & 19.6 & 10.6 & 25.0 + 25 & 3 & 13.1 & 12.7 & 26.3 + 30 & 1 & 23.2 & 17.4 & 52.8 + 30 & 3 & 37.2 & 20.1 & 55.0 + we have proposed two linearizations for bivariate polynomials . the first linearization does not involve any computation as the coefficients of the polynomials appear as ( block ) coefficients of the matrices @xmath80 , @xmath81 , and @xmath52 . this linearization is suitable for both scalar and matrix bivariate polynomials . the second linearization , useful for scalar polynomials , involves little computation and returns much smaller matrices . they are still larger than the theoretically smallest possible size @xmath16 , but their construction is very simple and fast . moreover , while the asymptotic order is @xmath438 , the order for small @xmath1 is about @xmath30 ; for polynomials of degree 3 and 4 we have presented determinantal representations of order 3 and 5 , respectively . as an application we have presented a method for finding roots of two bivariate polynomials . we show that an approach , where the polynomial system is first linearized into a two - parameter eigenvalue problem , which is later solved by a modified staircase method , is numerically feasible and gives good results for polynomials of degree @xmath29 , as well as for polynomials of higher degree but with few terms . any further results on even smaller determinantal representations that can be efficiently constructed numerically , could enlarge the above degree . * acknowledgment * the research was performed while the first author was visiting the casa group at tu eindhoven . the author wishes to thank the nwo for the visitor s grant and the casa group for the hospitality . , resultant - based methods for plane curves intersection problems , in : proceedings of the 8th international conference on computer algebra in scientific computing , vol . 3718 of lecture notes in computer science , springer - verlag ( 2005 ) 7592 . , phclab : a matlab / octave interface to phcpack . in : m. stillman , j. verschelde , and n. takayama ( eds ) , _ software for algebraic geometry _ , volume 148 of the i m a volumes in mathematics and its applications , springer new york ( 2008 ) 1532 . , a unified approach to computing real and complex zeros of zero - dimensional ideals , in : m. putinar , s. sullivant ( eds . ) , _ emerging applications of algebraic geometry _ , the i m a volumes in mathematics and its applications 149 , springer - verlag , new york ( 2009 ) 125155 . , lmi representations of convex semialgebraic sets and determinantal representations of algebraic hypersurfaces : past , present , and future , in h. dym , m. c. de oliveira , m. putinar ( eds . ) , _ mathematical methods in systems , optimization , and control : festschrift in honor of j. william helton . _ operator theory : advances and applications 222 , birkhuser ( 2012 ) 325349 .

phylogenetics is a study of how species are related to each other . evolutionary relationships are most conveniently represented by rooted leaf - labelled trees , where the leaves represent extant species and the root represents their most recent common ancestral species . similarly other internal vertices of evolutionary trees correspond to extinct ancestral species . the arrival of dna and protein sequence data in the last forty years led to explosive growth of phylogenetics . many of the modern phylogenetic methods consider sequence data under probabilistic models of sequence evolution . for such models to be useful for phylogenetic inference , it is important to establish their _ identifiability _ ( i.e. , to show that nonisomorphic trees or different model parameters can not induce the same distribution on the sequences at the leaves under a given model of sequence evolution ) . mathematical theory of phylogenetic trees , especially probabilistic models of sequence evolution , the associated questions of identifiability and statistical consistency ( especially of the maximum likelihood methods ) have been extensively studied @xcite , giving a firm statistical foundation to the study of phylogenetic trees . while phylogenetic trees represent relationships between species , population pedigrees represent how individuals within a population are related to each other . communities all over the world have long been curious about knowing their ancestral histories , and have often kept detailed records of their family trees . in fact this curiosity goes back much further in the past than the interest in constructing evolutionary relationships between species . an example of a fairly detailed record of family histories is the icelandic database slendingab ( the book of icelanders http://www.islendingabok.is ) of genealogical records that covers almost the whole icelandic population and goes back to nearly 1200 years . such ancestral histories are often compiled from a variety of sources such as church records , birth and death records , obituaries etc . that are prone to ambiguities or missing data beyond a few generations in the past . in the last several years large amounts of data on intra - population genetic variation have been recorded . for example , the icelandic biomedical company decode genetics has compiled genomic , genealogical and health data of more than 100000 individuals ( which is a significant proportion of the current icelandic population ) . such data offer promising opportunity to cross check and resolve ambiguities in historic genealogical records besides being useful for other studies such as , for example , genetic factors associated with medical conditions . therefore , there is a renewed interest in accurately inferring pedigrees from genomic data . the statistical and combinatorial foundation for studying reconstruction problems for pedigrees has not been as developed as in phylogenetics . a purpose of this paper is to continue our earlier attempts to develop such a foundation for the problem of reconstructing pedigrees from observations ( sequences ) on extant individuals . to develop such a foundation , we need to establish results along the following lines : developing a biologically realistic model for sequences undergoing mutations and recombinations , and identifiability results for such a model ; statistical consistency results ; and finally results that give estimates for the amount of genomic data necessary to reliably construct pedigrees . this paper is mainly about identifiability questions . in the rest of this section , we discuss the above theoretical motivation in more detail . we begin by informally sketching some well known reconstruction and identifiability results for phylogenetic trees that not only motivate the work in this paper but also are crucially useful in the proof of the main identifiability result of this paper . _ results of zarecki and buneman . _ it was shown in @xcite that a leaf - labelled tree can be uniquely constructed from the pairwise distances between its leaves . the result was strengthened slightly as follows @xcite . suppose that @xmath0 is an _ additive function _ on the family of subsets of cardinality 2 of the vertex set of a leaf - labelled tree . here _ additive _ means that for any two vertices @xmath1 and @xmath2 , we have @xmath3 , where the summation is over all edges @xmath4 on the ( unique ) path from @xmath1 to @xmath2 . buneman showed that knowing @xmath0 on all pairs of leaves of a leaf - labelled tree without vertices of degree 2 is sufficient to uniquely construct the tree and the function . it is not surprising that these results are quite useful in phylogenetics , where observations on extant species ( leaves of an evolutionary tree ) are used to infer a suitably defined distance ( or an additive function ) between pairs of species , and then their phylogenetic tree is constructed uniquely . _ results of steel and chang . _ now suppose the evolutionary process on a ( rooted ) tree is modelled as follows : first the root is assigned a random state from a finite alphabet @xmath5 ( e.g. , @xmath5 may be \{a , t , g , c } in the case of dna sequences ) . each state is assumed to have a nonzero probability of being assigned to the root . each edge of the tree has associated with it a @xmath6 matrix of substitution probabilities . these substitution probabilities determine how the vertex - states evolve away from the root , and induce a distribution of states at the leaves of the tree . the model was formulated in @xcite as described above , while a slightly more general formulation in terms of markov random fields on unrooted trees was given in @xcite . it was independently shown by chang and steel that when the matrices defining the substitution probabilities satisfy certain mild conditions , the ( unrooted ) tree can be uniquely recovered from the joint distribution of states at the leaves of the tree . in particular they showed that the negative logarithm of the determinant of the matrix of substitution probabilities between pairs of vertices is an additive function on the pairs of vertices , and can be computed from the probability distribution on extant sequences . it was further proved in @xcite that the substitution matrices are also identifiable from the marginal distributions on triples of leaves of the tree . special cases of these results for models more commonly used in phylogenetics were known in the phylogenetics literature earlier . how far can we generalise such results if the underlying structure is more general than a tree ? a recent result in this direction is due to @xcite where it is shown that under some mild non - degeneracy conditions the dependency structure of a markov random field can be obtained from sufficiently many independent samples . in this paper we present simple models for recombinations and mutations for population pedigrees , and generalise phylogenetic identifiability results for them . one difference between reconstructing markov random fields and reconstructing pedigrees is that for pedigrees we have observations only on the extant individuals ( e.g. , dna sequences derived from living individuals ) . moreover , in the problem of reconstructing pedigrees , the samples of data ( e.g. , columns in a sequence alignment ) are not _ i.i.d . _ ( independent and identically distributed ) as a result of recombinations . in @xcite , we studied some purely combinatorial reconstruction problems motivated by zarecki s result , for example , the problem of reconstructing a pedigree from the pairwise distances between its extant individuals or from its subpedigrees ( pedigrees of subsets of the extant population ) . in @xcite , we showed that a pedigree can not in general be reconstructed from the collection of its proper subpedigrees . such a result implies that knowing pairwise distances between extant vertices is in general not enough to reconstruct pedigrees . in @xcite , we considered models for sequences evolving on pedigrees , and showed that for certain simple markov models , pedigrees are not identifiable from the distribution of observed states at extant vertices . we did construct examples of processes for which pedigrees could be proved to be identifiable , but the processes lacked the markovian property , which informally states that the state observed at a vertex depends only on the states of its parents . moreover , it seems that pedigrees that are difficult to reconstruct in a purely combinatorial framework ( e.g. , from pairwise distances between extant vertices or from subpedigrees ) are also likely to be difficult to reconstruct in a stochastic framework . for example , if a pedigree can not be reconstructed from its proper subpedigrees , then the marginal distributions of extant sequences on proper subsets of the extant population might be insufficient to uniquely recover the pedigree . on the other hand , negative results in a combinatorial setting may not imply non - identifiability in a stochastic framework . it is therefore important to study combinatorial reconstruction problems ( e.g. , classification of pedigrees that may be difficult to reconstruct combinatorially ) , stochastic identifiability problems for idealised markov models of recombination and mutation , and relationships between these problems . reconstruction problems of purely combinatorial nature are well known to combinatorialists , the foremost among such problems being the vertex reconstruction conjecture @xcite . the conjecture states that all simple undirected unlabelled graphs can be constructed from their collection of unlabelled induced subgraphs . combinatorial reconstruction problems have also been studied in phylogenetics , for example , problems of reconstructing phylogenetic trees from subtrees @xcite . steel and chang proved their phylogenetic identifiability results in two parts : computation of the additive ` log determinant ' function on pairs of leaves from the joint probability distribution on leaf states , and the combinatorial problem of reconstructing a tree from the additive function , which had been solved by buneman and zarecki . similarly the problem of reconstructing pedigrees under a recombination - mutation model may be solved in two parts : in the first part , we would like to reduce the identifiability question to an appropriate combinatorial reconstruction problem , and then in the second part we would like to show that the combinatorial reconstruction problem has a unique solution . to ensure the uniqueness of reconstruction , we will have to compute sufficiently strong combinatorial invariants of the pedigree from the joint distribution of extant sequences . although we noted that distances between extant vertices in a pedigree are not sufficient to reconstruct a pedigree , we sketch a heuristic argument given in @xcite that shows how the distances between extant vertices in a discrete generation pedigree may be obtained from sequence data . suppose @xmath7 and @xmath8 are two extant individuals in a pedigree . suppose further that in the pedigree there are @xmath9 pairs of paths such that one of the paths in each pair ends on @xmath7 and the other ends on @xmath8 , and the two paths in each pair start at a common ancestor of @xmath7 and @xmath8 in the @xmath10-th generation , and the two paths in a pair do not share any other vertex . now if sufficiently many short recombination - free homologous segments of dna of @xmath7 and @xmath8 are compared , then we would expect about @xmath11 of them to have a common ancestor in the @xmath10-th generation . thus it may be possible to estimate @xmath12 . such calculations are theoretically possible for small values of @xmath10 assuming the population is large and the sequences are long . they then tried to use the numbers @xmath12 for all pairs @xmath13 for all @xmath10 to construct the pedigree . they computationally found many pairs of non - isomorphic pedigrees that have the same number of pairs of paths of each length . one such example is the pair of pedigrees shown in figure [ fig - hap ] , which was also mentioned in @xcite . but a more detailed analysis of sequence similarities ( between multiple sequences , if required ) under simple recombination and mutation models may give us more information than just pairwise distances between living individuals . in the main theorem of this paper ( theorem [ thm - paths ] ) , we show that the joint distribution on extant sequences determines a class of combinatorial invariants ( e.g. , certain types of subgraph sequences ) that supersedes pairwise distances between extant sequences and subtrees ( genealogical trees ) in a pedigree . we then show that pedigrees , such as the ones in figure [ fig - hap ] , that earlier seemed difficult to distinguish due to their combinatorial similarities ( including the non - reconstructible pedigrees constructed in @xcite ) are distinguished by the class of invariants . this paper is organised as follows . in section [ sec - prelim ] , we define pedigrees , alignments and subgraph sequences . in section [ sec - models ] , we give a schematic description of the recombination process , and formalise three models for sequences : model r ( a model in which there are recombinations but no mutations ) , model rm ( a model in which there are recombinations and mutations ) , and model m ( a model of mutations for sequences evolving on trees ) . we then formulate identifiability problems for these models . in section [ sec - model - r ] , we analyse pedigrees with two generations under model r. in section [ sec - model - rm ] , we prove the main theorem ( theorem [ thm - paths ] ) and demonstrate its applications . in the last section , we discuss a few open questions . a section on nomenclature follows the references where all symbols appearing in the paper and the number of the page on which they appear first are listed . we use the following notation for number systems and their subsets : @xmath14 for the set of integers , @xmath15 for the set of positive integers , @xmath16 for natural numbers , @xmath17 $ ] for the set @xmath18 . depending on the context , we write @xmath19 $ ] for the set of integers @xmath20 or real numbers @xmath21 ( and similarly @xmath22 , @xmath23 and @xmath24 $ ] for open or half - open intervals in integers and reals ) . the set of all @xmath10-tuples of elements of a set @xmath25 is written as @xmath26\}$ ] . the set of all functions from @xmath27 to @xmath25 is written as @xmath28 . next we introduce some graph theoretic notation . the vertex set and the edge set of a graph @xmath29 are denoted by @xmath30 and @xmath31 , respectively , and their cardinalities by @xmath32 and @xmath33 , respectively . the in - degree and the out - degree of a vertex @xmath34 in a directed graph are denoted by @xmath35 and @xmath36 , respectively . the degree of a vertex @xmath34 in an undirected graph ( or the total degree in a directed graph ) is denoted by @xmath37 . an arc from @xmath34 to @xmath38 in a directed graph and also an edge between @xmath34 and @xmath38 in an undirected graph is written as @xmath39 , and it will be understood from the context whether @xmath39 is meant to be a ( directed ) arc or an ( undirected ) edge . when any two objects @xmath40 and @xmath41 are isomorphic , we write @xmath42 . the isomorphism class of an object @xmath29 is written as @xmath43 . for a collection @xmath44 of labelled objects , we write @xmath45 for the set of isomorphism classes of objects in @xmath44 . let @xmath29 and @xmath46 be two directed or undirected graphs . we write @xmath47 ( or @xmath48 ) if @xmath29 is isomorphic to a subgraph of @xmath46 , and this notation may be used when @xmath29 or @xmath46 is unlabelled ( i.e. , they are just isomorphism classes ) . we write @xmath49 ( or @xmath50 ) when a labelled graph @xmath29 is a subgraph ( or a supergraph ) of a labelled graph @xmath46 . [ defn - general ] a _ general pedigree _ @xmath51 of a set @xmath27 is a directed acyclic graph on a vertex set @xmath52 and a set of arcs @xmath53 such that each vertex has in - degree 0 or 2 . the set @xmath27 is the set of vertices with out - degree 0 . the set @xmath54 is the set of vertices with in - degree 0 . the vertices in @xmath27 are called the _ extant vertices _ ( or the extant individuals in the population ) . the vertices in @xmath54 are called the _ founder vertices _ ( or the founders of the population ) . the _ order _ of the pedigree is @xmath55 . the _ depth _ of a pedigree is the length of ( i.e. , the number of arcs in ) a longest path in the pedigree . two pedigrees @xmath51 and @xmath56 are said to be _ isomorphic _ if there is a one - one map @xmath57 such that @xmath39 is an arc in @xmath58 if and only if @xmath59 is an arc in @xmath60 , and @xmath61 for all @xmath62 . we denote the natural partial order on @xmath63 by @xmath64 , i.e. , @xmath65 if there is a directed path from @xmath34 to @xmath38 ( or @xmath66 ) . we define isomorphism only between pedigrees of the same set of extant individuals and require that it fixes all extant vertices because ( informally speaking ) we would like to treat extant vertices to be labelled and other vertices to be unlabelled . throughout this paper , we will assume that all pedigrees have @xmath27 as their set of extant vertices . [ defn - diploid ] let @xmath51 be a pedigree . suppose that @xmath63 can be partitioned into unordered pairs of vertices such that the following conditions hold : the extant vertices are paired with extant vertices and the founder vertices are paired with founder vertices ; two non - extant vertices @xmath38 and @xmath67 are paired if and only if there are arcs @xmath68 and @xmath69 in @xmath53 ; no two paired vertices have a common parent . then @xmath58 together with one such pairing is called a _ diploid pedigree_. thus any general pedigree is a _ haploid pedigree _ ; a diploid pedigree is a haploid pedigree with a pairing of its vertices ( although not all pedigrees admit such a pairing ) . the pairing in a diploid pedigree is completely determined by the pairing of its extant vertices since all other pairs are determined by the definition . an advantage of a purely combinatorial definition of a diploid pedigree is that we can now assign just one sequence to each vertex . from this point of view , haploid pedigrees are pedigrees of sequences , not of individuals . pedigrees of individuals are diploid pedigrees obtained by pairing sequences in a haploid pedigree . figure [ fig - d2h ] illustrates this point of view . when vertices ( sequences ) @xmath70 and @xmath71 are paired , @xmath72 and @xmath73 must be paired , since they are parents of @xmath70 ; similarly , @xmath74 and @xmath75 must be paired . [ defn - char ] let @xmath76 be a finite set ( called an _ alphabet _ ) and @xmath63 a finite set . a _ character _ on @xmath63 is a map @xmath77 . for @xmath78 , an _ alignment _ of length @xmath79 on @xmath63 is a map @xmath80 . equivalently , an alignment is an @xmath79-tuple @xmath81 of characters on @xmath63 , or a two dimensional array of symbols from @xmath5 , with @xmath82 rows and @xmath79 columns . the rows of the array are called _ sequences _ , and are written as @xmath83 , @xmath84 to @xmath82 . individual entries in the array are written as @xmath85 , where @xmath84 to @xmath82 and @xmath86 to @xmath79 . the columns of an alignment are called _ usually @xmath63 will be the set of vertices of a pedigree , and we will be interested in alignments restricted to the set @xmath27 of extant vertices of the pedigree . the space of characters on @xmath27 is @xmath87 , which is also referred to as the space of _ site patterns_. the set of alignments of length @xmath79 on a set @xmath27 is @xmath88 , which we write as @xmath89 . let @xmath51 be a pedigree and let @xmath90 be an alignment on @xmath58 . if the sequences in @xmath91 have evolved under some process of recombination or mutation , then ( regardless of the details of the model of recombination or mutation ) we may suppose that each site in each sequence is inherited only from one of the two parent sequences . therefore , for each @xmath92 and each @xmath93 $ ] , there is a unique directed path @xmath94 from some founder vertex @xmath95 to @xmath34 that defines the genetic ancestry of @xmath96 . therefore , each site @xmath97 has associated with it a spanning forest @xmath98 , and each alignment of length @xmath79 has an underlying ( usually unknown ) sequence @xmath99 of spanning forests . similarly , each site @xmath97 has associated with it a _ directed subforest _ defined by @xmath100 , and we have a _ directed subforest sequence _ @xmath101 of the alignment . here we define these notions purely graph theoretically ( without reference to alignments or models ) . [ defn - xtree ] let @xmath27 be a finite set . a _ directed @xmath27-forest _ @xmath102 is a directed forest that satisfies the following conditions : the set of vertices with out - degree 0 is @xmath27 , and is called the _ leaf set _ of @xmath102 ; each component has a single vertex of in - degree 0 , called the _ root vertex _ of the component , and all other vertices have in - degree 1 ; all arcs are directed away from the root vertices . undirected @xmath27-forest _ is an unrooted forest with leaf set @xmath27 . suppose @xmath102 is a directed @xmath27-forest . it induces a natural partition of @xmath27 into _ maximal clusters _ @xmath103 such that vertices in cluster @xmath104 have a unique most recent common ancestor ( mrca ) @xmath105 in @xmath102 . we construct a subgraph of @xmath102 induced by @xmath106 $ ] and their descendants in @xmath102 , and then replace its ( directed ) arcs by ( undirected ) edges . the resulting undirected unrooted graph is called the _ undirected @xmath27-forest _ of @xmath102 , written as @xmath107 . it is the maximal undirected @xmath27-forest contained in @xmath102 . in the above definition , the term _ @xmath27-forest _ is meant to be analogous to the term _ @xmath27-tree _ that is commonly used in phylogenetics @xcite . in this paper we will adapt some of the phylogenetic identifiability results for undirected @xmath27-forests that appear as undirected graphs underlying subgraphs of pedigrees . therefore , in the following definition , we specialise the terms for pedigrees . [ defn - subtree ] let @xmath58 be a pedigree . a _ spanning forest _ of @xmath58 is a spanning subgraph @xmath29 of @xmath58 such that the in - degree of each vertex in @xmath29 is 1 unless it is a founder vertex in @xmath58 . a _ directed @xmath27-forest _ of @xmath58 is a subgraph @xmath102 of @xmath58 such that @xmath102 is a directed @xmath27-forest and the root vertex of each component of @xmath102 is a founder vertex in @xmath58 . each spanning forest @xmath29 in @xmath58 contains a unique directed @xmath27-forest of @xmath58 , and we denote it by @xmath108 . an _ undirected @xmath27-forest _ in @xmath58 is the unique undirected @xmath27-forest in any directed @xmath27-forest in @xmath58 . each spanning forest @xmath29 in @xmath58 contains a unique undirected @xmath27-forest of @xmath58 , and we denote it by @xmath109 . note that we use the term _ spanning forest _ in a specific sense : a spanning forest is not any spanning forest in the graph theoretic sense . we illustrate these terms in figure [ fig - subtrees ] , which shows a pedigree and a spanning forest @xmath29 with @xmath110 ( shown by bold arcs ) . in this example , the unique directed @xmath27-forest @xmath108 in @xmath29 has the arc set @xmath111 ; its clusters are @xmath112 and @xmath113 , and the root vertices of its components are @xmath114 and @xmath4 . the unique undirected @xmath27-forest @xmath109 in @xmath29 consists of vertex set @xmath115 and edge set @xmath116 . note that vertex 3 is isolated in @xmath109 since it is the mrca of its cluster , but we need it in our analysis . two ( directed or undirected ) @xmath27-forests @xmath102 and @xmath117 are said to be _ isomorphic _ ( written @xmath118 ) if there is a graph theoretic isomorphism @xmath119 from @xmath102 to @xmath117 such that @xmath61 for all @xmath120 in @xmath27 . the isomorphism class of a directed or an undirected @xmath27-forest @xmath102 , denoted by @xmath121 , is the set of all @xmath27-forests @xmath117 that are isomorphic to @xmath102 . the set of all directed @xmath27-forests of @xmath58 is denoted by @xmath122 . the set of isomorphism classes of directed @xmath27-forests of @xmath58 ( or the set of _ distinct _ directed @xmath27-forests in @xmath58 ) is denoted by @xmath123 . the set of spanning forests of a pedigree @xmath58 is denoted by @xmath124 . the set of undirected @xmath27-forests of @xmath58 is denoted by @xmath125 . the set of isomorphism classes of undirected @xmath27-forests of @xmath58 ( or the set of _ distinct _ undirected @xmath27-forests in @xmath58 ) is denoted by @xmath126 . [ prop - spanning - forests ] each spanning forest of a pedigree @xmath58 has @xmath127 arcs and contains exactly one directed @xmath27-forest . there are @xmath128 spanning forests . each directed @xmath27-forest @xmath102 is contained in @xmath129 spanning forests . each spanning forest of @xmath58 is obtained by selecting one of the two arcs that point to each non - founder vertex . the unique directed @xmath27-forest @xmath102 in a spanning forest @xmath29 is the subgraph of @xmath29 spanned by vertices in @xmath29 that are ancestral in @xmath29 to the extant vertices . finally , @xmath130 is the number of arcs not pointing to vertices in @xmath102 , therefore , @xmath131 is the number of non - founder vertices outside @xmath102 , at each of which we can choose one of the two incoming arcs to construct a spanning forest containing @xmath102 . in any model of recombination , the number of recombination events is determined by the spanning forest sequence underlying an alignment , but we define it for all spanning forest sequences without reference to alignments or models . [ def - rg - sg ] let @xmath58 be a pedigree and let @xmath132 be a spanning forest sequence in @xmath58 . if @xmath68 and @xmath69 are distinct arcs in @xmath58 such that @xmath68 is in @xmath133 and @xmath69 is in @xmath134 , then we say that a recombination has occurred at site @xmath135 at vertex @xmath34 or that @xmath135 is a recombining site . if @xmath68 is an arc in @xmath133 and @xmath134 then we say that there is no recombination at @xmath34 at site @xmath135 . we define the number of recombinations in @xmath136 to be @xmath137 where @xmath138 is the number of recombinations separating @xmath133 and @xmath134 . the number of points of no recombination is @xmath139 the directed @xmath27-forest sequence of @xmath136 is the sequence @xmath140 , and the undirected @xmath27-forest sequence of @xmath136 is the sequence @xmath141 . we assume that in any reasonable model of sequence evolution , sequences are first assigned to the founder vertices , and then subsequent generations of individuals inherit their sequences from their parents sequences subject to recombinations and mutations . we are then interested in the following types of _ identifiability _ questions . [ problem - main ] suppose sequences of equal length over a finite alphabet are assigned to the founder vertices of a pedigree . the sequences then evolve on the pedigree undergoing recombinations and mutations , giving a probability distribution on the space of alignments on the set of extant vertices . can we determine the pedigree uniquely ( i.e. , up to isomorphism ) - with or without the knowledge of the size or the depth of the pedigree or the various probability parameters defining the recombination and mutation processes , with or without restrictions such as discrete generations or constant population , and so on ? in the case of diploid pedigrees , we will be given the distribution on alignments on the set of extant vertices along with a pairing of extant vertices . in this paper , we study the above types of questions under two simple models of recombination and mutation . in _ , we assume that sequences evolve on a pedigrees under a process of recombinations without mutations . in _ model rm _ , we assume that sequences evolving on a pedigree undergo recombinations and mutations . for convenience , we also formalise the mutation part of model rm for the spanning forests of pedigrees and in general for directed @xmath27-forests , and call it _ model m_. in all these models , we assume that first all the founder vertices of a pedigree ( or the root vertices of a spanning forest or a directed @xmath27-forest ) are assigned sequences . these sequences are independently selected from a uniform distribution on @xmath142 , where @xmath76 is a known finite alphabet . then the sequences evolve on the pedigree ( or a spanning forest or a directed @xmath27-forest ) in a top - down manner , i.e. , a vertex is assigned a sequence only after its parents have been assigned sequences . we begin with a schematic description of the recombination process . our description is largely based on chapter 12 of @xcite . figure [ fig - meio ] schematically shows the process of gamete ( sperm or egg ) formation in eukaryotes . initially there is a parent cell with one pair of homologous non - sex chromosomes . each chromosome is then duplicated with the identical sister _ chromatids _ joined together at the _ centromere _ , forming a _ four - strand bundle_. then the two duplicated chromosomes exchange material between _ chiasmata _ ( _ recombination points _ ) . in the diagram , there are three recombination events . the first recombination is between strands 1 and 3 ( counted from top to bottom ) , the second is between strands 2 and 4 , and the third is between strands 2 and 3 . the four chromatids after the exchange of material are shown next . then the cell undergoes two cell divisions to create four haploid gametes , each receiving one of the four chromosomes . as shown in the diagram , at each recombination point a crossover occurs between one strand from the first pair and one strand from the second pair . at each crossover , a strand from the first pair and a strand from the second pair are chosen randomly with equal probability ( independent of other chiasma ) . this independence property is known as the lack of _ chromatid interference_. suppose that the locations of recombinations are modelled as a poisson point process along the sequence ( or on @xmath143 ) with the rate @xmath144 or a bernoulli process with probability @xmath145 ( so that a crossover occurs after a site on a sequence with probability @xmath145 independently of other sites or sequences ) . since exactly two of the four gametes - one from the first pair and one from the second pair - inherit any recombination , any given gamete inherits a recombination with probability @xmath146 . therefore , for the sequence of any fixed gamete , the locations of recombinations are still modelled by a poisson process , but with the rate @xmath147 , ( or bernoulli process with probability @xmath148 ) . therefore , a model may be formalised with just two parent sequences instead of four . a poisson process for the locations of chiasmata was first proposed in @xcite . based on the above description , we formalise models r and rm , in which we assume that crossovers in a finite sequence occur according to a bernoulli process . _ model r : _ consider three sequences @xmath83 of length @xmath79 over alphabet @xmath5 , where @xmath149 and @xmath38 and @xmath67 are parents of @xmath34 . the sequence @xmath150 is obtained by recombining sequences @xmath151 and @xmath152 as follows . let @xmath153 be a markov chain on the state space @xmath154 with transition probabilities @xmath155 if @xmath156 for @xmath157 , and @xmath158 . then , for @xmath159 , @xmath160 if @xmath161 . thus @xmath162 indicates a crossover from one sequence to the other . we refer to this model as _ model r_. _ model rm : _ consider three sequences @xmath83 of length @xmath79 over alphabet @xmath5 , where @xmath149 and @xmath38 and @xmath67 are parents of @xmath34 . the sequence @xmath150 is obtained from the sequences @xmath151 and @xmath152 by a process of recombinations and mutations as follows . let @xmath153 be a markov chain on the state space @xmath154 with transition probabilities @xmath155 if @xmath156 for @xmath157 , and @xmath163 . then if @xmath164 and @xmath165 , then @xmath166 is assigned @xmath1 with probability @xmath167 , and @xmath166 is assigned a state different from @xmath1 with probability @xmath168 . when a state different from @xmath1 is assigned to @xmath166 , each state in @xmath169 has equal probability @xmath170 of being assigned to @xmath166 . we refer to this process as _ model rm_. _ model m : _ this process is defined for spanning forests of a pedigree and directed @xmath27-forests . first each founder or the root vertex in each component of a directed @xmath27-forest is assigned independently and uniformly randomly a state from @xmath5 . suppose @xmath97 is the parent vertex of @xmath135 in a spanning forest of a pedigree or in a directed @xmath27-forest . let @xmath83 and @xmath171 be the sequences of @xmath135 and @xmath97 , respectively , both of equal length @xmath79 over alphabet @xmath5 . then for each @xmath172 $ ] , @xmath173 is assigned the same state as @xmath174 with probability @xmath167 , and @xmath173 is assigned a state different from @xmath174 with probability @xmath168 . when a state different from @xmath174 is assigned to @xmath173 , each state in @xmath175 has equal probability @xmath170 of being assigned to @xmath173 . we refer to this process as _ model m_. model m on a directed @xmath27-forest @xmath102 is equivalent to a similarly formulated model on the undirected @xmath27-forest of @xmath102 . we root each component of the undirected @xmath27-forest arbitrarily , and assign to it a state from @xmath5 uniformly randomly , independent of the roots of other components . the state then evolves away from the root in each component . if a component itself is an isolated vertex , it is simply assigned a state from @xmath5 uniformly randomly . since the mutation model described here is reversible , the same distribution on the site patterns is observed on @xmath27 in the undirected @xmath27-forest as in a directed @xmath27-forest for a given @xmath170 . therefore , when we try to construct a tree from the character distribution on its leaves , we can not construct the directed @xmath27-forest , but we can at best construct the undirected @xmath27-forest in it . therefore , we will consider model m only on undirected @xmath27-forests . thus model rm may be thought of as a synthesis of models r and m so that the recombination - free segments of sequences evolving on a pedigree may be examined under model m with phylogenetic methods . let @xmath58 be a pedigree on @xmath27 . for an alignment @xmath176 , we denote by @xmath177 the probability that sequences of length @xmath79 evolving on the pedigree @xmath58 under model @xmath178 give an alignment @xmath91 on @xmath27 . we use analogous notation when the model @xmath178 is replaced by the model @xmath179 , or when the pedigree @xmath58 is replaced by a directed or an undirected @xmath27-forest @xmath102 and the model under consideration is the mutation model @xmath180 . we denote the various probability spaces by @xmath181 , @xmath182 , @xmath183 , @xmath184 , and so on . for pedigrees @xmath58 and @xmath60 , we will write @xmath185 when @xmath186 for all @xmath187 , and analogously for other models . as in the case of alignments , we treat the spaces of spanning forest sequences , directed @xmath27-forest sequences , and undirected @xmath27-forest sequences as probability spaces , and denote them by @xmath188 , @xmath189 , and @xmath190 , respectively . the spanning forest sequences , and directed and undirected @xmath27-forest sequences , ( and the corresponding sequences of isomorphism classes of spanning forests and directed and undirected @xmath27-forests ) are defined by only the recombination events , therefore , the probability spaces are unchanged if @xmath179 is replaced by @xmath178 . for a spanning forest sequence @xmath191 , we will write @xmath192 for the probability of @xmath193 in the probability space @xmath194 . we will use analogous notation for other sequences and probability spaces . unless stated otherwise , when we will refer to alignments or sequences of spanning forests or other objects , we will mean alignments or sequences of spanning forests or other objects , respectively , from the appropriate probability spaces that are clear in the context . [ def - identifiability ] nonisomorphic pedigrees @xmath58 and @xmath60 in a class @xmath195 are said to be distinguished from each other under model @xmath178 if @xmath196 for some @xmath79 , ( i.e. , for some @xmath79 , there exists @xmath197 such that @xmath198 ) . a pedigree @xmath58 in a class @xmath195 is said to be identifiable under model @xmath178 if it is distinguished from every other pedigree @xmath60 in @xmath195 , ( i.e. , if there is a pedigree @xmath60 in @xmath195 such that @xmath199 for all @xmath200 , then @xmath60 is isomorphic to @xmath58 ) . pedigrees in a class @xmath195 are said to be identifiable under model @xmath178 if all pairs of pedigrees in @xmath195 are distinguished from each other under model @xmath178 . similar terminology will be used for other models and for undirected @xmath27-forests . stronger notions of identifiability may be defined , and correspondingly stronger variants of identifiability questions may be asked . the above definitions assume the model parameters to be fixed . but we may ask if there are nonisomorphic pedigrees @xmath58 and @xmath60 and model parameters @xmath201 such that @xmath202 for all @xmath203 . given the probability distribution a pedigree induces on the space of alignments , we may ask if the pedigree can be recognised to be in a class @xmath195 . we assume in all results in this paper that the model parameters and the class @xmath195 ( typically defined by the size of a pedigree ) to be fixed ( but possibly unknown ) . pedigrees of order 1 are in general not identifiable since for all pedigrees of order 1 , the extant sequence will be uniformly distributed . in fact for all pedigrees , all extant sequences will be uniformly distributed . but when there are more than one extant vertices , the joint distribution of extant sequences will contain some information about the pedigree on which they have evolved , because , for example , some vertices may have common ancestors , so their sequences will be correlated . we will therefore consider identifiability questions for pedigrees of order more than 1 . the models r and rm on a pedigree @xmath58 with @xmath204 arcs may be interpreted as hidden markov models with the set of hidden states @xmath124 and the set of observed states @xmath87 . this is illustrated in figure [ fig - chains ] . the initial probability for each hidden state is @xmath205 . the probability of an observed state conditional on a given hidden state @xmath206 can be easily computed ( and actually depends only on @xmath109 and @xmath170 ) . the probability of transition from state @xmath133 to @xmath207 is given by @xmath208 in most contexts in which hmms are used , one assumes that the set of hidden states in known , but here we do not know the set @xmath124 . we also note that the sequence of @xmath27-forests is not a markov chain . consider directed @xmath27-forests @xmath209 and @xmath210 at sites @xmath97 and @xmath211 , respectively . suppose a vertex @xmath34 in @xmath212 has parents @xmath38 and @xmath67 in the pedigree . the probability that the arc @xmath68 is in @xmath210 depends on the history before the @xmath97-th site . for example , if @xmath68 was in @xmath213 then the probability that it would also be in @xmath210 is @xmath214 . but if @xmath68 was not in @xmath213 the probability that it would be in @xmath210 is @xmath215 . therefore , @xmath216 is not a markov chain , i.e. , @xmath217 may not be the same as @xmath218 if the sequences @xmath219 and @xmath220 are different . therefore , we can not interpret the models as hmms with directed or undirected @xmath27-forests as hidden states . of characters at the extant vertices of a pedigree @xmath58 . among the intermediate chains , only @xmath221 is a markov chain.,width=445 ] in this section , we analyse diploid pedigrees and certain haploid pedigrees that obey many properties of diploid pedigrees under model @xmath222 . we show that , with some exceptions and mild conditions on @xmath145 and @xmath223 , diploid pedigrees of depth 2 are reconstructible from the probability distribution on the alignments on extant vertices . we use very basic techniques such as pairwise comparisons between extant sequences to exploit the correlation between them to reconstruct their pedigree . let @xmath58 be a pedigree and let @xmath102 be an undirected @xmath27-forest . we define @xmath224 . [ prop - ngt ] let @xmath58 be a pedigree with @xmath204 arcs . let @xmath225 be any character . then the probability that the @xmath10-th character in an alignment is @xmath226 is given by @xmath227 in particular , it does not depend on @xmath10 . let @xmath228 be a spanning forest sequence . it is a time - homogeneous markov chain on @xmath124 , with transition probabilities given by @xmath229 it follows that @xmath230 for all @xmath231 . let @xmath232 be a sequence of characters . then , under model rm , @xmath233 & = & \sum_{g \in { \mathcal{g}}_p } { \mathtt{pr}}\{c_k = c\mid g_k = g\ } { \mathtt{pr}}\{g_k = g { \mid}p , rm(p,\mu)\ } \nonumber \\[2 mm ] & = & \frac{1}{2^e}\sum_{g \in { \mathcal{g}}_p } { \mathtt{pr}}\{c_k = c\mid g_k = g\ } \nonumber \\[2 mm ] & = & \sum_{t \in \lvert { \mathcal{u}}_p \rvert } \frac{n(g > t : p)}{2^e } { \mathtt{pr}}\{c\mid t , m(\mu)\}. \nonumber \end{aligned}\ ] ] a similar result holds when model rm is replaced by model r. the above proposition implies that if two pedigrees have the same number of undirected @xmath27-forests of each type and the same number of arcs , then the character frequencies in alignments alone are not sufficient to distinguish the two pedigrees . for example , pedigrees in figure [ fig - hap ] can not be easily distinguished . but it turns out that , under model r , most diploid pedigrees of depth 2 are easily distinguished by making pairwise comparisons between extant sequences and computing the probability that they agree at a site . the above proposition implies that the probability that two sequences in an alignment agree at a site @xmath10 does not depend on @xmath10 in models r and rm . moreover , such a probability can be computed easily under model r. any given undirected @xmath27-forest @xmath102 of a pedigree induces a partition of its leaves so that the leaves within a component are in the same part . under model r , the probability that the sequences at two leaves @xmath135 and @xmath97 are in the same state at a site is 1 if they are in the same component of @xmath102 . otherwise , the probability is @xmath234 . under model rm , the probability depends on @xmath170 if they are in the same component , and is @xmath234 otherwise . [ prop-2seq ] let @xmath58 be a diploid pedigree of depth 2 . let @xmath235 . then the subpedigree of @xmath135 and @xmath97 is determined by the probability distribution induced on @xmath236 under model r. there are four possible ways in which any two vertices @xmath135 and @xmath97 are related in a diploid pedigree , which are shown in figure [ fig - grand ] . for each of them , we give the probability that the sequences @xmath70 and @xmath72 match at any site @xmath10 . in the following , we set @xmath237 . and @xmath97 may be related , width=445 ] _ case 1 : _ if @xmath135 and @xmath97 have the same parents , then @xmath238 . _ case 2 : _ if @xmath135 and @xmath97 have distinct pairs of parents but the same grand parents , then @xmath239 . _ case 3 : _ if @xmath135 and @xmath97 have distinct pairs of parents but exactly one pair of grand parents in common , then @xmath240 . _ case 4 : _ if @xmath135 and @xmath97 have no common parents or grand parents , then @xmath241 . therefore , unless @xmath242 , the above cases are distinguished by the marginal joint distribution on @xmath236 under model r. the assumption that @xmath58 is a diploid pedigree is essential . it implies that @xmath135 and @xmath97 are not related as in the haploid pedigree @xmath243 shown in figure [ fig - hap ] . we can verify that @xmath244 for both @xmath245 and @xmath243 , so they are indistinguishable by the above method , which was also pointed out as a consequence of proposition [ prop - ngt ] . the above probabilities do not depend on @xmath145 , therefore , we have a slightly stronger identifiability statement : if diploid pedigrees @xmath58 and @xmath60 of depth 2 and crossover probabilities @xmath145 and @xmath246 are such that @xmath247 then their subpedigrees of order 2 are correspondingly isomorphic . [ prop - n2 - 1 ] when @xmath248 , pedigrees of depth 2 in which no two vertices have exactly one common parent are identifiable under model r. in particular , when @xmath248 , diploid pedigrees of depth 2 are identifiable under model r. there are only 4 ways in which any two extant vertices @xmath135 and @xmath97 are related . they are illustrated in figure [ fig - grand ] . each of the possible relationships is recognised by proposition [ prop-2seq ] . we denote the 3rd and the 4th types of relationships by @xmath249 and @xmath250 , respectively . suppose that no two extant vertices @xmath251 in a pedigree are related to each other as @xmath249 or @xmath250 . then the pedigree is constructed by adding extant vertices one by one . on each step we add one extant vertex and join to previously added extant vertices as in figure [ fig - grand ] - ( i ) or ( ii ) , whichever is appropriate . therefore , we assume that at least two extant vertices @xmath252 are related as @xmath249 or @xmath250 . let @xmath253 be a ( nonempty ) maximal subset of @xmath27 such that for any two distinct extant vertices @xmath135 and @xmath97 in @xmath253 , either @xmath249 or @xmath254 . every other extant vertex @xmath10 not in @xmath253 is related to some vertex in @xmath253 as in figure [ fig - grand ] - ( i ) or ( ii ) . therefore , once the subpedigree of @xmath253 is constructed , there is only one way to extend it to the whole pedigree . to construct the subpedigree of @xmath253 , we first construct an edge labelled graph with edge set @xmath253 in which edges @xmath135 and @xmath97 are incident if and only if @xmath249 . this is a known problem in graph theory , namely , the problem of constructing an edge labelled graph from its line graph . it was proved @xcite that there are only 4 pairs @xmath255 of connected nonisomorphic edge labelled graphs that have the same line graphs . edge labelled graphs that can not be uniquely constructed from their line graphs must contain components isomorphic to @xmath133 or @xmath256 . we refer to @xcite ( in particular , chapter 15 , problem 1 ) for discussion about reconstructing graphs from their line graphs , in particular , for the complete list of pairs @xmath255 . the first pair is @xmath257 ( with edges of each of them labelled @xmath258 ) . based on the example @xmath257 , we construct pedigrees shown in figure [ fig - k3 ] , in which all pairs of extant vertices are similarly related . we distinguish the two pedigrees in figure [ fig - k3 ] by comparing the probabilities @xmath259 and @xmath260 for any site @xmath2 . in @xmath261 , there are 512 spanning forests . among them there are 192 spanning forests in which two extant vertices have a common grandparent , giving the first term on the rhs below . in the remaining 320 spanning forests , no two extant vertices have a common grandparent , which explains the second term on the rhs below . therefore , @xmath262 in @xmath263 , there are 512 spanning forests . among them there are 16 spanning forests in which @xmath258 have a common grandparent ( giving the first term on the rhs below ) , 144 spanning forests in which two extant vertices have a common grandparent at site @xmath2 ( giving the second term ) , and 352 spanning forests in which @xmath258 have distinct grandparents at site @xmath2 ( giving the third term ) . therefore , @xmath264 whenever @xmath248 , @xmath265 , therefore , @xmath261 and @xmath263 can be distinguished . the two expressions are equal when @xmath266 . similarly , other pedigrees constructed from @xmath267 are distinguished when @xmath248 . since there are only two types of site patterns for three sequences when @xmath266 ( either the three sequences agree at a site or exactly two of them agree ) , the two cases can not be distinguished by considering other site pattern probabilities . in diploid pedigrees , no two vertices have exactly one common parent , therefore , when @xmath248 , they are identifiable under model r. we used only site pattern probabilities in the above proofs . but because of recombinations , consecutive sites in an alignment are not independent . we use the dependence between sites to eliminate the restriction @xmath248 when the crossover probability @xmath145 is sufficiently small . [ prop - n2 - 2 ] when @xmath266 and @xmath145 is sufficiently small , haploid pedigrees @xmath261 and @xmath263 ( shown in figure [ fig - k3 ] ) are distinguished under model @xmath179 . we compute the probability that , in an alignment @xmath91 , there are long runs of sites at which all the three sequences @xmath70 , @xmath72 and @xmath74 are equal . in particular , we compute bounds on @xmath268\}$ ] on the two pedigrees . for @xmath261 , for any fixed @xmath269 , if any two of the three sites @xmath270 are inherited from the same grand parent , then @xmath271 , and if all of them are inherited from distinct grandparents , then @xmath272 is 1/4 . therefore , @xmath273 { \mid}p_2,r(p ) \ } \leq ( 1/2)^{t+1}.\ ] ] for @xmath263 , the probability that the first site of all sequences is inherited from a common grand parent is 1/32 . the probability that at each successive site the three sequences have a common grand parent is @xmath274 . therefore , @xmath275 { \mid}q_2,r(p ) \ } \nonumber \\[2 mm ] & \geq & \frac{\left((1-p)^6+p^3(1-p)^3\right)^t}{32}. \end{aligned}\ ] ] when @xmath145 is sufficiently small and @xmath276 is sufficiently large , the above probability for @xmath263 is more than that for @xmath261 . we do not analyse other examples of pedigrees based on graphs that are not reconstructible from their line graphs ( graphs @xmath277 mentioned in proposition [ prop - n2 - 1 ] ) , but they may be analysed similarly . in this section we develop ideas from section [ sec - model - r ] ( especially proposition [ prop - n2 - 2 ] ) in much more generality . earlier we observed that parts of alignments that are free of recombination may be analysed with phylogenetic methods . since we do not know where the recombinations have occurred in an alignment , we choose long segments of carefully chosen alignments ( or sets of alignments ) and show that they have higher probability of having evolved on one particular @xmath27-forest ( or a sequence of @xmath27-forests ) than any other @xmath27-forest ( or a sequence of @xmath27-forests ) . for example , the method of proposition [ prop - n2 - 2 ] works because @xmath263 contains a tree in which @xmath258 have a common ancestor and @xmath261 does not contain such a tree . therefore , a sufficiently long sequence of characters in which sequences @xmath278 and @xmath279 are in the same state will more likely have evolved on a pedigree such as @xmath263 than on a pedigree that does not contain such a tree . this argument may be generalised to count the number of @xmath27-forests of each type from the distribution of extant sequences . such a generalisation requires identifiability results for phylogenetic trees , which we state in a form suitable for our application . in this section , we state known results on identifiability and statistical consistency of maximum likelihood reconstruction of phylogenetic trees . we need to adapt them slightly since the @xmath27-forests in a pedigree differ from phylogenetic trees in three respects - they may have vertices of degree 2 , they may be unresolved ( i.e. , they may have vertices of degree more than 3 ) , and they may be disconnected ( two extant vertices may not have a common ancestor in a given directed @xmath27-forest in a pedigree ) . we address them in the following identifiability result , which was originally proved for phylogenetic trees in @xcite in the @xmath266 case . identifiability and statistical consistency of maximum likelihood reconstruction of phylogenetic trees were independently proved in full generality for all @xmath280 in @xcite . [ thm - identifiability ] for all @xmath281 and any two undirected @xmath27-forests @xmath282 and @xmath283 with bounded number of edges , if @xmath284 for all characters @xmath285 then @xmath286 . the result follows from the analogous results in @xcite for phylogenetic @xmath27-trees , but we have to clarify three issues : unlike the phylogenetic @xmath27-trees , the @xmath27-forests as defined in this paper may be disconnected , unresolved , and may have vertices of degree 2 . _ connectivity : _ any two extant vertices @xmath287 and @xmath288 are in different components of @xmath282 and @xmath283 if and only if @xmath289 for all @xmath290 , where @xmath291 is the state at the extant vertex @xmath135 in a character @xmath226 . if @xmath135 and @xmath97 are in the same component , then @xmath292 can not be arbitrarily close to @xmath234 if the number of edges ( and hence the distance between @xmath135 and @xmath97 ) is bounded . therefore , we can consider the identifiability question for each component separately . _ unresolved @xmath27-forests : _ given an unresolved phylogenetic tree , there are resolved phylogenetic trees with site pattern probabilities arbitrarily close to the site pattern probabilities for the unresolved tree . therefore , even though unresolved phylogenetic @xmath27-trees are identifiable , statistical consistency of maximum likelihood methods requires that the substitution probabilities on the edges of a phylogenetic tree are bounded below by a positive real number . in our model , corollary [ cor - consistency ] below is possible because the substitution probability @xmath170 is fixed on each edge . _ vertices of degree 2 : _ let @xmath34 and @xmath38 be any two vertices of an @xmath27-forest . suppose that @xmath34 and @xmath38 are of degree 1 or more than 2 . suppose all internal vertices on the path between @xmath34 and @xmath38 have degree 2 . since @xmath170 is fixed for all arcs , the distance between @xmath34 and @xmath38 on a tree is determined by the substitution probability on the @xmath39 path . the substitution probability on the @xmath39 path is determined by the distribution on the space of characters . in the above result , if @xmath282 and @xmath283 were directed @xmath27-forests , then we would be able to conclude that @xmath293 . let @xmath294 . suppose that @xmath295 . we associate with each undirected @xmath27-forest @xmath102 a vector @xmath296 in @xmath297 , where @xmath298 . therefore , the condition @xmath299 for all characters @xmath225 may be equivalently written as @xmath300 . given @xmath301 and a point @xmath302 , let the open ball of radius @xmath1 centred at @xmath303 be denoted by @xmath304 . here the radius may be taken to be in the 1-norm ( i.e. , the distance between points @xmath305 and @xmath306 is defined by @xmath307 ) . let @xmath91 be an alignment on @xmath27 . we define a vector @xmath308 , where @xmath309 are the _ fractional site pattern frequencies _ , i.e. , @xmath309 is the fraction of columns of @xmath91 of type @xmath310 . then the above identifiability result implies a statistical consistency result for maximum likelihood . it informally says that as the length of a random alignment @xmath91 goes to infinity , we expect @xmath311 to be arbitrarily close to @xmath312 with probability approaching 1 if @xmath102 is the true @xmath27-forest , and that the probability that @xmath311 is arbitrarily close to @xmath312 approaches 0 if @xmath102 is not the true @xmath27-forest . [ cor - consistency ] for all @xmath313 , @xmath314 and @xmath315 , there exists @xmath316 such that for any two undirected @xmath27-forests @xmath102 and @xmath117 such that @xmath317 , and an alignment @xmath318 , @xmath319 and @xmath320 we give bounds on the above probabilities in terms of @xmath79 , which we prove using bernstein s inequality . [ lem - bernstein1 ] let @xmath321 be i.i.d . bernoulli random variables with @xmath322 . then for all @xmath323 and @xmath324 , @xmath325 and ( equivalently ) @xmath326 [ lem - eps - l1 ] let @xmath313 and @xmath281 . let @xmath327 . let @xmath102 be an undirected @xmath27-forest . then @xmath328 there are @xmath329 distinct characters , with probabilities @xmath330 , @xmath331 . if @xmath332 , then @xmath333 for some @xmath135 therefore , we apply bernstein s inequality to each distinct character , and write the probability that @xmath334 . we then apply the union bound to get the result . if we set @xmath335 it will ensure that the @xmath27-forests are _ separated _ by open balls of radius @xmath336 in the space of site pattern probability vectors , i.e. , the open balls @xmath337 and @xmath338 are non - intersecting whenever @xmath339 and @xmath209 are non - isomorphic . we will use this value of @xmath336 unless specified otherwise . now for an undirected @xmath27-forest @xmath339 , we define @xmath340 and @xmath341 by selecting a sufficiently large value of @xmath79 we can make @xmath342 arbitrarily small as in lemma [ lem - eps - l1 ] . moreover , if @xmath343 , then @xmath344 hence @xmath345 can be made arbitrarily small as per lemma [ lem - eps - l1 ] . we set @xmath346 , which depends on @xmath79 and @xmath336 . in theorem [ thm - paths ] ( particularly in the proof of inequality [ eq - term1a ] ) we require a concentration inequality similar to the inequality in lemma [ lem - eps - l1 ] for the situation in which @xmath79 sites of an alignment have evolved on an @xmath27-forest @xmath339 and @xmath347 sites ( for a small @xmath348 ) have evolved on another @xmath27-forest @xmath209 . ( we will keep the notation simple by assuming that @xmath347 is an integer . ) therefore , we give the following variants of lemmas [ lem - bernstein1 ] and [ lem - eps - l1 ] . [ lem - bernstein2 ] let @xmath349 be i.i.d . bernoulli random variables with @xmath350 . let @xmath351 be bernoulli random variables . let @xmath352 , where @xmath353 is a positive constant . let @xmath354 and @xmath355 and @xmath356 . if @xmath357 and @xmath358 , then @xmath359 since @xmath360 , we have @xmath361 & \implies & \left(\frac{s_n}{n(1+c ) } - p \leq -r\right ) \;\text{or}\ ; \left(\frac{s_n}{n(1+c ) } - p \geq r\right ) \\[1 mm ] & \implies & \left(\frac{\sum_{i=1}^n x_i}{n(1+c ) } - p \leq -r\right ) \;\text{or}\ ; \left(\frac{cn + \sum_{i=1}^n x_i}{n(1+c ) } - p \geq r\right ) \\[1 mm ] & \implies & \left(\frac{\sum_{i=1}^n x_i}{n(1+c ) } \leq p - r \leq m \right ) \;\text{or}\ ; \left(\frac{cn + \sum_{i=1}^n x_i}{n(1+c ) } \geq p+r \geq m \right ) \\[1 mm ] & \implies & \left(\frac{\sum_{i=1}^n x_i}{n(1+c)}-\frac{p}{1+c } \leq m-\frac{p}{1+c}\right ) \text{or}\ ; \left(\frac{cn + \sum_{i=1}^n x_i}{n(1+c ) } - \frac{p+c}{1+c } \geq m-\frac{p+c}{1+c } \right ) \\[1 mm ] & \implies & \left(\frac{\sum_{i=1}^n x_i}{n}-p \leq -r^\prime \right ) \;\text{or}\ ; \left(\frac{\sum_{i=1}^n x_i}{n}-p \geq r^{\prime\prime}\right ) . \end{aligned}\ ] ] now we apply bernstein s inequality ( lemma [ lem - bernstein2 ] ) to each term and obtain the desired bound . let @xmath91 be an alignment of length @xmath362 . suppose that @xmath79 characters of @xmath91 evolved on an undirected @xmath27-forest @xmath102 and the remaining characters evolved on undirected @xmath27-forests @xmath363 . the following lemma states that if @xmath353 is sufficiently small , then @xmath311 is concentrated near @xmath364 for large @xmath79 . moreover , as in lemma [ lem - eps - l1 ] , if we require @xmath311 to be sufficiently near @xmath364 with probability at least @xmath365 , then the length of the alignment @xmath362 must be @xmath366 . in the following lemma , we do not specify the constants @xmath367 and @xmath368 precisely , but they can be chosen depending on @xmath336 . [ lem - eps - l2 ] let @xmath313 and @xmath281 . let @xmath369 for a suitably chosen positive constant @xmath353 . let @xmath370 be undirected @xmath27-forest . then @xmath371 & \leq & \sum_i^{|\sigma|^{|x|}}\left(\exp \left\{\frac{-l(r_i^\prime)^2}{2p_i(1-p_i ) + 2r_i^\prime/3}\right\ } + \exp \left\{\frac{-l(r_i^{\prime\prime})^2 } { 2p_i(1-p_i ) + 2r_i^{\prime\prime}/3}\right\}\right ) , \end{aligned}\ ] ] where @xmath372 and @xmath368 are positive constants as in lemma [ lem - bernstein2 ] . we apply lemma [ lem - bernstein2 ] for each component of @xmath311 and use the union bound as in the proof of lemma [ lem - eps - l1 ] . for each component , we use @xmath373 as before . constants @xmath372 and @xmath368 ( and @xmath374 , which are implicit ) depend on @xmath336 and the probabilities @xmath330 , @xmath375 . the constant @xmath353 may be taken to be the smallest among @xmath376 . a hidden markov model ( hmm ) is defined by two sequences @xmath377 and @xmath378 of random variables . the sequence @xmath379 takes values in @xmath380 $ ] and is a stationary markov chain with transition matrix @xmath91 and initial distribution @xmath381 to @xmath1 , which is also the stationary distribution of the markov chain . the random variables @xmath378 take values in @xmath382 $ ] , and are independent and identically distributed conditional on @xmath379 . the distribution of @xmath383 depends only on @xmath384 . let @xmath385 be the @xmath386 matrix of conditional probabilities @xmath387 , where @xmath388 $ ] and @xmath389 $ ] . the sequence @xmath378 are the _ observations_. identifiable hidden markov models were characterised in @xcite , where a precise description of conditions on @xmath91 and @xmath385 for which the probability distribution on observed sequences determines @xmath91 and @xmath385 ( up to re - labelling of hidden states ) was given . here _ identifiability up to a re - labelling of hidden states _ means the following : if @xmath25 is an @xmath390 permutation matrix , then the hmm with parameters @xmath391 ( where @xmath119 is treated as a column vector of length @xmath1 ) is equivalent to ( induces the same distribution on the space sequences of observed states as ) the hmm with parameters @xmath392 . therefore , identifiability only means computing the matrices and the initial distribution up to equivalence . we denote the class of models equivalent to @xmath391 by @xmath393 . earlier we noted that models r and rm for sequences evolving on a pedigree @xmath394 define a hidden markov model with the spanning forests in @xmath58 as hidden states and characters from @xmath87 as observed states . we call it @xmath395 and denote its matrices by @xmath396 and @xmath397 . the initial distribution on hidden states is uniform : each spanning forest has the probability @xmath398 if the pedigree has @xmath399 arcs . we informally look at some of the issues about its identifiability . the transition matrix @xmath396 is defined by transition probabilities given in equation ( [ eq - gi - gj ] ) . therefore , @xmath396 will be identical ( up to a permutation of rows and columns ) for all pedigrees with the same number of arcs , for a fixed @xmath145 . but the set of spanning forests ( hidden states ) is unknown . we now describe at a high level how we compute the rows of @xmath397 . suppose the pedigree contains an undirected @xmath27-forest @xmath339 . there are @xmath400 spanning forests @xmath29 of @xmath58 that contain @xmath339 as the unique undirected @xmath27-forest , and corresponding to each of them we have a row of @xmath397 that is equal to @xmath401 . now consider a set @xmath402 of sufficiently long alignments as defined in equation [ eq - mcai ] . we compute the probability of @xmath403 ( i.e. , the probability that a random alignment is in @xmath403 ) . suppose that @xmath404 is the probability that there are no recombination events . thus @xmath404 approaches 1 as @xmath145 approaches 0 . then one of the terms in the expression for the probability of @xmath403 will be @xmath405 , where @xmath406 is the probability that the first site evolved on a spanning forest @xmath29 that contained @xmath339 as the undirected @xmath27-forest . there will be terms for contributions from other undirected @xmath27-forests @xmath407 , but they will be much smaller than the above term because they will contain factors @xmath408 ( as in equation [ eq - p - mcai ] ) . there will also be terms to account for recombinations among the first @xmath79 sites , but they will be small as well since they will contain @xmath145 as a factor ( in contrast to @xmath404 , which is a power of @xmath409 ) . so let us say @xmath410 is @xmath411 . therefore if @xmath145 is sufficiently small and @xmath79 is sufficiently large , then @xmath412 will be roughly equal to the dominating term @xmath413 , which will uniquely determine @xmath414 . in other words , if @xmath60 is another pedigree such that @xmath415 , then @xmath410 and @xmath416 will differ roughly by a multiple of @xmath417 . in the proof of proposition [ prop - n2 - 2 ] , we used a similar idea : the pedigree @xmath263 contains a certain subtree @xmath102 in which @xmath258 have a common ancestor , while the pedigree @xmath261 does not such a subtree . as a result , the alignments that are close to @xmath364 are more likely to have evolved on @xmath263 than on @xmath261 . suppose now that @xmath397 is identified and each of its rows is labelled by the corresponding unlabelled undirected @xmath27-forest . that is , the matrices @xmath397 that appear among the triples in the equivalence class @xmath393 of hmms are constructed . as pointed out above , the matrix @xmath396 and the initial distribution are also known up to relabelling of hidden states . but the equivalence class @xmath418 is not known unless we are able to label the rows and the columns of @xmath396 by unlabelled undirected @xmath27-forests in a manner consistent with the labelling of rows of @xmath397 . in other words , for full identifiability of @xmath395 , we would like to construct an automaton with transition probabilities given by @xmath396 and with its states labelled by unlabelled undirected @xmath27-forests . identifying the pedigree from the _ labelled automaton _ will then be a purely combinatorial problem . in this paper we do not succeed in constructing matrix @xmath396 with rows and columns labelled by undirected @xmath27-forests , but we are able to count certain types of _ walks _ ( to be described next ) on the automaton with vertices labelled by undirected @xmath27-forests . suppose that @xmath419 is a sequence of undirected @xmath27-forests such that no two consecutive ones are isomorphic . analogous to @xmath420 , we define @xmath421 as the number of sequences @xmath422 of spanning forests in @xmath58 such that @xmath423 , where consecutive @xmath133 are separated by just one recombination . ( a single recombination at a site is more likely than multiple recombinations . moreover , if there is a recombination at a site @xmath135 , but the two spanning forests @xmath133 and @xmath134 contain isomorphic undirected @xmath27-forests , then such a recombination has no effect on the emitted characters . these are the reasons why we consider the sequences @xmath339 and @xmath133 as above . ) we then consider a set @xmath424 of alignments of length @xmath425 ( for a suitably large @xmath79 ) obtained by concatenating alignments from @xmath403 for @xmath84 to @xmath269 . we compute the probability of @xmath424 ( as we described for @xmath426 above ) , and show that the dominating term is proportional to @xmath427 , and other terms are of smaller order of magnitude for small @xmath145 . this allows us to compute @xmath428 . a more visual description of the walks may be given as follows . suppose the pedigree has @xmath399 arcs . we define a graph on the vertex set consisting of the spanning forests of the pedigree , with two spanning forests @xmath133 and @xmath207 being adjacent if there is exactly one recombination separating them ( i.e. , @xmath429 ) . the graph is a hypercube . the hidden markov chain on the set of spanning forests jumps on the vertices of the cube . if there is at most one recombination at any site ( which is more likely than more than 1 recombinations at a site ) , then we have a walk on the edges of the cube . we label each vertex @xmath133 of the cube by the undirected @xmath27-forests @xmath430 . our interest is to construct this object for a more complete understanding of the hmm . but problem is made difficult by the fact that the emission probabilities associated with @xmath133 and @xmath207 are identical if @xmath431 . therefore , we construct a weaker object , namely the number of walks of each length on the cube such that consecutive vertices have distinct labels . let @xmath58 be a pedigree . for @xmath432 , we define @xmath433 & : = & n(g_1 > t_1 , g_2 > t_2 , \ldots , g_m > t_m : p ) \\[2 mm ] & : = & |\{\mathbf{g } \in { \mathcal{g}}_p^m : t_u(g_i ) \cong t_i \forall i \in [ m ] \wedge |e(g_{i+1})\bigtriangleup e(g_{i})| = 2 \ ; \forall\ ; i \in [ m-1]\}|,\end{aligned}\ ] ] where the second condition in the last line says that there is exactly one recombination event between consecutive @xmath133 . in the rest of this section , we show how invariants @xmath434 may be computed from the probability distribution on the space of alignments under model rm . in the end , we demonstrate an application to pedigrees @xmath245 and @xmath243 shown in figure [ fig - hap ] . lem - pg ] let @xmath58 be a pedigree with @xmath204 arcs . let @xmath435 be a sequence of spanning forests in @xmath58 . then under model @xmath179 , the probability that @xmath136 is a sequence of site - specific spanning forests is given by @xmath436 where @xmath437 and @xmath438 are as in definition [ def - rg - sg ] . we have a factor @xmath145 for each recombination event and @xmath409 whenever there is no recombination ( i.e. , an arc is contained in two consecutive spanning forests in the sequence ) . the probability that the first spanning forest is @xmath40 is @xmath398 . let @xmath318 be an alignment of length @xmath79 . for an interval @xmath439 \subseteq [ l]$ ] , we write @xmath440 $ ] for the part of the alignment between columns @xmath441 and @xmath442 ( inclusive of columns @xmath441 and @xmath442 ) . for a sequence of alignments @xmath443 $ ] , let @xmath444 denote the alignment obtained by concatenating alignments @xmath445 ( in that order ) . let @xmath446 $ ] be sets of alignments . we define @xmath447\}.\ ] ] [ lem - pa ] let @xmath58 be a pedigree with @xmath204 arcs . the probability of an alignment @xmath318 on @xmath58 is given by @xmath448 & = & \sum_{k=1}^{l}\ ; \sum_{\mathbf{g } \in { \mathcal{g}}_p^k } \frac{(1-p)^{e(l - k)+s(\mathbf{g})}p^{r(\mathbf{g})}}{2^{e } } \sum_{\substack{\mathbf{l } \in { \mathbb{z}}_+^k : \\ \sum\limits_{i=1}^kl_i = l}}\ ; \prod_{i=1}^k{\mathtt{pr}}\{a[l_{i-1}+1,l_i ] { \mid}t_u(g_i),m(\mu ) \ } , \nonumber \\ & & \end{aligned}\ ] ] where @xmath449 , @xmath450 and @xmath451 , and the second summation is over @xmath136 such that consecutive spanning forests @xmath452 and @xmath134 are unequal for @xmath453 $ ] . the probability of an alignment of length @xmath79 is obtained by summing its probability over all spanning subgraph sequences of length @xmath79 . suppose that the recombinations in a spanning subgraph sequence occur only at sites @xmath454 , for @xmath455 , where @xmath456 . we write the spanning forest sequence of length @xmath79 as @xmath457 . then the probability of the alignment is written as a product of probabilities of its segments that have evolved on spanning forests @xmath133 ( i.e. , effectively on @xmath430 ) . this probability is summed over @xmath458 ( with the constraint @xmath459 ) , @xmath460 $ ] and @xmath461 . for a fixed @xmath462 , the probability of @xmath457 is given by lemma [ lem - pg ] . for @xmath463 , we will compute @xmath464 by summing ( [ eq - pa ] ) over @xmath465 . in such a calculation , we will sometimes use the following upper bound in evaluating the last summation in ( [ eq - pa ] ) for fixed values of @xmath10 , and fixed @xmath466 and @xmath136 . [ lem - pa - ub2 ] let @xmath463 . let @xmath467 . let @xmath468 : a \in { \mathcal{a}}\ } , i \in [ k]$ ] . then for any fixed @xmath469 @xmath470 we have equality if @xmath471 . the claim follows from the observation that @xmath472 . the following lemma is used in the proof of equation ( [ eq-3b ] ) , and the reader may skip it until then . [ lem - seq ] let @xmath473 be a finite set . let @xmath474 and @xmath475 be two sequences in @xmath473 , each of length @xmath476 , defined by @xmath477 where @xmath478 and @xmath479 $ ] , and @xmath480 where @xmath481 and @xmath482 $ ] . suppose that @xmath474 and @xmath475 satisfy the constraints : @xmath483 $ ] ; @xmath484 $ ] ; @xmath485 ; if @xmath486 , then @xmath487 for some @xmath135 . then there is at least one block @xmath488 of @xmath474 over which @xmath474 and @xmath475 mismatch everywhere ; hence there are at least @xmath489\}$ ] mismatches between the two sequences . suppose that the claim is false ; so in each block @xmath488 of @xmath474 , there is a matching symbol in @xmath475 . therefore , @xmath419 is a subsequence of @xmath475 and @xmath490 . this , together with @xmath483 $ ] and @xmath491 , implies that @xmath486 and @xmath492 $ ] , which contradicts the assumption that when @xmath486 , there is some @xmath135 for which @xmath487 . [ thm - paths ] let @xmath58 and @xmath60 be any two pedigrees with @xmath493 arcs . let @xmath494 be any sequence of undirected @xmath27-forests in which consecutive @xmath27-forests are non - isomorphic . then for all @xmath281 , there exists @xmath495 such that for all @xmath496 , the following statement is true : if @xmath497 , then @xmath498 . let @xmath499 , where @xmath500 and @xmath336 are as defined in equations ( [ eq - eps0 ] ) and ( [ eq - mcai ] ) , respectively . we will choose @xmath501 ( defined at the end of section [ sec - consistency ] ) and @xmath79 ( that depends on @xmath501 ) suitably later . the probability of @xmath424 on @xmath58 and @xmath60 is written by summing ( [ eq - pa ] ) over all @xmath465 . but based on theorem [ thm - identifiability ] , corollary [ cor - consistency ] and lemma [ lem - eps - l2 ] , we can make the following qualitative and somewhat informal statement : if @xmath79 is large enough , then @xmath502 will get significantly higher contribution from spanning forest sequences @xmath503 of length @xmath504 such that @xmath505 for all @xmath506 $ ] and @xmath507 are all close to @xmath79 , than from spanning forest sequences @xmath508 for which there are many mismatches ( in terms of isomorphism ) between sequences @xmath509 and @xmath510 or if they require more than @xmath511 recombinations . let @xmath512 , where @xmath513 is the joint probability of @xmath424 and the event that there are exactly @xmath10 recombinations . furthermore , we write @xmath514 , where @xmath515 is the contribution from spanning forest sequences @xmath516 of length @xmath517 such that @xmath505 for all @xmath506 $ ] , and @xmath518 is the remaining contribution to @xmath519 , i.e. , from spanning forest sequences @xmath520 of length @xmath517 such that either @xmath521 ( i.e. , the @xmath511 recombinations occur at fewer than @xmath511 sites ) , or @xmath522 and @xmath523 for some @xmath506 $ ] . we will show that only @xmath515 makes a significant contribution to @xmath524 . in particular , we will show that the various contributions to @xmath525 take the following form : @xmath526 \sum_{k < m-1}p_k(p ) + p_{(m-1)b}(p ) & < & \delta_1 \\[1 mm ] \sum_{k \geq m } p_k(p ) & < & \delta_2 , \end{aligned}\ ] ] where @xmath527 does not depend on the pedigree ( but depends on the undirected @xmath27-forest sequence @xmath474 ) . moreover , @xmath527 , @xmath528 and @xmath529 depend on @xmath530 and @xmath501 . we will show that when @xmath145 and @xmath501 are sufficiently small and @xmath79 is sufficiently large , @xmath528 and @xmath529 are very small compared to @xmath527 . it will imply that for @xmath502 and @xmath531 to be equal , @xmath532 and @xmath533 must be equal . ( otherwise , there would be a difference between @xmath502 and @xmath531 that is of the order of a multiple of @xmath527 . ) _ a lower bound on @xmath515 . _ since the consecutive @xmath27-forests in @xmath474 are nonisomorphic and @xmath505 for all @xmath506 $ ] , there is at least one recombination between consecutive spanning forests @xmath133 . and since there are exactly @xmath511 recombinations , there must be exactly one recombination between consecutive spanning forests . therefore , @xmath534 & & \times \left ( \sum_{a \in { \mathcal{a } } } \sum_{\substack{\mathbf{l } \in { \mathbb{z}}_+^m : \\ \sum\limits_{i=1}^ml_i = m^2l}}\ ; \prod_{i=1}^m{\mathtt{pr}}\{a[l_{i-1}+1,l_i ] { \mid}t_u(g_i),m(\mu ) \}\right ) \nonumber \\ & : = & n(\mathbf{g } \geq \mathbf{t}:p ) \delta(\mathbf{t } ) , \end{aligned}\ ] ] where @xmath449 , @xmath450 . the inner summation is evaluated for any fixed choice of @xmath535 such that @xmath505 for all @xmath506 $ ] , because for any fixed @xmath536 $ ] , and spanning forest sequences @xmath537 and @xmath538 , if @xmath539 for all @xmath506 $ ] , then for each alignment @xmath91 , we have @xmath540 { \mid}g_i^{l_i},m(\mu ) \ } = { \mathtt{pr}}\{a[l_{i-1}+1,l_i ] { \mid}(g_i^{\prime})^{l_i},m(\mu ) \ } = { \mathtt{pr}}\{a[l_{i-1}+1,l_i ] { \mid}t_u(g_i),m(\mu ) \}$ ] . therefore , the rhs is a product of @xmath532 and a factor that does not explicitly depend on the pedigree , but only on @xmath474 . we have @xmath541 to prove the lower bound , we sum ( [ eq - pa ] ) over spanning forest sequences of the type @xmath503 where @xmath505 for all @xmath506 $ ] and @xmath507 are such that the @xmath511 recombinations occur at sites @xmath542 , for a small positive constant @xmath353 , ( i.e. , we ignore contributions from the spanning forest sequences in which some of the recombinations are not near the boundaries of the blocks @xmath543 of @xmath424 ) . we can choose the recombination sites in @xmath544 ways . since @xmath545 and @xmath546 , the probability of @xmath136 is @xmath547 . we write @xmath548 , where @xmath549 are blocks of @xmath136 of length @xmath425 each . then we have @xmath550 for each @xmath136 and for all @xmath135 , provided we choose @xmath551 and @xmath501 appropriately according to lemma [ lem - eps - l2 ] . _ an upper bound on @xmath552 . _ we have @xmath553 & \leq & \sum_{k=0}^{m-1 } \frac{1}{2^e } \left(\sum_{l=0}^{k}n_{kl}\binom{m^2 l -1}{l } ( \epsilon_{max})^{m - l}\right ) p^k(1-p)^{e(m^2l-1)-k } \nonumber \\[2 mm ] & \leq & c_2 l^{m-1 } \epsilon_{max } : = \delta_1 , \end{aligned}\ ] ] where @xmath554 is the number of spanning forest sequences @xmath555 in which @xmath10 recombinations occur at @xmath476 sites , and @xmath556 is a constant that depends on @xmath4 and @xmath269 . we explain below how the bound is obtained . we fix @xmath557 such that @xmath558 . for @xmath559 , since we are only interested in the contribution @xmath560 , we fix @xmath136 with the additional restrictions in the definition of @xmath560 . suppose that the @xmath10 recombinations occur at @xmath476 distinct sites @xmath561 , where @xmath562 . so we write @xmath563 . there are @xmath554 choices for @xmath564 . for each choice of @xmath565 , there are @xmath566 choices for the recombining sites @xmath567 . each @xmath136 has a probability @xmath568 , where @xmath569 and @xmath570 . we show that @xmath571 is bounded above by @xmath572 for each @xmath136 with @xmath476 recombining sites that satisfies the above constraints . let @xmath573 be a spanning forest sequence of length @xmath504 . for @xmath506 , j \in [ m]$ ] , we refer to the subsequences @xmath574)$ ] as _ blocks _ of @xmath136 , and subsequences @xmath575)$ ] as _ subblocks _ of @xmath136 . we say that subblock @xmath576 is _ recombination - free _ if there are no recombinations between any two sites of the subblock . ( the subblock may have recombinations at its boundaries . ) by lemma [ lem - seq ] , there is at least one block , say the @xmath135-th block , over which the sequences @xmath474 and @xmath577 mismatch everywhere . since there are @xmath269 subblocks in each block and @xmath578 , there are at least @xmath579 recombination - free subblocks in the @xmath135-th block of @xmath136 . let these subblocks be denoted by @xmath580 . we have @xmath581 for each of them . therefore , @xmath582 _ an upper bound on @xmath583 . _ we use the following fact about binomially distributed random variables : if @xmath584 , then @xmath585 . ( this result is a consequence of the union bound . ) since there are @xmath586 points at which a recombination can possibly occur , we have @xmath587 where @xmath588 is a constant that depends on @xmath4 and @xmath269 . we write similar bounds for @xmath60 . now suppose that @xmath589 but @xmath590 . therefore , equations [ eq - term1a ] and [ eq - term1b ] imply that @xmath591 & = & |n(\mathbf{g } \geq \mathbf{t}:p ) - n(\mathbf{g } \geq \mathbf{t}:q)|\delta(\mathbf{t } ) \nonumber \\[2 mm ] & \geq & \frac{(2cl)^{m-1}(1-p)^{(em^2 l - m - e+1)}p^{(m-1 ) } ( 1-\epsilon_{max})^{(m^2)}}{2^{e}}. \end{aligned}\ ] ] but that is impossible if we can choose @xmath501 , @xmath79 , and @xmath145 so that @xmath592 , or @xmath593 in other words , the discrepancy @xmath594 can not be compensated for by the remaining terms in @xmath595 and @xmath596 . such a choice is possible . for example , we first choose @xmath597 , let us say , @xmath598 , where @xmath599 . we then set @xmath600 and @xmath601 so that @xmath602 the choice of @xmath79 must guarantee phylogenetic consistency as in corollary [ cor - consistency ] . moreover , we must set @xmath79 as small as possible so as to get a better bound on @xmath145 . by lemma [ lem - eps - l2 ] , @xmath79 is required to grow only logarithmically in @xmath603 , so let @xmath604 for a suitable choice of @xmath605 . now conditions ( [ eq - cond ] ) are satisfied for @xmath606 for @xmath607 . then ( [ eq - final ] ) is satisfied for sufficiently large @xmath608 . in the following , we illustrate theorem [ thm - paths ] with a general result on counting @xmath27-forests and a couple of specific examples . if the conditions of theorem [ thm - paths ] are satisfied , then @xmath58 and @xmath60 have the same number of undirected @xmath27-forests of each type . we apply theorem [ thm - paths ] for @xmath609 . suppose that @xmath58 is a pedigree with @xmath204 arcs . let @xmath102 be an undirected @xmath27-forest . suppose that we want to count the number of copies @xmath610 of @xmath102 in p. let @xmath611 be the distinct copies of @xmath102 in @xmath58 . for each @xmath339 , let @xmath612 be the undirected @xmath27-forests in @xmath58 that contain @xmath339 . ( to be precise , we have subgraphs @xmath613 and @xmath614 in @xmath58 such that when the ( directed ) arcs of @xmath613 and @xmath614 are replaced by ( undirected ) edges , we get the undirected @xmath27-forests @xmath339 and @xmath615 . but we do not make this distinction in the following . ) note that @xmath615 are all distinct @xmath27-forests . there are @xmath616 non - founder vertices in @xmath58 at which we can choose one of the two incoming arcs to construct a spanning forest @xmath29 that contains @xmath339 . therefore , for each @xmath339 there are @xmath617 spanning forests that contain @xmath339 . therefore , @xmath618 now by grouping terms on the rhs by isomorphism classes of @xmath615 , we obtain @xmath619 since @xmath620 for all undirected @xmath27-forests @xmath621 , we also have @xmath622 for all undirected @xmath27-forests @xmath102 . the examples of non - reconstructible pedigrees given in @xcite can be distinguished from the probability distribution on the space of alignments under model rm . pedigrees in these examples do not have the same number of @xmath27-forests of each type . for example , one pedigree in each pair contains a common ancestor of all extant vertices while the other does not . this may also be observed in the examples in figure [ fig - k3 ] . but knowing the number of @xmath27-forests of each type is not always enough to distinguish pedigrees , as verified by pedigrees @xmath245 and @xmath243 shown in figure [ fig - hap ] . they both have the same number of directed ( therefore , also undirected ) @xmath27-forests of each type : they have 4 directed @xmath27-forests in which 1 and 2 have a common ancestor . we denote their undirected @xmath27-forest by @xmath282 , which is a path of length 4 with end vertices 1 and 2 . there are 12 directed @xmath27-forests in which 1 and 2 do not have a common ancestor . we denote their undirected @xmath27-forest by @xmath283 , which consists of two isolated vertices 1 and 2 . moreover , @xmath623 and @xmath624 . also , for @xmath625 ( and for @xmath626 ) , we check by direct counting that @xmath627 . but @xmath245 and @xmath243 can nevertheless be distinguished as shown below . [ cor - pq ] pedigrees @xmath245 and @xmath243 in figure [ fig - hap ] are distinguished from the probability distribution on the space of alignments under model rm provided the crossover probability is sufficiently small . let @xmath282 and @xmath283 be the @xmath27-forests as described above . we apply theorem [ thm - paths ] for @xmath628 with @xmath629 . we can verify that @xmath630 and @xmath631 . therefore , @xmath245 and @xmath243 give different distributions on the space of alignments . in the following , we describe how @xmath632 and @xmath633 are counted . let @xmath634 denote the directed @xmath27-forests in @xmath245 and @xmath243 consisting of paths @xmath635 and @xmath636 . similarly , we write @xmath637 for other directed @xmath27-forests in @xmath245 and @xmath243 , rooted at @xmath8 , @xmath353 and @xmath114 , respectively . these are the 4 directed @xmath27-forests that have @xmath282 as their undirected @xmath27-forest . all other directed @xmath27-forests have @xmath283 as their undirected @xmath27-forest . _ counting _ @xmath638 : since @xmath639 , we count 16 different contributions to @xmath638 depending on the choices for @xmath640 and @xmath641 in @xmath642 . please refer to figure [ fig - p ] . when @xmath643 and @xmath644 ( left ) , and when @xmath643 and @xmath645 ( right),width=445 ] when @xmath643 and @xmath646 : this is possible only if @xmath647 . there are 4 choices of @xmath40 , and for each of them there are 4 choices for @xmath41 depending on which arc of @xmath634 is replaced to obtain @xmath41 . thus we have 16 sequences @xmath136 for which @xmath648 and @xmath646 . when @xmath643 and @xmath644 : the dashed arcs @xmath649 and @xmath650 in figure [ fig - p ] on the left must be obtained by replacing @xmath651 and @xmath652 , and there are two sequences @xmath136 that achieve this : either @xmath653 or @xmath654 . also , there are 4 possible ways to include arcs pointing to @xmath0 and @xmath655 in @xmath40 . therefore , there are 8 sequences @xmath136 such that @xmath643 and @xmath644 . when @xmath643 and @xmath645 : there are only 2 sequences @xmath136 for which this is possible , since the arcs @xmath656 and @xmath657 ( shown in bold in figure [ fig - p ] on the right ) must already be in @xmath40 . when @xmath643 and @xmath658 : this case is similar to the case @xmath645 . thus there are 28 choices of @xmath136 such that @xmath643 , and similarly 28 choices each for @xmath659 , @xmath660 , and @xmath661 . therefore , there are 112 sequences @xmath136 such that @xmath662 . _ counting _ @xmath633 : again we count 16 different contributions to @xmath633 depending on the choices for @xmath640 and @xmath641 in @xmath663 . please refer to figure [ fig - q ] . when @xmath664 and @xmath644 ( left ) , and when @xmath643 and @xmath658 ( right),width=445 ] when @xmath643 and @xmath646 : as in case of @xmath245 , we have 16 sequences @xmath136 for which @xmath643 and @xmath646 . when @xmath643 and @xmath644 : the dashed arcs @xmath650 and @xmath665 in figure [ fig - q ] on the left must be obtained by replacing @xmath652 and @xmath666 , and arc @xmath667 must already be in @xmath40 . there are two sequences @xmath136 that achieve this : either @xmath653 or @xmath668 . also , there are 2 choices for arcs pointing to @xmath0 , therefore , there are 2 choices for @xmath40 . therefore , there are 4 sequences @xmath136 such that @xmath643 and @xmath669 . when @xmath643 and @xmath645 : this case is similar to the case in which @xmath644 , therefore , there are 4 sequences such that @xmath643 and @xmath645 . when @xmath643 and @xmath658 : in this case , the arcs @xmath670 and @xmath671 must already be in @xmath40 . thus there are only two sequences @xmath136 that are counted depending on whether @xmath672 or @xmath668 . thus there are 26 choices of @xmath136 such that @xmath643 , and similarly 26 choices each for @xmath659 , @xmath660 , and @xmath661 . therefore , there are 104 sequences @xmath136 such that @xmath662 . thus we have verified that @xmath638 and @xmath633 are unequal for @xmath673 , implying that @xmath245 and @xmath243 are distinguished by the probability distribution they induce on the extant sequences under model rm . in this paper we have presented a rigorous mathematical framework for studying pedigree reconstruction problems under probabilistic models . we extended phylogenetic identifiability results to reconstruct pedigrees under an idealised model of recombination and mutation . the main result of this paper is the computation of a class of combinatorial invariants from the joint distribution of extant sequences . as a corollary , we were able to show that certain known examples of pedigrees that could not be distinguished from path lengths alone , could be distinguished by a more detailed analysis of subgraph sequences . here we identify a few open problems and directions for future investigation . _ identifiability of computationally tractable invariants : _ the invariants of a pedigree @xmath58 defined by @xmath434 may be quite difficult to apply in general . theorem [ thm - paths ] may be difficult to use computationally for bigger pedigrees . even for other reconstruction problems in graph theory , computational verifications are difficult . for example , ulam s reconstruction conjecture has been computationally verified to be true only for graphs on at most 11 vertices @xcite . even on restricted classes of graphs , computational reconstruction experiments are difficult to perform . it will be useful to derive from @xmath674 ( or independently ) other identifiable invariants that may be easier to use in computational experiments . theorem [ thm - paths ] only states that if two pedigrees induce the same joint distribution on extant sequences under model rm , then they agree on the invariants @xmath675 . but it will be important to prove a converse or a result of the type : two pedigrees induce the same distribution on extant sequences under model rm if and only if they take the same value for a class of combinatorial invariants . such a result would reduce the identifiability problem to a purely combinatorial problem of proving or disproving that the class of invariants is complete . such a class of invariants may be @xmath676 or it may be somewhat stronger than @xmath677 . it may be possible to compute , by the methods of theorem [ thm - paths ] , other invariants , for example , spanning forest sequences in which consecutive spanning forests are not necessarily separated by just one recombination . _ improving the bound on @xmath145 : _ in the lower bound on @xmath527 in equation ( [ eq - term1b ] ) , we have @xmath398 in the denominator , while @xmath678 is roughly @xmath679 , where @xmath4 is the number of arcs in a pedigree . therefore , it is possible to obtain better bounds on @xmath145 for the applicability of the main theorem for pedigrees with fewer arcs . therefore , we would significantly improve the upper bound on @xmath145 if we showed that a pedigree can be reconstructed from the collection of its subpedigrees ( pedigrees of subsets of the extant population ) of order @xmath10 for some small @xmath10 . in @xcite we made a conjecture about how small @xmath10 may be for pedigrees of order @xmath680 in which the population remains constant over generations . thus solving the purely combinatorial problem of reconstructing pedigrees from their subpedigrees , while important in its own right , will be useful for improving bounds on @xmath145 in theorem [ thm - paths ] . on the other hand it is also likely ( although we do not conjecture ) that theorem [ thm - paths ] is valid without restrictions on @xmath145 , or for all but finitely many values of @xmath145 , or for all @xmath145 except when @xmath223 takes small values . but we do not have good intuition as to why proposition [ prop - n2 - 2 ] ( with @xmath266 ) requires a more complicated argument and an upper bound on @xmath145 . _ maximum likelihood computation of the invariants @xmath681 _ : it will be of interest to derive statistical consistency results and bounds on sequence lengths , analogous to corollary [ cor - consistency ] and lemma [ lem - eps - l1 ] , for computing the invariants @xmath675 . for example , we would like to make the following qualitative statement precise . suppose @xmath58 is a pedigree with @xmath4 arcs . let @xmath607 be given . suppose @xmath474 is an undirected @xmath27-forest sequence of length @xmath269 with no two consecutive @xmath27-forests isomorphic . then there is a sufficiently large @xmath682 such that if a collection of sequences of length @xmath683 evolved on @xmath58 giving an alignment @xmath91 , then the likelihood ratio @xmath684 is large for all pedigrees @xmath60 such that @xmath685 . we expect that @xmath682 would be of the order of @xmath686 . i would like to thank jotun hein and mike steel for many valuable comments and for pointing me to many useful references relevant to this work . in particular , mike s comments on and references to the literature on phylogenetic identifiability results and results on the consistency of ml were very useful . this project was funded by the grant `` from complete genomes to global pedigrees '' from epsrc ( engineering and physical sciences research council ) , u.k . i would like to thank jotun hein and epsrc for supporting my postdoctoral stay at the university of oxford under this grant . i am currently working on the project `` tree - graphs and incidence matrices : theory and applications '' funded by cnpq , brasil ( processo : 151782/2010 - 5 ) at the instituto de matemtica e estatstica , universidade de so paulo , where i completed a significant revision of this paper and improvements on the bounds in the main theorem . i would also like to thank yoshiharu kohayakawa with whom i had very useful conversations during the revision of this paper . guy bresler , elchanan mossel , and allan sly . reconstruction of markov random fields from samples : some observations and algorithms . in a. goel _ et al . _ , editor , _ 11th international workshop , approx 2008 , and 12th international workshop , random 2008 , lncs 5171 _ , pages 343356 , berlin heidelberg , 2008 . springer - verlag .

intense laser - matter interaction provides a route to create energetic particles ( e.g. , electrons , ions , and photons ) using table - top equipment . clusters , possessing the transparency of gas targets and the high charge density of solid targets , proved to be very efficient absorbers of laser light . their small size , compared to laser wavelength and skin depth , avoids reflection of the laser beam at the cluster surface as well as the loss of hot electrons into the cold bulk . in fact , almost 100% absorption of the laser light was reported in experiments with rare gas clusters @xcite . useful reviews on the subject are refs . @xcite . upon irradiation of the rare gas clusters by intense laser light , electrons first absorb energy and leave their `` parent '' ions . this is known as inner ionization , meaning that the electrons are still bound to the cluster but not necessarily to their `` parent '' ions . the total electric field ( i.e. , laser plus space charge field ) inside the cluster leads to inner ionization up to high charge states not possible with the laser field alone ( ionization ignition @xcite ) . as the laser intensity during the pulse increases , these electrons absorb energy from the laser field and may leave the cluster leading to the positive charging of the cluster known as outer ionization . thus outer ionization leads to a non - neutral plasma . with the increasing outer ionization , the restoring force of the ions counteracts ionization ignition so that the latter stops at some point . the net positive charge left behind finally explodes due to the coulomb repulsion and hydrodynamic pressure , leading to the conversion of electron energy into ion energy . typically mev ions and kev electrons @xcite are measured in experiments . it is clear from the described scenario that the understanding of the relevant mechanism(s ) of laser energy absorption leading to the heating of cluster electrons and outer ionization is of great importance for the development of a complete theoretical description . laser energy absorption by electrons proceeds either through resonances ( linear or nonlinear ) or through non - adiabaticities ( all possible types of collisions ) . all of these processes lead to dephasing of the current density with respect to the laser field , which , according to poynting s theorem , is a prerequisite for absorption . collisional absorption via collisions of electrons with ions are of minor importance at near infrared wavelengths @xmath0 nm or greater @xcite whereas it is the dominant absorption mechanism at short wavelengths @xcite , not studied in this paper . the finite size of the clusters suggests that `` collisions with the cluster boundary '' may be responsible for the energy absorption . however , this viewpoint is misleading , as will be shown in sec . [ sec2b ] . during the expansion of the ionic core , the decreasing charge density @xmath1 leads to the decrease of the mie plasma frequency , @xmath2 ( atomic units @xmath3 are used unless noted otherwise ) . for very short near infrared laser pulses @xmath4 can not meet the linear resonance ( t)=[lr ] , unless the cluster has sufficiently expanded ( typically after a few hundred femtoseconds ) . linear resonance , well understood in theory , experiments , and simulations @xcite , is thus ruled out for very short pulses or during the early cycles of a long pulse laser - cluster interaction where ion motion is negligible . in this case , nonlinear resonance ( nlr ) , whose origin lies in the anharmonicity of the cluster potential , turns out to be the dominant collisionless absorption mechanism . in fact , for very short linearly polarized ( lp ) laser pulses , it was clearly shown @xcite that essentially _ all _ electrons that contribute to outer ionization pass through the nlr , which was unequivocally identified as _ the _ collisionless absorption mechanism in the absence of linear resonance . the eigenfrequency @xmath5 $ ] of a ( laser- ) driven oscillator in an anharmonic potential , being dependent on its excursion amplitude @xmath6 ( or the energy ) , may dynamically meet the nlr = . [ nlr ] due to the many body nature of the interaction , the identification , the separation , and the interpretation of the absorption mechanisms in molecular dynamics or particle - in - cell ( pic ) simulations are often difficult . recently , a method of identification of the nlr in many - body simulations of rare gas clusters has been proposed @xcite . the possible importance of nlr was also mentioned or discussed previously @xcite . the rigid sphere model ( rsm ) @xcite where electrons and ions are modeled by homogeneously charged rigid spheres oscillating against each other is clearly an oversimplification of a real many - particle system such as a cluster . however , it proves useful for estimating the order of magnitude of the absorbed energy as well as for the calculation of the laser intensity where energy absorption is most efficient , as will be shown in the present work . moreover , it provides physical understanding and clearly displays nlr @xcite . the heating of cluster electrons in circularly polarized laser fields has not yet received much attention , at least theoretically . experiments with rare - gas clusters show almost no effect on the x - ray emission @xcite and ion energy distribution @xcite when laser light of different ellipticity is used . theoretically , circular polarization is particularly interesting because the above mentioned `` collisions with the cluster boundary '' are strongly suppressed in this case . hence one may expect energy absorption being less efficient . nlr , on the other hand , occurs in both cases , and , in fact , the energy absorption turns out to be equally efficient . moreover , the absorbed energy compares well with the rsm predictions . the outline of the present paper is as follows : in sec . [ sec2a ] we briefly review the nlr and the rsm . in sec . [ sec2b ] the rsm is extended to circular polarization , where nlr is observed as well . in sec . [ sec2c ] the rsm threshold intensities are calculated . section [ sec3 ] is devoted to the pic @xcite results for both linear and circular polarization . finally , we summarize our results in sec . [ sec4 ] . throughout this paper we use @xmath7-cycle laser pulses of near infrared wavelength @xmath8 nm and a fixed cluster radius @xmath9 nm unless stated otherwise . nlr is a robust phenomenon that qualitatively is insensitive to cluster and laser parameters . the ion motion can be neglected in the study of energy absorption in very short laser pulses . thus the ions just form a static , positively charged background of spherical shape . for not too high laser intensity , the collective motion of the electrons can also be approximated by a homogeneous , rigid sphere of negative charge . in the simplest case the radii of ion and electron sphere are assumed to be equal . in a more realistic model the electron cloud expands @xcite . however , the method of identification of the nlr used in this paper is independent of the degree of electronic expansion . the center of mass of the electron - ion system is , in good approximation , located at the center of the ion sphere . in an oscillating laser field , the homogeneously charged electron sphere oscillates back and forth against the positively charged ion sphere . the equation of motion of the electron center of mass in a lp laser field , polarized along @xmath10 , can be written as + g(r ) = - , [ rsm_eom ] where @xmath11 is the excursion of the electron sphere , normalized to the cluster radius @xmath12 , @xmath13 , @xmath14 is the normalized time in units of the laser period , and the dimensionless electrostatic restoring force @xmath15 is g(r ) = ( ) ^2 \ { ll r-+ & 0r2 + & r2 + . . [ gr ] the first term in the upper line of is the linear force when the displacement of the electron sphere is small , the next two terms are the nonlinear terms which appear due to the partial overlap of the electron cloud with the ion cloud . the term @xmath16 in the lower line of is the coulomb force between the separated electron sphere and ion sphere . the quantity on the right hand side of is the normalized driver strength , which is the quiver amplitude @xmath17 of a free electron in the laser field @xmath18 normalized to the cluster radius @xmath12 . , a @xmath19 cycle laser field @xmath20 , and cluster radius @xmath21 nm . within a narrow field strength interval ( here @xmath22@xmath23 ) the absorbed energy per particle ( solid line ) increases by many orders of magnitude . the dashed line represents the absorbed energy by a purely harmonic oscillator driven by the same laser field . [ rsmi],scaledwidth=50.0% ] earlier work @xcite showed that absorption of laser energy in the rsm is characterized by a threshold driver strength below which absorption is negligible ( harmonic regime ) and above which absorption is almost constant . figure [ rsmi ] shows this threshold behavior for @xmath24 and a @xmath19-cycle @xmath25-pulse @xmath20 for @xmath26 . the dashed line is the absorption corresponding to a driven , purely harmonic oscillator + ( ) ^2 = - [ harmonicrsm ] in the laser field @xmath27 . the laser energy absorbed by a single electron in a @xmath28-cycle laser pulse of period @xmath29 is = - _ 0^n t v(t)e_(t ) dt . [ introdephase ] solving analytically for the velocity @xmath30 and integrating one finds for the absorbed energy per electron @xcite e_0 ^ 2 [ introabseng ] where @xmath31 ^ 2 [ \omegamie^2 - ( 1 - 1/n)^2 \omegalaser^2]^2 \simeq ( \omegamie^2 - \omegalaser^2)^4 $ ] ( for sufficiently large @xmath28 ) was used . the analytical estimate is plotted in fig . [ rsmi ] together with the absorbed energy obtained from the numerical solution of . one sees that the absorbed energy jumps by many orders of magnitude to a higher value after crossing a threshold driver strength . the higher the cluster charge density is , the higher is this threshold driver strength . the rigid sphere model shows this behavior of efficient absorption above the threshold driver strength at all cluster charge densities independent of the linear resonance condition @xmath32 , contrary to the nanoplasma model @xcite . since the rigid sphere model does not necessarily requires expansion of the cluster for the efficient absorption of laser energy , it permits us to understand the behavior of energy absorption and the underlying mechanism for very short laser pulses . equation can be formally rewritten as + ^2 = - . [ nlho ] equation yields the instantaneous , scaled effective frequency squared ^2= = [ efffrequ ] which passes through unity at the nlr . the right hand side of is the restoring force divided by the excursion of the electronic cloud , which in the case of harmonic motion would be the square of the characteristic frequency . ( red , upper solid line ) vs laser cycles above the threshold driver strength for @xmath33 . here we take @xmath34 corresponding to a laser intensity @xmath35 , a @xmath19-cycle @xmath25-pulse @xmath20 of wavelength @xmath8 nm , and a cluster radius @xmath9 nm . excursion @xmath36 ( green , middle solid line ) and energy of the electron sphere @xmath37 ( blue , lower solid line ) are included in the plot . outer ionization ( i.e. , @xmath38 ) and occurrence of nlr @xmath39 ^ 2=1 $ ] always coincide ( dashed vertical line ) . [ rsmii],scaledwidth=50.0% ] figure [ rsmii ] shows a typical example of the temporal behavior of @xmath39 ^ 2 $ ] above the threshold driver strength for @xmath33 . since @xmath40 , @xmath39 ^ 2 $ ] starts at this value and drops with increasing driver strength . it passes through unity at the time indicated by the vertical line , and it is exactly at that time where the electron sphere becomes free ( outer ionization ) . this incidence is clearly visible from the energy of the electron sphere , which passes through zero and the excursion as well . outer ionization and occurrence of nlr happens for all driver strengths above the threshold whereas the resonance is never met below the threshold . since the amplitude of the excursion of the electronic sphere depends upon the driver strength , the excursion amplitude should also be large enough so that the nlr is passed . the decrease in the effective frequency with the increase of the amplitude of excursion of the electronic sphere in the force field can be understood by analyzing its motion in the corresponding anharmonic potential v(r ) = ^2 r^2 \ { ll -+ & 0r2 + - & r2 . + . [ rsm_pot]the period @xmath29 of oscillation of the electronic sphere in the potential @xmath41 can be approximated by a perturbation series @xcite as @xmath42^k\ , \diff r}{\sqrt{e_\mathrm{tot } - v_0(r)}}. \label{timeperiod}\ ] ] here , @xmath43 is the total energy of the electronic sphere in the ionic field , @xmath44 is the harmonic oscillator potential and @xmath45 is the perturbation to the potential with @xmath46 and @xmath47 . the effective frequency is then @xmath48 . for the excursion @xmath49 and cluster radius @xmath9 nm ( @xmath50a.u . ) we can consider @xmath51 and the approximate potential @xmath52 . corrections up to @xmath53 yield @xmath54 . the unperturbed period is @xmath55 and the successive corrections are @xmath56 , @xmath57 , @xmath58 , @xmath59 , @xmath60 , @xmath61 with @xmath62 , @xmath63 , @xmath64 , @xmath65 , @xmath66 and @xmath67 , respectively . neglecting the higher order term of @xmath15 for @xmath68 , a simpler approximation to the effective frequency can be derived from by replacing the excursion @xmath69 with the excursion amplitude @xmath70 : _ ( ) ( 1 - /16)^1/2 ( 1 - /32 ) . [ rsm_freq2 ] vs the excursion amplitude @xmath70 of the electronic sphere for different cluster charge densities @xmath71@xmath72 . the solid lines are computed from the numerical solution of . the dashed lines are the corresponding analytical approximations from for @xmath73 up to 6 , and the dashed - dotted lines are from . for the charge density @xmath74 one expects nlr ( horizontal dashed line ) to occur at the excursion @xmath75 . [ rsmiii],scaledwidth=50.0% ] figure [ rsmiii ] shows the effective frequency vs the excursion amplitude @xmath70 of the electronic sphere in the potential for various cluster charge densities @xmath76@xmath72 as calculated from the numerical solution of together with the approximations and . the effective frequency as calculated from shows good agreement below the excursion @xmath77 and low charge densities ( e.g. , @xmath78 ) . for higher charge densities more corrections [ large number of @xmath73 values in ] are needed , which are very much cumbersome to calculate . although fits well with the numerical solution , @xmath79 becomes negative when @xmath80 . however , the results using agree with the exact ones in fig . [ rsmiii ] up to @xmath81 for the excursions of interest , i.e. , up to the point where the electronic sphere undergoes nlr . the variation of frequency with the excursion amplitude as shown in fig . [ rsmiii ] explains why a threshold driver strength is required for an appreciable laser energy absorption as well as for the crossing of the nlr in the fig . [ rsmii ] : only a driver exceeding a certain threshold field strength will lead to excursions compatible with the nlr condition @xmath82 . clusters in a circularly polarized ( cp ) laser field received less attention in the literature . it is not known _ a priori _ how the outer ionization and energy absorption by clusters depend on the laser polarization . in laser - atom interaction the laser polarization has dramatic effects : since in cp the free electrons do not return to their parent atom all the atomic effects relying on rescattering such as high - order harmonic generation , high - order above - threshold ionization , and nonsequential ionization are strongly suppressed . in the context of clusters , the study of the absorption efficiency as a function of the laser polarization can help to discriminate among different absorption mechanisms . for instance , if laser energy absorption was indeed due to `` collisions with the cluster boundary '' it would be suppressed in cp because the electrons mainly swirl around parallel to the `` cluster boundary '' rather than crossing ( and hence colliding ) with it . however , as we will show , the absorption of laser energy is largely independent of the laser polarization , thus ruling out `` collisions with the cluster boundary '' as a meaningful absorption mechanism . let us first extend the rsm to cp . in a cp laser field with electric field components in @xmath10- and @xmath83-direction , the equation of motion for the electronic sphere in the rigid sphere approximation of a cluster can be written as \ { c _ x + _ y } + \ { cr_x + r_y } = - \ { c e_^x ( ) + e_^y ( ) } . [ rsm_eomcpl]here , @xmath84 , @xmath85 , @xmath86 , and @xmath87 one identifies @xmath88 and @xmath89 as the two components of the restoring force in . note that we have divided the electric field components by a factor @xmath90 so that the ponderomotive energy @xmath91 , ( i.e. , the time - averaged quiver energy of a free electron in the laser field ) is the same as in the lp case with the same @xmath92 ( otherwise @xmath93 would be a factor of two higher in the cp case ) . the square of the effective , time - dependent oscillator frequency in the cp laser field can be written as ^2 = = [ efffrequcpl ] , which has the same form as in the lp case . ( green , top left solid line ) vs laser cycles above the threshold intensity @xmath94 for circular polarization , @xmath33 , and cluster radius @xmath9 nm . here , @xmath95 , corresponding to a laser intensity @xmath96 , @xmath19-cycle @xmath25-pulse with components @xmath97 , @xmath98 , and wavelength @xmath8 nm . excursion @xmath69 ( blue , middle left solid line ) and energy @xmath37 ( red , bottom left solid line ) of the electron sphere are also plotted . the energy @xmath37 is scaled down by a factor 10 to display within the excursion and the frequency range . outer ionization ( i.e. , @xmath38 ) and occurrence of nlr @xmath39 ^ 2=1 $ ] always coincide ( dashed vertical ) . [ rsmiicpl],scaledwidth=50.0% ] earlier , in fig . [ rsmii ] , it was shown that nlr and outer ionization in the rsm only occur when a threshold laser intensity is crossed . the same is true for the occurrence of nlr with cp light . figure [ rsmiicpl ] shows the temporal behavior of @xmath99 ^ 2 $ ] above the threshold driver strength for a @xmath19-cycle cp laser pulse of wavelength @xmath8 nm . here the cluster charge density is @xmath72 times over - critical , i.e. , @xmath40 at which @xmath100 starts and drops with increasing driver field during the pulse . the nlr @xmath101 ^ 2=1 $ ] is passed at the time indicated by the vertical line . as in the lp case in fig . [ rsmii ] , the electron sphere is set free at the time the nlr is passed : the energy of the electron sphere passes through zero , and the excursion sharply increases to a high value . it is also clear from fig . [ rsmiicpl ] , that once the electronic sphere is set free , the frequency drops to zero , and the total absorbed energy remains positive . a zero effective frequency implies an infinite period , i.e , the electron sphere does not return to the ion sphere . the main difference between fig . [ rsmiicpl ] and fig . [ rsmii ] is that in the case of the cp laser field the decrease of the effective frequency is smooth ( i.e. , no oscillations ) since the electric field vector rotates but its absolute value remains constant . as a consequence , the electron sphere spirals out , staying away from the potential center where @xmath102 . for lp instead , the electron sphere is driven through the origin and hence the effective frequency undergoes oscillations before it drops to the resonance value , as visible in fig . [ rsmii ] . nlr is clearly identified in both cases . nlr occurs above a threshold driver strength . beyond this driver strength the rigid electron sphere gains laser energy which is many order of magnitude higher than below the threshold ( see fig . [ rsmi ] ) . in an open potential such as the electron sphere is detached from the ion sphere above the threshold driver strength , i.e. , appreciable energy absorption and outer ionization occur simultaneously . we would like to remark that in closed potentials nlr occurs as well , as discussed in ref . @xcite . the threshold driver strength can be estimated . the dimensionless potential @xmath103 of the electron sphere in the ionic field can be written as @xmath104 application of a static electric field @xmath105 ( corresponding to the peak field strength of a low - frequency laser field ) , suppresses the potential in one direction by the amount @xmath106 . the effective potential seen by the electron sphere is @xmath107 with @xmath108 . the potential barrier vanishes if @xmath109 and @xmath110 , leading to @xmath111 , and the nlr threshold intensity is estimated to be for linearly ( lp ) and circularly ( cp ) polarized laser fields . results from the full numerical solution of with lp ( lower solid ) , cp ( top solid ) , the vanishing barrier approximation ( vba - lp , middle solid ) , and the vba corrected for cp ( upper dashed ) are shown ( see text for a discussion ) . the over - the - barrier approximation ( oba ) ( lower dashed ) underestimates the exact threshold intensities . [ rsmpert2],scaledwidth=50.0% ] @xmath112 we call this the vanishing - barrier - approximation ( vba ) . in atomic ionization the so - called over - the - barrier approximation ( oba ) or bethe - rule @xcite allows to estimate at which electric field strength a certain atomic charge state dominates . if one applies the oba to the rsm one obtains the two equations @xmath113 and @xmath114 with @xmath115 the barrier location . one finds @xmath116 . this gives the oba threshold intensity of the nlr @xmath117 which underestimates the numerically determined threshold intensity , as we will show now . figure [ rsmpert2 ] shows the threshold intensity as a function of the cluster charge density @xmath71@xmath118 ( corresponding to average charge states @xmath119 for xenon ) . the numerically determined threshold intensities for lp and cp laser light show that when the cluster charge density is low , the nlr occurs almost at the same value of the threshold intensity , irrespective of the polarization . as the charge density increases the nlr threshold intensity appears to be higher for cp than for lp . the vba of the threshold intensity is in good agreement with the numerical result for lp whereas the oba underestimates it . this fact might be related to the recently observed `` enhanced saturation intensities '' in the ionization of finite size systems such as c@xmath120 ( see , e.g. , @xcite and references therein ) , indicating that the latter might neither be a many - electron nor a quantum effect but just due to the finite size of the target . the difference of the threshold intensities for lp and cp is due to the definition of the cp field ( [ ex ] ) , ( [ ey ] ) where a factor @xmath121 has been introduced in order to render the ponderomotive potential equal for lp and cp . however , for the threshold intensity it is the electric field ( or the intensity ) that matters , not @xmath122 . for a given @xmath105 the laser intensity is @xmath123 in the lp case but only @xmath124 for cp . therefore , the upper black , dashed line in fig . [ rsmpert2 ] shows the vba threshold intensity multiplied by a factor of two , which is in good agreement with the numerical results for the cp laser field at higher charge densities . so far we have studied the nlr absorption of laser energy in a simplified model system assuming an anharmonic potential generated by the ions in which the homogeneous and rigid electron cloud moves . in reality , the potential builds up during the interaction with the laser pulse because of ionization . the delicate interplay of inner ionization , energy absorption by various mechanisms , and outer ionization can be simulated using methods such as pic or molecular dynamics . previous work @xcite , studying lp short laser pulses , showed that nlr can be clearly identified in such simulations as well . in this section we present results obtained from three - dimensional pic simulations . the cluster is exposed to @xmath19-cycle @xmath25-pulses @xmath125 of near infrared wavelength @xmath126 nm , i.e. , the total pulse duration is @xmath127 . since motion of ions does not play an important role during the entire pulse the ions are assumed fixed , which ensures a well defined , constant mie frequency @xmath128 . a pic electron has the same charge to mass ratio as a `` real '' electron , that is , @xmath129 in atomic units . each pic electron moves under the influence of the external laser field and the space charge field @xmath130 due to the potential @xmath131 that is created by all charges ( mapped to the numerical grid ) . hence the equation of motion of the @xmath132-th pic electron is _ i + ( _ i , t ) = -_(t).[eqpic ] equation is solved self - consistently for all pic electrons . clearly , @xmath133 depends on the position of _ all _ other particles @xmath134 as well . the pic simulation starts with the neutral cluster configuration so that @xmath135 . vs laser intensity for charge density @xmath72 times the critical density and a xe@xmath136 cluster ( @xmath137 electrons , mean ionic charge state @xmath138 , and a cluster radius @xmath9 nm ) . solid lines are pic results , dashed lines are the rsm results for a @xmath19 cycle pulse @xmath125 of wavelength @xmath139 nm . [ picrsm1],scaledwidth=50.0% ] figure [ picrsm1 ] shows pic results for the kinetic energy ( ke ) and the total energy ( te ) absorbed per electron in units of @xmath122 , vs the peak laser intensity . one sees that the absorbed energy per electron is on the order of @xmath122 . however , the absorbed energy is nonlinear in @xmath122 and displays a maximum before it drops because of the saturation of outer ionization . the depletion is due to the fact that at a given intensity most ( if not all ) electrons are removed from the cluster ( complete outer ionization ) . further inner ionization would be required to generate `` fresh '' electrons that could continue to absorb energy . the maxima in the pic absorption curves are located close to the threshold intensity predicted by the rsm . the total absorbed energy around the maximum of the pic absorption curves ( rsm as well ) is on the order of @xmath140@xmath141kev which has been also reported in experiments of intense laser clusters interactions @xcite . in pic simulations no sharp intensity threshold exists since each pic electron sees its own time - dependent field ( space charge field plus the laser field ) . therefore sharp jumps ( as seen in the rsm , e.g. , fig . [ rsmi ] ) are absent . figure [ picrsm2 ] compares the kinetic and potential energies of a single pic electron and the electronic sphere in the rsm . the energies are plotted vs the excursion @xmath36 ( along the direction of the laser field ) . the peak laser intensity is @xmath142 w/@xmath143 . we have checked that there are many pic electrons that leave the cluster similarly to the one presented in fig . [ picrsm2 ] . the potential and the kinetic energy gained by the electron sphere in the rsm is divided by the total number of pic electrons so that a quantitative comparison is possible . the main difference between pic and rsm results is that in the pic simulation all the electrons see initially a zero potential since the potential builds up from the neutral cluster configuration in the course of the interaction with the laser field while in the rsm the electronic sphere oscillates in a prescribed potential . those pic electrons that stay long inside the cluster experience almost the full ionic potential which then is similar to the rsm potential . in the laser polarization direction . the escaping pic electron resembles the equivalent electronic particle in the rsm . peak laser intensity @xmath142 w/@xmath143 , charge density is @xmath144 . other parameters as in fig . [ picrsm1 ] . [ picrsm2],scaledwidth=50.0% ] as mentioned earlier , the absorption of energy by a pic electron depends upon the self - consistent potential which develops during the laser pulse due to outer ionization . as a result different pic electrons move along different trajectories , `` see '' a different potential , and thus are set free at different times . from the pic simulation vs the excursion @xmath36 in laser polarization direction . yellow ( solid gray ) lines represent cuts of the potential @xmath145 at time @xmath146 for different @xmath83 and @xmath147 . the circles ( red ) represent the total energy @xmath148 of individual pic electrons located within the simulation box at that time . the dashed lines represent the kinetic ( green , dashed gray ) energy ( ke ) and potential ( dashed black ) energy ( pe ) of a pic electron that is outer ionized when the excursion @xmath149 meets the condition @xmath150 at the same time ( see fig . [ rsmiii ] for the charge density @xmath74 ) . the peak laser intensity is @xmath142 w/@xmath143 . other parameters as in fig . [ picrsm2 ] . [ picpot],scaledwidth=50.0% ] figure [ picpot ] shows a snapshot of the collective potential @xmath131 at time @xmath146 for a cut at @xmath147 ( and various @xmath83 throughout the cluster ) . the most lower curve represents the potential for @xmath151 . the red circles represent the total energy @xmath152 of individual pic electrons located within the simulation box at that time . one observes that many pic electrons are accumulated near the bottom - left of the potential well at this time . these pic electrons with @xmath153 remain bound since the laser field amplitude is already decreasing from the 4th cycle onward . the kinetic and the total energy of a pic electron which leaves the potential well when its excursion becomes @xmath149 at the same time is also shown . this excursion approximately satisfies the nlr condition @xmath154 for the charge density @xmath155 , as identified in fig . [ rsmiii ] with the rsm analysis . the nlr behavior exhibited by the pic electron in fig . [ picpot ] is not accidental . for the sake of an unequivocal and explicit identification of the nlr we now analyze the motion of all individual pic electrons in the same way as it has been done with the motion of the electron sphere in the rsm in sec . recalling , the equation for the effective , time - dependent oscillator frequency analogous to for the @xmath132-th pic electron reads ^2_,i(t ) = -= [ efffrequpic ] . earlier we have mentioned that @xmath133 depends on the position of _ all _ other particles @xmath134 and the simulation starts with the charge neutral cluster configuration i.e. , @xmath135 . hence , a pic electron `` sees '' initially an effective frequency @xmath156 . the laser field disturbs the charge equilibrium and @xmath157 becomes different from zero . @xmath157 may be even negative in regions of accumulated electron density ( repulsive potential ) . as the cluster charges up , @xmath158 quickly increases beyond unity ( where the rsm starts in the first place ) . the start from @xmath156 , the possibility of negative @xmath157 , and the three - dimensionality are the main differences to the rsm analysis above . , excursion @xmath159 , and the total energy @xmath152 for the pic electron of fig . [ picrsm2 ] vs time in laser cycles . the total energy becomes positive only when the nlr is crossed ( indicated by the vertical dashed line ) . this result resembles the rsm result in fig . [ rsmii ] when nlr is met . the charge density is @xmath74 , the peak laser intensity is @xmath142 w/@xmath143 . other parameters as in fig . [ picrsm2 ] . [ timevsomega ] , scaledwidth=50.0% ] figure [ timevsomega ] shows the effective frequency squared , the total energy @xmath160 , and the excursion @xmath159 vs time for the pic electron whose energy history is shown in fig . [ picrsm2 ] . we define the time when , for a particular electron , @xmath161 becomes @xmath162 as the ionization time of that electron . data are plotted shortly before the emission of this particle from the cluster potential . it is clearly visible in fig . [ timevsomega ] that the pic electron is escaped only when the resonance line @xmath163 is passed . figure [ timevsomega ] can be well compared with fig . [ rsmii ] showing ionization of the rsm via nlr . in the case of pic simulations the fulfillment of the nonlinear resonance condition @xmath163 is necessary but not sufficient for ionization . as the potential builds up , the pic electrons transiently meet the nlr condition , and , in fact , some electrons leave the cluster at that early stage when the potential is still shallow and the laser field is relatively weak . however , as the potential deepens , pic electrons `` dropping '' below the energy necessary for nlr to occur , behave from then on similar to the rsm and may finally escape only by climbing up in the potential and hitting the nlr @xmath164 . during the 0.4 laser cycles plotted in fig . [ timevsomega ] the @xmath158-curve displays artificial , short - time scale fluctuations inherent in pic simulations @xcite . however , we checked that macroscopic observables such as the absorbed energy or the degree of outer ionization are well converged . , ( b ) @xmath165 , ( c ) @xmath166 , ( d ) @xmath167 , ( e ) @xmath168 , ( f ) @xmath169 , ( g ) @xmath170 , and ( h ) @xmath171 laser cycles for lp , laser intensity @xmath172 , and @xmath33 . other parameters as in fig . [ timevsomega ] . the radial positions ( in units of @xmath12 ) are color - coded . electrons become free upon crossing the nlr , i.e. , @xmath173 . [ picweffvsenglp],scaledwidth=55.0% ] by following the dynamics of the electrons in the effective frequency vs energy - plane we identify the main pathway to outer ionization and efficient absorption . figure [ picweffvsenglp]a h shows the scaled effective frequencies squared @xmath158 of the individual pic electrons vs their energies @xmath174 they would have if the driver is switched off instantaneously at @xmath175 , @xmath176 , @xmath177 , @xmath178 , @xmath179 , @xmath180 , @xmath181 , and @xmath182 laser cycles , respectively . at the time when , for a particular electron , @xmath183 becomes @xmath162 the ionization occurs for that electron . the laser intensity is @xmath184 , and the pre - ionized cluster is @xmath72 times over - critical so that @xmath33 . as is clearly visible in fig . [ picweffvsenglp ] , each electron reaches positive energy close to the point @xmath173 . the radial position of each electron is color - coded , indicating that outer ionization occurs at radii around @xmath185 . during the early time of the laser pulse ( fig . [ picweffvsenglp]a , b ) when many electrons are still inside the cluster , @xmath158 spreads over a wide range , starting from the maximum value @xmath186 down to negative values due to the repulsive force exerted by the compressed electronic cloud . note that negative values in effective frequency occur mainly at early times where most of the electrons are still inside the cluster . electrons with positive but very small @xmath183 and @xmath187 represent low energetic electrons removed earlier during the pulse ( see fig . [ picweffvsenglp]a , b ) . the occurrence of nlr is less clear for these early leaving electrons . as mentioned above , these electrons move in a shallow effective potential with @xmath188 when they leave the cluster with ease and with rather low kinetic energy because the laser intensity is still low at the time of their emission . figures [ picweffvsenglp]c f show that most of the electrons escape from the cluster by passing through the channel @xmath173 at radii around @xmath185 . it is also visible that more and more electrons are driven to positive frequency before they leave the cluster by passing through @xmath173 . this is so because as more and more electrons are freed , the remaining electrons experience predominantly the force by the ionic background , and they move deep into the potential ( see their negative values in energy ) where they experience the full mie - frequency @xmath189 . in fig . [ picweffvsenglp]c , d , the few electrons with positive energy but small radii are those driven back to the cluster by the laser field . in figures [ picweffvsenglp]e , f , electrons are strongly aligned ( no scattered points ) since the laser field is approaching its maximum ( at @xmath190-cycles ) . after the peak of the laser pulse ( figs [ picweffvsenglp]g , h ) the restoring force of the ions on almost all electrons dominates the laser force . since many of the features of energy absorption and nlr in a lp laser field are common to the case of cp , we only point out the main differences . , excursion @xmath159 , and the total energy @xmath152 for a pic electron in a cp field vs time in laser cycles . the total energy becomes positive only when the nlr is crossed ( indicated by vertical , dashed line ) . this result resembles the rsm result in fig . [ rsmiicpl ] . the charge density is @xmath74 , the peak laser intensity is @xmath142 w/@xmath143 . [ timevsomegacpl ] , scaledwidth=50.0% ] equation holds in the cp laser field as well . figure [ timevsomegacpl ] shows the effective frequency squared vs time for one of the pic electrons , together with the total energy @xmath160 and the excursion @xmath159 . one can see that the pic electron is freed ( i.e. , its total energy becomes positive ) only when the resonance line @xmath163 is passed . , ( b ) @xmath165 , ( c ) @xmath166 , ( d ) @xmath167 , ( e ) @xmath168 , ( f ) @xmath169 , ( g ) @xmath170 , and ( h ) @xmath171 laser cycles for a laser intensity @xmath172and @xmath33 . other parameters as in fig . [ picrsm1 ] . the radial positions ( in units of @xmath12 ) are color - coded . electrons become free upon crossing the nlr , i.e. , @xmath173 . [ picweffvsengcpl],scaledwidth=55.0% ] figure [ picweffvsengcpl ] is the cp analogue of fig . [ picweffvsenglp ] . the results are very much similar to the lp case shown in fig . [ picweffvsenglp ] and the arguments made there apply here as well . nlr is clearly observed . the main difference is that the pic electrons are nicer aligned towards the resonance point , even at early times during the laser pulse ( see fig . [ picweffvsengcpl ] b , c , d ) . almost no scattered particles are visible because the dynamics mainly consist of swirling around the cluster center rather than oscillating through it . the number of electrons returning to the cluster is much less so that the recombination and rescattering probability is smaller in the case of cp . the same is observed in laser - atom interaction experiments , with important consequences for harmonic generation and non - sequential ionization . for sufficiently high charge density , the degree of outer ionization ( defined as the ratio of removed electrons to the total number of electrons ) is less compared to the case of lp laser pulses . the explanation is the same as in sec.[sec2c ] : the effective laser intensity for cp is by a factor of two lower than for lp because of our definition of the cp field ( [ ex ] ) , ( [ ey ] ) . one may object that , since the denominator in necessarily increases while the numerator decreases for an electron on its way out of the cluster potential , that the passage through a point @xmath191 with @xmath10 _ some _ value @xmath192 is rather the consequence of outer ionization than the mechanism behind it . however , nlr only occurs at @xmath193 , and the results in fig . [ picweffvsenglp ] and fig . [ picweffvsengcpl ] show only little spreading along @xmath158 at @xmath194 . moreover , the fact that _ both _ the single electron energies become positive _ and _ the radii exceed @xmath195 when @xmath196 indicates that nlr is indeed the responsible mechanism behind outer ionization accompanied by efficient absorption of laser energy . vs laser intensity for charge densities ( a ) @xmath197 ( linear resonance ) , ( b ) @xmath198 , ( c ) @xmath199 , and ( d ) @xmath200 . the absorbed energies for lp ( red , lighter gray ) and cp ( blue , darker gray ) using pic ( dashed ) and the rsm ( solid ) are shown . [ rsmengcpl],title="fig:",scaledwidth=48.0% ] + vs laser intensity for charge densities ( a ) @xmath197 ( linear resonance ) , ( b ) @xmath198 , ( c ) @xmath199 , and ( d ) @xmath200 . the absorbed energies for lp ( red , lighter gray ) and cp ( blue , darker gray ) using pic ( dashed ) and the rsm ( solid ) are shown . [ rsmengcpl],title="fig:",scaledwidth=48.0% ] finally , fig . [ rsmengcpl ] shows the average value of the total absorbed energy per electron vs the peak laser intensity for cluster charge densities between @xmath201@xmath72 in cp and lp laser fields . pic results are compared with the rsm absorption results . the absorbed energy per electron in figs . [ rsmengcpl]a d is plotted in units of @xmath202 whereas the same results are shown in figs . [ rsmengcpl]e h in units of the ponderomotive energy @xmath93 . the pic results in figs . [ rsmengcpl]a d show that the absorbed energy increases linearly in the log - log representation up to a certain intensity and then tends to saturate due to the saturation of outer ionization . one sees that the saturation in the pic results occur close to the rsm threshold intensity . since with increasing charge density the restoring force due to the ions increases , the saturation of energy absorption in the rsm and pic occur at higher laser intensities as the density increases from @xmath201 to @xmath74 in fig . [ rsmengcpl]a d . when outer ionization and the energy absorption saturate with increasing peak laser intensity , the average absorbed energy per electron divided by @xmath93 ( which is proportional to the so - called fractional absorption ) starts decreasing in figs . [ rsmengcpl]e f . figures [ rsmengcpl]a , e show that at linear resonance @xmath197 , the absorbed energy is already high at low values of the laser intensity ( @xmath203 ) , and absorption is very efficient as compared to higher charge densities @xmath198@xmath72 , presented in figs . [ rsmengcpl]b d and figs . [ rsmengcpl]f h . in fact , [ rsmengcpl]e illustrates that the absorbed energy is on the order of @xmath204 ( both in the rsm and in the pic ) before the saturation of outer ionization . however , one should bear in mind that at too low laser intensities inner ionization would not occur in the first place so that in reality there would be no absorption at all . energy absorption in cp and lp laser fields at all intensities and all charge densities are almost equally efficient . recalling that with our definition of the cp laser field the ponderomotive potential is equal for lp and cp while the electric field amplitude is not , we conclude that for the absorbed energy @xmath122 matters while the nlr threshold is determined by the field strength . the discrepancy between the pic and rsm results at lower intensities is , again , due to the fact that the rsm provides a very well defined potential , although a more and more shallow one as the density decreases , in which the rigid electron sphere moves . in pic , the potential is initially zero and builds up in the course of outer ionization . at higher intensities , the potentials for the remaining pic electrons are closer to the rsm potential ( see fig . [ picrsm2 ] ) , explaining the improving agreement between rsm and pic results as the laser intensity increases . in summary , two different approaches to study collisionless laser energy absorption by clusters , namely ( i ) the rigid sphere model and ( ii ) particle - in - cell simulations , were pursued in this work . the goal was to identify the dominant mechanism of energy absorption and outer ionization of the cluster electrons in near infrared , short laser pulses where collisional absorption is known to be inefficient . we showed that the cluster electrons contributing to efficient absorption and outer ionization undergo nonlinear resonance , meaning that the instantaneous frequency of their motion in a time - dependent , anharmonic , effective potential transiently meets the laser frequency . nonlinear resonance is the only possible absorption mechanism if the laser pulse is too short for the linear resonance to occur ( or during the early cluster dynamics in longer pulses ) and if electron - ion collisions ( inverse bremsstrahlung ) are negligible . in order to prove the occurrence of nonlinear resonance we used a method to analyze the results obtained from particle - in - cell simulations , namely the mapping of the system of electrons and ions that interact through their mean field onto a system of nonlinear oscillators . the occurrence of nonlinear resonance in the particle - in - cell simulations presented in this paper resembles the nonlinear resonance in the rigid sphere model . for a given cluster charge density , there is a threshold intensity around which the average electron energy displays a maximum conversion of laser energy . the threshold intensity can be calculated using the newly introduced vanishing barrier approximation . the common over - barrier approximation applicable to atoms fails in the case of finite - size potentials and underestimates the required laser field strength for ionization . this fact might be related to the experimentally observed `` enhanced saturation intensity '' in the ionization of finite - size systems such as c@xmath120 . the efficiency of energy absorption from the laser is almost the same for linear and circular polarization , indicating that `` collisions with the cluster boundary '' as an explanation are misleading . instead , nonlinear resonance is the main absorption mechanism in both cases . to illustrate this , the rigid sphere model has been extended for the circularly polarized laser pulses in this paper . t. ditmire , r.a . smith , j.w.g . tisch , and m.h.r . hutchinson , , 3121 ( 1997 ) . molecules and clusters in intense laser fields , edited by jan posthumus ( cambridge university press , cambridge , u.k . , 2001 ) . u. saalmann and j.m . rost , j. phys . b : at . phys . * 39 * , r39-r77 ( 2006 ) . krainov and m.b . smirnov , phys . 370 * , 237 ( 2002 ) . c. rose - petruck , k.j . schafer , k.r . wilson , and c.p.j . barty , , 1182 ( 1997 ) . d. bauer and a. macchi , , 033201 ( 2003 ) . t. ditmire , j.w.g . tisch , e. springate , m.b . mason , n. hay , j. marangos , and m.h.r . hutchinson , nature ( london ) * 386 * , 54 ( 1997 ) . v. kumarappan , m. krishnamurthy , and d. mathur , , 043204 ( 2003 ) . e. springate , s.a . aseyev , s. zamith , and m.j.j . vrakking , , 053201 ( 2003 ) . k. ishikawa and t. blenski , , 063204 ( 2000 ) . f. megi , m. belkacem , m.a . bouchene , e. suraud , and g. zwicknagel , j. phys . phys . * 36 * , 273 ( 2003 ) . d. bauer , j. phys . * 37 * , 3085 ( 2004 ) . c. siedschlag and j.m . rost , , 043402 ( 2004 ) . t. ditmire , t. donnelly , a.m. rubenchik , r.w . falcone , and m.d . perry , , 3379 ( 1996 ) . t. dppner , th . fennel , th . diederich , j. tiggesbumker , and k.h . meiwes - broer , , 013401 ( 2005 ) . l. kller , m. schumacher , j. khn , s. teuber , j. tiggesbumker , and k.h . meiwes - broer , , 3783 ( 1999 ) . s. zamith , t. martchenko , y. ni , s.a . aseyev , h.g . muller , and m.j.j . vrakking , , 011201(r ) ( 2004 ) . i. last and j. jortner , , 2215 ( 1999 ) . ulf saalmann and jan - michael rost , , 223401 ( 2003 ) . fennel , g.f . bertsch , and k.h . meiwes - broer , eur . j. d * 29 * , 367 ( 2004 ) . c. siedschlag and j.m . rost , , 031401 ( 2005 ) . t. martchenko , ch . siedschlag , s. zamith , h. g. muller , and m. j. j. vrakking , , 053202 ( 2005 ) . t. taguchi , t.m . antonsen , jr . , and h.m . milchberg , , 205003 ( 2004 ) ; thomas m. antonsen , jr . toshihiro taguchi , ayush gupta , john palastro , and howard m. milchberg , phys . plasmas * 12 * , 056703 ( 2005 ) . mulser and m. kanapathipillai , , 063201 ( 2005 ) ; p. mulser , m. kanapathipillai , and d.h.h . hoffmann , , 103401 ( 2005 ) . d. bauer and p. mulser , j. phys . ser . * 11 * , 169 ( 2005 ) . korneev , s.v . popruzhenko , d.f . zaretsky , w. becker , laser phys . * 2 * , 452 ( 2005 ) . parks , t.e . cowan , r.b . stephens , and e.m . campbell , , 063203 ( 2001 ) . fomichev , s.v . popruzhenko , d.f . zaretsky , and w. becker , j. phys . . phys . * 36 * , 3817 ( 2003 ) . lin , h. chu , m. shen , y. xiao , c.h . lee , s. chen , and j. wang , optics communications * 231 * , 375 ( 2004 ) v. kumarappan , m. krishnamurthy , d. mathur , and l.c . tribedi , 023203 ( 2001 ) . s. ter - avetisyan , m. schnrer , h. stiel , u. vogt , w. radloff , w. karpov , w. sandner , and p.v . nickles , , 036404 , ( 2001 ) . m. krishnamurthy , d. mathur , and v. kumarappan , , 033202 ( 2004 ) . birdsall and a.b . langdon , _ plasma physics via computer simulation _ ( iop , bristol , 1991 ) . g.l . kotkin and v.g . serbo , _ collection of problems in classical mechanics _ ( pergamon , oxford , 1971 ) . bethe and e.e . salpeter , _ quantum mechanics of one- and two - electron atoms _ ( plenum publishing corporation , new york , 1977 ) . hertel , t. laarmann , and c.p . schulz , adv . at . * 50 * , 219 ( 2005 ) . a. jaro - becker , a. becker , and f.h.m . faisal , , 143006 ( 2006 ) .

recently , there has been a wide interest in relativistic potentials involving mixtures of vector and scalar potentials with opposite signs . the interest lies on attempts to explain the pseudospin symmetry in nuclear physics . @xcite , using a dirac hamiltonian with scalar @xmath1 and vector @xmath2 potentials quadratic in space coordinates , found a harmonic - oscillator - like second order equation which can be solved analytically for @xmath3 , as considered before by kukulin _ @xcite , and also for @xmath4 . very recently , ginocchio solved the triaxial , axial , and spherical harmonic oscillators for the case @xmath5 and applied it to the study of antinucleons embedded in nuclei @xcite . the case @xmath6 is particularly relevant in nuclear physics , since it is usually pointed out as a necessary condition for occurrence of pseudospin symmetry in nuclei @xcite-@xcite . also the observed isospin dependence of pseudospin can be explained by the effect of the @xmath7 potential on the shape of the @xmath8 potential @xcite-@xcite . as shown by bell and ruegg @xcite , the @xmath9 and @xmath6 cases correspond to a @xmath10 symmetry of the dirac hamiltonian with only scalar and vector potentials which is independent of the particular shapes of these potentials . a generalized harmonic oscillator ( ho ) for spin 1/2 particles that includes not only those combinations of scalar and vector potentials , but also a linear tensor potential , giving rise to the so - called dirac oscillator , has also been considered in @xcite . the research of this kind of interaction was started by it @xcite and has been revived by moshinsky and szczepaniak @xcite ( see @xcite for a complete reference list ) . in ref . @xcite a special attention was paid to the case when the scalar and vector potentials are such that @xmath11 , which , as stated before , has been related to the nuclear pseudospin symmetry . there it has been concluded that only when the tensor potential is turned on one gets negative energy bound - state solutions along with the usual positive energy solutions . those negative energy solutions are important to describe the possible existence of antinucleons in nuclei @xcite . when the tensor potential is absent the special conditions relating the scalar and vector potentials ( @xmath12 ) are needed to have a standard positive ho potential which is only able to bind fermions , and therefore exclude the negative bound - state solutions from the spectra actually , as was discussed in another context in @xcite , with an appropriate choice of signs , potentials fulfilling the relations @xmath12 are able to bind either fermions or antifermions . as an example , we will see later that a ( negative ) harmonic potential , with no tensor potential , binds antifermions and does not allow positive bound - state solutions . closely related to this is the fact that , in the nucleus , the charge - conjugation transformation relates the spin symmetry of the negative bound - state solutions ( antinucleons ) to the pseudospin symmetry of the positive bound - state solutions ( nucleons ) @xcite . this has also been discussed recently in @xcite , analyzing the ho for antinucleons with spin symmetry ( @xmath5 ) and without the tensor interaction . again , the charge conjugation connects this problem with the ho for nucleons with pseudospin symmetry ( @xmath13 ) , but in this case the positive harmonic binding potential is @xmath14 . note that , because the harmonic potential does not vanish for large distances but is a confining potential , it is possible to obtain particle bound - state solutions in the exact pseudospin limit @xcite . therefore , we believe that this connection between spin and pseudospin symmetry obtained by charge conjugation deserves to be more explored . in addition , after the pseudospin symmetry has been revealed as a relativistic symmetry by ginocchio @xcite , a link of it to chiral symmetry has been suggested based on sum rules calculations in relativistic chiral models that support a small value for @xmath8 @xcite . however , until now , a clear connection of pseudospin and chiral symmetries has not been established . the main motivation of this paper is to shed some light in the relation between spin and pseudospin symmetries by means of charge - conjugation and @xmath0 chiral transformations . in order to do so , we solve the mixed scalar - vector - pseudoscalar ho potential in 1 + 1 dimensions , taking advantage of the simplicity of the lowest dimensionality of the space - time . this approach is equivalent to consider fermions in 3 + 1 dimensions which are restricted to move in one direction @xcite . we explore the spectra when it is possible to obtain analytical solutions , _ i.e. _ , in the particular cases when @xmath5 and @xmath6 . as referred before , if the pseudocalar term is turned off , these cases correspond , respectively , to have exact spin and pseudospin symmetry in 3 + 1 dimensions . we explore all the possible signs of the quadratic ( @xmath15 or @xmath16 ) and linear ( pseudoscalar ) potentials , thus paying attention to bound states of fermions and antifermions as well . we compare both cases @xmath9 and @xmath6 to establish the charge - conjugation connection discussed above in the presence of the pseudoscalar term . we also consider the case of zero mass to look for the connection between spin and pseudospin symmetries by means of the chiral transformation . an added motivation for this paper is given by the recent demonstration that the relativistic 3 + 1 ho with scalar and vector potentials can describe successfully the heavy nucleus spectrum @xcite . the parameters of the ho are determined by fitting the scalar and vector potentials derived from relativistic mean field nuclear calculations ( rmf ) . the @xmath17pb neutron single - particle levels obtained in the ho are very similar to the rmf results . the pseudospin symmetry is shown to be almost satisfied for a heavy nucleus in the ho description . the possibility of breaking perturbatively this symmetry in the 3 + 1 ho by a tensor interaction ( in our 1 + 1 case a pseudoscalar coupling ) has been discussed in @xcite and can be included in a calculation like the one of ref . @xcite in order to improve the results . thus , the study of all the possible eigenenergies of the 1 + 1 ho , presented in this paper , considering not only the positive energy solutions already obtained for the 3 + 1 case in @xcite but also the negative ones , can be applied to analyze the existence of antinucleons in nuclei and their spectrum . the particle and antiparticle spectra we obtain for the cases @xmath6 and @xmath9 demonstrate the charge - conjugation connection referred above . they also show that the spectra of massless @xmath5 states are degenerate with massless @xmath11 states . further imposing chiral symmetry to these states , such that they would become eigenstates of @xmath0 , would imply that all potentials would be zero and thus one would have just massless free fermions . our conclusions are valid in the case of 3 + 1 dimensions , because , as will be shown later , one basically has to add the angular momentum quantum numbers @xmath18 and @xmath19 to the coefficients of the functions appearing in the 1 + 1 eigenvalue equations in order to get the corresponding 3 + 1 equations . this paper is organized as follows . in sec . [ sec : dirac_1 + 1 ] we present the general dirac equation in 1 + 1 dimensions with a potential with a completely general lorentz structure . the effect of charge - conjugation and the @xmath0 chiral transformations upon the general 1 + 1 dirac hamiltonian is discussed in secs . [ subsec : charge_conj ] and [ subsec : chiral ] respectively . in secs . [ subsec : eq_motion ] and [ subsec : sturm - liouville ] we present the equations of motion and discuss the isolated solutions , _ i.e. _ , the solutions out of the sturm - liouville problem , for @xmath9 and @xmath6 . the nonrelativistic limits of the @xmath9 and @xmath6 cases are discussed in sec . [ subsec : nonrel_limit ] , where we show that when @xmath6 there is no nonrelativistic limit when the pseudoscalar potential vanishes or is very small . in sec . [ sec : rel_h_oscillator ] we introduce the general relativistic harmonic oscillator in 1 + 1 dimensions for @xmath9 by assigning a quadratic potential @xmath20 for @xmath8 and a linear potential @xmath21 for the pseudoscalar potential @xmath22 . there we also present the eigenvalue equation and the wave functions for bound states , discuss the dirac oscillator case ( @xmath23 and @xmath24 ) and show how one can use the charge - conjugation and chiral transformations to get the eigenenergies for the @xmath6 case from the @xmath9 case . the following section is devoted to a detailed analysis of the solutions of the @xmath9 relativistic ho for all signs of the parameters @xmath25 and @xmath26 . finally in sec . [ sec : sol_sym_arg ] we discuss the spectrum for @xmath6 and massless fermions and in sec . [ sec : conclusions ] we draw the conclusions . the two - dimensional dirac equation can be obtained from the four - dimensional one with the mixture of spherically symmetric scalar , vector and anomalous magnetic - like ( tensor ) interactions . if we limit the fermion motion to the @xmath27-direction ( @xmath28 ) the four - dimensional dirac equation decomposes into two equivalent two - dimensional equations with 2-component spinors and 2@xmath292 matrices @xcite . then , there results that the scalar and vector interactions preserve their lorentz structures whereas the anomalous magnetic - like interaction turns out to be a pseudoscalar interaction . furthermore , in the 1 + 1 world there is no angular momentum so that the spin is absent . therefore , the 1 + 1 dimensional dirac equation allows us to explore the physical consequences of the negative - energy states in a mathematically simpler and more physically transparent way . in this spirit the two - dimensional version of the anomalous magnetic - like interaction linear in the space coordinate has also received attention @xcite-@xcite . later this system was shown to be a particular case of a more general class of exactly solvable problems asc2 . let us begin by presenting the dirac equation in 1 + 1 dimensions . the dirac equation for a particle with mass @xmath30 and a general potential @xmath31 , to be defined later , is formally the same as its 3 + 1 counterpart : @xmath32here @xmath33 and the @xmath34 matrices @xmath35 obey the usual relations @xmath36 , where @xmath37 is the @xmath34 identity matrix . the metric tensor and the covariant derivative operator are , respectively , @xmath38 and @xmath39 . the hamiltonian is defined in the usual way , such that eq . ( eq_dirac_gen ) is written as @xmath40where @xmath41 . the positive definite function @xmath42 , satisfying a continuity equation , is interpreted as a position probability density and its norm is a constant of motion . this interpretation is completely satisfactory for single - particle states @xcite . the traceless matrices @xmath43 and @xmath44 are defined by @xmath45 and @xmath46 , and obey the relations @xmath47 , @xmath48 . we set @xmath31 to be @xmath49where @xmath50 . this is the most general combination of lorentz structures because there are only four linearly independent 2@xmath292 matrices . the subscripts for the terms of the potential denote their properties under a lorentz transformation : @xmath51 and @xmath52 for the time and space components of the 2-vector potential , @xmath53 and @xmath54 for the scalar and pseudoscalar terms , respectively . if the terms in the potential @xmath31 are time - independent , the dirac equation ( [ def_hamiltoniano_d ] ) becomes @xmath55 where @xmath56 and @xmath57 is the fermion energy . to have an explicit expression for the @xmath58 and @xmath59 matrices one can choose 2@xmath292 pauli matrices which satisfy the same algebra . we use @xmath60 , @xmath61 and thus @xmath62 . the hamiltonian ( [ time_ind_h ] ) is invariant under the parity operation , _ i.e. _ , when @xmath63 , @xmath64 and @xmath65 change sign , whereas @xmath66 and @xmath67 remain the same . this is because the parity operator is @xmath68 , where @xmath69 is a constant phase and @xmath70 changes @xmath27 into @xmath71 . since this unitary operator anticommutes with @xmath58 and @xmath72 , they change sign under a parity transformation , whereas @xmath37 and @xmath59 , which commute with @xmath73 , remain the same . when one writes down the explicit equations of motion in terms of the components of the spinor @xmath74 , the combinations @xmath75 and @xmath76 of the vector and scalar components arise naturally . therefore , it is convenient to rewrite the hamiltonian ( [ time_ind_h ] ) in terms of these potentials . we have @xmath77 the charge - conjugation operation changes the sign of the electromagnetic interaction , _ i.e. _ , changes the sign of the time and space potentials in ( [ def_v ] ) . this is accomplished by the transformation ( see , for example , itzykson and zuber @xcite ) @xmath78where @xmath79 denotes matrix transposition and @xmath80 is a matrix such that @xmath81 . in @xmath82 dimensions one matrix that satisfies this relation is @xmath83if we choose the phase factor @xmath84 equal to 1 , we have @xmath85after applying this charge - conjugation operation to the dirac equation ( def_hamiltoniano_d ) , the time - independent dirac equation becomes @xmath86where @xmath87 and @xmath88 is given by @xmath89 in terms of the potentials @xmath15 and @xmath16 , this hamiltonian reads @xmath90one sees that the charge - conjugation operation changes the sign of the energy and of the potentials @xmath91 , @xmath92 and @xmath93 . in turn , this means that @xmath16 turns into @xmath94 and @xmath15 into @xmath95 . therefore , to be invariant under charge conjugation , the hamiltonian must contain only a scalar potential . the chiral operator for a dirac spinor is the matrix @xmath0 , and we will call `` chiral transformation '' the transformation associated with it . thus , the transformed spinor is given by @xmath96 and the transformed hamiltonian @xmath97 . since @xmath98 anticommutes with @xmath59 , the time - independent chiral - transformed dirac equation is @xmath99 where @xmath100 is given by @xmath101 or @xmath102 in terms of @xmath8 and @xmath14 . this means that the chiral transformation changes the sign of the mass and of the scalar and pseudoscalar potentials , thus turning @xmath8 into @xmath14 and vice - versa . a chiral - invariant hamiltonian needs to have zero mass and @xmath103 and @xmath22 zero everywhere . the space component of the 2-vector potential in ( [ def_v ] ) can be absorbed into the wave function by defining a new spinor @xmath104 such that @xmath105 in which @xmath106 , since we have @xmath107 . from this point on , we will refer to @xmath91 as simply a vector potential , following the common usage of this term ( usually denoted by @xmath2 ) . if we now write the spinor @xmath104 in terms of its components , @xmath108the dirac equation gives rise to two coupled first - order equations for @xmath109 and @xmath110 : @xmath111where the prime denotes differentiation with respect to @xmath27 . in terms of @xmath109 and @xmath110 the spinor is normalized as @xmath112 , so that @xmath109 and @xmath110 are square integrable functions . it is clear from the pair of coupled first - order differential equations ( eq_mot_1 ) and ( [ eq_mot_2 ] ) that @xmath109 and @xmath110 have opposite parities if the dirac equation is covariant under @xmath63 . under the charge - conjugation and chiral transformations the spinor ( def_psi_comp ) becomes @xmath113 and @xmath114 respectively . using the expression for @xmath115 obtained from ( [ eq_mot_2 ] ) with @xmath116 , _ viz . _ @xmath117 and inserting it in ( [ eq_mot_1 ] ) one arrives at the following second - order differential equation for @xmath118 : @xmath119 \psi_{+ } = 0\ . \label{2_order_gen_psi+}\ ] ] in a similar way , using the expression for @xmath120 obtained from ( [ eq_mot_1 ] ) with @xmath121 , _ viz . _ @xmath122 and inserting it in ( [ eq_mot_2 ] ) one arrives at the following second - order differential equation for @xmath123 : @xmath124 \psi_{- } = 0 \label{2_order_gen_psi-}\ ] ] for @xmath5 with @xmath125 , ( [ psi-_gen ] ) and ( 2_order_gen_psi+ ) reduce to @xmath126 @xmath127 \psi_{+ } = \frac{e^{2}-m^{2}c^{4}}{2mc^{2}}\psi_{+}\ , \label{2_order_delta=0_psi+}\ ] ] and for @xmath11 with @xmath128 , ( [ psi+_gen ] ) and ( [ 2_order_gen_psi- ] ) reduce to @xmath129 @xmath130 \psi_{- } = \frac{e^{2}-m^{2}c^{4}}{2mc^{2}}\psi_{-}\ . \label{2_order_sigma=0_psi-}\ ] ] either for @xmath5 with @xmath125 or @xmath11 with @xmath131 the solution of the relativistic problem is mapped into a sturm - liouville problem in such a way that solution can be found by solving a schrdinger - like problem . if one considers potentials @xmath91 and @xmath1 quadratic in @xmath27 with @xmath132 and a potential @xmath93 linear in @xmath27 one obtains schr'odinger - like equations for the harmonic oscillator potential . the solutions for @xmath5 with @xmath133 and @xmath11 with @xmath134 , excluded from the sturm - liouville problem , can be obtained directly from the original first - order equations ( [ eq_mot_1 ] ) and ( eq_mot_2 ) . they are @xmath135 \\ \\ \displaystyle\psi_{- } ^{\prime } + \frac{v_{p}}{\hbar c}\,\psi_{- } = -\frac{i}{% \hbar c}\left ( \sigma + 2mc^{2}\right ) \psi_{+}% \end{array } \right\ } , \mathrm{\;for\;}\delta = 0\mathrm{\;with\;}e =- mc^{2 } \label{isol_sol_delta=0}\ ] ] and @xmath136 \\ \\ \displaystyle\psi_{+ } ^{\prime } -\frac{v_{p}}{\hbar c}\,\psi_{+ } = -\frac{i}{% \hbar c}\left ( \delta -2mc^{2}\right ) \psi_{-}% \end{array } \right\ } , \mathrm{\;for\;}\sigma = 0\mathrm{\;with\;}e = mc^{2 } \label{isol_sol_sigma=0}\ ] ] where @xmath137 and @xmath138 are normalization constants . since , by parity conservation , @xmath22 must be an odd function of @xmath139 , its integral must be an even function of @xmath27 . thus , if the integral of @xmath22 has a definite sign , its overall sign can determine whether the isolated solutions ( [ isol_sol_delta=0 ] ) and ( [ isol_sol_sigma=0 ] ) are bound states . in particular , if @xmath22 and @xmath140 or @xmath14 are powers of @xmath27 , one sees from each pair of equations ( [ isol_sol_delta=0 ] ) and ( [ isol_sol_sigma=0 ] ) that , in general , one of the spinor components has to be zero , either @xmath118 in the first case ( sign of the @xmath22 coefficient positive ) or @xmath115 in the last case ( sign of the @xmath22 coefficient negative ) . this is the case of the harmonic oscillator potentials , as we will see in the next section . however , it is possible find solutions with the @xmath22 coefficient negative in the first case ( @xmath5 , @xmath133 ) and with the @xmath22 coefficient positive in the second case ( @xmath11 , @xmath141 ) , provided a certain relation between the coefficients of all the potentials is satisfied , as we will see later . in the nonrelativistic limit , @xmath142 , eq . ( [ 2_order_gen_psi+ ] ) becomes @xmath143+\sigma\psi_{+}=\mathcal{e}\psi_{+ } \ , \ ] ] where @xmath144 . since , in the nonrelativistic regime , for the range of values of @xmath139 in which the wave function is not negligible , @xmath145 , from eq . ( [ psi-_gen ] ) we see that @xmath123 is of order @xmath146 relative to @xmath118 , provided , of course , one has also @xmath147 . on the other hand , since in these conditions @xmath148 and the @xmath149 term in eq . ( [ eq_mot_nr2a ] ) is suppressed relative to the other terms , we also see that @xmath118 obeys a schrdinger equation with binding energy equal to @xmath150 . we shall consider now the nonrelativistic limit in the special cases @xmath151 and @xmath6 , which we will also consider later in the paper when we solve the dirac equation for harmonic oscillator potentials . when @xmath9 , eq . ( [ eq_mot_nr2a ] ) becomes @xmath152 here we clearly see that @xmath8 plays the role of a binding potential in the nonrelativistic limit and that @xmath22 gives rise to effective binding potentials proportional to @xmath153 and @xmath154 . this means that even a pseudoscalar potential unbounded from below could be a confining potential . note that if @xmath155 is of the same order that @xmath156 in a @xmath157 expansion within the classically accessible region ( positive kinetic energy ) where the wave function is not negligible , and if its variation over this region is such that @xmath158 ( @xmath22 changes very little over a distance of the order of a compton wavelength ) , @xmath22 gets suppressed relative to @xmath140 . for the case @xmath6 and if @xmath159 is very small in the classically accessible region , eq . ( [ eq_mot_nr2a ] ) becomes @xmath160 if @xmath161 or if , as explained before , the @xmath154 and @xmath162 terms are higher - order terms in a @xmath157 expansion within the classically accessible region , eq . ( [ eq_mot_nr2a_sigma=0 ] ) becomes a free particle schrdinger equation . this is in agreement with what was found in ref . @xcite , namely that when @xmath163 and @xmath14 is a three - dimensional harmonic oscillator potential , there is no nonrelativistic limit . in this case we were able to show that this is true in a more general framework . a model for a relativistic particle in which the vector and scalar potentials are such that @xmath6 , and where the pseudoscalar potential is suppressed , is intrinsically relativistic , as far as bound states are concerned . this last result deserves some more comments . the fact that in a dirac equation one can have lorentz scalar and pseudoscalar potentials leads to remarkable results that are at odds with what is known from nonrelativistic quantum mechanics , where the lorentz structure plays no role , since one has only potentials which couple to the energy , _ i.e. _ , behave as time components of a relativistic vector . among these results is the fact that the dirac equation is not invariant under a simultaneous shift of the energy , the scalar potential , and the pseudoscalar potential . it has already been verified that a constant added to the pseudoscalar screened coulomb potential @xcite or to the pseudoscalar inversely linear potential asc100 is physically relevant and plays a crucial role in ensuring the existence of bound states . as an example of another notable result , it is well known that a confining potential for a nonrelativistic fermion can not confine it in the relativistic regime when the potential is considered as a lorentz vector ( see , _ e.g. _ , @xcite ) , simply because there is pair creation and the single - particle picture does not hold anymore . it is surprising that the converse is also true , _ i.e. _ , that relativistic binding potentials may not bind in the nonrelativistic limit . this is the case of @xmath164 ( @xmath11 ) referred before , and is basically due to the fact that vector potentials couple in a different way than scalar potentials in the dirac equation , whereas there is no such distinction in the schrdinger equation . actually , as we have seen before , @xmath8 , which is the sum of a vector and a scalar potential , plays the role of a binding potential in the nonrelativistic limit , whereas a weak enough @xmath14 can not bind a fermion with nonrelativistic energy if @xmath6 . note that for potentials going to zero at infinity , like mean - field nuclear potentials , there are no bound states when @xmath6 , even if @xmath14 is not small compared with fermion rest energy @xcite-@xcite . regarding antiparticles , an attractive vector potential for a particle is , of course , repulsive for its corresponding antiparticle , and vice versa . however , the attractive ( repulsive ) scalar potential for particles is also attractive ( repulsive ) for antiparticles . this is expressed in the change @xmath165 and @xmath166 under charge conjugation , as referred in sec . [ subsec : charge_conj ] . this means that an antiparticle with nonrelativistic energies ( @xmath167 ) will not be bound when @xmath9 , provided @xmath22 is small enough . let us consider @xmath168 as we have seen in the last section , the chiral transformation performs the changes @xmath169 , @xmath170 , @xmath171 and @xmath172 . morever , since @xmath0 interchanges the upper and lower components ( see ( [ psi_comp_chiral ] ) ) , the resulting pair of transformed equations of motion are formally the same , so that their solutions have the same energy eigenvalues . this symmetry can be clearly seen from the two equation pairs ( [ psi-_gen])-([2_order_gen_psi+ ] ) and ( [ psi+_gen])-([2_order_gen_psi- ] ) as well as from the isolated solutions ( [ isol_sol_delta=0 ] ) and ( [ isol_sol_sigma=0 ] ) which are converted into each other by this kind of transformation . this means that one can restrict the discussion to the @xmath8 case ( @xmath5 ) . the results for the case when @xmath173 can be obtained immediately by just changing the sign of @xmath30 and of @xmath26 in the relevant expressions . the dirac spinor corresponding to the isolated solution with @xmath133 is obtained from ( [ isol_sol_delta=0 ] ) . only for @xmath174 there is a normalizable dirac spinor and , as commented before , the upper component vanishes whereas the lower component is given by @xmath175 , regardless the value of @xmath26 . for @xmath125 , eq . ( [ 2_order_delta=0_psi+ ] ) takes the form @xmath176 where @xmath177 the well - behaved solution for ( [ 2_order_delta=0_oh_psi+ ] ) , with @xmath178 necessarily real and positive , is the well - known solution of the schrdinger equation for the nonrelativistic harmonic oscillator ( see , _ e.g. _ , @xcite ) : @xmath179 where @xmath180 , @xmath181 is a normalization constant , @xmath182 is a @xmath183-th degree hermite polynomial and @xmath184 the lower component of the dirac spinor is obtained from ( psi-_delta=0 ) . when one uses the recursion relations for the hermite polynomials ( see , e.g. , @xcite ) one gets @xmath185 e^{-\lambda ^{2}x^{2}/2 } \label{psi-_delta=0_oh}\ ] ] since the hermite polynomial @xmath186 has @xmath187 distinct zeros one may conclude that @xmath120 has @xmath183 nodes , and the expression for @xmath123 suggests that it has @xmath188 nodes , depending on the sign @xmath26 . in fact , if @xmath189 , this last expression reduces to @xmath190 e^{-\lambda ^{2}x^{2}/2 } \ , \label{psi-_delta=0_k1=0}\ ] ] showing that @xmath115 is proportional to @xmath191 if @xmath26 is positive . if @xmath192 , the recursion relations of hermite polynomials imply that @xmath115 is proportional to @xmath193 ( see @xcite ) . from eqs . ( [ def_k ] ) , ( [ def_ee ] ) and ( [ delta=0_oh_ee ] ) we can get the quantization condition for the dirac eigenenergies : @xmath194 the nonrelativistic limit is reached when @xmath195 and @xmath196 . in this case @xmath197 for small quantum numbers , and eq . ( [ eigen_eq_delta=0 ] ) becomes @xmath198 so that @xmath199 is restricted to @xmath200 . in general there is no requirements on the signs of @xmath199 and @xmath201 , except that , as stated before , @xmath202 , and therefore , from ( [ def_k ] ) , the dirac eigenenergies corresponding to the bound - state solutions must be within the limits @xmath203 for @xmath204 , when there is a pure pseudoscalar potential linear in @xmath27 ( the two - dimensional dirac oscillator ) , eq . ( [ eigen_eq_delta=0 ] ) reduces to @xmath205\ . \label{energy_eigen_k1=0}\ ] ] note that @xmath206 for @xmath207 and @xmath208 for @xmath192 , because for the latter case the lower component is proportional to a hermite polynomial of degree @xmath209 . this result , together with eq . ( [ energy_eigen_k1=0 ] ) , allows us to conclude that the dirac eigenvalues are given by @xmath210 , where @xmath211 , @xmath212 , irrespective of the sign of @xmath201 . this means that _ the spectrum is independent of the sign of @xmath26_. the solutions @xmath141 and @xmath133 correspond to isolated solutions for @xmath192 and @xmath207 , respectively ( see eqs . ( isol_sol_delta=0 ) and ( [ isol_sol_sigma=0 ] ) ) . however , the spinors do depend on the sign of @xmath26 , because different signs induce different node structures , as can be seen from ( [ psi-_delta=0_k1=0 ] ) . since the charge - conjugation transformation changes the sign of @xmath26 , it connects both kinds of solutions , but each can have both positive and negative energies . thus a complete set of states for a certain @xmath26 includes both particle ( positive energy ) and antiparticle ( negative energy ) states but they are not related to each other by a charge - conjugation transformation . the spectral gap between these solutions is @xmath213 and does not disappear when @xmath214 . therefore , there is no room for transitions from positive- to negative - energy solutions and klein s paradox never comes into play . there are no scattering states when @xmath204 , but for @xmath215 the continuum states are those ones which have energies beyond the limits expressed by ( limit_values_k1_p)-([limit_values_k1_n ] ) , for which @xmath216 . as far as the chiral transformation is concerned , one can note that @xmath0 acts much in the same way as the charge - conjugation transformation , except that it does not change the sign of the energy and changes the mass sign . therefore , the chiral transformation connects solutions with the same energy sign but that have opposite @xmath26 signs . when @xmath217 , _ i.e. _ , no pseudoscalar potential , there is no isolated bounded solutions and the quantization condition expressed by eq . ( [ eigen_eq_delta=0 ] ) takes the form @xmath218 where @xmath219 stands for the sign function . this means that when @xmath220 there are only bound states for fermions with @xmath221 and those bounded solutions are kept apart from the negative - energy continuum ( @xmath222 ) by an energy gap greater than @xmath223 . on the other hand , for @xmath224 there are only bound states for antifermions with @xmath222 and those bounded solutions also do not mix with continuum states with @xmath225 . therefore , the positive - energy bound states of fermions never sink into the dirac sea of negative energies , and the negative - energy bound states of antifermions never reach the upper continuum , so again there is no danger of reaching the conditions for klein s paradox . it is important to note that the solutions with @xmath226 _ are not _ the antiparticles of the solutions with @xmath227 because charge conjugation changes @xmath8 into @xmath228 and vice - versa . this means that the antiparticles of the states with @xmath229 and @xmath227 are the negative - energy solutions of the relativistic system with @xmath230 , with @xmath25 positive . to conclude , let us present , as an example of the rule stated at the beginning of the section , the eigenvalue equation for the relativistic harmonic oscillator when @xmath231 : @xmath232here @xmath233 is the number of nodes ( the degree of the hermite polynomial ) for the lower component , since the chiral transformation interchanges the upper and lower components , so that the second - order equation to solve in this case is the one for the lower component , eq . ( [ 2_order_sigma=0_psi- ] ) . the upper component would be obtained from the corresponding first - order equation eq . ( [ psi+_sigma=0 ] ) . as a final remark , one may note that for massless particles , the @xmath5 and @xmath234 oscillators have the same spectrum with the sign of @xmath201 reversed . in 3 + 1 dimensions the corresponding eigenvalue equations to eqs . ( eigen_eq_delta=0 ) and ( [ eigen_eq_sigma=0 ] ) have a very similar structure . from @xcite we see that , for @xmath9 , one just adds the dirac equation quantum number @xmath19 to the coefficient of @xmath26 and modifies the coefficient of the square root to include the orbital angular momentum @xmath235 of the upper component @xmath236 while for @xmath6 one uses the orbital angular momentum @xmath237 of the lower component @xmath238 where @xmath239 ( see @xcite ) . if we compare these equations with ( [ eigen_eq_delta=0 ] ) and ( [ eigen_eq_sigma=0 ] ) , we see that the main features are identical , and thus all the main results obtained in the next section remain valid for the relativistic harmonic oscillator in three spatial dimensions . we turn now to the solutions of eq . ( [ eigen_eq_delta=0 ] ) . in order to obtain a deeper insight on those solutions for arbitrary @xmath199 and @xmath201 one has to seek a convenient method for solving this transcendental equation . if we square eq . ( [ eigen_eq_delta=0 ] ) the resulting equation is in general a quartic algebraic equation in @xmath57 , which in principle can be solved analytically . the price paid is that some spurious roots can appear in this process , although , of course , these can be eliminated by checking whether they satisfy the original equation . a more instructive procedure is follow a graphical method , by which one seeks the intersection points of the functions of energy in eq . ( [ eigen_eq_delta=0 ] ) : a parabola on the left - hand side @xmath240 and a square - root function on the right - hand side @xmath241 in this way , it is easy to see that there are at most two acceptable solutions of ( [ eigen_eq_delta=0 ] ) , since the parabola and the square - root function can have at most two intersection points . the construction of the graph of @xmath242 is split into three distinct classes depending on the @xmath199 values . below we discuss these three classes of problems . it should not be forgotten , though , that the solutions with @xmath133 are not included in this discussion , since they correspond to isolated solutions for @xmath174 . this kind of analysis and the conclusions that will be drawn from it in the following subsections can be applied with appropriate modifications to the 3 + 1 case , and therefore to nucleon and antinucleon single - particle spectra . this is especially true as far the relations between the spectra of systems with @xmath9 and @xmath6 are concerned , since these are obtained from the general properties of charge - conjugation and @xmath0 transformations which hold in 3 + 1 dimensions . although we have already solved this case , it is presented here for the sake of completeness . for @xmath204 , @xmath242 is just a nonnegative constant function as can be seen from eq . ( [ energy_eigen_k1=0 ] ) , having two symmetric intersections with the parabola . [ fig1 ] and fig . [ fig2 ] present the results for the first four quantum numbers . in these and the following plots we use the scaled quantities @xmath243 , @xmath244 and @xmath245 . the first shows how the values of the energy are obtained from the intersection of the two curves and the second how the eigenvalues vary with @xmath26 . the energies for the zero mass case can also be computed from the intersections with the dashed parabola in fig . [ fig1 ] . from these figures it is clear the symmetry of the energy levels . note again that the isolated solution is not plotted . for the first four levels when @xmath246 and @xmath247 . the black dots represent the values of energies coming from the intersection of the parabola and the horizontal lines correspond to the scaled function @xmath248 for @xmath249 . the dashed parabola is the function @xmath250 for @xmath251 . in this case , the horizontal lines are just @xmath242 for @xmath249 , @xmath246 in appropriate units and the horizontal scale is in the units of the energy @xmath57 . , width=377 ] for the first four levels when @xmath252 as a function of @xmath253 . for @xmath251 the plots are qualitatively the same , except that the lines for the different levels begin at the origin and the vertical scale would be in units of the energy @xmath57.,width=377 ] when @xmath220 , @xmath242 is a monotonically increasing function of @xmath57 . for each @xmath183 , it can intersect the parabola at one or two points , depending on the particular values of @xmath25 and @xmath26 . to determine the conditions for this to occur , we have first to determine the minimum of @xmath242 , which occurs when the expression under the square root in ( [ def_f_s ] ) is zero : @xmath254 this value sets a lower bound for the energies of all the solutions of eq . ( [ eigen_eq_delta=0 ] ) . for @xmath255 , the function @xmath242 takes the value @xmath256 and @xmath257 takes the value @xmath258 when @xmath227 , the minimum of @xmath259 occurs for negative values of the energy , that is , on the left ( and decreasing ) side of the parabola . if @xmath260 there is only one solution of eq . ( [ eigen_eq_delta=0 ] ) and it has positive energy . from ( [ f_s_min ] ) , this can happen only for @xmath207 , since @xmath261 . if @xmath262 there are _ always _ two solutions , one negative and one positive , because @xmath263 is always positive , no matter how large is @xmath264 , for a positive @xmath25 . furthermore , the positive and negative solutions are such that @xmath265 , since , when @xmath266 and @xmath26 negative there is a solution with @xmath267 , but this is an isolated solution out of the sturm - liouville problem we are considering . the considerations of the previous paragraph can be further elaborated by analyzing the behavior of the function @xmath268 @xmath269 according to what was stated in the previous paragraph , if @xmath270 one has two solutions for each @xmath183 , and if @xmath271 there is just one solution for each @xmath183 . to determine the signs of @xmath268 for each value of @xmath26 , we will need to know the values of @xmath26 which are roots of @xmath272 , or of the equation @xmath273 excluding the solution @xmath274 . applying descartes rule of signs to this last equation , we see that it can only have a positive real solution . the rule states that the number of positive real roots @xmath275 of an algebraic equation with real coefficients @xmath276 is never greater than the number of changes of signs @xmath277 in the sequence @xmath278 ( not counting the null coefficients ) and , if less , then differs from @xmath277 by an even number . this means that @xmath275 can take the values @xmath279 , @xmath212 . in this case , for positive @xmath25 , there is only one such sign change and therefore there is only a positive and unique solution , which we shall call @xmath280 . the function @xmath268 changes sign at @xmath281 and @xmath282 , being negative in the interval @xmath283 and positive outside this interval . therefore , according to what we have said before , there is only one positive - energy solution when @xmath284 and two solutions ( one positive and one negative ) when @xmath192 or @xmath285 . if @xmath274 , the two functions intersect at @xmath286 , but again , since this solution was excluded from the sturm - liouville problem leading to the eigenvalue equation ( [ eigen_eq_delta=0 ] ) , we consider only the harmonic oscillator solutions with energy greater than @xmath287 . ) for @xmath288 and three values of @xmath253 . the circle , square and diamond symbols mark the energy solutions for @xmath289 and @xmath290 , respectively . in the case of @xmath291 , the square root function for @xmath292 is also represented and the respective energy solutions are marked by the triangle symbols.,width=377 ] the graphical representations of @xmath257 and @xmath293 are shown in fig . fig3 for @xmath294 . we plot the square - root function for three values of @xmath295 ( the scaled @xmath26 ) , for a positive @xmath25 . we plot @xmath259 with @xmath296 for @xmath297 and with @xmath289 and @xmath292 for @xmath298 . the value of @xmath299 was chosen such that @xmath268 has a positive value for @xmath300 . there is always an infinite and unbounded set of positive - energy states with @xmath301 for any value of @xmath26 , and there are always negative states when @xmath26 is negative . note that increasing the value of @xmath183 does not change qualitatively these features , as can be seen from the plots of @xmath259 when @xmath26 is negative . in fig . [ fig4 ] it is clearly seen that for values of @xmath26 between zero and @xmath302 there are only positive - energy solutions . moreover , one can see from eq . ( [ fs_min = fp_min ] ) that the value of @xmath302 increases with @xmath25 , meaning that when @xmath303 one reaches the limit of the harmonic oscillator ( @xmath304 ) where there are only positive bound - state solutions with @xmath305 . thus , the existence of the pseudoscalar potential gives rise to negative - energy solutions coming from the dirac sea . for @xmath306 these kind of solutions always exist . we excluded the isolated level with @xmath289 for negative @xmath26 . note also that for decreasing negative @xmath307 the positive - energy solution for @xmath289 goes to @xmath308 , which was the case we discarded in the discussion of the dirac oscillator in the previous section . when @xmath309 there is an infinite set of degenerate solutions with @xmath255 , because in this case the intersection point of @xmath259 with the negative - energy branch of @xmath250 is the initial ( minimum ) point of the square - root function , which of course does not change with @xmath183 ( see fig . fig3 ) . however , when @xmath255 , the value of @xmath178 defined in ( [ def_k ] ) is zero , and thus , from ( [ def_lambda ] ) , @xmath310 and therefore these solutions do not correspond to bound states ( see eqs . ( [ psi+_delta=0_oh ] ) and ( [ psi-_delta=0_oh ] ) ) . for @xmath311 the degeneracy is removed and we have a very high density of very delocalized states ( since @xmath312 ) which then decreases with increasing @xmath313 . in fig . [ fig4 ] we also plotted the parabola @xmath314 ( in scaled units @xmath315 ) which corresponds to the asymptotic values of the negative - energy states when @xmath316 and thus represents the set of lower energy bounds of the negative - energy solutions for each pair of @xmath25 and @xmath26 values . this parabola can be also seen as delimiting the region in the energy-@xmath26 plane in which @xmath216 , _ i.e. _ , where there are no bound - state solutions . this is related to the discussion in the previous section leading to eqs . ( [ limit_values_k1_p ] ) and ( [ limit_values_k1_n ] ) . as referred before , for a given energy @xmath317 and positive @xmath25 there are not bound states for values of @xmath26 in the interval @xmath318 . in fact , from what was said above in this section , one can be more precise and state that there are no negative bound states for @xmath319 . since @xmath302 is zero when @xmath189 and increases with increasing @xmath25 , this means that when @xmath189 ( the dirac oscillator ) there are no continuum states and when @xmath320 ( or when @xmath321 ) , all the negative - energy states are continuum states . as a function of @xmath253 . the @xmath289 level for negative @xmath26 and negative energy was not plotted because it corresponds to the isolated solution with @xmath322 ( @xmath267 ) . the value of @xmath323 is marked with a cross in the energy axis . the hollow circles mark the points ( @xmath274 , @xmath322 ) and ( @xmath282 , @xmath324 for which there are no bounded solutions . the dashed parabola represents the function @xmath325 , whose meaning is given in the text.,width=377 ] for @xmath224 the function @xmath242 decreases monotonically with the energy , reaching again its minimum when the expression under the square root in ( [ def_f_s ] ) is zero , setting now a maximum value for the energy of the solutions of eq . ( [ eigen_eq_delta=0 ] ) . the expressions for @xmath326 , @xmath327 and @xmath328 are the same as corresponding ones in the previous subsection , respectively eqs . ( [ e_min ] ) , ( [ f_s_min ] ) and ( [ f_p_min ] ) , except that , of course , now @xmath199 is negative . noting that @xmath293 at @xmath133 has a nonnegative value given by @xmath329 , we see that there is always an infinite and unbounded set of negative - energy solutions ( one for each @xmath183 ) with @xmath222 . for @xmath289 and @xmath330 , there is a solution with @xmath331 which was already discussed in the previous section . however , when @xmath224 , one can solve eq . ( [ isol_sol_delta=0 ] ) when @xmath133 and obtain a bound - state solution with both upper and lower components different from zero , if and only if @xmath332 . one can show that this corresponds to a @xmath289 solution given by the functions ( psi+_delta=0_oh ) and ( [ psi-_delta=0_oh ] ) taking the limit @xmath333 . in addition to that set of solutions , there can be another set of solutions only if the function @xmath334 ( see eq . [ def_f(k2 ) ] ) is zero or positive , as explained before . to determine the values of @xmath201 for which this happens , one has to compute the roots @xmath335 of @xmath336 . when @xmath224 , descartes rule of signs applied to @xmath336 states that one can have a positive root and two or zero negative roots ( this is done by changing @xmath337 in ( [ descartes_rule ] ) ) , in addition to , of course , the root @xmath338 . the two negative roots arise for a value of @xmath339 such that @xmath340 . in scaled units , this value is @xmath341 , corresponding to the double negative root @xmath342 . for @xmath343 there is only a positive root , while for @xmath344 there are two negative roots and a positive root . because @xmath334 changes sign at @xmath345 ( except when @xmath346 is a double root ) and @xmath334 is always positive when @xmath347 , this means that when @xmath343 there is only one range of values of @xmath201 where @xmath348 whereas when @xmath349 there are two ranges of @xmath201 values where this happens . within these ranges , only negative - energy solutions exist . when @xmath350 , the value @xmath351 corresponds to an infinite set of degenerate negative - energy solutions with scaled energy @xmath352 , which , as mentioned before , are not bound - state solutions . these features of the spectrum can be seen in figs . [ fig5 ] and [ fig6 ] . as a function of @xmath253 . as in fig . fig4 the three non - zero values of the roots of @xmath268 , @xmath353 , are marked with a cross in the energy axis . as in that figure , the hollow circles mark the point ( @xmath354 , @xmath322 ) and the three points where ( @xmath282 , @xmath355 ) . the dashed parabola represents the function @xmath356 ( see text).,width=377 ] as a function of @xmath253 . the conventions and meaning of the symbols are the same as in figs . [ fig4 ] and [ fig5 ] . the dashed parabola represents the same function as in fig . [ fig5].,width=377 ] in fig . [ fig5 ] the solutions with @xmath357 with @xmath358 are basically degenerate ( they follow closely the parabola @xmath359 ) and the @xmath289 solution is very close to them . since , as remarked before , on the parabola the value of @xmath178 is zero , these are very weakly bounded solutions . note that the @xmath289 positive - energy solution with @xmath360 goes to @xmath287 when @xmath361 , thus becoming the isolated solution of the dirac oscillator ( @xmath189 ) . this can also be seen in fig . [ fig6 ] , where one may also remark that there is a value of @xmath362 around @xmath363 for which the solutions for distinct values of @xmath183 and @xmath357 are almost degenerate . this point corresponds to the local minimum of @xmath268 , which , for @xmath364 , is very close to the double root @xmath365 . similarly as in the case of @xmath227 , the solutions with @xmath366 have infinite degeneracy , but they do not correspond to bounded solutions . again , for @xmath26 near @xmath302 there is a high density of very delocalized states . the parabola @xmath359 sets now an upper bound ( @xmath316 ) for the positive - energy states . the limits @xmath367 and @xmath347 correspond , respectively , to the harmonic oscillator solutions for negative - energy ( antiparticle ) states and to the dirac oscillator . in the first case there are no bound states for particles ( positive - energy states ) . as @xmath368 increases , positive bound states start to appear . to complete the study of solutions of equation ( [ eigen_eq_delta=0 ] ) we present now a plot of its solutions when @xmath26 is kept fixed and let @xmath25 vary . note that the existence of one or two solutions is governed by the sign of the function @xmath369 identical with @xmath268 , except that now @xmath26 acts as a parameter and @xmath25 is the variable . the critical value is now @xmath370 . in fig . fig7 we present a plot of the solutions for @xmath207 , when there are two possible values for @xmath371 , the roots of @xmath372 . apart the factor @xmath373 , this is just a quadratic equation in @xmath25 , so that it is easily shown that when @xmath374 it has no real solutions . this means that , since in this case @xmath375 , there are always two sets of solutions . note that now the points in the energy-@xmath25 plane for which @xmath376 are on the hyperbole @xmath377 , such that only in the region between the hyperbole branches , where @xmath178 is positive , bound states may exist . as a function of @xmath378 . as before , the hollow circles mark the energy solutions for @xmath379 . the dashed hyperbole represents the function @xmath356 ( see text).,width=377 ] as before , we can recognize the dirac oscillator and the harmonic oscillator limits , as one sets @xmath189 and @xmath380 , respectively . in this last case , one has the harmonic oscillator solutions for particles or antiparticles depending whether @xmath25 goes to @xmath381 or @xmath382 . one can again see that the energy levels of the dirac oscillator are symmetric about @xmath383 and that there are no scattering states . here this follows simply from the fact that @xmath189 is an asymptote of the hyperbole branches which separate the continuum region from the bound - state region . as referred before , the case @xmath6 can be obtained from the @xmath9 case by applying the charge - conjugation transformation . we recall that this transformation performs the changes @xmath384 , @xmath385 and @xmath386 , or , for harmonic oscillator potentials , the changes @xmath387 and @xmath388 . this can be seen if we solve the eigenvalue equation for @xmath6 , eq . ( [ eigen_eq_sigma=0 ] ) , for @xmath389 and plot the solutions for several values of @xmath362 . the result is shown in fig . [ fig8 ] . as a function of @xmath253 and @xmath6 . the conventions are the same as in fig . the dashed parabola represents now the function @xmath390 . @xmath391 stands for the quantum number of the solutions of eq . ( [ eigen_eq_sigma=0]).,width=377 ] if we compare this figure with fig . [ fig5 ] ( @xmath392 ) we see that the plots are identical if we reverse both the vertical and horizontal axes , _ i.e _ , if we reverse the sign of the energy and of @xmath26 , respectively . of course , the identification of the particle and antiparticle levels is also reversed , as it should be , since we are applying the charge - conjugation transformation . the existence of both positive- and negative - energy solutions in a system with @xmath6 can be relevant for nuclei . in the 3 + 1 dirac equation of relativistic nuclear mean - field theories there is a connection between the ( isoscalar ) vector and tensor potential . the latter is proportional to the derivative of the former @xcite . for the harmonic oscillator potentials used in this paper , this would give , using the notation of ref . @xcite , the relation @xmath393the constant @xmath394 can take values up to around 1.3 in nuclei using relativistic mean - field potentials of woods - saxon type @xcite . for harmonic oscillator mean - field potentials , there should be also an upper value for @xmath395 , and therefore the relation ( [ k_2_k_1_f_v ] ) sets a maximum for @xmath396 in nuclei . as was seen above , this is relevant to know whether there can be simultaneously nucleon and antinucleon bound states in nuclei . this would depend very much on the strength of the @xmath397 potential , which is given by @xmath25 . as is apparent from fig . [ fig8 ] , for small values of the scaled @xmath25 , @xmath398 , the maximum value of @xmath399 allowed by eq . ( [ k_2_k_1_f_v ] ) could be such that there would exist only positive - energy states . on the other hand , for stronger @xmath14 potentials , the relation ( [ k_2_k_1_f_v ] ) could allow both positive- and negative - energy bound states , as can be seen by applying the axis reversal described above to the @xmath9 spectrum shown in fig . [ fig6 ] . in sec . [ sec : rel_h_oscillator ] we have seen that the @xmath0 chiral transformation can also be used to get formally @xmath6 from @xmath151 solutions , involving , however , the unphysical transformation @xmath400 . this objection is overcome when we deal with ultrarelativistic particles , for which we may consider @xmath401 . in this case , the chiral transformation relates massless physical states with @xmath402 and a pseudoscalar potential @xmath22 to massless states with @xmath403 and @xmath404 , where @xmath405 can be an arbitrary potential . for harmonic oscillator potentials , one has @xmath406 and @xmath388 . since , as explained in sec . [ sec : rel_h_oscillator ] , the chiral transformation does not change the equations of motion , the energy of their solutions is not changed . this can be seen in figs . [ fig9 ] and [ fig10 ] , where are plotted , respectively , the solutions of the @xmath9 and @xmath407 eigenvalue equations , eqs . ( [ eigen_eq_delta=0 ] ) and ( eigen_eq_sigma=0 ) , when @xmath251 , for @xmath408 and several @xmath26 values . the basic features of these plots were already discussed in the previous section . as before , the @xmath383 isolated solution was excluded . as a function of @xmath253 for @xmath9 and @xmath251 . the conventions are the same as in fig . the dashed parabola represents the function @xmath409.,width=377 ] as a function of @xmath253 for @xmath6 and @xmath251 . the conventions are the same as in fig . the dashed parabola represents the function @xmath409.,width=377 ] we see that the spectra are the same apart from the @xmath388 change , which shows up as a reversal of the horizontal axis . for @xmath274 , the pure harmonic oscillator solution , both spectra are _ identical _ , with no negative - energy solutions and , of course , no nonrelativistic limit . note , however , that the @xmath9 and @xmath6 energy eigenstates _ are not the same _ , since @xmath0 interchanges the upper and lower components . this amounts to say that these states are not eigenstates of @xmath0 , which of course they could not be , since , because of @xmath1 , it does not commute with the hamiltonian ( [ hamiltonian_d_sigma_delta ] ) even with @xmath410 and @xmath161 . if @xmath103 would be zero , then , because one has either @xmath411 or @xmath6 , @xmath412 would have also to be zero , and we would have a free ultrarelativistic particle , which has a good chirality , _ i.e. _ , is an eigenstate of @xmath0 . in this work we studied the solutions of 1 + 1 dirac equation with a potential with the most general lorentz structure which leads to harmonic oscillator potentials in the second - order differential equation for the spinor components . this is achieved when the vector @xmath412 and scalar @xmath103 potentials are quadratic and satisfy either @xmath3 or @xmath4 , and the pseudoscalar potential is linear . we showed that there are no bound - state solutions in the nonrelativistic limit for @xmath11 and @xmath22 very small , meaning in this case we have an intrinsic relativistic problem . we analyzed in great detail , by use of a graphical procedure , the solutions of the transcendental eigenvalue equation in the case of @xmath9 , for all signs of the potentials and for both particle and antiparticle states . as referred in the introduction , there is an increasing interest in studying the behavior of antinucleon in nuclei , and in particular how well pseudospin and spin symmetries hold for nucleons and antinucleons respectively . we discussed in detail the isolated solutions , _ i.e. _ , solutions out of the sturm - liouville problem , as well as the critical points in the parameter space of @xmath25 and @xmath26 , the coefficients of the quadratic and linear potentials . near these points there is a high density of bounded states which are very extended in space . these critical points are also relevant to determine which values of @xmath25 and @xmath26 allow to have a spectrum of both fermion and antifermion bounded solutions simultaneously or just one of these kind of solutions . we established a connection to 3 + 1 dimensional relativistic oscillator harmonic potentials applied to nuclei , where there is a relation between those coefficients and concluded that , for a sufficiently strong @xmath14 potential , both kinds of solutions ( _ i.e _ , for both nucleons and antinucleons ) can coexist . we showed how to obtain the solutions for @xmath6 from the @xmath9 case , using the charge - conjugation and chiral transformations . in particular , we used the chiral transformation to show that , for massless fermions and zero pseudoscalar potential , the @xmath5 and @xmath11 solutions have the same spectrum . these conclusions are quite general and do not depend on the number of dimensions so that they remain true in 3 + 1 dimensions . also , most of the features of the 1 + 1 spectra presented in this work are also present in the 3 + 1 relativistic harmonic oscillator . indeed , since the 1 + 1 eigenvalue equations for @xmath9 and @xmath6 are very similar to the corresponding 3 + 1 equations , we can apply the same graphical methods to the 3 + 1 case , and therefore most of the analysis of the 1 + 1 case remains true with appropriate changes . for instance , the shape of the parabola in the energy-@xmath26 plane related to @xmath413 would depend now on the value of the spin - orbit quantum number @xmath19 . the function @xmath268 would also depend on @xmath19 , and therefore the roots @xmath302 and critical value @xmath414 would also depend on @xmath19 . for @xmath274 there would be the known harmonic oscillator @xmath415 degeneracy , produced by the u(3 ) symmetry studied very recently by ginocchio @xcite in connection to spin and isospin symmetry . finally , the chiral transformation enables us to switch between the so - called pseudospin ( @xmath6 and @xmath161 ) and spin symmetries ( @xmath5 and @xmath161 ) @xcite _ for any potential_. in the context of a study of the symmetries of a 3-dimensional relativistic harmonic oscillator , ginocchio already showed that one can transform the generators of the su(2 ) spin symmetry into the pseudospin generators by a @xmath0 transformation @xcite . we showed that this transformation allows us to relate the @xmath402 and @xmath403 problems for any potential @xmath405 , and , for massless fermions and zero pseudoscalar potential , also to conclude that the two energy spectra are the same .